June 2008
Subject to priorities defined by the Technical Council and the Presidium, the results of fibs work in
Commissions and Task Groups are published in a continuously numbered series of technical publications
called 'Bulletins'. The following categories are used:
category minimum approval procedure required prior to publication
Technical Report approved by a Task Group and the Chairpersons of the Commission
StateofArt Report approved by a Commission
Manual, Guide (to good practice) approved by the Technical Council of fib
or Recommendation
Model Code approved by the General Assembly of fib
Any publication not having met the above requirements will be clearly identified as preliminary draft.
This Bulletin N 45 was approved as an fib Stateofart report by Commission 4 in April 2004.
This report was drafted by Task Group 4.4, Computer based modelling and design, in Commission 4, Modelling
of structural behaviour and design:
Koichi Maekawa 10, 1, 3, 5, 6 (Univ. of Tokyo, Japan, CoConvener), Frank Vecchio 1, 3, 5, 6, 10 (Univ. of
Toronto, Canada, Coconvener)
Stephen Foster 2, 6, 3, 4, 5, 8 (Univ. of New South Wales, Australia, Editor)
Oguzhan Bayrak 4, 6 (Univ. of Texas at Austin, USA), Evan Bentz 3, 4 (University of Toronto, Canada),
Johan Blaauwendraad 9 (Delft Univ. of Technology, The Netherlands), Jan Cervenka 5 (Cervenka
5, 6, 10
Consulting, Czech Republic), Vladimir Cervenka (Cervenka Consulting, Czech Republic), Tetsuya
Ishida 6 (Univ. of Tokyo, Japan), Milan Jirasek 6 (Czech Technical Univ. in Prague, Czech Republic),
Walter Kaufmann 2 (dsp Ingenieure & Planer AG, Switzerland), Johann Kollegger 5 (Technische Univ.
7, 8 7
Wien, Austria), Daniel Kuchma (Univ. of Illinois, USA), Ho Jung Lee (SC Solutions, Inc., USA),
Giuseppe Mancini (Politecnico Torino, Italy), Giorgio Monti (Sapienza Universit di Roma, Italy) 4, 6,
3
Josko O!bolt 5, 6 (Univ. Stuttgart, Germany), Clemens Preisinger 5 (Technische Univ. Wien, Austria),
Enrico Spacone 4 (Univ. of ChietiPescara, Italy), Tjen Tjhin 8 (Buckland and Taylor Ltd. Bridge
Engineering, USA)
1, 2, 3 ...
Chapter number for which this member was the main preparing author
1, 2, 3 ...
Chapter number for which this member provided contributions
Full address details of Task Group members may be found in the fib Directory or through the online services on
fib's website, www.fibinternational.org.
Cover image: FE modelling of high strength squat shear walls (image courtesy of S. Foster)
Although the International Federation for Structural Concrete fib  fderation internationale du bton  does its
best to ensure that any information given is accurate, no liability or responsibility of any kind (including liability
for negligence) is accepted in this respect by the organisation, its members, servants or agents.
All rights reserved. No part of this publication may be reproduced, modified, translated, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without prior written permission.
First published in 2008 by the International Federation for Structural Concrete (fib)
Postal address: Case Postale 88, CH1015 Lausanne, Switzerland
Street address: Federal Institute of Technology Lausanne  EPFL, Section Gnie Civil
Tel +41 21 693 2747 Fax +41 21 693 6245
fib@epfl.ch www.fibinternational.org
ISSN 15623610
ISBN 9782883940857
Preface
In September 2000, fib Commission 4 established a task group (TG4.4) with the objective of
preparing a report for guiding design engineers in the safe use of computerbased analysis
procedures for design of reinforced concrete structures. The working party that first met in
Berlin 2001 brought together a group of highly regarded researchers from Europe, the
Americas, East Asia and Australasia with the objective of producing a document for use by
engineers with some background in numerical modelling. In the six years since work started
on this report, advanced models have continued to be developed; however, this report is not
about picking one model over another but, rather, how designers can use existing and future
models as tools in design practice, in benchmarking of their models against established and
reliable test data and in selecting an appropriate safety factor as well as recognising various
pitfalls.
Nonlinear computer analysis methods have seen remarkable advancement in the last half
century with much research activity in the manner of constitutive modelling of reinforced
concrete behaviour and in the development of sophisticated analysis algorithms. These
advancements are well documented in various stateoftheart reports and remain the subject
of intensive research today. Linear and nonlinear analysis methods, combined with plasticity
design processes, and with local detailing methods such as strutandtie modelling, can form
the basis of design of new, complex, structures that are not easily dimensioned using other
rational design methods.
The stateoftheart in nonlinear finite element analysis of reinforced concrete has progressed
to the point where such procedures are close to being practical, everyday tools for design
office engineers. No longer solely within the domain of researchers, they are finding use in
various applications; many relating to our aging infrastructure. Nonlinear computer analysis
procedures can be used to provide reliable assessments of the strength and integrity of
damaged or deteriorated structures, or of structures built to previous codes, standards or
practices deemed to be deficient today. They can serve as valuable tools in assessing the
expected behaviour from retrofitted structures, or in investigating and rationally selecting
amongst various repair alternatives. Nonlinear finite element analysis procedures are also
proving particularly valuable in forensic analyses. In the near future, they will likely form the
main engine in computerbased automated design software, although in a form likely invisible
to the user.
This report provides an overview of concepts and techniques relating to computerbased
modelling of structural concrete. It attempts to provide a diverse and balanced portrayal of the
current technical knowledge, recognizing that there are often competing and conflicting
viewpoints. The report is written primarily for the benefit of the practicing engineer, rather
than as a stateoftheart for researchers, concentrating more on practical application and less
on subtleties in constitutive modelling. To the members of the working party, our sincere
thanks for the extensive and voluntary work undertaken over an extended period to get this
report competed.
14 November 2007
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures iii
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Contents
1 Introduction 1
1.1 Preamble 1
1.2 Notation 2
1.3 Sample applications 2
(1.3.1 Kimberley!Clark warehouse 1.3.2 Sleipner A offshore platform
1.3.3 Frame corner 1.3.4 Base slabs in LNG storage tank)
1.4 The question of accuracy (1.4.1 Reasons for caution) 20
1.5 Challenges remaining 27
1.6 Objectives 29
1.7 Scope of report 30
1.8 References 30
2 Design using linear stress analysis 33
2.1 Introduction 33
2.2 Membrane structures 34
(2.2.1 Notation 2.2.2 General 2.2.3 Reinforcement in one direction
2.2.4 Isotropically reinforced panels 2.2.5 The general solution
2.2.6 Some comments on the angle ! 2.2.7 The design concrete
compression strength, fcd. 2.2.8 Example Design of a reinforced concrete
squat shear wall)
2.3 Slabs and shells 52
(2.3.1 General 2.3.2 Stress resultants 2.3.3 Equilibrium, stress
transformation and boundary conditions for slabs 2.3.4 Normal moment
yield criterion for slabs 2.3.5 Sandwich model for the dimensioning of shell
elements 2.3.6 Dimensioning of slab and shell elements in design practice
2.3.7 Example 1 2.3.8 Example 2)
2.4 3D solid modelling 70
(2.4.1 Introduction 2.4.2 Background 2.4.3 Application to reinforced
concrete 2.4.4 Reinforcement dimensioning for 3D stresses ! example 1
2.4.5 Reinforcement dimensioning for 3D stresses ! example 2)
2.5 References 78
3 Essential nonlinear modelling concepts 83
3.1 Introduction 83
3.2 Nonlinear concrete behaviour 84
(3.2.1 Concrete in compression 3.2.2 Concrete in tension
3.2.3 Modelling of tension stiffening 3.2.4 Modelling of concrete cracks
3.2.5 Modelling of reinforcement)
3.3 Nonlinear concrete modelling framework 98
(3.3.1 Elasticity 3.3.2 Plasticity 3.3.3 Damage 3.3.4 Mixed models
3.3.5 Discrete modelling frameworks)
3.4 Solution methods 102
(3.4.1 Newton!Raphson method 3.4.2 Modified Newton!Raphson
method)
3.5 Precision of nonlinear concrete FE analyses 104
3.6 Safety and reliability 105
3.7 Statistical analyses 114
3.8 Concluding remarks 115
3.9 References 115
4 Analysis and design of frame structures using non!linear models 121
4.1 Introduction 121
4.2 Notation 122
iv fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures v
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
vi fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
1 Introduction
1.1 Preamble
Computerbased analysis procedures for reinforced concrete structures, and finite element
analysis procedures in particular, have seen tremendous advancement in the last halfcentury.
Much research activity has occurred in the manner of constitutive modelling of reinforced
concrete behaviour and in the development of sophisticated analysis algorithms. These
advancements are well documented in various stateoftheart reports, and still remain the
subject of intensive research.
Occurring at the same time, and no less significant, has been the accelerated development of
computing technology and hardware. Data compiled by Bentz (2006), shown in Figure 1.1,
provide a clear measure of the exponential growth in computing power in recent years. Shown
is the time required to conduct a nonlinear shear analysis of a prestressed Tbeam using a
layered beam element algorithm. It is seen from the graph that, in 25 years, computing speed
has increased by five orders of magnitude. Analyses that required several days of CPU time
on supercomputers two decades ago run in minutes on personal desktop computers today. The
advent of powerful and relatively inexpensive computers has greatly expanded the size and
complexity of problems that can be analysed, and has greatly reduced the computer time
required for their solution.
Computer Performance
Pentium 4
100 15 sec
Pentium III
Pentium Pro
10 3 min
Estimated SPECint95
Pentium
1 30 min 80486
80386
0.1 5 hours
80286
0.001 20 days
8080 CFT MCFT Kobe DSFM
published published Earthquake published
7 months
0.0001
1970 1975 1980 1985 1990 1995 2000 2005
Year of CPU introduction
The stateoftheart in nonlinear finite element analysis (NLFEA) of reinforced concrete has
thus progressed to the point where such procedures are close to being practical, everyday
tools for design office engineers. No longer solely within the domain of researchers, they are
finding use in various applications; many relating to our aging infrastructure. NLFEA
procedures can be used to provide reliable assessments of the strength and integrity of
damaged or deteriorated structures, or of structures built to previous codes, standards or
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 1!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
practices deemed to be deficient today. They can serve as valuable tools in assessing the
expected behaviour from retrofitted structures, or in investigating and rationally selecting
amongst various repair alternatives. In situations that have not turned out well, NLFEA
procedures are finding applications to forensic analyses and litigations that follow. In the near
future, they will likely form the main engine in computerbased automated design software,
although in a form likely invisible to the user.
1.2 Notation
Also discussed below are two additional examples in which nonlinear finite element analysis
procedures were used to aid in the design of structures. The first involves a prestressed
concrete frame; the second, the base slab in an underground liquid natural gas (LNG) storage
tank. Again, both examples serve to illustrate the usefulness of advanced analysis procedures
in solving difficult design problems.
!
2 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The KimberleyClark building was built in 1944, in Niagara Falls, Canada, in accordance
with thencurrent building codes. The building was a simple fourstorey structure with
basement, having plan dimensions of approximately 38 x 36 m (see Figure 1.2). The structural
system employed was primarily a reinforced concrete flat slab system with six bays in each
direction. The centretocentre column spacing was approximately 6.25 m in the NS
direction, and 5.85 m in the EW direction. Exterior columns were rectangular with haunches,
while interior columns were circular with capitals. Column and capital diameters decreased
with increasing elevation; the columns supporting the third floor were 450 mm in diameter
with 1.5 m diameter capitals. The floortofloor height ranged from 3.35 m to 3.65 m. Exterior
walls were constructed of brick masonry, and stair/elevator shafts were located at various
points around the perimeter of the structure.
The floor slabs were typically 200 mm thick. At the third floor level, the slab was thickened
by 150 mm around the perimeter, over a width of 1.3 m. The floor slabs were reinforced with
No.4 (13 mm dia.) and No. 5 (16 mm dia.) deformed bars. The reinforcement details were
consistent with a columnstrip/middlestrip design method, as shown in Figure 1.3. Similar
reinforcement patterns and amounts were used in both directions.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 3!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The thirrd floor of the warehoouse becam me the site for the storrage of druums of nick kel pellet
beginninng in midD December 1977. The drums, on wooden paallets, weree transporteed to the
third flooor by a freiight elevatoor, and then moved to the
t storage location
l by forklift trucck. Each
drum weighed
w 2255 kg, and there were eiight drums to t a pallet. The palletss measured 915 mm
square. In general, the pallets were stackked side by side, two high,
h giving a live floorr load of
about 43.1 kPa. Cooincidentallly, an alternnatebay loaading patterrn evolved as one bay was left
free to facilitate foorktruck acccess. Anotther perhap ps relevant detail is that the floorrs below
!
4 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
housed a paperprroducts mannufacturingg plant. Wiith such opperations, thhe dangers of dust
explosioons and firee are an everrpresent cooncern.
F
Figure 1.4: Looading and co
ollapse patternn.
The conntention off the paper manufacturring firm was w that thee stored nicckel overloaaded the
floor, resulting
r inn its collappse and coonsequently y sparking an explossion and fire.
fi The
propriettors of the nickel opeeration arguued that an explosion and fire inn the plant was the
primaryy cause of collapse;
c thaat is, the inttense heat from
f the firre instantly weakened the
t floor
above and
a led to itss failure.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 5!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
column capitals and reduced clear spans, beam elements representing the drop panels were
used. The four stairwells, located roughly at the corners of the warehouse floor plan
(Figure 1.2), likely influenced the behaviour of the floor slab as will be discussed. To estimate
their lateral stiffness, the stairwells were considered to be cantilevered from the basement
floor. All constitutive modelling was based on the models of the Modified Compression Field
Theory (Vecchio and Collins, 1986). Vecchio and Collins (1990) provide complete details of
the modelling.
The nonlinear analyses indicated that the ultimate capacity of the floor in the vicinity of the
collapse was approximately 45.5 kPa (superimposed load), as shown in Figure 1.5a. Including
the dead load, the total capacity was 50.3 kPa. Note that the estimated floor capacity is only
slightly higher than the estimated floor loading at the time of collapse, but is 7.6 times the
value of the specified floor capacity.
The most significant difference between linear and nonlinear analyses, in this situation, lies in
the fact that the nonlinear analyses accounted for the net elongation that typically occurs in
reinforced concrete elements in flexure (see Figure 1.5b). Tensile strains on the cracked face
of an element are normally much larger than the compressive strains on the opposite face,
particularly if the reinforcement is yielding. Hence, the tendency is for a slab to elongate as it
flexes under the applied transverse loads. When the slab is prevented from elongating, as it is
in this case by the columns, stairwells, elevator shafts, and floors above and below, a
compressive thrust is induced in the slab. For reinforced concrete elements, axial compressive
forces generally serve to increase flexural and shear capacity. This behaviour, known for
many years, is commonly referred to as compression membrane action.
In the case of the KimberleyClark Warehouse, a linear elastic analysis of the third floor
subsystem predicts very little gain in the axial compressive force as the floor load is
increased. The flexural capacity at the midspan of the floor is exceeded at approximately
21.5 kPa live load. Allowing for partial moment redistribution, as permitted by ACI 318
(1983), the estimated ultimate floor capacity increases to 25.4 kPa. If complete moment
redistribution is considered, the calculated floor capacity increases to approximately 31.1 kPa.
This remains well below the observed capacity at failure. The nonlinear analyses, on the other
hand, show a much more pronounced increase in axial force with increasing floor load; see
Figure 1.5c. The proper consideration of membrane action, only possible through nonlinear
analyses, was instrumental in achieving an accurate estimate of the load capacity of this
structure.
Further analyses were conducted considering the effects of elevated temperatures if a fire
were present below the heavily loaded third floor slab. A temperature of 1000!C was assumed
to be acting instantaneously on the bottom surface of the slab, and standard unidirectional
heat flow analyses were used to compute the progression of the thermal gradient through the
slab. Reductions in materials strength and stiffness, arising from the elevated temperatures,
were considered for both concrete and steel. The analyses indicated that collapse would have
occurred within 2 to 3 minutes from the onset of the fire. Hence, it was concluded that the
collapse could as likely have been triggered by a fire below weakening the floor as by floor
overload.
!
6 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 7!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The Condeep offshore platforms, of which Sleipner A was one, serve in the hostile
environment of the North Sea in water depths of up to 300 m. The construction of a typical
Condeep platform starts in a large drydock area, and is moved to a sheltered fjord as the
construction progresses. In later stages, the base structure is floated out to sea, then partially
submerged and mated with the deck structure. After deckmating, the completed structure is
towed to its offshore site and lowered to a final position on the sea floor. A critical factor in
the design of the base structure is the thickness of the cell walls. If the walls are too thin, they
may fail under the very high water pressures to which they are subjected during deckmating.
If the walls are too thick, the structure will not float, or will not be hydrostatically stable
during the towing operations. The margin between the two is narrow.
The gravity base structure of the Sleipner A offshore platform, which was 110 m high,
consisted of a cluster of 24 buoyancy cells. Four of these cells extended upward to form shafts
connecting to the deck structure. Details are given in Figure 1.6. The exterior walls of the
cells were circular, with radii of 12 m, whereas the interior walls separating the cells were
straight. The intersection of the interior walls formed a small triangular void called a tricell.
There were a total of 32 such tricells. Because the tricells were open at the top, they filled
with water once the tops of the cells were submerged. Hence, the walls of the tricells had to
resist a substantial hydrostatic pressure.
On August 23, 1991, the gravity base structure for Sleipner A was slowly being submerged as
part of the deckmating operation. The intention was to lower the structure until the base was
104 m below the ocean surface. However, when a depth of 99 m was reached, a loud
rumbling noise emanated from one of the drill shafts and water could be heard entering.
Within minutes, the structure began to sink in an uncontrollable manner. Moments after
disappearing below the surface, a series of implosions were felt as the buoyancy cells
collapsed. Evidence showed that the loss of the structure was attributed to the shear failure of
the wall of Tricell 23 adjacent to drill shaft D3. At failure, this 550 mm thick wall was
resisting a 65 m head of seawater, resulting in a pressure of about 655 kPa.
The reinforcement details in the tricell walls near the failure location, described in detail by
Collins et al. (1997), are summarized in Figure 1.7. In general, a grid of horizontal and
vertical bars was provided near each face of each wall of the tricell. For approximately the
bottom third of the height of the cell walls, 12 mm diameter stirrups were provided at a 170 x
170 mm spacing. In the middlethird height of the walls, the stirrup spacing was increased to
170 x 340 mm spacing. Just below the location of the failure, the stirrups were terminated and
the walls contained no shear reinforcement thereon up. Another detail of note is the provision
of a Theaded bar placed across the throat of the tricell joint. This 25 mm diameter bar was
about 1.0 m long and had steel plates welded on its ends to provide anchorage.
The failure of the Sleipner base structure, which involved a total economic loss of about $700
million USD, was probably the most expensive shear failure on record.
!
8 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 9!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
10 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
To deveelop a betteer understannding of thee factors in nfluencing thhe failure oof the tricell wall, a
series of
o threedim mensional nonlinear
n fiinite element analysess were unddertaken. Th he finite
elementt mesh usedd for the innvestigation is shown in i Figure 1.8. Because of symmettry, only
onesixtth of the trricell needed to be moodelled. Thee one elem ment thick m mesh contaiined 342
brick ellements (244 dof) and 338
3 wedge elements
e (1
18 dof). All constitutivve modelling g was in
accordaance with thhe Modifiedd Compressiion Field Th heory. In the analyses, the horizon ntal axial
force inn the tricell wall was heeld constantt at 5000 kNN/m while the
t hydrosttatic pressurre on the
inner faace of the wall
w was inccreased until failure. In n addition, a constant vvertical commpressive
stress of 7 MPa waas assumed acting in thhe walls. Full details of o the finitee element modelling
m
are disccussed by Coollins et al. (1997).
F
Figure 1.8: Fiinite element model
m of tricelll.
The dessigners of thhe structuree were interrested in leaarning how the strengthh of the triccell wall
would have
h changeed if the stirrrups, whichh had been terminated just below the failure location,
l
had beeen continuedd up the wall. They weere also keen n to know howh the lenggth of the Theaded
T
bar in thhe throat off the joint reegion influenced the faiilure load. To
T answer tthese questions, two
series of nonlinear finite elem ment analysees were cond ducted. In one
o set the wwalls were assumed
a
to contaain 12 mm diameter
d stiirrups at 1700 x 340 mm
m spacing, while
w for thee other set the
t walls
were asssumed to contain
c no stirrups. Within
W each set, the lenngth of thee Theaded bar was
varied from
f zero (i.e., no bar)) to the maaximum posssible lengthh (1.5 m). T The results of these
studies, summarizeed in Figuree 1.9, indicaate that the tricells couuld have resisted an ad dditional
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 11 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
20 m off water heaad if either the stirrupss had been continued up u the walll or if the Theaded
T
bars haad been aboout 500 mm m longer. The structu ure would have survivved in eith her case.
Changes were madde accordinngly to the design of the replaceement struccture. On April A 29,
1993, thhe new conncrete gravvity base strructure wass successfuully mated w with the deeck, and
Sleipnerr A2 was reeady to be toowed to seaa. It remainss in service today.
t
Figgure 1.9: Influuence of stirruups and lengthh of Theaded bar on prediccted failure prressure for triccell.
1.3.3 F
Frame corn
ner
Problem
m descriptioon
Frame corners
c are a typical deesign probleem where reinforcemennt detailingg plays an im mportant
role. Thhe region ofo a frame corner is subjected to t a compleex stress sttate and caannot be
analysedd by beam theory. Connsequently, it is design ned by emppirical rules which typiically do
not cover specific situations. The
T followiing analysiss was perfoormed withiin the project of the
newly built
b Brixeen Universiity (structuural designeed by Berggmeister, B Brixen). Th he frame
structurre is a part of the lectture auditorrium buildiing and is visually exxposed on the t back
elevatioon with a laarge glass facade
f creaating a signiificant aesthhetic impreession. Therrefore, a
stringennt requiremeent on the crack
c width limit was im mposed. Thhe girder is pprestressed
d and the
cable annchors signnificantly innfluence thee stress statee in the fraame corner. The reinfoorcement
detailingg was an immportant isssue in this project.
p Thee model is reduced to a section in ncluding
the symmmetrical haalf of the giirder with the
t floor slaab and partss of columnns; see Figu ures 1.10
and 1.111. Numerical analysiis based onn the work k of Cervennka (1999) was used for the
reinforccement detaailing optimiisation.
Numeriical model
The moodel developpment was based on thhe analysis of the wholle 5storey frame, Figu ure 1.10.
At firstt the question of model complexxity was co onsidered: How
H far doo we have to
t go to
capture the structurral behaviouur realisticaally? In this case two models
m weree considered
d; a two
!
12 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
dimensional model, based on a plane stress state, and a three dimensional model with a fully
3D stress state. Obviously, the choice of model has a strong impact on calculation time and
objectivity of results. While a 2D model is simpler and less time demanding, it cannot
appropriately capture the confinement effect in the anchoring region of the prestressing
cables. On the other hand, the 3D model is more realistic, but it requires more time for data
preparation and calculation. In this case it was decided to use the 2D model for global
information on the structural behaviour and for parametric studies relating to structural
optimisation. The 3D model was used for a detailed analysis of selected cases.
The numerical model reflected in a realistic way the geometry and reinforcement. It included
three types of reinforcing models: smeared reinforcement (stirrups and slab), bars (main
girder and column reinforcement), prestressing cables, see Figure 1.11.
Figure 1.10: Frame structure (left) and frame corner (right) dimensions (in m) and loading.
Figure 1.11: Reinforcing by ordinary bars (left) and prestressing cables (right).
Constitutive models used for concrete were of two types: For plane stress analysis the
damagebased model (Cervenka, 2002), and for 3D analysis the fractureplastic model
(Cervenka and Cervenka, 1999) were used. Both models employ the same approach for
tensile cracks, which is based on smeared cracks, crack band and fracture energy. Hordijks
exponential function is applied for the crack opening law. Compressive response in plane
stress is based on orthotropic damage with the failure function according to Kupfer. In 3D
analysis, the plastic flow theory with a MenetreWillam surface is utilized. The postpeak
softening is employed in both tension and compression responses.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 13 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Serviceability analysis
The prime objective of the analysis was to determine the crack width under service load in the
frame corner. It was difficult to calculate these cracks with usual beam theory and therefore a
FEbased simulation was employed. Crack patterns resulting from the analysis are shown in
Figure 1.12 for both models. In the 2D model the maximum crack width in the frame corner
was 0.28, mm and in the 3D analysis it was 0.15 mm. Both were calculated assuming a crack
spacing of 100 mm. Both crack patterns are very similar in locations and directions. It is
interesting that the 3D model gives smaller crack width. This is explained by the capability of
the model to capture the variable crack width in the third direction and thus producing smaller
crack width near the surface, where the reinforcement is located. The distribution of the
principal compressive stress in the corner in 2D analysis is shown in Figure 1.13.
2D model 3D model
Figure 1.12: Crack patterns under service load.
!
14 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The crack spacing must be also examined. The crack model is based on a fracture mechanics
approach, and the crack width within this frame is independent of the element size. The crack
spacing develops naturally as the strains localize. This works well for single discrete cracks.
However, in reinforced concrete with several parallel cracks of similar thickness, one main
crack need not develop due to the restraining effect of reinforcement. Then the crack width is
a product of crack strain and element size, while the latter represents the assumed crack
spacing. In the above case this spacing was 100 mm and represented a conservative estimate
of a real spacing. Thus, another crack width could be obtained by a simple recalculation for a
different spacing.
It should be realized that the above crack model includes the effects of strain softening in
cracks and also the socalled tension stiffening due to the contribution of uncracked and
partially cracked concrete regions, while considering the real geometrical arrangement of the
reinforcement. In the given case of design, the accuracy of the model cannot be easily
examined and we have to rely on a separate validation through known experimental data, for
which the tests of Hartl (1977) and Braams (1990) were used.
Ultimate analysis
After service load conditions were examined, the loading was further increased and the
ultimate failure load and corresponding failure mode were evaluated. The full analysis was
performed for two types of concrete postpeak behaviour in compression. The first type,
described as brittle, utilizes linear softening according to van Mier (1986), which can be
considered as a lowerbound behaviour. The second type was a perfect plasticity with
horizontal plateau after peak, which clearly overestimates ductility. Response for both models
and two types of compression softening can be seen from loaddisplacement diagrams in
Figure 1.14.
The brittle model in 2D shows a local instability at the load of 3.2 times the service load. At
this stage the compressive failure of the cracked concrete in the frame corner is initiated. Note
that this region is under a strong local effect of the prestressing tendon anchor. The corner
rotation due to brittle failure causes a temporary load decrease, then the load is again
increased until the bending capacity in the midspan is exhausted. The ultimate load of 63.8
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 15 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
kN/m only slightly exceeded the first peak load of 58.6 kN/m before the corner fails. The
redistribution of internal forces was also confirmed by changes in moment distribution in
girder.
Similar behaviour was observed in the case of the 3D model. However, in this case the local
instability was less pronounced and the load could be significantly increased to the maximum
75 kN/m after the partial compressive failure of the frame corner at 46.6 kN/m. The higher
ultimate load in the 3D model is explained by the better ability of the corner to resist moments
when lateral confinement is included. In spite of the partial brittle failure of concrete in the
frame corner, the structure showed relatively large ductility. The midspan defection of 355
mm at failure in the 3D analysis was 14 times that at service load. The failure at the frame
corner can be observed in Figure 1.13b, where the localized compressive strains are indicated
near the tendon anchor.
In order to identify the source of limited ductility, both models were also analysed by models
with perfectly plastic behaviour in compression. In this case the response of the 2D and 3D
models are almost identical and the ultimate load is increased to 97 kN/m with very ductile
behaviour. It shows that if brittle failure of concrete in the frame corner is avoided, the
bending capacity can be fully utilized. However, the brittle response is considered more
realistic.
Interesting is also information about the computer time required for analyses as shown in
Table 1.1. Calculations were performed on a Pentium 4 (2.2 MHz). The loading was imposed
in load steps with a NewtonRaphson iteration scheme up to service load. Then the solution
method was changed to arclength and loading continued to failure. The arclength method
was capable of capturing the local instability. The convergence criterion was 1% error in
equilibrium, and was satisfied in all steps until the global failure.
Conclusions
The nonlinear analysis has proven as an efficient tool for optimising reinforcement detailing.
The threedimensional analysis was more helpful for assessment of crack widths in the frame
corner and of the failure mode. It better described the effects of confinement on concrete
behaviour. The twodimensional analysis could be considered as a safe estimate and, in the
view of considerably lower computer time requirement, it was used for parameter study. The
2D and 3D analyses showed different results in case of brittle response. In case of ductile
response the results were almost identical.
!
16 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Problem description
The nominal punching shear strength of RC circular slabs (3D) is larger than that of beam
shear (2D) because hoop reinforcement creates confining radial compression under flexure
and the shear spantodepth ratio becomes substantially smaller compared with 2D beams and
columns. As the nominal punching shear strength is sizedependent, this factor leads to higher
cost performance with reduced hoop and transverse reinforcement when RC circular slabs of
large scale are designed in shear.
When a circular slab is supported with some deformational constraint, the hoop effect is
thought to be reduced and the shear capacity is probably diminished. On the contrary, the
effective shear span to depth ratio will be smaller and the capacity will be increased. Since the
design formulae are based upon the experiments of simply supported slabs, they cannot be
directly applied to more generic cases without any special discussion. For designing the world
largest underground LNG tank, a new structural type was adopted, i.e., the base slab and
sidewall shells are monolithically combined; see Figure 1.15. Here, nonlinear finite element
analysis was applied for estimating the punching shear capacity of the wall confined slabs,
because FEM can be freely applied to any boundary condition; see Figure 1.16.
Figure 1.15: Circular slab in punching shear and axisymmetric 3D analysis (Akiyama et al., 1996; Maekawa
et al., 2003)
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 17 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 1.16: Conventional punching shear design formulas and FEM simulation (Kawabara, 2002).
First, a 3D nonlinear analysis was made of a largescale conventional circular slab in shear
and checked by the available shear strength formulas. As the hydrostatic uniform force is
applied to the slab, a volumecontrol method or arclength method are effective in computing
the softened failure states. After checking the validity of nonlinear analysis, a parametric
study was performed with different peripheral boundary conditions; see Figure 1.17. It was
found that the nominal punching shear strength does not drop but increases due to more
confined boundary conditions and hoop reinforcement effects expected in these cases. The
most current LNG storage tank under construction has been designed with the hoop effect and
successful cost savings were achieved. In this practical case, computer based modelling was
utilized to extend the applicability of design formula to more general boundary conditions.
!
18 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Detailing
Thus, several other design details were considered and the nonlinear finite element analysis
was conducted for verification of its safety performance under static design loads
(underground water and soil pressure); see Figure 1.18. Practitioners in both design office and
construction site could have the modified dimensioning and steel arrangement which satisfies
the design requirement of safety and waterproofing serviceability.
Figure 1.18: RC slabsidewall connection: linear analysis based arrangement of rebars (left) and nonlinear
analysis aided design (right) (Nakano, 2002).
Discussion
In the four case studies presented, the use of advanced analytical techniques served to greatly
aid in the understanding of the factors and mechanisms influencing behaviour. With the
warehouse structure, the puzzle from an analysis perspective was not what event finally
triggered collapse but rather how could the floor sustain such high overloads in the first place.
In the case of the offshore structure, the challenge was to rigorously and accurately capture
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 19 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
the influence of local reinforcement details in an analysis of predicted response. In both cases,
as was shown, conventional analysis/design calculations were largely inadequate in providing
an accurate assessment of load capacity. Advanced nonlinear analysis procedures, on the
other hand, provided much improved modelling capability.
The four structures examined, while in themselves insufficient to form the basis for a
definitive statement, also point to another finding. That is, as design procedures become more
refined, and as economic and technologic pressures increase, the margin for error diminishes.
In the bulky concrete structures constructed a halfcentury ago, there was often much
redundancy in the structural system and a large margin of safety inherent in the design and
construction procedures. Hence, a structure such as the KimberleyClark warehouse was able
to withstand service floor loads almost an order of magnitude higher than designspecified.
On the other hand, with modern precisionengineered structures of today, based on more exact
methods of analysis and design, the margin for error is small and the penalty for
miscalculation is high. In such an environment, the value of sophisticated analysis procedures
is great provided they have been adequately verified and are properly applied.
At the 1981 IABSE Symposium in Delft, a blind competition was organized involving four
panels tested in a comprehensive research program then underway at the University of
Toronto. The test panels were orthogonally reinforced, and subjected to uniform, proportional
and monotonically increasing stress conditions; seemingly a very simple problem to model
and analyse. The results of these panel tests were not disclosed prior to analysts submitting
their predictions of strength and loaddeformation response. Approximately thirty entries
were received, many from the leading researchers in the field at the time. Shown in
Figure 1.19 is the range of responses for Panel C, one of the better predicted of the four
panels. The analysis results submitted showed a wide variation in predictions of the panels
shear strength, and an even wider divergence in computed loaddeformation responses
(Collins, Vecchio and Mehlhorn, 1985). Clearly the collective ability to model nonlinear
behaviour of reinforced concrete, particularly in shearcritical conditions, was not well
advanced.
More recently, in 1995, the Nuclear Power Engineering Corporation of Japan (NUPEC)
staged a prediction competition involving a largescale 3D shear wall subjected to dynamic
cyclic loading (NUPEC, 1996). The flanged shear wall exhibited highly nonlinear behaviour
before sustaining a sliding shear failure along the base of the web. One facet of the
competition called for estimates of the ultimate strength, and corresponding displacement, of
the wall as determined from static pushover analyses. Again, over thirty sets of predictions
were received; the results are summarized in Figure 1.20. The predictions of strength, as a
group, showed better correlation than was seen with the Toronto panels; however, the
deformation estimates still showed large scatter. Nevertheless, it could be concluded that the
ability of NLFEA to accurately capture the behaviour of reinforced concrete had measurably
advanced. It should be noted, however, that this was not a completely blind competition since
some of the test results had been disclosed to analysts prior to the competition.
!
20 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 1.19: Delft Panel competition results (Panel C): a) Variations in predicted shear strength; b) Variations
in predicted loaddeformation response.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 21 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
1.4.1 R
Reasons foor caution
Figure 1.20: NUPEC Wall predictiion results: a)) Variations in n predicted laateral load cappacity; b) Varriations in
predictedd lateral displlacement at ulltimate load.
!
22 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Reinforced concrete structures, particularly in their cracked states, are dominated in their
behaviour by a number of secondorder mechanisms and influencing factors. Depending on
the particular details and conditions prevailing, a structures strength, deflection, ductility and
failure mode may be significantly affected by mechanisms such as: compression softening
due to transverse cracking, confinement effects, tension stiffening, tension softening,
aggregate interlock and crack shear slip, rebar bond slip, rebar dowel action, rebar
compression buckling, scale effects, and creep and shrinkage, to name a few. For each of
these, a number of diverse formulations can exist. In the case of tension stiffening, for
example, the stiffening effect can be ascribed to a postcracking average tensile stress in the
concrete or, quite differently, to the loaddeformation response of the reinforcement. The user
of a NLFEA software must be aware of what mechanisms are likely to be significant in the
problem at hand, be certain that it is included in the analysis model, and have some
confidence that the model being used is reasonably accurate.
Experience required
Use of NLFEA for modelling and analysis of reinforced concrete structures requires a certain
amount of experience and expertise. Unlike, say, the use of plane section analysis techniques
to calculate the flexural strength of a beam cross section, the application is rarely straight
forward. Decisions made with respect to modelling of the structure and selection of material
behaviour models will have significant impact on the results obtained. Again, unlike sectional
analysis techniques, two analysts may obtain widely diverging results when modelling the
same structure using the same analytical model and the same software. Decisions made
regarding mesh layout, type of element used, representation of reinforcement details, support
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 23 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 1.21: Influence of compression softening model on computed loaddeformation response of Panel PV19.
As an illustration, consider the two simplesupported beams shown in Figure 1.22, tested by
PodgorniakStanik (1998); one with shear reinforcement and one without. Ten analysts, all
using the same software program and all previously experienced in its use, where asked to
independently provide predictions of the expected load capacity (Pu) and corresponding
midspan deflection (!u) for these shearcritical beams. Also requested were: the theoretical
section moment capacity (Mu) determined using hand calculations based on the common
rectangular stress block approach; the sectional shear capacity (Vu1) determined using the
Simple Method of the Canadian code specifications, which is essentially the standard 45
degree truss model; and the sectional shear capacity (Vu2) determined using the General
Method of the Canadian code specifications, ostensively a more accurate calculation
involving the consideration of compatibility conditions, inclination of the stress field, and
reduction in concrete strength due to transverse cracking. The results are summarized in
Table 1.2.
The calculations of moment capacity, using simple hand methods, were consistent amongst
analysts for both beams, showing a coefficient of variation (COV) of less than 3.5%.
Calculations of the shear capacity using code specifications produced larger variations. It is
interesting to note that the calculation of Vu1 (Simple Method) wasn't any less scattered than
!
24 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 25 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
that of Vu2 (General Method), due to some ambiguity present in the current code formulations
making interpretation of the code provisions subjective in certain situations. Also note that the
results were significantly more scattered for the beam without shear reinforcement (Beam
No. 2), as one might expect. Finally, let us examine the results obtained from the finite
element analyses. The estimates of load capacity (Pu) were relatively close to the code
calculated shear capacities, but showed significantly more scatter with COVs of about 17%
for the beam containing shear reinforcement, and 28% for the beam containing no web steel.
The estimates of deflection at ultimate load (!u) were widely divergent, in part because some
analysts were predicting a brittle shear failure and others a ductile flexural failure. It is worth
repeating that the same finite element program was used by the analysts, all experienced in its
use. Differences arose primarily from the selection of material behaviour models available
within the program library, and from the modelling of details such as reinforcement (smeared
or discrete) and support/loading conditions. Incidentally, the experimentally observed peak
load and corresponding deflection for Beam No. 1 were 672 kN and 20.7 mm; for Beam
No. 2, 370 kN and 5.9 mm.
NLFEA investigations invariably produce large quantities of data, typically spawning output
files of several megabytes in size or larger for each load stage. Information may be provided
on: stresses and strains at each integration point of each element, both with respect to local
and principal axes, both for the element and for the concrete component; nodal displacements;
sectional forces per unit width at each integration point; reactions; reinforcement stresses and
strains; stiffness matrix coefficients, and more. Considering that typical problems can involve
tens of thousands of degrees of freedom, the total amount of data quickly escalates to the
point where the use of postprocessors is virtually essential. Even then, the analyst must have
an awareness of what to look for and how to interpret it. Despite the use of sophisticated post
processors and graphics capabilities, there remains the possibility for misinterpretation.
Incomplete knowledge
We must accept that we still do not understand well, let alone have accurate models for, all
aspects of reinforced concrete behaviour (see Section 1.6 Challenges Remaining).
Applications of NLFEA should be done with a healthy degree of caution and scepticism.
Wherever possible, analysis software and models should be validated or calibrated against
benchmark tests involving specimens of similar construction and loading details, dependent
on mechanisms anticipated to be significant in the analysis problem at hand (as far as this can
be done). Wherever possible, results should be supported by analyses based on different
models or approaches.
Research philosophy
Lastly, it may be said that the research community, and associated technical committees, may
have failed to meet the needs of the profession in some respects. Many working in the area
have directed their efforts to developing sophisticated models and methods of analysis, in
many cases basing their work on esoteric models or rigorous application of classical
mechanics approaches not directly suited to reinforced concrete. Consequently, advancements
have been made in developing the NLFEA concepts and methodologies, but often with
reinforced concrete being merely the application. Unfortunately, reinforced concrete is a
complex and stubborn material that sometimes refuses to act according to accepted rules of
!
26 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
mechanics. Researchers might do well to refocus some efforts towards better understanding
and modelling reinforced concrete behaviour, with finite element analyses being merely the
tool. Certainly, however, there is room and need to advance on both fronts.
The first series involved five beams; all were 6000 mm in length, 1000 mm in depth, 300 mm
wide, reinforced with approximately 1.0 percent longitudinal reinforcement, containing no
transverse reinforcement, and subjected to a monotonically increasing load applied at the
midspan (Beams DB120, DB130, DB140, DB165 and DB180; see Figure 1.23 for beam
details). The only variable was the compressive strength of the concrete, ranging from 20
MPa to 80 MPa. These beams failed in a shearcritical manner upon the formation of the first
web shear crack. Current design code formulations would say that the shear strength of these
beams is directly proportional to a concrete contribution related to the tensile strength of the
concrete, which in turn is normally related back to the compressive strength (typically, the
tensile strength is taken as proportional to the square root of the compressive strength).
Hence, the 80 MPa beam would be expected to have a shear strength of close to double that of
the 20 MPa beam. Finite element analyses could also be expected to produce similar trends in
predicted strength, since the concrete tensile strength is the overriding parameter in most
analyses of such cases. Shown in Figure 1.24 are the loaddeformation responses measured
In the experiments. Note that there is little difference in the strengths and prepeak deflection
responses observed; certainly nothing approaching the doubling of strength anticipated. In
fact, the 80 MPa beam exhibits a shear strength lower than the 20 MPa beam. At work are
mechanisms related to smoothness of the fracture plane, aggregate interlock mechanisms, and
crack slip mechanisms.
A second series of beams from the same test program involved four beams similar in
dimensions and loading to the first (beams DB120M, DB140M, DB164M and DB 180M; see
Figure 1.23). The principal difference was that these beams contained the nearminimum
amount of shear reinforcement (0.08 %). Again, the compressive strength of the concrete
ranged from 20 MPa to 80 MPa. Shown in Figure 1.25 are the loaddeformation responses
recorded. Note that here, a small amount of shear reinforcement had a substantial influence on
the strength and failure mechanisms observed. Although the higher strength concrete beams
did exhibit a higher shear strength, there was still a good deal of perplexing behaviour
observed. See Angelakos et al. (2001) for a more complete description of the test program,
and a more thorough discussion of results and significance.
These two series of test results provide a stringent test of any NLFEA model. Wouldbe
analysts are encouraged to formulate all structural models, and select all constitutive models
and analysis parameters in advance of any preliminary computations, and to conduct the
analyses in a group and only once (i.e., forego any calibration or finetuning work).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 27 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
28 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
1.6 Objectivees
In addittion to ongooing work in n the develo
opment of im
mproved co onstitutive m
models and analysis
procedu ke NLFEA procedures more ameenable to
ures, researcch effort iss also requiired to mak
practicaal applicatio
on and to rem move somee of the conccerns regard
ding accuraacy and conssistency.
Specificc shortterm
m objectives should incllude the following:
" Providee design en
ngineers and
d nonexperrts guidancee in the appplication of NLFEA
proceduures to com
mmon and prractical prob
blems.
" Establish databank ks and ben nchmark prroblems, fo or various sstructure ty ypes and
loading
g conditionss, to facilitaate the valid
dating or callibrating off material beehaviour
modelss and analyssis procedurres.
" Recogn nize that divversity in th
he analysis procedures
p a
available is itself valuaable, and
encouraaged it. At the same tiime, work towards
t thee harmonizaation of con nstitutive
modelss and anaalysis apprroaches must m proceeed where appropriatte. The
identifiication and
d description n of relevaant behavioour mechannisms, and suitable
modelss for their reepresentatioon, should be accentuateed.
" Providee guidance on modelliing issues relating
r to assessmentt, rehabilitation and
forensic engineering; areas where
w NLFE EA procedu ures are findding signifiicant use
today.
" Work towards
t dev
veloping acccurate, con
nsistent and easytousee automated
d design
softwarre, with NL
LFEA being g one of thee viable means of perfo
forming bacckground
calculaations.
Task Grroup 4.4 (ComputerBaased Modellling and Deesign), of fib
b Commission 4 (Moddelling of
Structurral Behavioour and Deesign) was established to work towards theese objectiv
ves. This
report constitutes
c a initial efffort towardss achieving them.
an
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 29 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Chapter 2 discusses design procedures based on linear elastic modelling, the level at which
most codebased design specifications are formulated. Chapter 3 examines basic material
properties and behaviour models for concrete and reinforcement. Nonlinear modelling
concepts are explored in their application to plane frames and to two and threedimensional
structures in Chapters 4 and 5, respectively. Chapter 6 provides a more detailed and
comprehensive discussion of nonlinear constitutive models and behaviour mechanisms, and
their application to modelling. The use of benchmark tests and calibration procedures, a
necessary step in the use of any new modelling tool, is discussed in Chapter 7. Strutandtie
modelling concepts are reviewed in Chapter 8, special models for design of planar elements in
Chapter 9 and conclusions are discussed in Chapter 10.
1.8 References
ACI Committee 318. (1983), Building Code Requirements for Reinforced Concrete,
American Concrete Institute, Detroit, 111 pp.
Angelakos, D., Bentz, E. C., and Collins, M. P. (2001), The Effect of Concrete Strength and
Minimum Stirrups on the Shear Strength of Large Members, ACI Structural Journal, (in
press).
Akiyama, H., Goto, S. and Nakazawa, T. (1996), Shear strength of Large Reinforced
Concrete Circular Slabs Under Uniformly Distributed Load, Proc. of JCI, Vol.18, No.2,
pp.10971102.
Braam C.R. (1990), Control of Crack Width in Deep Reinforced Beams, Heron, Vol. 4,
No. 35.
Bresler, B., and Scordelis, A.C. (1963), Shear Strength of Reinforced Concrete Beams,
American Concrete Institute J., Vol. 60, No. 1, pp. 5174.
!
30 1 Introduction !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Collins, M.P., Vecchio, F.J., Selby, R.G., and Gupta, P.R., (1997). The Failure of an
Offshore Platform, Concrete International, Vol. 19, No. 8, pp 2835.
Hartl, G. (1977), Die Arbeitslinie Eingebetete Stbe bei Erst und Kurzzeitbelastung,
Dissretation, Universitt Innsbruck.
Kawabara, Y. (2002), Shear Capacity of Rigid Baseslab Connection for LNG Storage
Tanks, JSCE annual review.
Nakano, M. (2002), Inground LNG Storage Tanks Technological Trends and Latest
Technological Developments, Concrete Library of JSCE, No.39, June.
Okamura, H., and Maekawa, K. (1991), Nonlinear Analysis and Constitutive Models of
Reinforced Concrete, University of Toyko, ISBN 765515060, pp 182.
van Mier J.G.M (1986), Multiaxial Strainsoftening of Concrete, Part I: Fracture, Materials
and Structures, RILEM, Vol. 19, No.111.
Vecchio, F.J., and Collins, M.P. (1986), The Modified Compression Field Theory for
Reinforced Concrete Elements Subjected to Shear, Journal of the American Concrete
Institute, 83, No.2, pp. 219231.
Vecchio, F.J., and Collins, M.P. (1990). Investigating the Collapse of a Warehouse,
Concrete International, Vol. 12, No.3, pp 7278.
Vecchio, F.J. (2000), Disturbed Stress Field Model for Reinforced Concrete: Formulation,
ASCE Journal of Structural Engineering, 126, No. 9, pp. 10701077.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 31 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
[Blank Page]
!
!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
" Multiple load cases can be accommodated quickly and with a minimal change in the
input data.
" The greatest quantity of reinforcement is placed in the hightension regions. That is in
regions corresponding to the initial crack locations thus helping to control crack
propagation.
" No information is attained as to the collapse load of the structure (provided that
ductility demands are met, elastic analyses give a lower bound to the strength limit
design load)
" Detailing guidelines need to be established and followed to ensure ductility and
serviceability demands are met.
The above aside, however, it is not intuitively obvious how to interpret output from a finite
element analysis, in particular a linear analysis and how to detail the reinforcement. For
example, how do we detail the results of a shell element with 8 stress resultants into 4 layers
of longitudinal reinforcement plus transverse reinforcement? It is the objective of this chapter
to address this issue and to give guidance to the practicing engineer on some of the strengths
and on the limitations of using linear finite element modelling for reinforcement detailing.
Models for loaddeformation analyses are treated later in this report with the focus of this
section on classical yield conditions for membrane, slab and shell elements and structures
derived from plasticity theory. This allows for a straightforward reinforcement design based
on the applied loads. Furthermore, a consistent method to account for transverse shear forces
in membranes and slabs (Marti, 1990) will be outlined.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 33 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
34 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
2.2.2 General
In the output of linear elastic FE modelling of membranes the analyst is generally given
stresses and strains in a Cartesian coordinate system, which, for the analysis that follows are
chosen as collinear with the envisaged reinforcement directions. Whilst principal stresses are
of interest as they give the elastic load path, it is usually preferable to detail the reinforcement
along axes orthogonal to the axes of the structure or structural element being designed. For
example, Figure 2.1a shows normal and shear stresses on a 200 mm thick plane stress
membrane element, stress output that is typical for a linear FE analysis. From the normal
stresses it is not directly evident what quantity of steel reinforcement is needed to carry
tension stress or in what direction the steel should be placed. Plotting Mohrs circle the
principal stresses and their directions are calculated and given in Figures 2.1b and Figure 2.1c.
Assuming that the strength of the concrete is sufficient to carry the full compression stress but
carries no tension, steel reinforcement is needed such that *1 f y ' 8.5 MPa and is placed at
 ' 70.7 0 degrees to the global X axis. This requirement is satisfied with f ytd ' 435 MPa
and with 2layers of 20mm diameter bars at 160mm centres.
5 MPa ,
Y 8.5
M Pa
(20,10) 23 . 5
10 MPa o MPa
.'/0/12
20 MPa +n 8.5 MPa
23.5 MPa
X
3413 o
(5,10)
Whilst with this solution equilibrium is satisfied, it is not usually convenient to place
reinforcement aligned with the principal directions. Rather it is normally preferred to place
the steel orthogonal to the global axes. In this case the concrete and reinforcement stress
resultants must sum to the global stresses on the membrane, as depicted in Figure 2.2. In the
compression field approaches (such as for example the diagonal compression field theory of
Mitchell and Collins, 1974) the concrete is taken to carry no tension, that is + 1c ' 0 .
Figure 2.3 shows the relationship between the applied stresses, the concrete stresses and the
reinforcement. Using this approach it is seen that there lies an infinite number of solutions to
the stress field. Some of the more interesting solutions are discussed in brief, below.
Looking closely at the example presented in Figure 2.1, the designer may choose a stress
regime such that * x+ s ' 0 , that is to provide reinforcement solely in the global Y direction.
In this case the Mohrs circles are as shown in Figure 2.4. From the resulting geometry it can
be shown by equating the radii of the Mohrs circle of concrete stresses calculated from two
points on the circle (see Figure 2.4) that
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 35 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Y +y
1
*55+
+
y sy
1
+3
+x
= +
3
*55+
x sx

, xy X
*55+
x sx
,
$+556,555%
x xy
. .F
+3c +1c
555'4
+n
applied
stresses
concrete $+5567,555%
y xy
stresses
*55+
y sy
+y ,
Y
(+556,555)
x xy
,xy
R
+x +3c R +1c +n
+o
X
,xy (+5567,555)
y xy
*55+
y sy
Figure 2.4: Mohrs circles for reinforcing steel placed solely in the Ydirection
!
36 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
+ x2 8 , xy
2
+o ' (2.1)
2+ x
where + o is the normal stress at the centre of the Mohrs circle of concrete stresses and + x
and , xy are the Xnormal and shear stresses, respectively (shown in Figure 2.2). Given that
+ 1c ' 0 , the principal compression stress in the concrete may be obtained from + 3c ' 2+ o .
The geometry of the circles also shows that the stress in the Ydirection reinforcement
required for equilibrium is
* y+ sy ' + x 8 + y 7 2+ o (2.2)
Further examination of Figure 2.4 shows that Eqs 2.1 and 2.2 only give a solution provided
that the inequality + x 9 0 holds true.
Similarly it can be shown that for the case where reinforcement is placed only in the global X
direction (that is * y+ s ' 0 ), then provided that + y 9 0
+ 2y 8 , xy
2
+o ' (2.3)
2+ y
* x+ sx ' + x 8 + y 7 2+ o (2.4)
For the example presented in Figure 2.1, if it is desired to reinforce the element solely in the
Ydirection then the resulting reinforcement requirement (by Eqs. 2.1 and 2.2) is
* y+ sy ' 10 MPa . This requirement is satisfied with f ytd ' 435 MPa and with 2layers of
20mm diameter bars at 130mm centres. The minor principal stress + 3c ' 2+ o ' 725 MPa .
Thus, in the example presented, the solution of providing reinforcing steel in the Ydirection
only is less efficient than when reinforcing in the major principal stress direction. This is to be
expected but likely to be more practical.
One of the more interesting solutions is for the case of an isotropically reinforced panel. The
Mohrs circle for this case is shown in Figure 2.5. It is seen that the circle representing the
concrete stresses is a translation of the global stresses and that * x+ sx ' * y+ sy ' + 1 . The area
of reinforcement in each of the global X and Y directions is equal to that for the case where
the steel is placed in the direction corresponding to the major principal stress. Equilibrium can
be maintained by replacing the force in the major principal direction $F1 % with any two
concurrent forces $Fa and Fb % . Letting F1 act over a unit length and taking the forces Fa and
Fb as orthogonal, as shown in Figure 2.6, then
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 37 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
*55+555'5*55+
x sx z sy
,
$+556,555%
x xy
'F
$+5567,555%
y xy
*55+555'5*55+
y sy x sx
: Fb : F1
s:
1. c o
+b +1
1. sin
:
+a
Fa
F1 sin : F1 cos :
+a ' ' +1 +b ' ' +1 (2.6)
t. sin : t. cos :
where the angle : is as shown in Figure 2.6 and t is the thickness of the element. That is
+ a ' + b ' + 1 . As the angle : is arbitrary, all orthogonal reinforcement sets are valid
solutions with the steel stress in each direction equal the major principal applied stress $+ 1 % .
Looking again at the example of Figure 2.1, assuming that the strength of the concrete is
sufficient to carry the compression but carries no tension then steel reinforcement may be
placed in any two orthogonal directions (such as, for example, in the global X and Y axis
directions) such that * a+ sa ' *b+ sb ' 8.5 MPa . This requirement is satisfied with
f ydt ' 435 MPa and with 2layers of 20mm diameter bars at 160mm centres placed in each
of the X and Y directions.
The example of the isotropically reinforced plate may appear to contradict the solution
obtained when the steel is placed in the major principal stress direction without an equivalent
quantity of orthogonal reinforcement. Examination of Figure 2.6 shows that placing the steel
!
38 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
in the major principal direction corresponds to the particular solution : ' 0o giving Fa ' 0 ,
sin : ' 0 and leads to a breakdown of Eq. 2.6. This is the only condition for which this
occurs.
Using the theory of plasticity, Nielsen (1964, 1971) derived a set of general solutions for
isotropically and orthotropically reinforced plates. Clarke (1976), Clyde (1977),
Mller (1978), Marti (1979, 1980), Morley (1979) and many others have also made
significant contributions to the field. The cases examined above are particular solutions of a
general failure criterion. For the most general case the strength of a membrane is limited by
the yield surface expressed by the following criteria (Kaufmann and Marti, 1998):
2
Y1 ' , xy $ %$
7 * x f ytd .x 7 + x * y f ytd . y 7 + y ' 0 % (2.7a)
2
Y2 ' , xy $ %$
7 f cd 7 * x f ytd .x 8 + x * x f ytd .x 7 + x ' 0 % (2.7b)
2
Y3 ' , xy $ %$
7 f cd 7 * y f ytd . y 8 + y * y f ytd . y 7 + y ' 0 % (2.7c)
2
Y4 ' , xy 7 f cd 4 ' 0 (2.7d)
2
Y5 ' , xy $ %$
8 f cd 8 * x f ycd .x 8 + x * x f ycd .x 8 + x ' 0 % (2.7e)
2
Y6 ' , xy $ %$
8 f cd 8 * y f ycd . y 8 + y * y f ycd . y 8 + y ' 0 % (2.7f)
2
Y7 ' , xy $ %$
7 f cd 8 * x f ycd .x 8 + x f cd 8 * y f ycd . y 8 + y ' 0 % (2.7g)
where f cd is the design value of the concrete compression strength f ytd .x and f ytd . y are the
design strengths of the steel reinforcement in tension in the global X and Y directions,
respectively, and f ycd .x and f ycd . y are the design strengths of the steel reinforcement in
compression in the global X and Y directions, respectively. The yield surface given by Eqs
$
2.7a to 2.7g is plotted in the threedimensional stress space of + x , + y ,, xy in Figure 2.7 for a %
panel with * x ' 0.01 , * y ' 0.005 and f yd ' 500 MPa . In Eqs. 2.7ag, Y1 corresponds to
under reinforced elements where the steel in both the X and Y directions yields before failure
of the concrete, Y2 and Y3 correspond to web crushing failure and Y4 represents the noyield
condition of crushing in pure shear. The criteria Y5 , Y6 and Y7 are for plates in biaxial
compression.
Whilst the application of Eq. 2.7 is relatively straightforward it is not as easy to picture the
failure surface in the 3dimensional stress space. As the yield criterion is based on the
principal stresses we again revert to Mohrs circles of stress for an alternative graphical
solution. The case of an isotropically reinforced plate is used to demonstrate the relationship
between Mohrs circle and the general failure criterion given by Eq. 2.7. In Figure 2.8ad
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 39 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
10 10
, xy (MPa) 5
5 , xy (MPa)
0
0 25
20
0
5 15
10 10
+ y (MPa) 15 5 + x (MPa)
20 0
25
(a)
+ y (MPa)
5
+ x (MPa)
30 10
2 Regime 1
5
Regime 4
3
6
Regime 7
25
(b)
Figure 2.7: Yield surface for a reinforced concrete membrane with *x =0.01, *y = 0.005 and fyd = 500 MPa;
a) 3D View; b) plan view identifying yield regimes.
!
40 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
regime 4
regimes 2 & 3
;,; regime 1
;,; regimes
2&3
expanding
fcd /2
sliding
fcd /2
Mohr's
Mohr's
circle
circle
+3 *5f yd +n +3 +1 +n
fcd *5f yd
fcd
(a) (b)
sliding
Mohr's fcd /2
circle
+3 +1 +n +1 +n
*5f yd fcd fcd *5f yd
(c) (d)
.5*5f yd
,
R = fcd /2
R R
Region I
+n
Region II
(e)
Figure 2.8: Isotropically reinforced plate: ad) Mohrs circles defining the failure surface; e) admissible stress
envelope.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 41 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Mohrs circles are plotted for an isotropically reinforced plate. Starting from the case of equi
biaxial tension (that is +1 ' + 3 ' *f yd ) the circle expands with reducing + 3 until the point
where the radius reaches the limiting shear strength of f cd 2 (Figure 2.8a) with the
$
corresponding minor principal stress being +3 ' 7 fcd 7 * f yd . This phase represents the %
regime 1 shown in Figure 2.7. In the second phase the circle shifts towards the negative,
whilst maintaining a constant radius, until + 1 ' 0 and + 3 ' 7 f cd (Figure 2.8b). This
represents regimes 2, 3 and 4 in Figure 2.7. In the third phase, corresponding to regimes 4, 5
and 6 in Figure 2.7, the circle slides further to the negative and the reinforcing steel moves
into compression (Figure 2.8c). Finally, the circle collapses with the minor principal stress
remaining constant until the equibiaxial compression point is reached, that is the point
$ %
+1 ' +3 ' 7 fcd 8 * f yd , as shown in Figure 2.8d.
From the Mohrs circles shown in Figures 2.8ad the admissible stress field is plotted
(Figure 2.8e). For any solution to be admissible two criteria must be met. Firstly, the Mohrs
circle of the applied stresses must lie wholly within the circular region I, where region I is
centered on the normal stress axis and has a diameter of f cd . Secondly, region I lies wholly
within region II (refer Figure 2.8e). If these two criteria are met then the membrane is safe.
Note that in practice it is sufficient to show that any two diagonally opposed points on the
Mohrs circle of stress meet these criteria to demonstrate admissibility of the solution (for
example the stress pairs ( + x ,, xy ) and ( + y ,7, xy )).
For membranes subject to tension and/or shear the zone of greatest interest is that of regime 1
in Figures 2.7 and 2.8a and the design equation given by Eq. 2.7a. Plotted in Figure 2.9 is the
Mohrs circle of stress for the applied stresses and the stresses in the concrete for a membrane
in biaxial tension. From the upper shaded triangular region it is seen that the concrete stress
in the X direction is + xc ' + x 7 *x f ytd.x and therefore steel reinforcement is required such
that
where 7 90 o 9  9 90 o and  = 0 .
Similarly, from the lower shaded triangular region + yc ' + y 7 * y f ytd. y and therefore to
provide sufficient capacity the Y direction reinforcement is required such that
In Figure 2.9b the feasible domain of solutions to Eq. 2.7a is plotted. In this form the
relationship between Eq. 2.7a and Eqs. 2.8 and 2.9 is clear.
!
42 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
*555f
x ytd.x
,
$+5556,5555%
xc xy $+556,555%
x xy
*y fyd
+y585;,xy5cot
feasible domain
+x585;,xy tan
+y
0
0 +x *x fyd
Figure 2.9: a) Mohrs circle of stress for an underreinforced concrete plate; b) solution domain for under
reinforced element.
Equations 2.8 and 2.9 are identical to those derived by Nielsen (1964) using the theory of
plasticity 1. For Eqs. 2.8 and 2.9 to govern the behaviour (that is for the panel to be under
reinforced) the strength of the concrete must be such that f cd < 7+ 3c , that is the concrete
must not crush. Applying the first invariant of stresses the compression failure criterion can be
written as
Taking + 1c ' 0 (as per the compression field theory) and with f cd < 7+ 3c crushing failure is
not critical provided that
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1
Orthotropic membrane theory can be traced to Nielsen (1920) and Eqs. 2.8 and 2.9 to Leitz (1923) for
tan ' 1 [according to Nielsen (1964, 1971)].
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 43 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
For strength limit design of reinforced concrete membranes the strength criteria are fully
satisfied by Eqs. 2.8, 2.9 and 2.11 and are identical to the equations presented by Kaufmann
(1999) based on the yield surface defined by Eq 2.7.
Taking once more the example presented in Figure 2.1 and with  ' 65o , Eq. 2.8 gives
*x f ytd.x ' 1.45 MPa and * y f ytd. y ' 9.67 MPa. This requirement is satisfied with
f ytd ' 435MPa and with 2layers of 12mm diameter bars at 330mm centres placed in the
Xdirection and 2layers of 20mm diameter bars at 140mm centres placed in the
Ydirection. The required design concrete strength is given by Eq. 2.11 and is
f cd < 26.1 MPa .
In developing solutions using linear FE modelling it is important that the designer respect
the limitations of the concrete material. In a panel subject to a constant ratio of normal and
shear stresses (with at least one tensile principal stress) before cracking the stress field in the
concrete remains relatively elastic and the stresses in the steel reinforcement are negligible.
After cracking the tension stresses in the concrete rapidly reduce while those in the
reinforcing steel increase. Assuming that the concrete does not fail in compression then the
crack directions will remain relatively stable until yield of the steel in onedirection. After
yield in one direction the forces in the structural element are continuously redistributed to
balance the applied tractions until yield in all directions has occurred. Concrete panels,
however, have a limit on the amount of redistribution that is capable of being achieved. This
is referred to as ductility demand. As a rule concrete elements such as membranes and panels
should not be pushed far beyond that which is natural. Designers must critically examine
the load path being assumed to satisfy themselves that the element has a sufficient level of
ductility to meet the demands of the structure. To this end plots of principal stress directions
from linearelastic modelling can give the designer valuable insight. Care and engineering
common sense needs to be applied when designing any structural element but even more so
for the case of nonflexural membranes.
For a membrane element with its principal stresses in biaxial tension, steel reinforcement is
required in at least two nonconcurrent directions to carry the tension stresses with the
concrete taking the shear stresses. For the case of orthogonal reinforcement  ' 45o leads to
the lowest concrete strength requirement.
Many structural elements fall into the category of tensioncompression elements. Again
 ' 45o provides the lowest concrete strength requirement and in many cases provides for a
good solution. In some cases, however,  ' 45o may not be desirable. Figure 2.10a shows an
element in tensioncompression where taking  ' 45o leads to tension reinforcement in the X
direction and compression reinforcement in the Y direction. This, in some circumstances, can
lead to a significant requirement for reorientation of the cracks requiring a high ductility
demand. Figure 2.10b shows that there are limits on  if compression in the steel is not
wanted. Adding to Eq. 2.8 the condition *x f yd < 0 we get the limits
!
44 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
*55+
x sx tension *55+
x sx tension
, steel , steel
.5'5A4! $+556,555%
x xy $+556,555%
x xy
+3c +3c .
+1c
555'4 +n +1c
555'4 +n
$+5567,555%
y xy $+5567,555%
y xy
applied applied
stresses stresses
*55+
y sy
compression steel
(b)
(a)
o
Figure 2.10: a) Example showing the requirement of compression reinforcement for  = 45 ; b) no
compression steel solution by limiting .
for+ y 9 0 : 0 9 tan  ? 7 , xy + y
(2.12b)
for+ y < 0 : 0 9 tan  9 >
Choosing a value of  to meet the requirements of Eq. 2.12 ensures that the stresses in the
reinforcing steel do not go into compression. It is also observed for Eq. 2.12 that the lower
limit on Eq. 2.12a for + x 9 0 and the upper limit on Eq. 2.12b for + y 9 0 correspond to the
angles for unidirectional reinforcement in the x and y directions, respectively. In Figure 2.11
the solutions to Eq. 2.7a are plotted for the stresses of the example shown in Figure 2.1. The
figure shows that for  9 63.43o the X direction steel is in compression and for  @ 63.43o
the steel is in tension. The angle  ' 63.43o corresponds to the solution of reinforcement only
in the Y direction.
Lastly, for the case of elements subject to biaxial compression, the angle  should be matched
to that of the principal stress directions due to the applied loads. In this case reinforcing steel
may be used, if desired, to carry a proportion of the compression forces.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 45 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
40
+x = 20 MPa
+y = 5 MPa
30 , xy= 10 MPa
*y fyd (MPa)
Eqn. 2.7a
20
 = 63.43o
10
X dirn. reo. in X dirn. reo.
compression in tenision
 < 63.43o  > 63.43o
0
20 10 0 10 20
*x fyd (MPa)
Figure 2.11: Solution domain for element subject to plane stress.
It has been shown by a number of researchers (Robinson and Demorieux, 1977, Vecchio and
Collins, 1982, 1986, Miyakawa et al., 1987, Belarbi and Hsu, 1991, Pang and Hsu, 1992,
amongst others) that the disturbing effect of passing tension reinforcement through concrete
in compression weakens the concrete. In addition, as concrete strength is increased the
concrete becomes more brittle. To account for the imperfect assumption that concrete is a
rigid plastic material and that the ductility demands can be met an efficiency factor is
introduced to ensure that the concrete is not overstressed. There is considerable debate on the
efficiency factor for concrete in compression in Dregions and much has been written (see for
example the quite contradictory articles of section 4.6 Nielsen, 1999, and Foster and Malik,
2001). For a history of various efficiency factors methodologies the reader is referred to the
report on recent approaches to shear design by ASCEACI Committee 445 (1998).
where the 0.85 factor accounts for the variation between the insitu and cylinder strengths and
( is a disturbance factor. Equation 2.13 is of a generally suitable form for the design of
concrete plates subject to inplane stresses.
Whilst there are a number of variants of the efficiency factor relationship, the model by
Collins and Mitchell (1986) has generally withstood the test of time and has been
incorporated into the Canadian code of practice (CSA84, 1984). Based on the panel tests of
Vecchio and Collins (1982), Collins and Mitchell proposed that
!
46 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
1
(' ? 1.0 (2.14)
0.8 8 170)1
where ) 1 is the major principal strain normal to the direction of the compression field. The
transverse strain is required to be sufficiently large for the ductility demand to be met.
Adopting a minimum transverse strain of twice the yield strain of the steel reinforcement
gives
1
(' (2.15)
0.8 8 340 f y Es
where f y and Es are the yield strength and the elastic moduli of the steel reinforcement,
respectively. For 500 MPa grade reinforcing steel ( ' 0.6 and f cd ' 0.5 f c# .
Not all members or subelements of members are subject to significant transverse strains and
a reduction to the degree indicated by Eq. 2.15 is unjustified. For the case where transverse
strains are small and the concrete is essentially in uniaxial compression (transverse tensions
can be carried without cracking of the concrete) and for the case of biaxial compression then
the concrete is undisturbed by crossing tension stress fields. As a guide, if the major principal
stress due to the applied loads is such that +1 9 0.33 fc# then the disturbance factor maybe
taken as
Examples of members and regions of members where Eq. 2.16 applies include beams and
columns outside the shear zones and at edges of members where principal stresses and strains
are aligned with the boundary of the element.
Shear walls are common elements in buildings that are subject to inplane stresses. In this
example the squat shear wall shown in Figure 2.12 is designed to carry of factored design load
H * ' 4000 kN . The load may be applied in either direction as indicated in the figure. The
wall is 200 mm thick and is bounded by 500mm square stiff elements on both sides and a
1200mm by 400mm beam at the top of the wall.
A linearelastic finite element model of the wall was set up with the mesh shown in
Figure 2.13. The mesh consists of 30 by 8node isoparametric elements defined by 113 nodes.
The material properties were taken as E ' 25 GPa and B ' 0.18 for all elements. The model
was analysed with integration on a 3x3 Gaussian quadrature with the stress output on a 1
gauss point quadrature (that is at the centre of the element). The resulting normal stresses for
load case 1 are shown in Figure 2.13. For load case 2, the element stresses are a reflection of
the results of load case 1 about the YY axis (refer Figure 2.12).
For the reasons of ductility demand, discussed above, it is important to make an assessment of
the flow of forces in the wall and adjacent boundary elements. In the design that follows, for
the wall elements the minor principal stress direction for the concrete is taken as aligned with
the diagonal of the wall (refer Figure 2.12) giving  ' 58o . For the side face boundary
elements the principal stresses are taken as aligned with the global stress directions as shown
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 47 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
in Figure 2.12. The resulting stresses in the X and Y direction reinforcing steel are shown in
Figure 2.14.
For detailing of the reinforcement, the wall is divided into vertical and horizontal strips, as
shown in Figure 2.15. Reviewing the steel stresses in the horizontal (x) direction it is seen that
the maximum stress in each of wall strips 1X, 2X and 3X are of similar magnitude (5.45, 5.92
and 5.54 MPa, respectively). It is deemed therefore appropriate to maintain a single steel
stress in the X direction of 5.92 MPa for the full height of the wall. The area of steel
reinforcement per unit height of wall is given by
* f yd t h
As ' (2.17)
f yt & s
where t is the wall thickness, h is a unit height, f yt is the yield strength of the steel and & s is
the partial safety factor for the steel reinforcement (taken as & s ' 1.15 ). For 500 MPa grade
reinforcing steel and *x f yd ' 5.92 MPa, Eq. 2.17 gives As ' 2720 mm2 m . This is satisfied
with 2 layers of 16mm diameter bars at 150mm centres.
In the vertical (Y) direction the greatest stresses occur in design strips 1Y and 4Y (noting that in
strip 4Y the greatest stress occurs for load case 2). For 500 MPa grade reinforcing steel and
* y f yd ' 4.22 MPa, Eq. 2.17 (with h replaced by the unit width, b) gives an area of reinforcing
steel of As ' 1940 mm2 m . This is satisfied with 2 layers of 16mm diameter bars at 200mm
centres. In wall strips 2Y and 3Y the requirement is for * y f yd ' 2.22 MPa giving
As ' 1020 mm2 m and leading to 2 layers of 12mm diameter bars at 200mm centres.
For the boundary strips 1Y and 2Y the concrete stress angle is taken as 60 degrees giving a
maximum stress in the longitudinal (Ydirection) steel of * y f yd ' 7.5 MPa and with
f ytd ' 435MPa this gives As ' 4310 mm2 . This is satisfied with 8C28mm bars. The stress
in the x direction of *x f yd ' 1.81MPa is carried three legged ties at 160 mm centres. The
reinforcement details for the shear wall are shown in Figure 2.16.
The minor principal stresses in the concrete $+ 3c % are obtained from Eq. 2.11 and given in
Figure 2.14. In this case the + 1c axis is taken as coinciding with the major principal axis and
thus varies through the length of the element. In the compression boundary element the
transverse tension strains are low and the disturbance factor is taken as that given by Eq. 2.16.
The maximum compression stress is 5.66 MPa and when the effect of the axial reinforcement
in the section is considered it is seen that the compression strength of the boundaries does not
control the design.
In the web the compression stresses are reasonably uniform, indicating a degree of efficiency
in the design as the whole of the wall is utilised. The minimum minor principal stress in the
concrete wall is + 3c ' 9.0 MPa and as the transverse tensions are significant the disturbance
factor is taken as that given by Eq.2.15. The minimum specified concrete strength is obtained
from Eq. 2.13 and is f c# < 26.5 MPa . A concrete strength of f c# ' 30 MPa is adopted.
!
48 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
120
00 mm x 400 mm
bou
undary element
Y Loadd Case 2
Load Case 1 400
1c
500 mm squaare
2c
boundary eleement
58
3500
2c
1c
t = 200 mm
m 1c
2c
varies
60
Y
500 5500 500
F
Figure 2.12: Finite
F element analysis and
d design of squuat shear walll.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 49 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
+555'5
x 7F1/E5MPa +555'5
x 721A35MPa +555'5
x 7.1425MPa +555'5
x 741DA5MPa
+555'5
y 741435MPa +555'5
y 7414/5MPa +555'5
y 7414/5MPa +555'5
y 541445MPa
,5555'55
xy 41.E5MPa ,5555'55
xy 41.A5MPa ,5555'55
xy 41.05MPa ,5555'
xy 5541/F5MPa
4000 kN
+555'5
x 7/10.5MPa +555'5
x 7.1325MPa +555'5
x 7.1DE5MPa +555'5
x 7/13D5MPa +555'5
x 7413F5MPa +555'55
x 41445MPa
+555'5
y 5/1A/5MPa +555'5
y 7414.5MPa +555'5
y 741.05MPa +555'5
y 7412E5MPa +555'5
y 741345MPa +555'5
y 741F.5MPa
,5555'55
xy 41E05MPa ,5555'55
xy 213.5MPa ,5555'55
xy 21A05MPa ,5555'55
xy 21205MPa ,5555'
xy 55.1./5MPa ,5555'
xy 5541225MPa
+555'
x 7414/5MPa +555'5
x 741235MPa +555'5
x 741A25MPa +555'5
x 7/14/5MPa +555'5
x 741D.5MPa +555'5
x 7414.5MPa
+555'55
y 21E.5MPa +555'55
y 41AF5MPa +555'5
y 7412A5MPa +555'5
y 741EF5MPa +555'5
y 7/1F/5MPa +555'5
y 7/1EE5MPa
,5555'5
xy 541045MPa ,5555'5
xy 521.F5MPa ,5555'55
xy 014D5MPa ,5555'55
xy 21345MPa ,5555'5
xy 5.13A5MPa ,5555'5
xy 5410.5MPa
+555'55
x 41425MPa +555'55
x 410.5MPa +555'5
x 741425MPa +555'5
x 741DE5MPa +555'5
x 741F45MPa +555'5
x 7414.5MPa
+555'5
y 5D12.5MPa +555'5
y 5/1AD5MPa +555'5
y 741205MPa +555'5
y 7/1/25MPa +555'5
y 7.1./5MPa +555'5
y 721D35MPa
,5555'5
xy 5410F5MPa ,5555'5
xy 5214.5MPa ,5555'55
xy 213.5MPa ,5555'55
xy 21FD5MPa ,5555'55
xy 21/05MPa ,5555'55
xy 41DD5MPa
+555'55
x 41205MPa +555'55
x 41AF5MPa +555'55
x 41/D5MPa +555'5
x 7412/5MPa +555'5
x 741AE5MPa +555'5
x 741235MPa
+555'55
y 31445MPa +555'55
y .10F5MPa +555'5
y 741/F5MPa +555'5
y 7/1.E5MPa +555'5
y 7.1D45MPa +555'5
y 7D10A5MPa
,5555'55
xy 41ED5MPa ,5555'55
xy .1E/5MPa ,5555'55
xy 21245MPa ,5555'55
xy 21235MPa ,5555'55
xy 214A5MPa ,5555'55
xy 41A05MPa
Figure 2.13: Finite element mesh for squat shear wall example and stress output.
4000 kN
*55f
x yd '521..5MPa *55f
x yd '5213.5MPa *55f
x yd '521DA5MPa *55f
x yd '5.1335MPa
*55f
x yd '5/12/5MPa *55f
y yd '5.12/5MPa *55f
y yd '5.1..5MPa *55f
y yd '5/1345MPa *55f
y yd '541FE5MPa
*55f
y yd '5.12A5MPa
+3c '5741335MPa
+555
3c '7E1.E5MPa +555
3c '7E13F5MPa +555
3c '7310.5MPa +555
3c '701A.5MPa
*55f
x yd '501E05MPa *55f
x yd '5D1D05MPa *55f
x yd '501A.5MPa *55f
x yd '521AD5MPa
*55f
x yd '541FE5MPa
*55f
y yd '521445MPa *55f
y yd '5.1/05MPa *55f
y yd '5/10F5MPa *55f
y yd '541/05MPa +555'5
3 c 7/1A35MPa
*55f
y yd '5014D5MPa
+555
3c '731.05MPa +555
3c '7A1445MPa +555
3c '7E1.25MPa +555
3c '7F1./5MPa
*55f
x yd '5D1.D5MPa *55f
x yd '5D1A.5MPa *55f
x yd '5D1.F5MPa *55f
x yd '5501025MPa
*55f
x yd '541E25MPa *55f
y yd '521E25MPa *55f
y yd '5/1AA5MPa *55f
y yd '5/1/D5MPa
*55f
y yd '5D1DA5MPa
+555'5
3 c 721FD5MPa
+555
3c '7F13/5MPa +555
3c '7E1.35MPa +555
3c '7E1/.5MPa +555
3c '7F1AA5MPa
*55f
x yd '5D10D5MPa *55f
x yd '5D10.5MPa *55f
x yd '5D14E5MPa *55f
x yd '5521AF5MPa
*55f
x yd '5/1E/5MPa
*55f
y yd '501..5MPa *55f
y yd '5/1A45MPa *55f
y yd '541E25MPa +555'5
3 c 7D1FF5MPa
*55f
y yd '5310A5MPa
+555
3c '7F1.D5MPa +555
3c '731225MPa +555
3c '7310A5MPa +555
3c '7F1EF5MPa
Figure 2.14: Reinforcing steel stresses for shear wall orthogonally reinforced in XY.
!
50 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Wall Strip 4X
Wall Strip 3X
Wall Strip 2X
Wall Strip 1X
Boundary Strip 1Y
Boundary Strip 2Y
Wall Strip 1Y
Wall Strip 2Y
Wall Strip 3Y
Wall Strip 4Y
Figure 2.15: Design strips in the X and Y directions for detailing of the reinforcement.
1375 2750 1375
8 C.E
2 x C16@150
3500
8 C.E
3 x C12@160
stirrups
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 51 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Similar to that for membranes, the dimensioning of plate, slab and shell elements for stresses
determined from a FEanalysis is typically based on yield conditions derived from plasticity
theory. While careful consideration of the limited ductility of concrete is important in the
dimensioning of membrane elements, there is generally much less concern with this aspect in
slabs and shells as such structures are typically underreinforced. That is, failure is governed
by yielding of the reinforcement while the concrete does not crush. Important exceptions to
this observation are concentrated transverse forces, which may result in brittle punching
failures in slabs and shells without transverse reinforcement.
Over the past decades, several investigations have been carried out on the loaddeformation
analysis of shell elements, taking into account general nonlinear material behaviour (see for
example Khalifa, 1986, Kirschner and Collins 1986). The resulting approaches, which
typically use a subdivision of the shell elements in finite layers, are suitable for general load
deformation analyses of single elements, however, they do not allow for a straightforward
dimensioning. Thus, nonlinear FE modelling is less suitable for practical design purposes
where many elements must be dimensioned for different loadcases and the reinforcement
quantity and arrangement is yet to be determined. The required amount of reinforcement
could be minimized by aligning the reinforcement with the principal (tensile) stress directions.
This type of reinforcement has been applied to numerous slab and shell structures in the past,
and even today, some designers prefer to provide this socalled trajectory reinforcement,
particularly in shells. However, reinforcement layouts following the principal stress
trajectories are usually impractical, requiring a high amount of complex work on site.
Moreover, if several different loadcases must be considered, principal stress directions will
vary, making the provision of reinforcement aligned with them strictly speaking
impossible. For these reasons, orthogonal reinforcement is provided in almost all slab and
shell structures, and the following considerations will thus focus on this type of
reinforcement.
Integrating the stresses acting along the boundaries of a shell element, Figure 2.17 (a), one
obtains the elements stress resultants shown in Figure 2.17(b).
In the general case of a plane shell element, there are eight independent stress resultants, they
are, the bending and twisting moments
h2 h2 h2
mx ' G + x z dz , m y ' G + y z dz , mxy ' m yx ' G , xy z dz (in kNm/m = kN) (2.18)
7h 2 7h 2 7h 2
h2 h2
vx ' G , zx z dz , v y ' G , zy z dz (in kN/m) (2.19)
7h 2 7h 2
!
52 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
h2 h2 h2
n x ' G + x dz , n y ' G + y dz , nxy ' n yx ' G , xy dz (in kN/m) (2.20)
7h 2 7h 2 7h 2
(a) (b)
h/2
x
y z z h/2 z
dz
1 x
1 nx
+x dz y myx
ny nyx mx
my nxy vx
,yx dz
,xy dz ,zx dz vy
+y dz
,zy dz mxy
Figure 2.17: Plane shell element: (a) Stresses; (b) Stress resultants in the general case of combined membrane
and bending action.
The stress resultants of Eqs. 2.18 to 2.20 can be subdivided into two groups: the inplane axial
H I
forces nx , ny , nxy (membrane action) and the outofplane forces bending and shear forces
Hmx , my , mxyI Hvx , vyI (bending action). For plane elements, membrane and bending action
are independent, that is, two complete sets of equilibrium conditions involving only inplane
or outofplane forces can be formulated.
Curvature of the shell middle plane will result in a coupling of membrane and bending action,
such that equilibrium conditions of a curved element will generally involve all eight stress
resultants. The definition of stress resultants is also affected by curvature; in the case of thin
curved shell elements a good approximation can be obtained by multyplying the integrands in
Eqs. 2.18 to 2.20 by a factor of 1 7 z rx (stress resultants with last subscript y) or 17 z ry
(stress resultants with last subscript x). Thus, strictly speaking, for rx = ry one obtains that
mxy = myx and nxy = nyx (although , xy ' , yx still applies). However, for thin shells
z rx J 0 and z ry J 0 and thus, Eqs. 2.18 to 2.20 can usually be applied without
modification.
Slabs are probably the most important application of structural concrete. Thus, while this
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 53 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
chapter treats the dimensioning of shells, special attention is given to slabs, including
equilibrium and boundary conditions, stress transformation and reinforcement design.
For stresses and stress resultants, the sign convention illustrated in Figure 2.17 will be used
throughout this chapter. Accordingly, positive stresses are acting in the direction of positive
coordinates on boundaries with positive outer normal; with regard to normal stresses, this
means that tensile stresses are positive. Positive membrane and transverse shear forces are
those corresponding to positive stresses, and positive bending and twisting moments
correspond to positive stresses, according to the above definition, for positive values of
coordinate z. With regards to the double subscript indices, the first index stands for the
direction of the stress, while the second denotes the normal direction of the boundary on
which the stress or stress is acting.
Equilibrium conditions
Formulating equilibrium of the stress resultants acting on a slab element, Figure 2.18, one gets
where subscripts after a comma denote partial derivatives with respect to the corresponding
variable. Evaluating Eqs. 2.21bc for v x , x and vy, y and substituting into Eq. 2.21a gives the
elastic plate equation
which consists of the sum of equilibrium conditions for pure bending of unit strips in the x
and ydirection plus an additional coupling term consisting of the twisting moments.
(a) ! dx
vy dx
mxy dx
vx dy q dxdy
dy mx5dy
my5dx
myx dy x
z
(my+my,y dy) dx
(myx+myx,xdx) dy
(mx+mx,xdx) dy
y
(vx+v x,xdx) dy
(mxy+mxy,ydy) dx
(vy+v y,y dy) dx
!
54 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
$ %
mtn ' my 7 mx sinK cosK 8 mxy cos2K (2.23c)
my sinK my cosK X
K x K x
mxcosK mt T
n
myxsinK 2K 2K1
n mnt mn
myxcosK mxsinK 2 1
mtn t
mn 1 1 N
y
t Y K1 K Q (Pol)
y
(c) (d)
vx
vn K0 K vn
vy sinK vy cosK
K x K x
v0 vy v
n t
vt vxsinK
vn n
vxcosK vx v0 vy
1 t K
1 K0 L L 3L 2L
y 2 2
t
y
Figure 2.19: Stress transformation: (a), (b) Bending and twisting moments; (c), (d) Transverse shear forces.
Equations 2.23ac can be interpreted as equations for the transformation of bending and
twisting moments acting on any boundary perpendicular to the direction n, where the
orientation is determined by the angle K . This may be represented graphically using Mohrs
circle, Figure 2.19(b); in this case, twisting moments must be taken as positive if the
corresponding positive moment arrow is pointing towards the boundary under consideration.
The principal directions, for which twisting moments are zero, mtn ' 0 , are determined by the
condition
2m xy
tan 2K1 ' (2.24)
mx 7 m y
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 55 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
mx 8 m y $mx 7 m y % 2 8 4mxy2
m1,2 ' M (2.25)
2 2
Equilibrium of the forces acting on the slab elements shown in Figure 2.19(c) yields
These relationships can be interpreted as equations for the transformation of transverse shear
forces acting on any boundary perpendicular to the direction n, whose orientation is
determined by the angle K . The trigonometric functions can be represented graphically using
Thales circle, Figure 2.19(d). At any point of the slab, the principal transverse shear force
vo ' v x2 8 v 2y (2.27)
vy
tan K o ' (2.28)
vx
Along the boundary of a slab, there may exist, generally, a bending moment mn , a twisting
moment mnt and a transverse shear force vn (Figure 2.20a). However, the fundamental
equation for thin elastic slabs with small deflections (Kirchhoff, 1850) cannot satisfy more
than two boundary conditions. Thus, one may introduce a further condition for simply
supported and free edges as follows: replace the twisting moments mtn by a continuous
distribution of vertical forces, Figure 2.20(b), which, at the boundaries of infinitesimal
elements of length dt , compensate each other with the exception of the increase mtn,t dt . The
increase per unit distance, mtn ,t , is now combined with the transverse shear force vn , giving
a resultant edge force of
as shown in Figure 2.20(c). At a slab corner, the twisting moments mtn of the two joining
edges result, according to Figure 2.20(d), in a corner force of
2mtn (2.30)
!
56 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
mtn mtn5
dt
n
t +
mtn dt
m+n dt
5
5
n
1 vn+5dt
Vt5=5mtn
t Vt +Vt,t dt
25mnt
Figure 2.20: Static boundary conditions and discontinuities in slabs: (a) Stress resultants at slab edge;
(b) Forces corresponding to twisting moment; (c) Resultant edge orce; (d) Corner force; (e) Edge
shear force; (f) Force flow at corner; (g) Discontinuity.
The treatment of twisting moments along slab edges outlined above, first proposed by
Thomson and Tait (1883), can be justified by Saint Venants principle. However, it is not
entirely satisfactory due to its foundation on Kirchhoffs theory and the restrictive conditions
for the application of the latter. Figure 2.20(e) illustrates an alternative explanation of the
behaviour near slab edges, based only on equilibrium considerations. In order to satisfy
equilibrium of the edge strip, there must exist a transverse shear force transferred along the
edge such that, provided the slab boundary is stressfree and stresses + t within the edge strip
do not vary along the direction t (Clyde, 1979)
Note that the edge shear force (Eq. 2.31) is a force (for example in kN), while the resultant
edge force (Eq. 2.29) is a force per unit width (kN/m).
The condition (Eq. 2.30) for corner forces and the contribution mtn ,t of the twisting moments
to the resultant edge force (Eq. 2.29) are directly obtained from the existence of the edge
shear forces (Eq. 2.31). The corresponding boundary conditions can be summarised as
follows:
" simply supported edge: mn ' 0 , resultant edge force vn 8 mtn,t ' mn, n 8 2mnt ,t
" free edge: mn ' 0 , zero resultant edge force vn 8 mtn,t ' mn, n 8 2mnt ,t ' 0 .
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 57 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
These boundary conditions follow directly from equilibrium considerations and are thus valid
for any material behaviour. For thin elastic slabs more restrictive conditions can be
formulated (Timoshenko, 1959); however, these are not relevant if detailing and dimensioning
are carried out according to plasticity theory.
The edge shear forces (Eq. 2.31) must be taken into account when dimensioning free edges of
concrete slabs and shells with transverse reinforcement provided capable of resisting Vt . This
is illustrated using Figure 2.20(f), which visualises the force flow at the corner of a slab under
pure torsion. Two sets of inclined concrete compression struts, perpendicular to each other,
form at the top and bottom faces of the slab with their components normal to the slab edges.
These forces are resisted by corresponding reinforcement. The components parallel to the slab
edges are transferred to inclined concrete compressive struts along the edge strips. The
vertical components of the struts correspond to the edge shear forces (Eq. 2.31) and transverse
(vertical) reinforcement must be provided to resist these forces. This can be achieved by
providing hairpins or corresponding bends in the inplane reinforcement (providing full
anchorage of the bentback bar ends is achievable).
By joining two slabs along their free edges it follows, from the equivalence of twisting
moments along slab edges and edge shear forces according to Eq. 2.31, that while bending
moments mn must be continuous (Figure 2.20g), twisting moments mnt and transverse shear
forces vn may jump across statical discontinuities within a slab. At a statical discontinuity,
along which a shear force Vt is transferred, the conditions mn7 ' mn8 , Vt ' mnt
8 7
7 mnt and
Vt ,t ' vn7 7 vn8 must be satisfied (see Figure 2.20g).
The ultimate load of concrete slabs can be investigated on the basis of plasticity theory by
considering local stresses and strains inside the slab and corresponding yield conditions and
flow rules for concrete and reinforcement. This general procedure gives correct results in all
possible cases but its application is cumbersome and can only occasionally be justified. In
most cases it is sufficient to express statical and kinematical conditions as well as yield
conditions and flow rules in terms of generalized stresses and strains. That is, to consider the
stress resultants Eq. 2.18, Eq. 2.19 and Eq. 2.20 and corresponding generalised deformations,
neglecting local distributions of stresses and strains.
Yield conditions and flow rules in terms of generalised stresses and strains can be derived
from both the kinematic or upper bound method as well as the static or lower bound method
of plasticity theory. According to the kinematic method, deformations are restricted, by means
of kinematic assumptions, to a limited class of theoretically possible deformations that can be
described by a finite set of kinematic parameters. In the next section, the normal moment
yield criterion for slabs is derived on this basis. According to the lower bound method,
statically admissible stress distributions inside an element which do not violate the yield
conditions for concrete nor reinforcement are investigated. Strictly speaking, however, the
relationships obtained in this way (such as for the sandwich model presented later in this
chapter) are no yield conditions since, generally, no compatible mechanism can be found.
!
58 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Yield lines
Consider the shell element given in Figure 2.21(a), with reinforcement in directions x and y,
loaded by bending and twisting moments as well as membrane and transverse shear forces.
Figure 2.21(b) illustrates an element of a yield line in arbitrary direction t where, N! n and O!n
are the relative rotation rate of the rigid slab parts and the relative extension rate at the level of
the middle plane. For this type of kinematically admissible mechanism, only the bending
moments mn and the normal forces nn contribute to the energy dissipation,
For any given value of nn , the depth of the compressive zone c ' h 2 7 O!n N! n can be
determined from equilibrium of stresses in direction n, and one obtains an expression for the
ultimate moment mnu in direction n and, consequently, for the energy dissipation.
(a) 1 (b)
K x
1
h /2 c mn
nn
x t n
z nx n
myx h /2 .
ny On
y nyx mx
nxy vx .
my Nn
vy
mxy
y
(c) myusinK
fc cx fc cy z x
' fy
asx ' fy
asy K
mxu myu mxucosK n
x y nn
z asx fy z asy fy t ntn mtn
mn 1
y
(d) mxy
(e)
myu
k= 1
m'yu
mx
mxy
myu
my
mxu
mxu m'xu
my mx mxy
Figure 2.21: Normal moment yield criterion: (a) Shell element; (b) Yield line; (c)Superposition of ultimate
moments in directions x and y; (d) Yield condition; (e) Dimensioning.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 59 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In the following, only yield lines with O!tn ' 0 are treated, and membrane forces are neglected,
that is nn ' 0 . Thus, only the bending moments mn contribute to the energy dissipation.
By superimposing the ultimate moments mxu and myu in the reinforcement directions while
setting mxy ' nx ' ny ' 0 , a statically admissible state of stress is obtained (Figure 2.21c).
For an arbitrary direction n it follows that nn ' ntn ' 0 , mn ' mxu cos 2 K 8 m yu sin 2 K and
$ %
mtn ' myu 7 mxu sinK cosK (see Eq. 2.23). Generally, the depths of the compression zones
in the two reinforcement directions do not coincide, that is cx = cy and there is no compatible
mechanism according to Figure 2.21(b). Thus, the value of mn determined in this way is a
lower bound for the ultimate moment mnu in direction n, mxu cos 2 K 8 m yu sin 2 K .
The differences between mn and mnu obtained for cx = cy are generally negligible and the
inequality can be suppressed, leading to the conditions
Equation 2.33b, which is valid for negative moments, follows from the same considerations as
for Eq. 2.33a.
A state of stress mx , my and mxy corresponds, according to Eq. 2.23, to bending and
twisting moments in the direction n of
# ? mn ? mnu
7 mnu (2.35)
!
60 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
2
Y ' mxy $ %
7 $mxu 7 mx % m yu 7 m y ' 0
(2.36)
2
Y # ' mxy 7 $m#xu 8 mx %$m#yu 8 m y % ' 0
where mxu 7 mx < 0 , myu 7 my < 0 , m#xu 8 mx < 0 and m#yu 8 my < 0 . The conditions Y ' 0
and Y # ' 0 can be represented by two elliptical cones in the mx , my , mxy space, as shown in$ %
Figure 2.21(d).
Similarly to the yield conditions for orthogonally reinforced membrane elements, it is possible
to represent the yield conditions defined by Eq. 2.36 in parametric form:
Equation 2.37 is suitable for straightforward dimensioning of the four reinforcement layers,
Figure 2.21(d) where, typically, k = 1 is used.
Skew reinforcement
Yielding reinforcement layers in arbitrary, skew, directions can always be substituted by an
equivalent orthogonal reinforcement if it is assumed that all reinforcing bars are located,
approximately, in the same plane.
The effect of several reinforcement layers oriented in directions differing by angles : i from
$ %
the xaxis and with resistances nis ' as f sy per unit width corresponds to a fictitious
i
reinforcement in directions x and y with resistances
nxs ' P nis cos 2 : i , n ys ' P nis sin 2 : i , nxys ' P nis sin : i cos : i , (2.38)
i i i
The resistances nxs , nys and nxys can be transformed using Eqs. 2.23ac (setting ns instead
of m) just like bending and twisting moments and there exist two perpendicular principal
directions $R ,Q % , differing by an angle ( from the x and yaxis, for which nRQs ' 0 . The
ultimate moments mRu and mQu corresponding to the resistances nRs and nQs can be
substituted in the yield condition Eq. 2.36 together with the bending and twisting moments
mR , mQ and mRQ acting in the same direction, giving
2
Y ' mRQ $
7 mRu 7 mR mQu 7 mQ ' 0 %$ % (2.39)
where mRu 7 mR < 0 and mQu 7 mQ < 0 . An alternative procedure consists of transforming the
ultimate moments mRu and mQu into the directions x and y using Eq. 2.23, leading to
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 61 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
$ %
mxyu ' mQu 7 mRu sin ( cos ( (2.40c)
Comparing Eqs. 2.40ac with the bending and twisting moments mx , my and mxy leads to
2
Y ' mxy $
7 $mxu 7 mx % m yu 7 m y ' 0 % (2.41)
where mxu 7 mx < 0 and myu 7 my < 0 . The conditions Y ' 0 and Y # ' 0 (which has not
been given but follows from the same considerations as outlined for positive moments) can be
$
graphically represented as elliptical cones in the mx , my , mxy space, similar to the yield %
conditions of Eq. 2.36 but with the vertices no longer lying in the plane mxy ' 0 .
Assuming that the effective depth is constant in all directions, it is possible to evaluate mxu ,
myu and mxyu directly from Eq. 2.38 without having to determine the principal directions
$R ,Q % and substitute them into Eq. 2.41.
The normal moment yield criterion has been established and confirmed by various researchers
(for example Johansen, 1962, Nielsen, 1984, Park and Gamble, 1980, Wolfensberger, 1964,
Wood, 1961) and corroborated with a wide range of experimental data. Due to its simplicity,
it is widely used for the dimensioning of slabs in design practice. While its application is
generally unproblematic due to the typically ductile behaviour of slabs, it does have some
weaknesses of which the designer should be aware.
Supposing that sections perpendicular to the element middle plane remain straight and
perpendicular to the deformed middle plane, the deformations of thin slabs with small
deflections are fully determined by six kinematic parameters ) x 0 , ) y 0 , & xy0 , S x , S y , S xy
. These six parameters come from elongation in x and y, distortion of middle plane, bending in
directions x and y and twist (see Figure 2.22). The consideration of yield lines is kinematically
more restrictive and the yield conditions given by Eqs. 2.36 and 2.37 are thus generally in
spite of the slight inequality suppressed in the deduction of Eq. 2.33 upper bounds for the
ultimate load. This observation holds particularly true for elements subjected to high twisting
moments with respect to the reinforcement directions (reinforcement directions deviating
significantly from the directions of the principal moments) and high reinforcement ratios
(Marti, 1980, Nielsen, 1984). Special care is required in the dimensioning of elements and
excessive reinforcement ratios should be avoided unless a more sophisticated analysis is
undertaken, for example using to the sandwich model outlined below.
Furthermore, the normal moment yield criterion does not take into account transverse shear
forces which are always present in slabs. In the case of potentially shearcritical slabs (high
shear forces, thick slabs or slabs with horizontal construction joints) it is advisable to use a
model which allows for a consistent treatment of bending and twisting moments as well as
transverse shear forces. The sandwich model approach outlined below is particularly suitable
for such situations.
!
62 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
x
y z
Figure 2.22: Kinematics of a thin slab element: (a)(c) elongations )x0, )y0, and distortion &xy0 of element middle
plane; (d)(f) curvatures Sx, Sy and twist Sxy .
The eight stress resultants acting on a shell element, Figure 2.23(a), can be resisted by a
sandwich model. In this model the bending and twisting moments mx , my , mxy as well as H I
H I
the inplane forces nx , ny , nxy are attributed to the sandwich covers, while the sandwich
core resists the transverse shear forces vx , vy H I (illustrated in Figure 2.23b). Thus, the
sandwich model represents a lower bound solution with its basis in plasticity theory.
Figure 2.23: Sandwich model: (a) Shell element; (b) Layer forces; (c) Transfer of transverse shear force in
uncracked and cracked core.
The sandwich core transfers the principal transverse shear force v0 ' v x2 8 v 2y in the
$ %
direction K 0 ' tan 71 v x v y . If the transverse shear forces are small, that is, nominal shear
stresses v0 dv are below the nominal shear cracking resistance , c, red , the core will remain
uncracked and the forces in the bottom and top sandwich covers follow directly as
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 63 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
m n
n x inf,sup ' M x 8 x
dv 2
my ny
n y inf,sup ' M 8 (uncracked core, v0 d v 9 , c, red ) (2.42)
dv 2
m xy n xy
n xy inf,sup ' M 8
dv 2
If the principal transverse shear force v0 ' v x2 8 v 2y is high enough to produce cracking of
the sandwich core, the latter can be treated like the web of a girder of flanged crosssection
running in direction K o (Figure 2.23c). The corresponding tensile forces in the element plane
must be resisted by the sandwich covers in addition to the forces given by Eq. 2.42, resulting
in the following total forces in the sandwich covers [Marti (1990)]
mx nx v x2
n x inf,sup ' M 8 8
dv 2 2v0 tan 
my ny v 2y
n y inf,sup ' M 8 8 (cracked core, v0 d v < , c, red ) (2.43)
dv 2 2v0 tan 
m xy n xy vxv y
n xy inf,sup ' M 8 8
dv 2 2v0 tan 
According to Eqs. 2.42 and 2.43, a state of plane stress is obtained in the sandwich covers,
and they can be dimensioned as membrane elements. For the case of failure governed by
yielding of the reinforcement with the concrete remaining elastic in both sandwich covers
(defined as Regime 1 and given by Eq. 2.7a), the force per unit width in the reinforcement in
the x and y directions are determined with
mx nx v x2 mxy n xy vxv y
asx f y < 8 8 8 k 8 8
dv 2 2v0 tan  dv 2 2v0 tan 
my ny v 2y mxy n xy vxv y
asy f y < 8 8 8 k 71 8 8
dv 2 2v0 tan  dv 2 2v0 tan 
(2.44)
m n v x2 m xy n xy vxv y
a#sx f y < 7 x 8 x 8 8 k# 7 8 8
dv 2 2v0 tan  dv 2 2v0 tan 
my ny v 2y m xy n xy vxv y
a#sy f y < 7 8 8 8 k #71 7 8 8
dv 2 2v0 tan  dv 2 2v0 tan 
where as and a#s are the bottom and top reinforcement areas per unit width of the slab,
respectively.
For an uncracked core, that is v0 d v 9 , c, red , the terms in Eq. 2.44 containing vx or vy can
be omitted and no transverse shear reinforcement is required. That is * z ' 0 .
!
64 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
v0 tan
*z ' (2.45)
dv f y
is required for the core to carry transverse shear force. In Eqs. 2.42 to 2.45, dv is the effective
depth of the membrane forces in the sandwich covers, k and k # are arbitrary positive factors
and  is the inclination of the diagonal compression field in the sandwich core. Different
values of k and k # can be adopted for each element of a shell or slab, avoiding abrupt changes
or providing sufficient development length in order to anchor the differential forces in the
reinforcement while in choosing the inclination angle,  . In this regard similar considerations
apply as in the design of membranes. For simplicity in design practice using k ' k # ' 1 and
 ' 45o often provides for a good solution.
Equation 2.44 is valid only if the concrete in the sandwich covers does not crush. In order to
satisfy this condition, the thickness of the sandwich covers, and consequently the effective
depth dv , should be chosen such that
$
f c tinf < a sx f y 8 a sy f y 7 n x inf 8 n y inf %
$
f c t sup < a #sx f y 8 a #sy f y 7 n x sup 8 n y sup % (2.46)
where tinf , tsup = thickness of bottom and top sandwich cover and f c = effective concrete
compressive strength, respectively. Substituting Eqs. 2.43 and 2.44 in Eq. 2.46, we write
$
f c tinf < k 8 k 71 % mxy
dv
8
nxy
2
8
vxv y
2v0 tan 
(2.47)
$
f c t sup < k 8 k 71
%7 dmxy
8
nxy
2
8
vxv y
2v0 tan 
v
As a first and often sufficient approximation, dv can be chosen such that the middle planes of
each sandwich cover coincides with the centre between its two reinforcement layers.
The threshold value , c, red for considering cracked or uncracked behaviour of the sandwich
core should be carefully selected. Basically it could be taken as the nominal shear strength of
the slab or shell without transverse reinforcement, which, due to size effect, varies with the
element thickness; for example , c , red ' f ctd for membrane shear according to [CEBFIP
(1990)]. Since the provision of transverse reinforcement is quite timeconsuming, it is
advisable to choose the element thickness such that, apart from zones where concentrated
forces are introduced, no transverse reinforcement is required.
In cases with high axial compression nx , ny or high inplane shear forces nxy , it is possible
to attribute part of these forces to the sandwich core and then, strictly speaking, the concrete
compression in the sandwich core must be checked taking into account the combined action
$ %
of nx , ny , nxy and v0 . The case of high axial compression is likely to occur if prestressing is
treated as forces acting on the structure. However, this is beyond the scope of this report.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 65 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The design equations derived from the sandwich model (Eqs. 2.44 and 2.45) are simple and
transparent, allowing for a straightforward dimensioning of inplane as well as transverse
reinforcement in the general case of a shell element subjected to eight independent stress
resultants. In the special case of slabs and neglecting transverse shear forces, Eq. 2.44 is
equivalent to the normal moment yield criterion (Eq. 2.37) if exact values are taken for dv .
The incorporation of Eqs. 2.44 and 2.45 into a finite element program for shells does not any
present more difficulties than the implementation of the normal moment yield criterion
(Eq. 2.37) for the dimensioning of reinforcement, a postprocessing option which is already
offered by a number of commercial programs today (often called WoodArmer moments).
Usually, no explicit check of the assumed failure mode is made in slabs when applying
Eq. 2.37, assuming implicitly that the reinforcement ratios obtained are small enough to avoid
concrete crushing. While this condition is usually satisfied in slabs, a check of the assumed
failure mode should always be carried out for shells, particularly for elements subjected to
significant axial compression or cases with high reinforcement ratios. Thus, it is advisable to
implement the corresponding check of concrete resistance according to Eq. 2.46 together with
Eqs. 2.44 and 2.45.
Transverse shear forces obtained from a finite element calculation will generally be less
accurate than bending and twisting moments, since, as can be seen from Eqs. 2.21bc, which
can be rewritten as vx ' mx, x 8 mxy, y and v y ' my, y 8 myx, x , transverse shear forces are
derivatives of the bending and twisting moments. Thus, in order to get reasonable results, a
finer element mesh is required when using Eq. 2.44 and 2.45 in their general form than in
cases where transverse shear forces can be neglected a priori.
2.3.7 Example 1
Consider the slab element shown in Figure 2.24(a); according to the normal moment yield
criterion, Eq. 2.37, using a value of k = 1, the required bending resistances in the
reinforcement directions are
Note that the required reinforcement areas have been determined for uniaxial bending
according to CEBFIP (1993).
!
66 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In many practical cases, one or more of the reinforcement areas resulting from the application
of Eq. 2.37 will be negative. This means that no reinforcement is required in the
corresponding direction. Also, in such cases it is basically possible to reduce the
reinforcement area in the other direction by changing the value of k such that the
reinforcement area which was negative for k = 1 equals zero or corresponds to the minimum
reinforcement to be provided. However, this is rarely done in design practice.
Consider next the shell element shown in Figure 2.24(b). With a principal transverse shear
force of vo ' v x2 8 v 2y ' 719 kN , see Eq. 2.27, cracking of the core must be expected since
vo dv ' 2.25 MPa @ , ctd . The resulting forces in the bottom and top layers follow thus from
Eq. 2.42 as nx,inf ' 7274 kN m , ny,inf ' 736 kN m , nxy,inf ' 502 kN m (bottom layer)
and nx,sup ' 7274 kN m , ny,sup ' 696kN m , nxy,sup ' 7329 kN m (top layer).
Substituting these membrane forces into Eqs. 2.8 and 2.9, or directly using Eq. 2.44, gives
2 2
The transverse reinforcement resulting from Eq. 2.45 is * z ' 0.52% ' 5170 mm m , which
can be provided, for example, using 26 stirrups 16mm per square meter (16@150 x 250).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 67 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
$
asx,inf f y 8 asy ,inf f y 7 nx,inf 8 n y ,inf % ' 12.6 MPa ? f cd (2.50a)
tinf
$
asx, sup f y 8 asy , sup f y 7 nx, sup 8 n y , sup % ' 8.2 MPa ? f cd (2.50a)
t sup
If the reinforcement of the slab element of Figure 2.24(a) is dimensioned as if it were a shell
element using the sandwich model, one obtains reinforcement areas of asx,inf ' 1650,
asy,inf ' 120, asx,sup ' 270 and asy,sup ' 1800 mm2/m as opposed to that given by Eqs.
2.48ad obtained using the normal moment yield criterion, Eq. 2.37. It is seen that
reinforcement dimensioning according to the sandwich model is slightly on the safe side. The
differences are due to the larger lever arms used in the calculation of the bending resistances
for the normal moment yield criterion as compared with the averaged value of dv used in the
sandwich model.
2.3.8 Example 2
While a comprehensive treatment of the design process and its results are beyond the scope of
this report, some basic considerations will be outlined in order to illustrate the potential of the
sandwich model. The example selected is the shell roof of the fish market at the Port of
Santander, Spain, designed by the engineer Prof. Juan Jos Arenas and built 20012002. The
roof is shown in Figure 2.25.
The roof consists of 23 modules measuring approximately 5.1 x 45 m each, with a typical
thickness of 0.12 m. The geometry of each module is defined by a circular arch of 30 m span
complemented by two cantilevers of 7.5 m on either side in the transverse direction, and by
5.10 m wide parabolas of variable height in the longitudinal direction. Thus, globally, each
module is acting as an arch, tied together by a posttensioned, suspended floor or by post
tensioned concrete ties depending on the module. Construction was carried out using two
complete scaffolds, each for an entire module, starting at the two extremes of the roof. The
repeated use of the scaffolds reduced the impact of their construction costs on the overall
economy of the structure to a reasonable level.
Due to the high live loads on the suspended floors, a reinforcement following the principal
stress trajectories was not possible since these directions shift considerably depending on the
loadcase. Therefore, and because it was judged to be more practical, an orthogonal
reinforcement layout (in plan), aligned with the longitudinal and transverse axes of the
building, was adopted.
The reinforcement was dimensioned using the sandwich model outlined in 2.3.6, above. In a
first step, based on the different load combinations considered according to the pertinent
building codes, envelopes of all eight stress resultants were computed with a conventional
linear finite element program using eightnode shell elements (six degrees of freedom per
node). Each of the envelopes calculated consisted of the maximum and minimum values of
the stress resultant under consideration along with the concomitant values of the remaining
seven stress resultants. The results obtained in this way were then analysed according to
!
68 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 2.25: Fish market Port of Santander, Spain (designer Prof. Juan Jos Arenas).
Eqs. 2.44 to 2.47, with dv such that the middle planes of each sandwich cover coincided with
the centre between its two reinforcement layers for an assumed maximum bar diameter of
12 mm in each direction.
Using Eq. 2.45, the envelopes of the transverse shear forces were analysed to determine
whether or not the core would crack. In this case, as a result of the double curvature, loads are
transferred primarily through inplane actions, such that transverse shear forces remain small
and the provision of transverse shear reinforcement was only required in the edge modules.
Next, the required crosssections of the four reinforcement layers were determined using
Eq. 2.44 for a bar spacing of 150 mm. Finally, the assumed thickness of the sandwich covers
was checked according to Eq. 2.47.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 69 !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Each of the steps outlined above was carried out for all of the sixteen envelopes (maximum
and minimum of each stressresultant) at each of the roughly 1100 nodes per module. Thus, it
was decided to implement the entire design procedure in a specially written program which
was then run as a postprocessor to the finite element program used in the global analysis. The
postprocessor output consisted of numerical and graphical results, containing a CAD file
with all the data required for drafting of the reinforcement plans. Based on these results,
reinforcement plans such as the one shown in Figure 2.26 were produced and constructed
(Figure 2.27).
2.4.2 Background
In 3D space the stresses at a point are completely defined by the tensor (using von Karmans
notation)
Z+ x , xy , xz W
X U
+ ij ' X, xy + y , yz U (2.51)
X, xz , yz + z U
Y V
where x, y, z are any set of orthogonal axes and the stresses are defined as shown in
Figure 2.28a. For every point in a body there exist three stresses + x# , + y # and + z # on the
local x# y # z # axis system such that , x#y# ' , x#z# ' , y#z# ' 0 . These stresses are known as the
principal stresses and x# , y # , and z # the principal axes. It is well established that the principal
stresses are equal to the eigenvalues of the stress tensor. Perhaps less well recognised,
however, is that the direction cosines to the principal axes are given by the norms of the
eigenvectors of the stress tensor.
!
70 2 Design using linear stress analysis !
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
!
!
Figure 2.26: Reinforcement details for one module of the fish market roof Port of Santander, Spain.
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 71
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 2.27: Construction of the fish market roof at the Port of Santander, Spain.
For any oblique plane (refer Figure 2.28b) having a unit normal n ' nx , ny , nz passing H I
through a point P, the stresses at P can be resolved into a component normal to the plane $+ n %
and a shear component parallel to the plane $Sn % . For a stress to be principal S n ' 0 which,
from Eq. 2.51, implies that
+y
,xy +n
,yz ,xy
,yz sn
+x
,xz
,xz
+z
(a) (b)
Figure 2.28: a) 3D stresses at a point defined in the orthogonal xyz axis system; b) normal and shear stress for
an arbitrary plane passing through a point.
+ x nx 8 , xy n y 8 , xz nz ' + n nx
, xy nx 8 + y n y 8 , yz nz ' + n n y (2.52)
, xz nx 8 , yz n y 8 + z nz ' + n nz
Rewriting Eq. 2.52 of the form +.n = 0 it is seen that the equations are homogeneous. As all
three components of n cannot be zero the solution is nontrivial only if the determinant of the
coefficients + ' 0 , that is
+ x 7+ n , xy , xz
, xy + y 7+ n , yz ' 0 (2.53)
, xz , yz + z 7+ n
where I1 , I 2 and I 3 are the invariants of the stress tensor and are given by
$2
I 2 ' + x+ y 8 + x+ z 8 + y+ z 7 , xy 2
8 , xz %
8 , 2yz ' + 1+ 2 8 + 1+ 3 8 + 2+ 3 (2.55b)
and where + 1 , + 2 and + 3 are the principal stresses. By common definition the principal
stresses are ordered such that + 3 ? + 2 ? +1 .
As for twodimensions, stresses at a point in 3D can be plotted in the form of Mohrs circles
(shown in Figure 2.29) where the normal stress is plotted on the horizontal axis and the shear
stress plotted on the vertical axis. Three principal circles are possible between the principal
stress pairs + 1 7 + 2 , + 2 7 + 3 and +1 7 + 3 . In failure theorems the principal stress pair 13 is
regarded as the most important and the circle generated through this stress pair is referred to
as the major principal stress circle.
Sn
+55?5+55?5+
3 2 1
+n
+3 +2 +1
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 73
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In xyz space + x , + y and + z are, by definition, normal to the YZ, XZ and XY planes,
respectively. The shear stresses on these planes is then given by either the positive or negative
roots of
2 2 2
S x ' , xy 8 , xz ; S y ' , xy 8 , 2yz ; S z ' , xz
2
8 , 2yz (2.56)
As the orientation of the xyz axis system is arbitrary, its interpretation represents all possible
planes. It can be then shown that all points of $+ n , Sn % must line on or between the principal
stress circles and, thus, the feasible domain of solutions lies with the hatched region of
Figure 2.29.
In the applications that follow the xyz axes are taken to correspond with reinforcing directions.
As for two dimensions, normal stresses applied at a point in a reinforced concrete solid
element are carried by reinforcing steel and/or the concrete whilst shear stresses are carried by
the concrete alone. Given that the applied stress tensor has been determined, for example by
3D finite element solid modelling, the Mohrs circles of applied stress can be plotted, as
shown in Figure 2.30. Within the circles the stress points $+ i , Si % are also plotted where
i ' x, y, z . As the reinforcing steel cannot carry shear stress it follows that the points
$+ ci , Sci % must fall within the hatched region of the concrete stress circles where
+ ci ' + i 7 *i+ s and Sci ' Si and where *i is the volumetric ratio of reinforcement in the ith
direction and + s is the stress in the steel.
+555'5+557
ci i 5*5+
i s
Sn
*5+
i s
$+555,
ci Sci) $+55,
i S i)
+n
+c3 +c2 +c1 +3 +2 +1
Applying Eq. 2.51 to the stresses defined in Figure 2.30, the tensor of the concrete stresses is
written as
$
Z + x 7 * x f yd .x % , xy , xz W
X
+ cij ' X , xy $+ y 7 * y f yd. y % , yz
U
U (2.57)
X
Y , xz , yz $+ z 7 * z f yd.z %UV
where f yd.x , f yd. y and f yd.z are the design yield strengths of the reinforcing steels in the x
y and z directions, respectively.
The invariants of the concrete stress tensor are given by Eq. 2.55 with the appropriate
substitutions for + x , + y and + z . Comparing Eqs. 2.57 with Figure 2.30 it is seen that there
are 5 unknowns + c 2 , + c 3 , * y f yd. y , *x f yd.x and *z f yd.z (with + c1 ' 0 or another
prescribed limit such as + c1 ' fct ). As the solution to Eq. 2.53 provides for a maximum of
three real roots an infinity of solutions exist to Eq. 2.57. The designer then has the freedom to
apply two constraint equations with the three invariant equations making up the five equations
required for a solution.
The design process outlined above is demonstrated in the examples below. As the three
resulting stress points $+ ci , Sci % are not constrained to the boundary of the stress circles, as is
the case for two dimensions, graphical solutions for threedimensional stresses tend to be
somewhat more complex than for 2D. For this reason the examples presented below are
limited to analytical solutions.
The results of a stress analysis on a concrete structural element give the stress tensor in the xyz
dimensions as
Z2 6 7 4W
X
+ ij ' X 6 7 2 2 UU MPa (2.58)
XY 7 4 2 5 UV
It is desired to reinforce the element in the orthogonal directions of xyz. For the stresses
defined by Eq. 2.58, the magnitudes of the shear stresses are S x ' 2 13 [ 7.21 MPa ,
S y ' 2 10 [ 6.32 MPa and S z ' 2 5 [ 4.47 MPa and the principal stresses are
+ 1 ' 8.28 MPa , + 2 ' 4.32 MPa and + 3 ' 77.60 MPa . The Mohrs circle of stress for the
tensor of Eq. 2.58 is shown in Figure 2.31a. Viewing the stress plot (Figure 2.31a) it is
decided to seek the solution that gives the lowest demand on the concrete strength. This is
established by selecting the smallest diameter for the major principal stress circle. As only the
concrete carries shear stress the radius of the major stress circle $R173 % is constrained such
$
that R173 < max S x , S y , S z %
and thus for our example R173 ' Sx = 2 13 MPa.
Therefore, for the absolute minimum compression stress in the concrete the constraint
equations are given by
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 75
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Sn  (MPa)
$+565;
y S y ) $+565;
x S x )
$+565;
z S z )
+n (MPa)
+355= 7.60 +255= 4.32 +155= 8.28
$+556;
y S y)
$+556;
z S z )
$+5556;
cy S y )
*z +sz = 9.21
$+5556;
cz S z )
+n (MPa)
+555=
c3 14.42 +555
c2 = 2.88
Substituting Eqs. 2.59a and b into the stress invariant equations given by Eqs. 2.55ac we
write
I1 \ + c 2 7 4 13 ' 3 7 2 13 7 * y f yd . y 7 * z f yd .z (2.60a)
$ % $ %$
I 2 \ 74 13+ c 2 ' 72 13 3 7 * y f yd . y 7 * z f yd . z 7 2 8 * y f yd . y 5 7 * z f yd .z 7 56 % (2.60b)
I3 \ $ %$ %
0 ' 16 * y f yd . y 8 36 * z f yd .z 7 2 13 7 2 7 * y f yd . y 5 7 * z f yd .z 7 244 8 8 13 (2.60c)
Solving Eqs. 2.60ac gives + c 2 ' 72.88 MPa , * y f yd. y ' 3.88 MPa and *z f yd.z '
9.21 MPa. The final solution is plotted in Figure 2.31b.
Z 7 3 6 7 4W
+ ij ' XX 6 7 7 2 UU MPa (2.61)
XY 7 4 2 0 UV
and it is required to dimension the reinforcing steel. For the tensor of Eq. 2.61 the magnitude
of the shears are Sx ' 7.21MPa, S y ' 6.33 MPa and Sz ' 4.47 MPa and the principal
stresses are + 1 ' 3.28 MPa , + 2 ' 70.68 MPa and + 3 ' 712.60 MPa . The Mohrs circle of
stress for the applied tractions is plotted in Figure 2.32a. After reviewing the stress circles it is
decided to seek a solution such that no reinforcing steel is required in the ydirection, that is
* y f yd. y ' 0 . Substituting this constraint into Eqs. 2.55ac gives
I2 \ + 2c+3c ' 7358 7*x f yd.x 810*z f yd.z 8 *x f yd.x *z f yd.z (2.62b)
I3 \ 0 ' 288 4*x f yd.x 815*z f yd.z 7 7*x f yd.x *z f yd.z (2.62c)
$+565;S
x x )
Sn  (MPa)
$+565;S
y y )
$+565;S
z z )
+n (MPa)
+355= 12.60 +255= 0.68 +155= 3.28
(a) Applied Stresses
$+565;
x Sx ) Sn  (MPa)
$+5565;
cx Sx ) *x +sx = 3.77
*z +sz = 3.77
$+5565;
cy Sy ) $+565;
z Sz )
$+5565;
cz Sz )
+n (MPa)
+555
c3 = 14.57 +555
c2 = 2.98
Figure 2.32: Mohrs stress circles for example 2: a) applied stresses and b) concrete stresses.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 77
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Solutions to Eqs. 2.62ac are plotted in Figures 2.33a and 2.33b for the intermediate principal
concrete stress + c 2 versus the stress in the x and z reinforcement and for + c 2 versus + c 2 ,
respectively. From the first stress invariant it is seen that the minimum volume of
reinforcement for a unit element (for f yd.x ' f yd. y ' f yd.z ' f yd ) occurs at the point where
I1 is a minimum. In general, the optimal solutions lie at the upper end of + c 2 as shown in
Figures 2.33a and 2.33b for the example at hand. After consideration a solution is chosen such
that *x f yd.x ' *z f yd.z = 3.77 MPa, + c 2 ' 72.98 MPa and + c3 ' 714.57 MPa . The stress
circles for the chosen solution are shown in Figure 2.32b.
20 13 15
Ic1
*x fyd + *z fyd 14
15 20
15
(MPa)
(MPa)
Ic1 (MPa)
16
10 25
*z fyd
* fyd
+ c3 17
5 *x fyd 18
+c3 30
19
0 20 35
10 9 8 7 6 5 4 3 2 10 9 8 7 6 5 4 3 2
+c2 (MPa) +c2 (MPa)
(a) (b)
Figure 2.33: Solutions to Eqs. 2.62ac for +c2 versus a) reinforcement ratios and b) principal concrete stresses.
2.5 References
ASCEACI Committee 445 on Shear and Torsion, (1998), Recent approaches to shear design
of structural concrete, Journal of Structural Engineering, ASCE, Vol. 124, No. 12,
December, pp. 13751417.
Belarbi, A., and Hsu, T.T.C. (1991), Constitutive Laws of Reinforced Concrete in Biaxial
Tension Compression, Research Report UHCEE 912, Department of Civil Engineering,
University of Houston, Houston, Texas.
Cagley, J.R. (2001), Changing from ACI 31899 to ACI 31802 Whats New?, Concrete
International, Vol 23, No. 6, June, pp. 69183.
CEBFIP (1993), CEBFIP Model Code 1990 for Concrete Structures. Comit Euro
International du Bton, Bulletin dInformation No. 213/214, Lausanne, May 1993, 437 pp.
Clark, L.A. (1976), The Provision of Tension and Compression Reinforcement to Resist
InPlane Forces, Magazine of Concrete research, Vol. 28, No. 94, March, pp. 312.
Clyde, D.H. (1979), Nodal Forces as Real Forces, Final Report, IABSE Colloquium on
Plasticity in Reinforced Concrete, Copenhagen, International Association for Bridge and
Structural Engineering, IABSE Vol. 29, pp. 159166.
Clyde, D.H., (1977), A General Theory for Reinforced Concrete Elements, Australasian
Conference on the Mechanics of Structures and Materials, Christchurch, New Zealand,
August.
Collins, M.P., and Mitchell, D., (1986), Rational Approach to shear Design  The 1984
Canadian code Provisions, ACI Structural Journal, Vol. 83, No. 6, NovDec, 925933.
Foster, S.J., and Malik, A.R., (2002), Evaluation of Efficiency Factor Models used in Strut
and Tie Modelling of NonFlexural Members, ASCE, Journal of Structural Engineering, Vol.
128, No. 5, May, pp. 569577.
Foster, S.J., Marti, P. and Mojsilovi!, N. (2003), Design of Reinforced Concrete Solids
Using Stress Analysis, ACI Structural Journal, V100, N6, NovDec, pp. 758764.
Johansen, K.W. (1962), Yield Line Theory, Cement and Concrete Association, London,
181 pp.
Kaufmann, W., and Marti, P. (1998), Structural Concrete: Cracked Membrane Model,
Journal of Structural Engineering, ASCE, Vol. 124, No. 12, pp. 14671475.
Khalifa, J. (1986), Limit Analysis and Design of Reinforced Concrete Shell Elements,
Ph.D. Thesis, University of Toronto, Toronto, 314 pp.
Kirchhoff, G. R. (1850), ber das Gleichgewicht und die Bewegung einer elastischen
Scheibe, A. L. Crelles Journal fr die reine und angewandte Mathematik, Berlin, Vol. 40,
No. 1, pp. 5158.
Kirschner, U., and Collins, M.P. (1986), Investigating the Behaviour of Reinforced Concrete
Shell Elements, University of Toronto, Department of Civil Engineering, Publication
No. 8609, Toronto, 210 pp.
Leitz, H., (1923) Eisenbewehrte Platten bei allgemeinem Biegungszustande, Die Bautechnik,
Vol. 1, pp.155157, 163167.
MacGregor, J.G., (1997) Reinforced Concrete  Mechanics and Design, 3rd Edition, Prentice
Hall, New Jersey.
Marti, P. (1980), "Zur Plastischen Berechnung von Stahlbeton, Institut fr Baustatik und
Konstruktion, ETH, Zrich, Bericht Nr. 104, Birkhuser Verlag Basel, 176 pp.
Marti, P. (1980), Zur Plastischen Berechnung von Stahlbeton, Institut fr Baustatik und
Konstruktion, ETH Zrich, IBK Bericht Nr. 104, 176 pp.
Marti, P. (1990), Design of Concrete Slabs for Transverse Shear, ACI Structural Journal,
Vol. 87, No. 2, pp. 180190.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 79
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Mitchell, D., and Collins, M.P. (1974), Diagonal Compression Field Theory A Rational
Model for Structural Concrete in Pure Torsion, ACI Journal, Vol. 71, August, pp. 396408.
Miyakawa, T., Kawakami, T., and Maekawa, K. (1987), Nonlinear Behavior of Cracked
Reinforced Concrete Plate Element Under Uniaxial Compression, Proceedings of the JSCE
No. 378, August, pp. 249258.
Morley, C.T. (1979), Yield Criteria for Elements of Reinforced Concrete Slabs, Plasticity in
Reinforced Concrete, IABSE Colloquium, Kopenhagen, 2123 May, Introductory Report,
Reports of the Working Commission, Vol. 28, pp. 3547.
Mller, P., (1978), Plastische Berechnung von Stahlbetonscheiben und Balken, Dissertation
Nr. 83, Birkuser Verlag Basel und Stuttgart.
Nielsen, M.P. (1963), Yield Conditions for Reinforced Concrete Shells in the Membrane
State, NonClassical Shell Problems, Procedings of the I.A.S.S. Symposium, Warsaw,
25 September, pp.10301040.
Nielsen, M.P. (1971), On the Strength of Reinforced Concrete Discs, Acta Polytechnica
Scandinavica, Civil Engineering and Building Construction Series No. 70, Copenhagen,
261 pp.
Nielsen, M.P. (1984), Limit Analysis and Concrete Plasticity, PrenticeHall, Englewood
Cliffs, 420 pp.
Nielsen, M.P., (1999), Limit Analysis and Concrete Plasticity, 2nd Edition, CRC Press LLC,
908 pp.
Pang, X.B., and Hsu, T.T.C. (1992), Constitutive Laws of Reinforced Concrete in Shear,
Research report UHCEE921, Department of Civil Engineering, University of Houston,
Houston, Texas, 188 pp.
Park, R., and Gamble, W.L. (1980), Reinforced Concrete Slabs, John Wiley & Sons, New
York, 618 pp.
Robinson, J.R., and Demorieux JM. (1977), Essais de Modeles dame de Poutre en Double
T, Annales de lInstitut Technique du Btiment et des Traveaux Publics, No. 354, October,
Serie: Beton No. 172, pp. 7795.
Thomson, W., and Tait, P.G. (1883), Treatise on Natural Philosophy, Vol. 1, Part 2,
Cambridge University Press.
Vecchio F.J., and Collins M.P (1986), The Modified Compression Field Theory for
Reinforced Concrete Elements Subjected to Shear, ACI Journal Proceedings, Vol. 83,
No. 22, Mar.Apr., 219231.
Vecchio, F., and Collins, M.P. (1982), The response of Reinforced Concrete to Inplane
Shear and Normal Stresses, The Department of Civil Engineering, University of Toronto,
Canada, March, 332 pp.
Wood, R.H. (1961), Plastic and Elastic Design of Slabs and Plates, Thames and Hudson,
London, 344 pp.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 81
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
[Blank Page]
!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
For some design and analysis problems, however, a linear analysis may not be sufficient. As a
simple example, consider the requirement of satisfying a serviceability limit states such as
calculating deflections and crack widths. For such a designcheck, the local extent of cracking
is important and a linear analysis may not be sufficient to confirm that this limit state has been
satisfied. While it is practical to design structures with linear analyses, it is sometimes helpful
to check additional limit states by nonlinear analysis methods. For new structures a nonlinear
analysis may be performed on the structure (or its elements) after initial proportioning using a
plasticitybased design procedure based on a linear elastic analysis (Chapter 2). In addition,
nonlinear analysis can also assist in the evaluation of complex geometry or poorly detailed
structures where the effects of localized cracking, for example, may be poorly modelled by
linear analysis. The following list provides some general cases where a nonlinear analysis
may prove useful:
1) Confirmation of safety for complex design details: Design codes offer methods to check
safety for standard members such as beams and columns but are often not sufficient for
more complex design geometries such as complicated walls systems with openings. While
the inclusion of strutandtie methods into modern design codes has assisted greatly in
expanding the generality of design codes, nonlinear analysis can be helpful to allow an
independent check of whether the development of cracking, for example, will be
consistent with the details of the reinforcement placement.
2) Assessment of safety of existing structures: Some structures, such as those built to older
design standards, may contain reinforcement quantities and details that are not consistent
with modern standards. In such situations there may be very significant costs associated
with strength upgrades of the structure and a nonlinear analysis can assist in providing a
better estimate of the safety factor of the asbuilt structure against collapse. This can be
particularly important in seismic regions where design earthquake forces have tended to
increase in severity through the years as knowledge of seismic hazards have improved.
3) Pushover analyses for structural capacity computation: Many codes of practice require a
pushover analysis of structural strength to estimate failure modes and to determine if the
collapse mode conforms to capacity design principles. That is, it is important to be able to
predict how a structure is expected to collapse so that it may be designed to do so in a
ductile mode such as flexure, rather than a brittle mode such as shear.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 83
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
behaviour. This allows a higher level of confidence to be placed in the calculated safety
factor than that obtained from a blind analysis.
5) Estimation of PO effects: The influence of geometric nonlinearity, in the form of the PO
effect, can be important in the moment and axial load design of frames. As concrete
material science continues to improve, it can be expected that columns of higher
slenderness will become more common and these will be more sensitive to the effect. The
combination of material nonlinearity, in the form of cracking, and geometric nonlinearity,
is a complex phenomenon that can require nonlinear finite element analysis.
7) Resistance to fire: Structures that must be designed to resist significant fire can show
major effects resulting from the thermal expansion of parts of the structure that may not be
captured by linear analysis. Considering the effects of fire and structural collapse is a
nonlinear problem that requires complex analysis.
As reinforced concrete is a nonlinear material, the salient nonlinear effects that concrete
shows are described next. Afterwards, the various frameworks with which these nonlinear
effects can be incorporated into finite element solutions are described followed by a short
discussion concerning the expectation of precision in concrete nonlinear analyses. Finally, a
short discussion of safety reliability calculations is presented. In each case, the description of
behaviour or implementation is not written to be exhaustive but simply to demonstrate to the
reader some of the more important aspects that are required for the engineer to use nonlinear
finite element tools intelligently.
Energy : Energy balance may need to be considered to fully model listed effect
Stiffness: Elements must include variable stiffnesses to fully model the listed effect
Bifurcation: Indicates a discrete change in behaviour rather than a smooth transition
Plastic strains: Indicates that simple nonlinear elasticity may not provide a full solution
Size dependent: Behaviour cannot be fully modelled with strain terms alone. Modelling also requires an
absolute distance relationship such as a crack width
Size dependent
Plastic strains
Bifurcation
Stiffness
Plain Concrete Behaviour Energy Examples of behaviour
Tension Macrocracking X X X
Tension softening * X X X
Cyclic response X X
Creep X X
Crack closing effect X X X
Shrinkage X X X
Compression Crushing X X X
Nonlinearity at high strains X X (Reinhardt and Xu, 1999)
Postpeak unloading X X X X
Cyclic response X X
Creep X X
Rate of loading X X
Bi or triaxial confinement X X
Poisson's ratio X
Thermal effects X X
(Sakai and Kawashima, 2006)
Reinforcement Behaviour
Tension Yielding X X
Strain hardening X X
Thermal effects X X
Rate of loading X X
Rupture X
Compression Buckling X X X
Shear Dowel action X X X
(Ozcebe, Ersoy, Tankut, 1999)
Combined Concrete & Reinforcement Behaviour
Tension Bond X X
Tension stiffening * X X X
Tension splitting X X X
Compression Compression Softening X
Shear Aggregate interlock X X
* Tension softening is distinct from tension stiffening behaviour. See text sections 3.2.2, 3.2.3.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 85
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
plain concrete in tension will produce different loaddeformation curves after cracking if
meshes with different element sizes are used and special measures are not taken (Baant and
Oh, 1983, Baant and Planas, 1998).
While plain concrete can be subjected to localization effects under increasing deformations,
these effects are often not of significant importance in structures reinforced with a minimum
quantity of ductile reinforcement. With such structures, the ductile properties of the
reinforcing steel dominate the global behaviour. Examples of behaviour where localization
effects can have a significant influence on behaviour include punching shear, shear in
members without transverse reinforcement, overreinforced columns and beams and large
unreinforced concrete structures such as dams. In each of these cases the behaviour at failure
is more controlled by that of plain concrete than that of the reinforcement.
In using Table 3.1, and nonlinear finite element procedures in general, engineers must check
their finite element discretization and modelling assumptions used in their analyses. Note that
many of the attributes in the table are contentious in the research community and, as such, are
open to debate. Responsible engineers must use experience, judgment and intuition as to
whether, for example, localization issues will be important. While it may seem that nonlinear
finite element analyses could allow nonengineers to produce reasonable and accurate results,
this is not the case; the experience and judgment of engineers is more important in nonlinear
finite element analysis than it is with linear analysis due to the significantly higher complexity
level.
The following sections expand on a subset of the important behaviour in Table 3.1 explaining
the particularly important experimentally observed characteristic nonlinear concrete
behaviours.
Concrete subjected to compression acts in a nonlinear way as has been recognized for many
years (e.g. Turneaure and Maurer, 1908). Three important aspects of concrete compressive
behaviour are described in more detail here: localization in compression, confinement of
concrete, and compression softening.
As concrete shows a decrease in stress for increasing strains beyond the strain associated with
the compressive strength, localization of concrete in compression can be expected. That is, a
damage band forms where all the deformation associated with the postpeak region occurs
while the remainder of the specimen unloads with decreasing strain. Two primary effects will
result from this; firstly larger specimens failing primarily in a compressive mode can be
expected to show less ductility in the postpeak region than smaller specimens and, secondly,
a size effect is predicted for some member types whereby larger elements in compression will
be weaker in terms of stress than smaller elements. Both of these effects may be of particular
importance in practice as most laboratory testing is done on smallscale models and, thus,
may not be representative of the real safety of large structures in the field.
It is important to note, however, that there remains much debate on the importance of
localization effects for practical structures that contain realistic quantities of reinforcement
even within the laboratory testing field. For example, some tests on specialized structural
components confirm the presence of a size effect in unreinforced flexural compression zones
(Kim, et al. 2000). In contrast, other carefully performed tests on actual laboratory beams
subjected to constant moment have not always shown the decrease in ductility expected with
size (Alca et al. 1997).
It is well known that concrete can carry higher compressive stresses with larger deformations
when it is laterally confined as illustrated in Figures 3.1 and 3.2. Finite element models can
require either biaxial or triaxial confinement relationships. Early experimental investigations
on biaxially loaded RC elements focused on the strength. Kupfer et al. (1969) reported on the
strength, deformational characteristics and microcracking of concrete under biaxial stress
conditions. Figure 3.1 illustrates the stressstrain relationships under biaxial compression. As
can be seen in both Figure 3.1a and Figure 3.1b, the maximum compressive strength of
concrete increases for the biaxial compression state. Kupfer et al.s (1969) results suggest that
a maximum strength increase of about 25% is achieved at a stressratio of +2/5+1 = 0.5. In
addition, concrete ductility increases under biaxial compression.
(a) (b)
Figure 3.1: Concrete in biaxial compression: (a) stress versus strain; and (b) strength envelope (Kupfer et al.,
1969)
It is interesting to note that during the latter stages of the loading, an increase in overall volume
occurs (dilatancy); this is typically attributed to the progressive growth of micro cracking in
concrete.
For triaxial compression, Richart et al. (1928) conducted tests at low and moderate confining
stresses with typical results shown in Figure 3.2a. Balmer (1949) conducted triaxial tests at
very high confining stress levels (Figure 3.2b). As can be observed from the figures,
depending on the confining stress, concrete may act as a quasibrittle, plasticsoftening or
plastic hardening material. This is mainly due to the fact that under higher confining stresses,
the possibility of bond failure between cement paste and aggregates is significantly reduced,
and the failure takes place through the crushing of the cement paste. As can also be seen,
pronounced increases in strength with increasing confining pressure were observed by both
Richart et al. and Balmer. In fact, experiments show that concrete has a fairly consistent
failure surface that is a function of the principal stresses under triaxial loading.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 87
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
(a) (b)
Figure 3.2: Triaxial compression stress versus axial strain: (a) low confinement (Richart et al., 1928); and (b)
high confinement (Balmer, 1949).
One of the earlier 3D failure surface models that incorporated this strength enhancement was
that of Ottosen (1977); shown in Figure 3.3. While there are many models that have been
published, the important point is not which of the models is used but whether or not the model
produces a fair representation of the failure surface. Many commercial packages will include,
for example, a MohrCoulomb failure model. While the MohrCoulomb failure criteria can be
made to represent concrete strength at very high levels of confinement, as at these levels the
failure surface for the concrete approaches circular, the model does not reproduce well the
failure surface for concrete under the levels of confining pressure typical of that in common
reinforced concrete structures. If the MohrCoulomb approach is to be used it must be tuned
for the problem being investigated. The better software packages for concrete analysis will
include a specific failure model for concrete but even then some care is needed as some
failure surface models require different calibration for biaxial compression than for other
triaxial compressive stress conditions.
Figure 3.3: Ottosen (1977) failure surface for triaxial (+1, +2 +3) stresses (plotted in the HaighWestergaard
coordinate space).
Many different stressstrain models exist to account for either the biaxial or triaxial behaviour
of concrete in compression. These depend on many parameters and are generally specific to
continuum modelling of triaxially loaded materials or are specific to the behaviour of
columns, for example. One model which is often used in nonlinear elasticity is the Kent and
Park (1971) model later modified by Scott et al. (1982) to include the strength and ductility
enhancement due to confinement effects and the effect of strain rate (Figure 3.4). Such
models, however, ignore size and compression localization effects and can lead to erroneous
(but usually conservative) results when considering post peak behaviour. Such models are
usually sufficient, however, in adequately determining the response of a member for
increasing loading up to the maximum load. Nevertheless, as for all nonlinear modelling, the
results should be scrutinized with due caution. In analyses of cracked concrete in
compression, confinement may often be ignored with little penalty as the confining stress is
almost completely released normal to the cracks.
Kfc#
)50h
0.5Kfc# Confined concrete
0.2Kfc#
Unconfined
concrete
) 0=0.002 ) 50u )50c ) 20c )c
Figure 3.4: Modified Kent and Park (1971) model (Scott et al., 1982).
A related but not identical concept to confinement is that of weakening and softening in
compression due to the effects of transverse tensile strains beyond cracking (Collins 1978).
Figure 3.5 shows the results from many shear tests as reported by Vecchio and Collins (1993)
to quantify this effect. The specimens in this figure were subjected to combinations of shear
and axial loads and were reinforced in such a way that they were able to resist large tensile
strains transverse to the direction of the applied compression. As can be seen in the figure,
some of these specimens failed by crushing of concrete in compression at only 2040% of the
strength of concrete control specimens. As strains perpendicular to the direction of applied
compression increase, this strength reduction effect, often called compression softening,
becomes more severe. This effect can be thought of as an extension of the confinement effect
into the tensile domain and is particularly important in understanding of the behaviour of
concrete in shear.
The one type of nonlinearity that is expected in all concrete structures is cracking. This
phenomenon is difficult to account for in a simple fashion, yet is vital in making realistic
estimates of structural stiffness.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 89
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 3.5: Peak principal compressive stresses versus the principal tensile strain ratios (Vecchio and Collins,
1993)
When plain concrete cracks, it is able to resist residual, direct, tensile stresses across the crack
for small crack widths. These stresses, sometimes called bridging stresses or cohesive stresses
are sufficiently small that they are often ignored in traditional beam design. The decrease in
these bridging stresses as the crack width increases is called tension softening and models
based on the energy of fracture (Gf) are usually used to quantify the effect. Overall the
behaviour of cracking in plain concrete is reasonably well modelled by methods such as those
based on fracture mechanics.
In concrete that is reinforced, the process of cracking is more complex. Shown in Figure 3.6
are the multiple cracks that form in reinforced concrete around the bars (bond cracking) as
well as macrocracking that extends to the surface of the member (Goto 1971). Each of the
cracks shown in Figure 3.6 can be expected to behave consistently with the tension softening
behaviour mentioned above for plain concrete. When the effects of all these cracks are
integrated together, including the bond cracks, and the entire concrete section is considered as
a whole, behaviour consistent with that shown in Figure 3.7 is obtained and the concrete
component is termed tension stiffening.
This tension stiffening is the difference in behaviour between the response of a barebar and
the observed response of the composite reinforced specimen. This tension stiffening can be
thought of as a macroscopic property applied on a largescale compared to tension softening
which is a property on a smallerscale. As Figure 3.8 shows, an analysis with very small
elements that considers the bond cracking and a tensionsoftening model does produce the
same macroscopic tension stiffening behaviour as that observed in experiments.
Tensile force
d
Y
Member response
c
Tensile force
a Uncracked
R b S b Crack formation
c Stabilized cracking
Bare bar d Postyielding
a R First crack
Reinforcement S Last crack
(unembedded) Y Yielding
%"#
& 61)78+/01//+49+
31)3:8+34.310
%
Macro tension stiffness model
$"#
!"#
Computed from micro analysis

!
Plain concrete! #!! $!!! $#!! %!!!
softening & '()*+, ).+/01)(.+2, (3145
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 91
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
maintained ignoring all concrete stresses beyond those expected by tension softening. This
equilibrium criterion is why the macroscopic concrete stresses are called tension stiffening
and not tension strengthening: they cannot increase the strength of the tension specimen in
Figure 3.9, only increase its stiffness before yield at a crack.
Considering the differences noted in the previous section between tension stiffening and
tension softening, it is perhaps not surprising that there is more than one way to include
tension in finite element analyses of concrete structures. It is helpful to divide up the level of
analysis complexity into the following three levels:
With smallscale tension modelling, as shown in Figure 3.8, it is necessary to use extremely
small elements to capture the extensive cracking that occurs due to bond and slip effects and
the analysis should be fully threedimensional. These models may use the assumption of
perfect bond whereby the nodal displacements of the concrete adjacent to the reinforcement
are assumed to be the same as that of the reinforcement itself. Due to the number of elements,
this type of modelling is rarely practical for realistically sized structures at the present time.
With mediumscale tension modelling, the bond forces are accounted for with a macroscopic
bond model that provides a relationship between the relative displacements of the concrete
compared to the reinforcement. While this simplification means that more nodal degrees of
freedom are required, it has the advantage that much larger elements are allowable and two
dimensional analysis becomes possible. Larger elements are possible as it is no longer
necessary to be able to resolve the extensive bond cracking that the smallscale tension
modelling requires. If a concrete tension softening relationship is used, however, it will still
be necessary to use small elements to capture the change in concrete stresses between the
cracks. If, instead, a concrete tension stiffening effect is included, then still larger elements
may be used, though with the cost that special routines may be needed (depending on the
formulation) to ensure that the stress in the reinforcement at a crack does not exceed the yield
stress.
Largescale tension modelling is perhaps the most practical method of including concrete
tension in nonlinear analyses currently. Perfect bond between the concrete and reinforcement
again becomes appropriate and much larger elements may be used in two or three dimensions.
Care is required to ensure that the tension stiffening relationship used is consistent with the
bond properties of the reinforcement and that local conditions at a crack are checked for
equilibrium. Due to the use of perfect bond, effects caused by insufficient development of
reinforcement must also be explicitly considered. A reasonable way to account for this effect
with perfect bond programs is to progressively increase the cross sectional area of the
reinforcement over the development length of the bar to properly model the total force that
can be transferred to the concrete at every location. Finally, again, it is necessary with all
analyses based on tension stiffening to ensure that the stresses in the reinforcement at cracks
are not allowed to exceed the yield stress with some sort of crack check. For further
discussion, see Vecchio and Collins (1986), Petrangeli and Obolt (1996), Maekawa and An
(2000) and Foster and Marti (2002, 2003).
Modelling of tension stiffening can be undertaken in one of two ways: the first is to modify
the stiffness of reinforcing bars; the second is to modify the concrete stiffness to carry the
tension force after generation of cracks. In the first model, the stress transfer carried by both
steel and concrete due to bond action can be expressed by changing the stiffness of
reinforcing bars, as per Figure 3.7b (Gilbert and Warner, 1978). The CEBFIP Model Code
(1993) gives the following phases: uncracked concrete, crack formation, stabilized cracking
where only crack opening occurs and postyielding. While such a treatment has an advantage
in terms of simple 1D calculation of response, this is not generally applicable to 2D or 3D
problems since the stiffness is unchanged even if the direction of cracks relative to the axis of
the bar changes (Okamura and Maekawa, 1991).
The second approach of modifying the concrete properties can be subclassified as i) methods
based on a macroscopic smeared crack model, which describes spatially averaged behaviour
of RC containing multiple cracks in a defined control volume, or ii) methods based on
mesoscopic discrete crack models.
Tension stiffening models based on smeared crack concepts use a spatially averaged
constitutive relationship, which involves multiple cracks and nonuniform local stress
between the cracks in a control volume. Figure 3.10 gives three examples of models
representing average stressstrain relationships of concrete in tension. Since this type of
modelling is generally assumed independent of the spacing of cracks, direction of reinforcing
bars and reinforcement ratio, the stabilization and distribution of propagating cracks is assured
in a control volume. For this reason, it is one of the most versatile and applicable constitutive
relationships in executing structural analyses in cracked reinforced concrete.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 93
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 3.10: Examples of stressstrain models of concrete in tension (from Bentz 2005).
1 1
+' G +( x, y )dv Local stress +' G +( x, y )dv Local stress
VV VV
1
)'
VG)(x,y)dv
v
Yield plateau jump!
1
)'
V G
)(x,y)dv
v
(a) (b)
1 1 Local stress
+' G +( x, y )dv Local stress +' G +( x, y )dv
VV VV
Strain distribution
Stress distribution
Stress distribution
Strain distribution
Plastic zone
1
)'
V G
)(x, y)dv
v
1
Plastic zone )'
V G
)(x, y)dv
v
(c) (d)
Figure 3.11: Average stressstrain relationship of a steel bar in concrete: a) at crack initiation; b) at beginning
of yield; c)developing yield; d) full yield (after Maekawa and An, 2000).
even though the parts of the bar away from the cracks are still linearelastic (Figure 3.11a). As
the strain increases, the yielded portion of the bar is extended (Figure 3.11b and c). Then,the
average yield stress becomes lower than that of bare bar. For accurate prediction of
deformational behaviour of RC members after yielding, the constitutive relationship capable
of considering this phenomena is needed. Usually a bilinear or trilinear relationship is
assumed for the average response of reinforcing bars.
For normal concrete combined with twoway reinforcement with a ratio of 0.1 to 2 percent in
each direction, the constitutive law based on a smeared crack approach is independent of the
control volume (Maekawa et al., 2003). In other words an average stress can be evaluated
uniquely for an average strain history regardless of density of macroscopic cracks within the
control volume and, therefore, there is no size effect in the constitutive law for reinforced
concrete in a uniform strain field. Similarly it has been shown that mean stressstrain
relationships of reinforcement in RC plates are also not influenced by the number of cracks
within an element, including in the postyielding regime. This observation is convenient for
many structural analyses as it is not necessary to define individual macroscopic cracks within
the control volume for members that are provided with minimum, or greater, reinforcement.
This is not the case, however, where minimum reinforcement is not provided and cracks pass
through an unreinforced (or underreinforced) concrete section. The case where the
distributed cracking assumption is not appropriate and cracks are localized is discussed in
detail in the following section.
A relatively new approach to the modelling of tension stiffening has been developed by Marti
et al. (1998) known as the tension chord model (Figure 3.12). The significance of this
approach is that an equivalent plastic bond stressslip model (Figure 3.12c) is explicitly
included in the formulation allowing for calculations of crack spacing, crack widths and
tension stiffening directly from the model. The model was further developed into a 2D
formulation (the cracked membrane model) by Kaufmann and Marti (1998), shown in Figure
3.13, and included in a FE model by Foster and Marti (2002, 2003). With adoption of the
stepped rigidperfectly plastic bondslip relationship (Figure 3.12c), the stresses in the steel
and in the concrete can be determined for any point within the differential element between
cracks (Figure 3.12b). The stresses in the concrete and in the reinforcing steel at and between
cracks are then calculated directly from equilibrium.
(a) srm
] (b) N N
Ac
dx
x
,b dx
(c)
+c+ d+c +c
, b0
,b +s+ d+s ,b +s
, b1
Oy O
Figure 3.12: Tension chord model: a) reinforced tension chord; b) differential element; c) bond shear stress
slip relationship (Marti et al., 1998).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 95
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
m
sr
m
cft
sr
srmy
r
srmx
X
Tension Stiffening
x ct
Stresses
srmx srmx
^555f
Figure 3.13: 2D tension stiffening stresses in: a) material axis directions; b) equivalent orthogonal tension
chords (Kaufmann and Marti, 1998, Foster and Marti, 2003).
If a discrete modelling approach is preferred over that of a smeared crack approach, for the
modelling of reinforced concrete members, it is necessary to explicitly deal with: each
individual crack that propagates around the reinforcing bars; the distribution of local stress
and strain in steel bars; and the interaction at the interface between concrete and bar (that is,
stress transfer through bond, bond deterioration, and so on). The crack propagation around a
steel bar is caused as a result of threedimensional strain fields and bond transfer mechanisms
and, since the fracture process in 3D stress field differs from that in 2D, a 3D analysis is
appropriate for modelling of reinforced concrete elements using discrete crack models (Morita
and Kaku, 1979, Maekawa et al. 1999). In adopting this approach, fine meshes are needed
with many degrees of freedom and, thus, the use of discrete crack models for the analysis of
reinforced concrete members has generally not gained favour over that of the more general
distributed smeared crack approach. For problems that involve tensile fracture of the concrete,
however, discrete, crack band or nonlocal models are needed to capture localized behaviour
and is discussed in more detail in the next section.
An important issue with the modelling of tension in finite element models is the way that
cracks are considered. In general they can be taken either as smeared throughout the element
or only present at finite element boundaries which is also called discrete cracking.
The second option is to model the cracks discretely by allowing gaps to form between
different elements. This is the option taken by discrete element and lattice models discussed
below and these models are generally based on tension softening relationships.
Just as care is required when considering the stressstrain relationship between tension
stiffening and tension softening, it is also important to consider that localization problems can
occur for smeared cracking problems as well. Consider the finite element mesh shown in
Figure 3.14b. The structural cracks are shown within the elements rather than between them
indicating that this is a smeared crack analysis. However, the cracks are only within one line
of elements indicating that virtually all the displacement is localized within this region of the
member. As such, this extent of cracking in the model is consistent with a tension softening
type of formulation which is designed to account for localization. In contrast, the mesh in
Figure 3.14a shows cracking smeared throughout the tension region and thus no localization
problems are present, consistent with a tension stiffening relationship.
Element to Model
a) no localization b) localization
Figure 3.14: Localization in Reinforced Concrete FE Analysis.
A final issue to consider with the modelling of cracking in concrete relates to the effects of a
change in principal average stress directions that can often results from redistribution. Initial
cracks are usually assumed to form at the angle of principal tensile stress in the continuous
material before cracking, but after cracking the average stress directions may no longer align
with this axis. In general there are two approaches for dealing with this: fixed angle cracks
and rotating angle cracks. Both are approximations.
Experiments on large elements subjected to pure shear have shown that the angle of principal
strains, averaged over a sufficient gauge length to include several cracks, can rotate
significantly between first cracking and ultimate limit states. In practice, new cracks form at
different angles to the previous cracks as the loading increases. This behaviour is also
observed in beams with stirrups as new shear cracks form at different angles to that of initial
shear cracks. Rotating crack models, therefore, allow the angle of cracking to change as the
analysis progresses to capture this effect. By contrast, a fixed crack approach forces the angle
of the cracks to remain constant after initial cracking with the shear stresses calculated on that
crack surface usually checked against aggregate interlock relationships.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 97
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Introduction of steel into a reinforced concrete analysis, particularly one that includes smeared
concrete tension stiffening stresses, should consider the effects of those concrete stresses on
the overall behaviour. In general, this can be done by adjusting the concrete tensile behaviour
at first yield at a crack or by adjusting the steel behaviour at first yield. Figure 3.15a) shows
the option of using barebar stressstrain relationships for the steel summed with a smeared
concrete tension stiffening stress. At first yield at the crack, the concrete tensile forces are
explicitly reduced to zero to ensure that the steel stress at a crack does not exceed yield. In
this example, it is conceptually assumed that strain hardening at a crack will not occur. Fig
3.15b considers the opposite option of leaving the concrete stressstrain relationship
unchanged and deriving new equations for the embedded steel behaviour. Both of these
methods produce the same overall answer in this example. In any case, it can be seen that
simply adding the average concrete tension stiffening stresses to the bare bar steel stresses
would violate equilibrium at a crack after first yield at a crack. Finally, special care must be
taken in analyses that involve nonductile reinforcement such as glass or carbon fibre or low
ductility steel reinforcement. In such situations, the increase in reinforcement stresses that
occur at the cracks will result in rupture when the average strain is still well below the rupture
strain of the material. Figure 3.16 shows a comparison of the observed components
contributing to the tension force for a steel bar embedded in concrete for three tests of Shima
et al. (1987).
The most familiar continuum model is the elasticity model where loading and unloading are
pathindependent and follow Hookes law (Figure 3.17). Structural steel before yielding and
structural concrete before cracking can be modelled well by such a linear elastic model.
Purely elastic models can be of arbitrary complexity, and need not require isotropic
behaviour. Orthotropic modelling of concrete has been found to work well when
consideration is given to the observed tension behaviour being dramatically different from
compressive behaviour in principal directions. The most general anisotropic modelling allows
all 21 independent elastic moduli in three dimensions to be defined.
stress
fc fc
average steel stress
fcr fcr less than yield
fs fs
strain strain
(a) Bare bar (b) Embedded steel
0.3
Specimen no.4 Specimen no.5 Specimen No.6
Load (MN)
Bare bar
0.2
Bare bar
Bare bar
0.1
Tensile force Tensile force Tensile force
0
0.8 Bare bar
Stress (GPa)
Bare bar
Bare bar
0.4
2
Average stress of Average stress of Average stress of
concrete concrete concrete
1
0 1 0 1 0 1 2
Average strain (102) Average strain (102) Average strain (102)
Figure 3.16: Contributions to the tension force in a reinforced concrete element and corresponding stresses
(Shima et al., 1987).
One direct result from the analysis of elastic bodies is that the stress distributions around the
tips of cracks appear as real values in the output. Simple elastic theory for a fine crack
predicts a stress singularity at the tip of the crack, which is clearly impossible. Concrete
responds to this by the formation of a fracture process zone just ahead of the crack and
cohesive bridging stresses after cracking. Thus, even linear elastic models contain
approximations to reality after cracking.
The accuracy of deformation by linear analysis is far from the reality for cracked concrete
structures but in many cases of ultimate limit state verification, computed section forces by
linear analysis provide reasonably close approximations to the exact nonlinear solution. This
is because the accuracy of section forces does not primarily depend on the absolute values of
section stiffness but on the relative stiffness profile. This is the reason why the linear analysis
is often adopted as a computational tool for design so to obtain the section forces even for
ultimate limit state examination.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 99
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
stress
stress
stress
elastic strain
linear elastic strain plastic hardening strain recoverable strain but
(recoverable strain) (irrecoverable) irrecoverable stiffness
RC modeling
+Y Average shear stress
7
, (MPa ) Average shear stress
7
, (MPa )
,YX 6
C 6
C
5 5
4 4 Y
+X
3 3
Y 2 2
LX 1 1
,XY 0
0 2 4 6 8
0
0 2 4 6 8
Average shear strain ) ( 10 3 ) Average shear strain ) ( 10 3 )
X examples of computed stressstrain
Figure 3.17: Basic concepts of elasticity, plasticity, damage and their combination.
3.3.2 Plasticity
Perfect plasticity with no hardening is useful when solely determining a maximum capacity
for a given structural element. This capacity is derived by assuming that all the steel yields
and the solution is therefore relatively easy to obtain. In general, the ultimate strengths
obtained from the nonlinear analysis methods in this practitioners guide will be below or at
most equal to the strengths obtained from an assumption of perfect plasticity. Note that due to
the general inclusion of an elastic component coupled with a plastic component, the concepts
of stresses at crack tips, perhaps modelled with fracture mechanics of concrete, fit equally
well with plasticity models as with purely elastic methods.
3.3.3 Damage
Methods based on damage mechanics allow the elastic stiffness of the material to be reduced
as a result of accumulation of damage from strain excursions or repeated loadings. This
nonlinearity may represent the loss of mechanical volume or degraded capacity to absorb the
elastic energy, which is thermodynamically reversible in nature. Concrete shows an
irreversible reduction of elastic unloading/reloading stiffness when small lateral confinement
is provided. When higher lateral confinement is applied to concrete in compression, damage
of the elasticity properties is observed in softening of the unloading and reloading paths.
Damage is also seen in concrete tension after cracking as described above.
Models based partly on elasticity with some basic concepts from plasticity and damage
mechanics can provide simple models that can capture many important nonlinear effects in
reinforced concrete materials and reinforced concrete structural elements. One of the simplest
examples is a notension linear elastic concrete model coupled with an elastoplastic steel
model. This is a well known simple approach, which has been popularly used for flexural
behavioural computations of reinforced concrete sections based on EulerKirchoffs inplane
assumption. For both confined and cracked concrete in compression, combinations of
elasticity, plasticity and damage can be consistently applied in structural analysis.
The second way of utilizing a discrete element model is for a continuous body where the
locations of the discontinuities (cracks) are not known a priori when the mesh is created. One
option is then to explicitly remesh the analysis space at each load step to follow crack
propagation in a more explicit way. For fracture simulation, Ingraffea and Saoumu (1984)
first introduced the discrete approach to concrete structures. The discontinuities of the
displacement field resulting from the failure processes are introduced directly into the
numerical model. New elements are added at the crack surface to explicitly account for
tension softening and aggregate interlock. In this approach, the crack location is instantly
recognizable but it can be a lengthy procedure especially for threedimensional analyses.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 101
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
An alternate approach is to implicitly assume that all elements have cracklike gaps between
them from the start that may or may not be active. Methods such as lattice models fall into
this category (e.g., van Mier et al., 1995, Bolander and Saito, 1998). The discrete crack gap
can be numerically embedded in finite element volumes and the connectivity of these
embedded crack gaps can be satisfied on the element boundaries based upon hybrid stress and
strain hypotheses. Thus, differences between smeared and discrete approaches is becoming
less distinguishable for geometrically continuous domains and, while the methods were first
applied to the analysis of material level problems, large structural problems can also be
modelled as has been demonstrated by Bolander et al. (2000) and Bolander and Hong (2002).
Let the quantities (qo , & p) denote the displacement and current load vector at a given point
on the equilibrium path, as shown in Figure 3.18, where & is a load magnification parameter
and p is a reference vector of applied loads. The equation governing equilibrium can then be
written as
or
where f is a vector of internal forces and is a function of the current displacement state.
Eqs. 3.1 and 3.2 describe the equilibrium state of the discretized structure. While the solution
of the linear equation system
Kq f = 0 (3.3)
can be calculated directly, this is not possible for nonlinear systems. Solution techniques for
nonlinear systems typically require the solution of linear systems repeatedly until
convergence is obtained with the most frequent schemes used based on a Newton method.
The most frequently used iteration scheme for the solution of nonlinear equations is some
form of the NewtonRaphson procedure. In the case where Eq. 3.1 cannot be solved exactly, it
can be shown that (refer Chapter 6) the solution may be progressively calculated by adding
the change in displacements O qi to the current displacement state calculated from
71
Oqi ' K Ti r ( qi ) (3.4)
i
where KT is the tangent stiffness determined for the current state and r (qi ) are the out of
balance forces. The NewtonRaphson solution process is illustrated in Figure 3.18, noting that
for every step the current stiffness matrix is formed and the linearized equations solved for
O qi .
From the above calculations it is seen that the formulation of a new tangent stiffness for each
iterative cycle, and the solution of a new system of equations, must be undertaken. In
computing terms, this can be time consuming and costly. To overcome this cost the
approximation K Ti ' K To is often made. This modifies Eq. 3.4 to
71
O qi ' K To r ( qi ) (3.5)
and resolution of the same equation set is repeatedly used. The solution at each iteration is
sped up; however, more iterations to convergence are required, as can be seen in Figure 3.19.
The overall economy of the solution procedure is dependent on the problem size and non
linear behaviour. An updated mNR approach may be adopted where the tangent stiffness
matrix, K Ti , is updated if convergence is not obtained after a predetermined number of
iterations, n, and continues to be updated every n iterations following.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 103
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
As an example of the variability that may be expected for some analyses, consider the data
shown in Figure 3.20 of measured stiffness values of concrete as a function of compressive
strength. Note the wide scatter in the observed values. Consider that at a concrete compressive
strength of 25 MPa, limestone aggregates have can produce concretes with stiffnesses varying
from 20,000 MPa up to 35,000 MPa. Calculations sensitive to the uncracked stiffness of
concrete, such as the displacements of a beam before cracking, can be expected to be similarly
variable. No structural analysis should be expected to provide extreme precision in results
whatever the complexity of the modelling.
Engineers should perform sensitivity analyses to ensure that a given set of results is not
simply a result of mesh sensitivity or unduly controlled by a parameter in which the engineer
has low confidence. For concrete, one also should take into account the difference of strength
between control specimens and structural concrete realized in the structures concerned. The
difference is caused by size effects, concrete placing work, shrinkage, temperature hysteresis
and so on.
When seeing a figure like Figure 3.20, proponents of high precision in nonlinear finite
element analyses may counter that such scatter simply indicates that the analysis was not
properly calibrated to the experiment that was to be modelled. Of course practicing engineers
know that for most design or analysis problems, there is no experiment to calibrate to. Claims
that extremely precise fits between experiment and analysis are guaranteed should be treated
with caution and not used to justify dramatically lower safety factors, for example.
Lastly, it is important that the model being adopted be verified against benchmark tests of the
problem type being considered. Benchmarking testing is discussed in detail in Chapter 7.
Figure 3.20: Modulus of elasticity as a function of compressive strength (Rashid et al. 2002).
The validation of material models are generally done by modelling of selected standardized
benchmark or control tests (for example, tests on reinforced panels or fracture of control sized
specimens). Alternatively, validation of the material models can be performed as a part of a
standardized numerical simulation task. An assessment of the capability of the material
models to perform, as required to simulate the behaviour of a structure, is normally
undertaken when selecting and assessing the appropriateness of the software package.
A validation of structural model is undertaken for a particular model and software package in
a specialpurpose study undertaken by means of benchmark calculations. In this process,
specific types of structural and material behaviour are tested. Such studies form a rational
basis for choosing appropriate material models and software for modelling of a particular
structure, structural member or element loaded under similar conditions. For example, if a
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 105
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
shear wall is to be simulated under a particular loading regime for a given set of boundary
constraints, a validation of the finite element software against shear wall experiments should
be performed using test data identified as being reliable. In this process of verification, it is
vital that all the potential modes of failure are identified and the model verified to show that
these failure modes are captured with accuracy.
For performance assessment, some indicators are necessary to suitably characterize the
structural responses. In general, there can be three types of indicators; force/stress related
indicators, deformation/strain related indicators and energy related indicators (force
multiplied by displacement or stress by strain). The first two indicators are practically useful
for static design and the energybased indicator for seismic design or damage control
(Akiyama, 1985). If our problem is a perfectly linear one, the same safety allowance can be
expressed no matter what sorts of indicators are used for the safety check, because force and
deformation are linearly correlated.
The safety checking scheme in most modern codes is based on limit states and associated
partial safety factors. In this scheme, the resistance or response is based on design (extreme)
parameters, which are derived from nominal ones using a reduction by the partial safety factor
#p. By this way, uncertainty of materials, geometry and other properties is included in input
data. Similarly, the design loads or actions are considered by extreme characteristic values,
where the load factors #load are considered. In a typical static situation, material characteristic
parameters are reduced, while loads are increased, both to design values. Safety of design is
checked locally at particular points, for example in beam sections, by the limit state condition
as
where R(&p) is the extreme local resistance or capacity of the structural response and E(&L) is the
extreme local action of load or response indicator. This condition assures the safety of each
local point. It does not indicate a direct measure of global safety of the structure but it
generally results in conservative and safer assessment under static loads. This method is
popular in practice, since it is easy to apply and current partial safety factors based on linear
analysis are available. However, when we work with nonlinear problems, linear correlation of
force and displacement does not hold in nature. Some codes such as Eurocode 2 (2002)
explicitly mention that the partial safety factor method is not applicable to nonlinear analysis.
The problem of applying the method of partial safety factors to nonlinear modelling is
demonstrated using two extreme cases. Nonlinear forces versus deformation relationships (left
side of Figure 3.21) are common in reinforced concrete mechanics. For example, we can
imagine relationship between the transverse shear force and the strain in the stirrup
reinforcement. If the strength of concrete and amount of web reinforcement are changed, we
have different static force versus deformation relationships, as shown in Figure 3.21. Let us
consider the steel yield as a limit state safety check and an allowable safety margin in terms of
strain (i.e., a deformation related indicator). Here, the corresponding safety allowance in terms
of shear (force related indicator) differs greatly. Since a force indicator is roughly
proportional to external static loads, it is thought reasonable to specify the safety allowance
(safety factor) by the forcebased index for expressing reliability of nonlinear analysis to
identify the limit state concerned. In fact, limit state design methods do not adopt stirrup strain
but apply sectional shear force as the limit state indicator for the static safety assessment.
Rcap
Rcap
static force, stress : R strain, displacement based index strain, displacement based index
For the case of elastoplastic problems like steel structures, however, any indicator identifies
the same safety allowance (see Figure 3.21) when the initiation of plasticity is selected as the
limit state because of the proportionality of internal/external forces and deformations. Then,
in the discussion of this chapter, the safety allowance or variation is depicted by force related
indicators. There exists similar feature on fatigue limit state design for service life.
Nonlinear analysis is typically a simulation of a loading test for an entire load history up to
failure, including the serviceability and ultimate limit states. In current codes there are
different partial safety factors for different load stages and failure modes. It is difficult to
apply current safety factors according to their definitions in nonlinear analysis. A material
model based on the partial safety factors represents an imaginary, not real, material. It does
not represent an extreme behaviour with a certain probability of failure. Automatic reduction
of all parameters may cause much weaker material than required by the safety concept. In
some cases, it can be an unsafe model. Thus, considering the design values of material
parameters in nonlinear analysis does not guarantee that the target safety values, as prescribed
and calibrated by standards, are achieved.
Let us first consider a simply supported beam under uniform loading, where the bending
moment in the midsection is completely independent of material behaviour. If we use a beam
finite element with plane section hypothesis, the nonlinear analysis is identical with the
current design methods (based on elastic assessment of internal forces) and the partial safety
factors can be applied. If we consider a statically indeterminate structure, the actions in
sections or material points depend on material behaviour and need not be proportional to the
magnitude of the external load. In general, a redistribution of internal forces due to nonlinear
behaviour can produce either positive or negative effects on local failure. An example of such
situation is concrete under a confining pressure where higher lateral pressures provide
increases in strength. In such cases, the application of partial safety factors is not justified.
Since the strong nonlinear behaviour can be expected in most of reinforced concrete
structures, it is concluded that the partial safety factor approach should be used with care in
statically indeterminate structures, especially when local postyield states are allowed in the
design.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 107
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Recently, static pushover analysis has been carried out in use of nonlinear FE analysis; not
only for seismic design but also for static safety design of statically indeterminate structures
composed of several members. The pushover analysis may present the whole structural
capacity under static loads and the scenario of collapse after the capacity limit state. Figure
3.22 shows a schematic of pushover analysis and the design value of loads. Up to the peak
capacity of the global structural system, some component members may fail or locally exceed
their capacity. Thus, the computed characteristic capacity of the global structural system
should be factorized by a global safety factor with consideration to the accuracy of the
nonlinear structural analysis. This factorized design value of the global system capacity has to
be greater than the overall design static load (vector). The performance assessment in terms of
the global system safety is advantageous rather than the conventional methods where all
constituent elements are required not to exceed their ultimate limit state. This means that
failure of any member is not accepted, even if the global structural stiffness remains positive.
When section failure of structural members is itemized as the limit state for structural safety,
some indicators to judge the occurrence of flexural and shear failures are needed. The
combined axial and flexural failure can be detected by monitoring the longitudinal
compressive strains of the most extreme fibre of the section, as shown in Figure 3.23. When
2D or 3D solid elements are used, pointwise local strains tend to be dependent on the sizes of
the finite elements and detailed patterns of mesh discretization (Chapters 3 and 6). Then, a
space averaged mean strain or stress indicator (sectional forces) is preferable as a design
index because the spaceaveraged indicator is not sensitive to the detail of the mesh
discretization.
Regarding the restorability of structures after extreme loads, material damage based indicators
are also proposed, especially in seismic designs for practice (JSCE, 2002). Damage induced in
structural concrete has a close correlation with repair/strengthening costs. Figure 3.24 shows
an example of design modelling for concrete used in fibre modelling (Chapter 4). The local
damage indicator denoted by F is assumed to be a function of maximum experienced
compressive strain in a load or timestep run of a nonlinear structural analysis with nonlinear
material models. The sectional averaged F value is used as a damage indicator in regards to
restorability after the application of extreme loads. The transverse displacement of the
columns corresponding to F = 0.5 is almost the same as the allowable ductility level specified
in the past design codes based on empirical macroscopic models of members. Confinement
(Chapters 3, 4 and 6) by lateral ties and steel is taken into consideration by modifying fibre
stressstrain relationships. As the lateral confinement moderates the damage intensity, the
damage evolution indicator F can be modified as shown in Figure 3.24(3). The damage
indicator was formulated to represent reduction ratio of the elastic stiffness on unloading
paths and is a measure of how much reversible strain energy is stored in the damaged
continuum.
computed characteristic
structural capacity
force (vector)
global safety
factored in terms of
design value of analysis model used
design value capacity
of loads
partial safety margin factored
in terms of failure/collapse modes
importance of structures, etc.
displacement
computed characteristic
structural capacity
force (vector)
design value of
capacity
global safety factored
design value
in terms of analysis
of loads
partial safety model used
factored in terms of
failure/collapse modes, etc.
displacement
Figure 3.22: Safety factor and the whole structural limit state by pushover analysis.
curvature
moment
curvature
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 109
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
(2)
(3)
(4) (5)
(6) (7)
Figure 3.24: Structural concrete damage based indicator (Tsuchiya and Maekawa, 2006).
In case of shell structures subjected to coupled inplane and outofplane actions, cracking of
concrete and yielding of reinforcement are likely to develop over a wide domain of the
structures because the reinforcement is uniformly distributed and flow of stress tends to
expand over the structure. Then, the structural response level at the start of reinforcement
yield is well below the capacity. In these types of structures, compressive localization can be a
design limit state for safety and serviceability performance (Chapter 6). Figure 3.25 shows a
transient process of compressive localization after cracking of the concrete and yielding of the
reinforcement in shells. When the principal compressive strain averaged within the control
volume reaches the limit value (nearly 5000B)), some elements exhibit high strain rate while
the strain rate of neighbouring elements decreases. In case of circular shells subjected to shear
actions, a localized band accompanying shearcompression kinematics can be seen when the
principal compressive strain reaches this limit state. Then, for underground tank design, the
averaged principal compressive strain is now used as the indicator for damage control with
regard to compressive localization. This limit state almost corresponds to the structural
capacity as shown in Figure 3.25. The rationale of this criterion was checked with
reinforcement volumes from 0.4% to 2.0% (Harada et al., 2001).
where Rm is the resistance or the capacity limit obtained by nonlinear analysis, as stated
above, based on the mean material characteristic parameters, &R is the global safety factor of
the corresponding resistance and E(&L) is the factorized external action or the response
indicator of the structural system concerned.
The global safety factor describes the safety of the system on a global level. Thus, it should
cover the uncertainties of all components of the structural system and analysis. Since safety is
related to the average resistance, it can be described by a global central safety factor. This
represents a generally accepted safety margin of usual structures produced according to
general standards. This covers wide and rather unspecified range of uncertainties with the
safety not related to specific random properties of a given structure or product. The practical
determination of global safety factor is more difficult comparing to partial safety
specification, where it is related to known random variations of dimensions and material
properties.
Therefore, the values of global safety factors are not yet proposed in the majority of codes.
For example, Eurocode 2 (2002) states that the partial safety factor concept is not applicable
to the nonlinear analysis, and does not propose a format for safety check. DIN 1045 (1998)
proposes the global safety factor for the system resistance &R = 1.3 valid only for ductile
modes of failure. Higher values should be considered in case of brittle models of failure such
as concrete shear and diagonal tension, or compression. The nonlinear analysis considered in
practical codes is mostly limited to the models based on beamcolumn systems. The mean
material parameters are based on nominal values that, typically, represent 5% probability.
e qud b e b
& Rd E $& G G 8 & QQ % ? Rc $
` or & Sd & Rd E & g G 8 & q Q ? Rc qud % ` (3.8)
c & gl ` c & gl `
d a d a
where, &G, &Q are the global safety coefficients respectively for permanent and variable loads,
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 111
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
270
0
;
<<
1.0 0.0 1.0 2.0 3.0 4.0 =
E E
%!!! " F): !" F):
Diagonal Cracking
931kN Collapse Load
C4)8+2:D 5
$!!! 2352kN
Flexural Cracking
735kN
!
>%"! >$"! !"! $"! %"! ?"! @"! #"!
>$!!!
>%!!! Test
Analysis COM3
>?!!!
A 941, )0(4.+&.7*+2B5
Figure 3.25: Structural concrete damage based indicator (Harada et al., 2001).
&Rd is the model uncertainty coefficient on the resistance side (suggested value gRd = 1.06), &Sd
is the model uncertainties coefficient on the action side (suggested value gSd=1.15), &gl is the
global structural safety factor (suggested value gg = 1.20, but gRd ggl = 1.27 = gGl), qud is the
ultimate level of the internal actions path, reached in the incremental process of nonlinear
analysis. For the case where no model uncertainties are considered, the inequalities in Eq. 3.8
are modified to
eq b
$
E & G G 8 & QQ ? Rcc ud % `` (3.9)
d & Gl a
The JSCE (1999) LNGTank design code presents full 3D nonlinear analysis of soilRC shell
interacting systems and proposes a factor of 1.3 for the global safety factor in regard to the
global deformational indicator, as shown in Figure 3.25. Here, we input the most probable
values for material characteristic parameters of structural concrete and soil foundation.
According to current practice, at serviceability limit states, we can use directly for the mean
response &R = 1.0, and unfactored loads or actions. At ultimate limit states, the global safety
factor should be &R < 1.3, depending on the ductility of the response.
Concerning the ultimate limit state, Mancini (2002) and Bretagnoli et al. (2004) proposed for
nonlinear static analyses that for the evaluation of qud, the analysis should be stopped when
the ultimate strength and the corresponding deformation are reached within the most critical
region and where the whole structure is incapable of supporting any further load increments.
An example of deep beam capacity is shown in Figure 3.26. During the analysis, the first
critical element initially crushed but the structure was able to carry further load increments up
to the crushing of a second element. At this point the model was unable to reach the
equilibrium for any further load increments. This last step has been considered as the final
point of the internal actions path. Figure 3.26 shows the resisting interaction surface for the
critical second element, and the internal action path in the same element, up to the intersection
with the resisting surface. Figure 3.26 also illustrates the procedure for the application of
safety format in a vectorial combination of internal actions and, by means of definition of a
safety interaction surface, derived by the limit one by linear transformation referred to the
axes origin. If the critical element and its internal paths are not clearly identified, the external
loads magnified by the global safety factor can be applied to the structural concrete. If the
magnified loads would not cause failure of the structure, it implies the design satisfies the
safety requirement.
P [kN]
Critical 450
Element 1
400 Numerical
Experimental
350
300
250
200
x 150
A Critical
Reinforcement 100
Element 2
50
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
x [m]
Figure 3.26: Application of safety format of strength limit state of internal actions of a deep beam (Mancini
2002, Bretagnoli et al., 2004).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 113
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In the draft Australian concrete structures standard (DR 05252, 2005), for new designs and/or
assessment of existing structures, a system safety approach is used to ensure that the structure
(or structural element) meets expected safety demands. The draft adopts the following stress
check procedure for use with nonlinear frame and stress analysis:
It shall be confirmed that the design resistance of the structure or the component
member is equal to or greater than the design action 
Rd < Ed
where Ru.sys is the capacity of the structure and is determined for the same
combination of actions adopted for Ed with mean values of material properties,
and Csys is a system reduction factor and is dependent on the mode of failure.
The values for Csys adopted by the draft standard are dependent on the type of failure. For
structural systems in which the deflections and local deformations at high overload are an
order of magnitude greater than those for service conditions; and yielding of the
reinforcement occurs well before the peak load is reached, Csys = 0.7. Otherwise, Csys = 0.5
(although larger values than 0.5 may be used if it can be shown that, at high overload,
adequate warning is given of impending collapse).
This method allows an assessment of the response under given loading conditions and with
consideration of random nature of the input parameters and can be used in a rationally based
assessment of global safety. Solution of this problem can be performed numerically
combining the structural and statistical analyses. The numerical model of structure is based on
a deterministic nonlinear analysis using the finite element method. The probability of
distribution of the response can be obtained by numerical methods based on random
sampling. In these methods, the random variables of samples are generated by statistical
methods and sample response is realized by a nonlinear solver. Finally, the statistical
parameters of the response are analyzed again by statistical methods. An example of this
approach, applied to bridges, is presented in Bergmeister (2002).
The advantage of the above approach is that the reliability can be rationally evaluated by
failure probability or by a safety index. Both safety measures are well justified in reliability
engineering and prescribed by standards (eg., Eurocode 1, 2002). The resulting safety margin
(global safety factor) is based on actual input parameters, their random variation and their
mechanical relevance.
3.9 References
Akiyama, H. (1985), EarthquakeResistant LimitState Design for Buildings, UTPress,
Tokyo.
Alca, N., Alexander, S.D.B., MacGregor, J.G., (1997) Effect of size on flexural behavior of
highstrength concrete beams, ACI Structural Journal, 94(1), pp. 5967.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 115
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Bergmeister, K., et. al. (2002), Structural analysis and safety assessment of existing concrete
structures, Proceedings of fib Congress  Concrete Structures in 21 Century , Osaka, Japan,
Session 11, 4754.
Bretagnoli, G., Carbone, V. I., Giordano, L. and Mancini, G. (2004), Safety format for non
linear analysis, Proceeding of the fib Symposium, Avignon, France.
DR 05252 (2005), Concrete Structures, Draft for Public Comment Australian Standard,
Standards Australia, Revision of AS36002001.
Balmer, G.G. (1949), Shearing strength of concrete under high triaxial stresscomputation of
Mohr's envelope as a curve." Tech. Rep. No. SP23, Structure Research Laboratory, Denver.
Baant, Z.P. and Oh, B.H. (1983), Crack band theory for fracture of concrete, Materials and
Structures, 16 (1983) 155177.
Baant, Z.P. and Planas, J. (1998), Fracture and Size Effect in Concrete and Other
Quasibrittle Materials, CRC Press, Boca Raton.
Bentz, E.C. (2005), Explaining the riddle of tension stiffening models for shear panel
experiments, ASCE J. of Struct. Engng., 131(9), pp. 14221425.
Bolander, J.E., and Saito, S. (1998) Fracture Analysis Using Spring Networks with Random
Geometry, Engineering Fracture Mechanics, 61, pp. 569591.
Bolander, J.E., Hong, G.S., and Yoshitake, K., (2000). Structural Concrete Analysis Using
RigidBodySpring Networks, ComputerAided Civil and Infrastructure Engineering, 15, pp.
120133.
Bolander, J.E. and Hong, G.S. (2002). RigidBodySpring Network Modeling of Prestressed
Concrete Members, ACI Structural Journal, 99(5), SeptOct, pp. 595604.
Cervenka, V. and Gerstle, K. (1971, 1972), Inelastic analysis of reinforced concrete panels:
(1) Theory, (2) Experimental verification and application, Publications IABSE, Zrich, V.31
00, 1971, pp.3245, and V.32II, 1972, pp.2639.
CEBFIP Model Code 1990 (1993), Thomas Telford Services, Ltd., London, for Comit
EuroInternational du Bton, Bulletin dInnformation No. 213214, Lausanne, pp.437.
Collins, M.P., (1978), Towards a rational theory for RC members in shear, ASCE J. of
Struct. Engng., 104(4), pp. 649666.
DIN 10451 (1998), Tragwerke aus Beton, Stahlbeton und Spannbeton. Teil 1: Bemessung
und Konstruktion.
Eurocode 1 (2002), Actions on structures  Part 11: General actions densities, self weight,
imposed loads for buildings. CEN/TC 250/SC1.
Eurocode 2 (2002) Design of concrete structures  Part 1: General rules and rules for
buildings. CEN. 2nd draft. 2002.
Foster, S.J., and Marti, P. (2002). FE Modelling of RC Membranes Using the CMM
Formulation, Proceedings of the Fifth World Congress on Computational Mechanics
(WCCM V), July 712, Vienna, Austria, Editors: Mang, H.A.; Rammerstorfer, F.G.;
Eberhardsteiner, J., Publisher: Vienna University of Technology, Austria, ISBN 395015540
6, http://wccm.tuwien.ac.at
Foster, S.J., and Marti, P. (2003). Cracked Membrane Model: FE Implementation, ASCE,
Journal of Structural Engineering, V129, N9, September, pp. 1155 1163.
Gilbert, R.I., and Warner, R.F. (1978), Tension stiffening in reinforced concrete slabs,
ASCE J. of Struct. Engng., 104(ST12), pp. 18851900.
Goto, Y. (1971), Cracks formed in concrete around deformed tension bars, ACI Journal
Proceedings, 68(4), pp. 244251.
Harada, M., Onituka S., Adachi, M. and Matsuo, T. (2001), Experimental study on
deformation performance of cylindrical reinforced concrete structure, Proc. of Japan
Concrete Institute, 23(3).
Harajli, M.H. (2004), Comparison of bond strength of steel bars in normal and high strength
concrete, J. of Materials in Civil Engng., 16(4), pp. 365374.
Kaufmann, W. and Marti, P. (1998), Structural concrete: cracked membrane model", ASCE
J. of Struct. Engng., ASCE, 124(12), pp.14671475.
Kent, D.C., and Park, R. (1971), Flexural members with confined concrete, ASCE J. of
Struct. Engng., 97(ST7), pp. 19691990.
Kim, J.K., Yi, S.T., Yang, E.I. (2000), Size effect on flexural compressive strength of
concrete specimens, ACI Structural Journal, 97(2), pp. 291296.
Kupfer, H., Hilsdorf, H.K., and Rusch, H. (1969), Behavior of concrete under biaxial
stresses, ACI Journal Proceedings, 66(8), pp. 656666.
Marti, P., Alvarez, M., Kaufmann, W., and Sigrist, V. (1998). Tension Chord Model for
Structural Concrete, Structural Engineering International, IABSE, 4/98, 287298.
Maekawa, K., An, X., and Tsuchiya, S. (1999), Application to fracture analysis of concrete
structures, recent development in mechanics of fracture for concrete structures, Concrete
Journal, Japan Concrete Institute, 37(9), pp.5460.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 117
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Maekawa, K. and An, X. (2000), Shear failure and ductility of RC columns after yielding of
main reinforcement, Engineering Fracture Mechanics, 65, pp. 335368.
Manacini, G. (2002), Nonlinear analysis and safety format for practice, Proceeding of the
First fib Congress, Osaka, Japan, pp. 5358.
Morita, S. and Kaku, T. (1979), Splitting bond failure of large deformed reinforcing bars,
ACI journal, Proceedings Vol.76, pp.93110.
Okamura, H. and Maekawa, K., Nonlinear Analysis and Constitutive Models of Reinforced
Concrete, GihodoShuppan Co. Tokyo, 1991.
Ozcebe, G., Ersoy, U., Tankut, T. (1999), Minimum flexural reinforcement for Tbeams
made of higher strength concrete, Canadian J. of Civil Engng., 26(5), pp. 525534.
Richart, F.E., Brandtzaeg, A., and Brown, R.L. (1928), A Study of the Failure of Concrete
Under Combined Compressive Stresses, University of Illinois Engineering Experimental
Station, Bulletin No.185, 104 pp.
Rots, J. G. and Blaauwendraad, J. (1989), Crack models for concrete: discrete or smeared?
Fixed, multidirectional or rotating?, HERON, 34(1).
Scott, B.D., Park, R., and Priestley, M.J.N. (1982), StressStrain Behavior of Concrete
Confined by Overlapping Hoops at Low and High Strain Rates, ACI Journal Proceedings,
79(1), pp. 1327.
Shima, H., Chou, L. and Okamura, H. (1987), Micro and macro model for bond behavior in
reinforced concrete, Journal of The Faculty of Engineering, The University of Tokyo (B),
39(2), pp.133194.
Turneaure, F.E., Maurer, E.R. (1908), Principles of Reinforced Concrete Construction, 1st
Edition, John Wiley and Sons, Inc, New York.
van Mier, J. G. M., Schlangen, E. and Vervuurt, A. (1995), Lattice type fracture models for
concrete, Chapter 10 in Continuum Models for Materials with Microstructure (Mhlhaus,
H.B., ed.), John Wiley & Sons, Chichester, pp.341377.
Vecchio, F. and Collins, M. P. (1986), The modified compression field theory for reinforced
concrete elements subjected to shear, ACI Journal, 3(4), pp. 219231.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 119
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
[Blank Page]
!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
These models can be used in order to determine the reserve strength and deformation capacity
of existing healthy or damaged buildings. Once existing capacities are known along with
anticipated strength and ductility demands, then a structural repair or upgrade that is
consistent with the existing deficiencies of concrete structures and expected loading
conditions can be devised.
When performing accurate nonlinear analysis that leads the structure to its ultimate state, it is
fundamental to rely on detailed models that can capture the strong nonlinear behaviour of
concrete and steel reinforcing bars. As an example, reinforcing bar buckling models,
presented in this chapter, can be used in conjunction with confinement models for concrete to
quantify the available strength and ductility of a reinforced concrete element, also under
cyclic action. In these cases, both the behaviour of buckled bars and the response of confined
concrete must be modelled accurately, so that the remaining rotation capacity of a plastic
hinge and strength of earthquakedamaged columns can be correctly evaluated.
Retrofit of a corrosiondamaged concrete column would be another typical example where the
models presented in this chapter can be used. The capacity reduction in concrete columns can
be evaluated based on the amount of reinforcing bar loss due to corrosion. Once this reduction
in the capacity is determined, the models presented in this section can be used in determining
the feasibility of various repair techniques that can be used.
This Chapter is organised in two sections: Section 4.3 Nonlinear Models of Frame Elements,
and Section 4.4 Interpretation of Results. For material modelling, the reader could refer to
Chapters 3 and 6, especially to the section where issues relevant to modelling of the materials
concrete and steel are treated. Also, longterm effects of ageing and distress in both concrete
and steel should be looked at as well as buckling of steel bars and details of lap splices.
Section 4.3 Nonlinear Models of Frame Elements is devoted to the comparison among
different approaches in modelling the response at the element level. Fibre elements are
compared to strutandtie type of models. Modelling of shear and of bondslip problems are
given particular attention.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 121
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In Section 4.4 Interpretation of Results, important issues related to the localisation problem
are treated, both for displacement and forcebased frame elements.
4.2 Notation
! $ x% array containing the derivatives of the shape functions NU(x)
D bi $ x % bond interface force between beam and bar i
E Youngs modulus
e ( x) section deformations vector
F element flexibility matrix without rigid body modes
f ( x) section flexibility matrix
G shear modulus
I moment of inertia
K element stiffness matrix
LIP length pertaining to an integration point
Lp plastic hinge length
M bending moment
M B $ x% beam section bending moment
NB $ x% vector of axial forces in the beam
Ni $ x % vector of axial forces in bar i
NP $x% force interpolation functions array
NU $ x % shape functions
P element resisting forces
py $ x % transverse distributed load
V shear
VB $ x % beam section shear force
s ( x) force distributions vector along the element
U nodal displacement vector
u(x) section displacements vector
ub i $ x % bond slip between beam and bar i
u0 longitudinal displacement at the crosssection reference axis
v0 transverse displacement at the crosssection reference axis
wh weight factor
yi distance of bar i from element reference axis
:0 crosssection deformation due to bending
& crosssection deformations due to shear
)0 axial strain at section centroid
f section curvature
 section rotation
Because the inelastic behaviour of reinforced concrete frames often concentrates in certain
sections (typically the ends of girders and columns for seismic loads and midspan for
distributed, static loads), one option for modelling the nonlinear behaviour of reinforced
concrete frames is to used a lumped plasticity model, where nonlinear zerolength springs are
placed at the critical points, connected by linear elastic elements. Depending on the
formulation these models, they may consist of several springs that are connected either in
series or in parallel.
The earliest parallel component element was introduced by Clough and Johnston (1966) and
allowed for a bilinear momentrotation relationship: the element consists of two parallel
elements, one elasticperfectly plastic to represent yielding and the other perfectly elastic to
represent strainhardening. The stiffness matrix of the member is the sum of the stiffnesses of
the components. Takizawa (1976) generalized this model to multilinear monotonic behaviour
allowing for the effect of cracking in reinforced concrete members.
Filippou and Issa (1988) subdivide the element in different, parallel, subelements. Each sub
element describes a single effect, such as inelastic behaviour due to bending, shear behaviour
at the interface or bondslip behaviour at the beamcolumn joint. The interaction between
these effects is then achieved by the combination of subelements. This approach allows the
hysteretic law of the individual subelement to be simpler, while the member still exhibits a
complex hysteretic behaviour through the interaction of the different subelements.
The series model was formally introduced by Giberson (1967), although it had been
reportedly used earlier. Its original form consists of a linear elastic element with one
equivalent nonlinear rotational spring attached to each end. The inelastic deformations of the
member are lumped into the end springs. This model is more versatile than the original
Clough model, since it can describe more complex hysteretic behaviour by the selection of
appropriate momentrotation relationships for the end springs. This makes the model
attractive for the phenomenological representation of the hysteretic behaviour of reinforced
concrete members.
Several other lumped plasticity constitutive models have been proposed. Such models include
cyclic stiffness degradation in flexure and shear (Clough and Benuska, 1966, Takeda et al.,
1970, Brancaleoni et al., 1983), pinching under reversal (Banon et al., 1981, Brancaleoni et
al., 1983) and fixed end rotations at the beamcolumn joint interface due to bar pullout
(Otani, 1974, Filippou and Issa, 1988). Typically, axialflexural coupling is neglected.
Nonlinear rate constitutive representations have also been generalized from the basic
endochronic theory formulation in Ozdemir (1981) to provide continuous hysteretic
relationships for the nonlinear springs. An extensive discussion of the mathematical functions
that are appropriate for such models is given by Iwan (1978).
The dependence of flexural strength on the axial load under uniaxial and biaxial bending
conditions has been explicitly included in the modelling of beams and structural walls. In
most lumped plasticity models, the axial forcebending moment interaction is described by a
yield surface for the stress resultants and an associated flow rule according to the tenets of
classical plasticity theory (Prager and Hodge, 1951). The response is assumed to be linear for
stress states that fall within the yield surface in which case the flexural and axial stiffness of
the member are uncoupled and independent of the end loads. With the introduction of
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 123
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
multiple yield and loading surfaces and corresponding hardening rules, multilinear
constitutive representations that include cracking and cyclic stiffness degradation are possible
for the springs, as was originally suggested by Takayanagi and Schnobrich (1979).
A lumped model is a simplification of the actual behaviour that involves the gradual spread of
inelastic deformations into the member as a function of loading history. This modelling
deficiency was recognized in several correlation studies, particularly those related to large
resisting elements of flexural wallframe structures (Charney and Bertero, 1982, Bertero et al.,
1984). The basic advantage of the lumped model is in its simplicity in that it reduces storage
requirements and computational cost and improves the numerical stability of the
computations. Most lumped models, however, oversimplify certain important aspects of the
hysteretic behaviour of reinforced concrete and are, therefore, limited in their applicability.
One such limitation derives from restrictive a priori assumptions for the determination of the
spring parameters. Parametric and theoretical studies of girders under monotonic loading by
Anagnostopoulos (1981) demonstrate a strong dependence between model parameters and the
imposed loading pattern and level of inelastic deformation. Neither factor is likely to remain
constant during the dynamic response. The problem is further accentuated by the fluctuation
of the axial force in the columns. Because of this history dependence, damage predictions at
both the local and the global level may be grossly inaccurate. Such information can only be
obtained with more refined models capable of describing the hysteretic behaviour of the
section as a function of axial load. Another limitation of most lumped plasticity models,
proposed to date, is their inability to describe adequately the deformation softening behaviour
of reinforced concrete members. Such deformation softening can be observed as the reduction
in lateral resistance of an axially loaded cantilever column under monotonically increasing
lateral tip displacement. Again, more advanced models are needed in this case.
The generalization of the rigid plastic theory concepts by Prager and Hodge (1951) to
reinforced concrete column stress and strain resultant variables, such as bending moment and
rotation and axial force and extension, limits the applicability of these models to well detailed
members with large inelastic deformation capacity at the critical regions. For a reinforced
concrete column section, the yield surface of the stress resultants is actually a function of a
reference strain that couples the corresponding displacement components. This contradicts
classical plasticity theory which does not account for deformation softening and assumes that
the section deformability is unlimited.
A refined resultant model has been proposed by ElTawil and Deierlein (2001) that developed
a bounding surface plasticity model implemented in the stressresultant space. The model is
generally applicable to steel, reinforced concrete and composite members. Two variations of
the plasticity model are considered: a finitesurface and a degeneratesurface version. The
former explicitly considers a fully elastic response region to exist within the inner surface and
is, thus, applicable to steel members that typically have such behaviour. The degenerate
surface model shrinks the elastic region to a point and the section behaviour starts out as
inelastic in any loading direction. This version is suitable for sections that have little or no
elastic response region such as reinforced concrete and composite sections. Stiffness
degradation is accounted for as a function of the plastic strain energy absorbed by the
composite member.
To overcome some of the limitations of classical plasticity theory in the description of the
interaction between axial force and bending moments, Lai et al. (1984) proposed a fibre hinge
model that consists of a linear elastic element extending over the entire length of the
reinforced concrete member and has one inelastic element at each end. Each inelastic element
is made up of one inelastic spring at each section corner, representing the longitudinal
reinforcing steel, and a central concrete spring that is effective in compression only. The five
spring discretisation of the end sections is capable of simulating the axial forcebiaxial
bending moment interaction in reinforced concrete members in a more rational way than is
possible by classical plasticity theory. In Lai's model, the forcedeformation relationship for
the effective steel springs follows that of Takeda et al. (1970) but the parameters that define
the envelope are established from equilibrium considerations.
In distributed models, material nonlinearity can take place at any element section and the
element behaviour is derived by a weighted integration of the section response. In practice,
since the element integrals are evaluated numerically, only the behaviour of selected sections
at the integration points is monitored. Either the element deformations or the element forces
are the primary unknowns of the model and these are obtained by suitable interpolation
functions from the global element displacements or forces, respectively.
Discrete cracks in distributed models are represented as smeared over a finite length rather
than treated explicitly. The constitutive behaviour of the crosssection is either formulated in
accordance with classical plasticity theory, in terms of stress and strain resultants, or is
explicitly derived by discretisation of the crosssection into fibres, as is the case in the spread
plasticity fibre models. Frame models are usually based on either:
i) EulerBernoulli beam theory in which plane sections remain plane and normal to the
longitudinal axis of the beam; that is, there are no shear deformations (see Figure
4.1a); or
ii) Timoshenko beam theory of plane sections remaining plane but not normal to the
longitudinal axis with the difference between the normal and the plane section
rotations being the shear deformation (see Figure 4.1b).
Biaxial bending is a simple extension of the case of uniaxial bending, shown in Figure 4.1.
y y : 0 ( x)
a'
&
a'
b' Deformed
Deformed b'
dv0 dv0
dx v0 ( x) dx v0 ( x)
a a
Undeformed Undeformed
x
x
b b
u0 ( x ) u0 ( x )
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 125
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In the development of the theory that follows, the following notation is used with an element
of length dx (Figure 4.2):
dv0
= total crosssection rotation
dx
dV
p'
M M+dM dx
dM
V'
dx
V V+dV
dx
Figure 4.2: Beam segment of infinitesimal length.
Based on the forces acting on a segment of infinitesimal length (Figure 4.2), the following
differential equations are derived:
EulerBernoulli beam
d 2M
Equilibrium: 'p (4.1)
dx 2
Section constitutive law: M ' EIf (4.2)
d 2v0
Section compatibility: f ' (curvature) (4.3)
dx2
d 2 e d 2 v0 b
Differential equation: c EI 2 ` ' p (4.4)
dx 2 d dx a
du du0 d 2v
Strain at any point of the crosssection: ) x ' ' 7 y 20 (4.5)
dx dx dx
Timoshenko beam
g dV
hh dx ' 7 p
Equilibrium: i (4.6)
hV 7 dM ' 0
hj dx
Section constitutive laws: V ' GAs& M ' EIf (4.7)
d: 0
Section compatibility: f' (flexural deformation) (4.8)
dx
dv0
&' 7 :0 (shear deformation) (4.9)
dx
g e d 2 v0 d: 0 b
h GAs c 2 7 ` ' 7p
h d dx dx a
Differential equations: i (4.10)
h e dv0 b d 2: 0
GA
h s cd dx 7 : 0` 7 EI '0
j a dx 2
In Eqs. 4.7 and 4.10, As is the shear area. To find the expression for As in the shear stress
expression according to the Timoshenko beam theory, assume that the strain energy
1 1
U) =
2AG , xy& xy dA ' G G , xy2 dA
2 A
(4.11)
is the same as that found using the stresses and strains of the exact solution for an elastic
section. For a rectangular crosssection it can be shown that As ' 1.2bh ' 1.2 A . Figure 4.3
compares the sectional shear stresses from Timoshenko beam theory for a rectangular,
linearly elastic, section with that of the exact solution.
, xy , xy
y
h x
6V e h 2 2b V
, xy ' c 7y ` , xy ' const '
bh3 d 4 a GAs
"Exact" Theory Timoshenko Beam Theory
Two formulations are presented for frame elements, one based on the classical displacement
based approach, the other based on the forcebased approach.
Displacementbased formulation
The displacementbased beam formulation uses the classical finite element approach to derive
the element stiffness matrix and the element restoring force vector. The central step is the
assumption of displacement fields along the element that are defined in terms of nodal
displacements.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 127
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
If U is the nodal displacement vector defined in Figure 4.4, the section displacements
u $ x % ' Hu0 $ x % v0 $ x %I are defined by:
T
u $ x % ' NU $ x % U (4.12)
where NU $ x % are the shape functions that define a linear axial displacement u0(x) and a cubic
transverse displacement v0(x):
Z x x W
X1 7 L 0 0
L
0 0 U
NU $ x % ' X 2 3 2 2 3 2
U
X exb exb e xb exb e xb e xb e xb U
X 0 1 7 3c ` 8 2 c ` x c1 7 ` 0 3c ` 7 2 c ` 7x c ` 8 x c ` U
Y dLa dLa d La dLa d La d La dLa V
v1 v2
u1 u2
1 2
U ' Hu1 v1 1 u2 v2  2 I
T
Using the principle of minimum potential energy, or other equivalent variational principles,
the following expressions are obtained for the element stiffness matrix K and for the element
resisting forces P:
K = G ! T $ x % k $ x % ! $ x % dx (4.13)
L
P = G ! T $ x % s $ x % dx (4.14)
L
where ! $ x % is an array containing the derivatives of the shape functions NU(x) (in particular
the first row of B is the first derivative of the first row of NU, and the second row of B is the
second derivative of the second row of NU) and k(x) is the crosssection stiffness matrix. The
matrix k(x) depends on the crosssection model selected for the analyses and s(x) is the cross
section force vector, that is:
s $ x % = H N $ x % M $ x %I
T
(4.15)
Forcebased formulation
The forcebased beam formulation assumes force fields rather than displacement fields along
the element. The element is developed here without rigid body modes (Figure 4.5), for
reasons that will become clear later on in the chapter.
k1, 1 k 2,  2
N, u
P = HM 1 NI
T
M2
If P is the nodal force vector defined in Figure 4.5, the force distributions along the element
are written as:
s $ x% ' NP $ x% P (4.16)
where P is the previously defined vector containing the crosssectional forces and N P $ x % is
the force interpolation functions array given by:
Z 0 0 1W
NP $ x% ' e x b e x b U
X (4.17)
Xc 7 1 ` c ` 0 U
YXd L a d L a UV
Again, using the principle of minimum complementary potential energy, or other equivalent
variational principles, the following expressions are obtained for the element flexibility matrix
without rigid body modes F and for the corresponding nodal deformations U:
F = G N TP $ x % f $ x % N P $ x % dx (4.18)
L
U = G N TP $ x % e $ x % dx (4.19)
L
where f(x) is the section flexibility matrix and e(x) is a vector containing the section
deformations. That is:
e $ x % = H) 0 $ x % f $ x %I
T
(4.20)
The motivation and interest for forcebased elements stem from fact that the equilibrium
relationship (Eq. 4.16) is exact within the assumptions of the EulerBernoulli beam theory.
In other words, the axial load and the bending moment remain constant and linear,
respectively, irrespective of the beam crosssection variation and material behaviour. Element
loads are included in the forcebased formulation by modifying the element force distributions
in Eq. 4.16.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 129
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
One problem encountered with forcebased elements is with their implementation in a general
purpose finite element program. The element state determination, that is the computation of
the element stiffness matrix K and resisting forced P, become complex. As for the stiffness
matrix, this is easily computed by inverting the flexibility matrix:
K = F 71 (4.21)
and by adding the rigid body modes to K to obtain K. Computing the element resisting forces
is a much more complex problem. The complexity stems from the fact that there is no way to
directly relate section resisting forces and element resisting forces, as is the case with the
displacementbased elements. Spacone et al. (1996) propose an iterative method to compute
the exact element forces and deformations. The procedure is basically a NewtonRaphson
iteration loop under imposed nodal displacements that adjusts the element forces and section
deformations until there is compatibility between section deformations and imposed nodal
deformations, as expressed by the weak statement of Eq. 4.19. The element is computationally
more involved and expensive than the corresponding twonode displacementbased element
but its precision leads to the use of a single element per structural member, thus leading to
large savings in the global number of degrees of freedom to be solved for.
Similarly to the EulerBernoulli beam element, two formulations are presented for the
Timoshenko beam element, one based on the classical displacementbased approach and the
other based on the forcebased approach.
Displacementbased formulation
The displacementbased formulation is similar to that of the EulerBernoulli beam but there
are three independent fields: horizontal displacement, u0, vertical displacement, v0, and axis
rotation, :0. A typical choice for these elements is to assume parabolic displacement fields for
the three displacements. The resulting threenode element is shown Figure 4.6.
:1 :2 :3
v1 v2 v3
u1 u2 u3
1 2 3
In this case
U ' Hu1 v1 :1 u2 v2 : 2 u3 v3 : 3I
T
(4.22)
and
u $ x % ' Hu0 $ x % v0 $ x % : 0 $ x %I
T
(4.23)
Eq. 4.12 still holds true with the displacement interpolation vector NU(x) defined by
Z N1 $ x % 0 0 N2 $ x % 0 0 N3 $ x % 0 0 W
X U
NU $ x % ' X 0 N1 $ x % 0 0 N2 $ x % 0 0 N3 $ x % 0 U (4.24)
XY 0 0 N1 $ x % 0 0 N2 $ x % 0 0 N 3 $ x % UV
The explicit expressions of the shape functions are given in natural coordinates in Figure 4.7.
1
N1 $R % ' R $R 7 1% N 2 $R % ' $1 7 R 2 % 1
N 3 $R % ' R $R 8 1%
2 2
1 1 1
1 0 1 R
Figure 4.7: Displacement interpolation functions for threenode 2D Timoshenko beam element in natural
coordinates ( R ' 718 2 x L ).
The remainder of the element formulation is formally identical to that of the EulerBernoulli
beam. The section force vector now becomes:
s $ x % = H N $ x % M $ x % V $ x %I
T
(4.25)
e $ x % = H) 0 $ x % f $ x % & $ x %I
T
(4.26)
and the vector B(x) changes accordingly. The resulting element is approximate even in the
linear elastic case and several elements must be used to obtain a satisfactory approximation of
the solution. Also, these elements may lock in the sense that too much energy goes into the
shear deformation mode, leading to a stiff element.
Forcebased formulation
The forcebase formulation for a Timoshenko beam is identical to that of the EulerBernoulli
beam with the exception that the expressions for the section forces s(x) and section
deformations e(x) change according to the definitions given in the displacementbased
formulations. The force interpolation functions become:
Z 0 0 1W
X U
ex b exb U
N P $ x % ' Xc 7 1 ` c ` 0 (4.27)
Xd L a dLa U
X U
Y 1/ L 1/ L 0 V
The flexibility matrix F is once again exact within the assumptions of the Timoshenko
beam theory, irrespective of the variation in beam section or material behaviour.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 131
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The definitions of the section stiffness matrix k(x) and flexibility matrix f(x) for both the
EulerBernoulli and Timoshenko beams depend on the section models used in the
formulations. Some of the models most commonly used in the published literature are
discussed in this section.
Once the behaviour and existing capacities of structural components are established, the
deficient components can be upgraded in order to meet the capacity demands. Fibre
(sectional) analysis is an appropriate tool that can be used to estimate behaviour of the
Bregions of reinforced and prestressed concrete members. In Dregions, beam theory does
not apply as the plane sections remain plane principle of the beam theory is not accurate or,
necessarily, safe. In these regions strutandtie modelling (STM) can be used to estimate the
capacity. It is important to appreciate that STM can be used to model Bernoulli beams in the
way Ritter (1899) and Mrsch (1902) used concrete compression struts and tension ties
(stirrups and longitudinal reinforcement) to explain the force flow and stress fields in a typical
reinforced concrete beam.
In this section fibre analysis procedures and strut and tie modelling techniques are treated in
order to:
" perform (conventional and seismic) assessment on existing buildings,
" assess rehabilitation interventions on inadequate constructions,
" identify potential causes of damage or collapse in buildings.
Structural members can be divided into Bregions (i.e. Bernoulli regions or beam regions)
and Dregions (i.e., disturbed regions or discontinuity regions.) In Bregions the beam
theory applies and flexural behaviour of reinforced and prestressed concrete members can be
established by using equilibrium, compatibility and constitutive relationships. Fibre
(sectional) analysis is an appropriate tool that can be used to establish behaviour of the
Bregions of reinforced and prestressed concrete members. In addition, through the use of the
modified compression field theory (MCFT), it is possible to extend the use of fibre models to
shear critical elements. In Dregions, strutandtie models, such as those proposed by
Marti (1985a, 1985b) and Schlaich et al. (1987), can be used in order to estimate the capacity.
Ritter and Mrsch planted the first seeds of what is currently known as STM by idealizing
reinforced concrete beams with a series of diagonal concrete struts and reinforcing bars (ties).
It is equally important to note that a significant amount of research has been conducted in this
area since the original research of the aforementioned pioneers. Strutandtie modelling is a
detailing and ultimate strength calculation procedure for discontinuity regions within
structures. Strutandtie modelling represents a design method for complex structural details
that has a basis in mechanics and is derived from the theory of plasticity but is simple enough
to be readily applied in design. The method involves the idealization of a complex structural
member into a simple collection of struts, ties, and nodes representing, in a general manner,
the flow of stress paths within the member.
Fibre analysis and STM procedures presented in this section can be used to evaluate the
existing capacity of structures with or without damage. Once the behaviour and existing
capacities of structural components are established, the deficient components can be upgraded
in order to meet the capacity demands. The methods presented in this section are of particular
significance for this reason.
Fibre analysis
Flexural behaviour of reinforced and prestressed concrete members can be estimated by using
equilibrium, compatibility and constitutive relationships. Compatibility in fibre analysis is
achieved through the plane sections remain plane hypothesis (Hooke, 1678, Bernoulli,
1705, and Navier, 1826). This geometric assumption forms the basis of the engineering beam
theory that is used in sectional analysis of concrete members. Equilibrium is achieved by
integrating stresses at any given section and equating them to the required sectional forces.
Appropriate stressstrain relationships for concrete and reinforcing and/or prestressing steel
need to be used in order to obtain the stress distribution through the depth of the section for a
given strain profile.
Concrete strain distribution can be defined by two variables (for example, two strains, a strain
and curvature, etc.). If the strain profile is known, realistic stressstrain relationships for
materials can be used to determine the stress distribution, which can subsequently be
integrated to evaluate the resultant forces for concrete and reinforcing steel. In this way, the
moment and axial load acting on the section can be determined.
Conversely, if we know the strain profile we can evaluate the axial force and bending moment
that caused these strains. The only detail that makes the evaluation of the response of flexural
members a somewhat cumbersome task lies in the integration of the stress over the part of a
concrete section in compression. Numerical integration can be performed to simplify this task
with the section idealized as a series of rectangular (or trapezoidal) layers and with the stress
in a given layer assumed as constant over the width of that layer. Figure 4.8 illustrates a
typical example where the concrete section is subdivided into multiple layers to facilitate the
numerical integration process or fibre analysis.
The modified compression field theory (MCFT) of Vecchio and Collins (1982, 1986) has
proven to be a simple, yet powerful, computational tool that can be used in FE analysis of
reinforced concrete structures and in the sectional analysis for concrete members to predict
the deformation and load carrying capacity. The latter will be discussed in this section.
yci ysj
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 133
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The MCFT was based on reinforced concrete panel tests subjected to inplane forces.
Equilibrium, compatibility and stressstrain relationships based on test observations were used
in the formulation of the models. Constitutive laws were used to link average stressstrain
relationships both for the reinforcement and concrete. The MCFT is based on a rotating crack
model that assumes that there is a gradual reorientation of principal stresses (and strains) in
the concrete parallel to, and normal to, the direction of the cracks. Furthermore, the MCFT
assumes enforced alignment of principal stress and strain directions and links the principal
compressive stresses to principal compressive strains considering the effect of coexisting
principal tensile strains.
The use of MCFT within a layered section (fibre) analysis is discussed in Section 4.3.4. In this
way the fibre analysis, discussed above, can be extended to include rational models to obtain
shear force, axial force and bending moment interaction for reinforced and prestressed
concrete sections and members.
Strutandtie modelling
Strutandtie modelling (STM) has proven to be an efficient tool for analyzing structural
members. The efficiency arises only in areas of discontinuity where plane sections do not
remain plane, strains are not linear, and transverse strains may not be conservatively
neglected. In these areas, typical layered sectional analysis does not provide a reasonable
estimate to a sections capacity due to violation of the basic assumptions made when
analyzing the section. It should be noted, however, that any structural member may be
analyzed to find its ultimate capacity using a STM. This is possible because the method
results in a lower bound solution. For example, a typical Bernoulli beam may be analyzed
using struts and ties as shown by Ritter (1899) and Mrsch (1902) in Figure 4.9. In his classic
1902 text, Mrsch explained the truss model shown in Figure 4.10 in more detail.
BRegions
DRegions
h1.5h typ
h
h
h11.5h1
h
h21.5h2
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 135
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
According to the theory of plasticity, any statically admissible stress field that is in equilibrium
with the applied loads and in which stress levels are on or within the material yield surface,
constitutes a lower bound solution. However, the strain capacity of the materials is a
fundamental requirement to fully satisfy that a lower bound solution occurs and, in that concrete
is not an elastoplastic material; some care is needed in determining an appropriate model.
To begin a STM, the designer must first discretize the member into B and Dregions. The
Bregions are areas of the member where plane sections remain plane, strains are linear and
transverse strains are negligible. Dregions are those in which sectional analysis does not
apply because the aforementioned assumptions are not reasonable. These are areas under
point loads, points of reaction, openings, reentrant corners, frame joints, etc., as illustrated in
Figure 4.11. The Dregion is assumed to be 1 to 1.5 times the member height to either side of
the disturbance, as per St. Venants principle.
Once the Dregions are isolated they can be decoupled from the rest of the member for
analysis purposes. It is essential that the forces applied to the Dregion are accurate. In some
cases this step might prove to be most challenging to designers. A clear force path must be
determined to find all external loads acting on the member. Once the external loads are
determined then structural analysis must be completed to find external reactions and internal
forces. If the member is not entirely comprised of single a Dregion, then sectional analysis
must be undertaken to determine internal forces from the adjacent Bregions, as shown in
Figure 4.12.
D region B region
Once the forces are accurately placed on the decoupled Dregion, a STM or truss model may
be applied. The truss model should accurately depict the flow of tensile and compressive
forces within a statically determinate truss. The model may be chosen by experience and
judgment of the designer, by following the elastic principal stresses produced by a finite
element model (see Chapter 8), or by guidelines given by code bodies and/or the literature.
The following considerations should be taken into account when choosing an appropriate
STM for a Dregion:
" Separate models may have to be conceived for different load cases. For example
one model may be used when gravity loads control, and a separate model may be
used when lateral loads control, with the region detailed to accommodate both
models (Figure 4.13).
Tie Strut
" Serviceability of the member must be taken into account. The model chosen should
account for serviceability by limiting cracking and provide sufficient stiffness to
control deflections (Figure 4.14).
" The physical geometry of the truss should be practical and free of congestion, with
no overlapping struts or nodes.
" When analyzing insitu structural members, consideration shall be given to the
existing condition of the concrete and physical placement of steel reinforcement.
For example, if concrete is severely cracked in a location of a modelled strut, the
strut capacity should be considerred negligible or zero. The centroid of modelled
ties should coincide with the reinforcing steel placed in the member.
In general, optimizing the model used for final design and detailing of a Dregion is a
subjective and iterative process. Schlaich et al. (1987) proposed to follow the principles of
minimum strain energy after cracking. That is, the model with the least and shortest amount of
ties is the most appropriate based on the assumption that cracked concrete struts will deform
little compared to steelreinforced ties. However, as stated in FIP (1998) and mentioned
above, this ultimate load model may not be valid when evaluating service conditions.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 137
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Once the nodal geometry is determined and, hence, the tie geometry, concentration is now
given to determining the strut geometry. The struts at midheight can be idealized using one
of the stress fields shown in Figure 4.15.
bursting forces
dc
d c /4
d c /4
Tb /2
l b {bursting zone}
C/2
C/2
Tb /2
:
d c /4
d c /4
C
(b) Bursting forces in bottle shaped struts
The designer must exercise judgment in selecting strut geometry within the model. If a strut is
located in the pure compression field of a member, or two nodes are modelled as close
together, producing confined compression fields, a prismatic strut geometry should be
considered. If the designer wishes to model a compression field at a beam support, in
Mrschs fashion, then a fan of compression struts should be considered. If a strut is in a
location that allows the compressive stresses to spread laterally then bottle shaped geometry
needs to be considered.
In the case of bottle shaped struts, further model refinement is needed to account for the
transverse tension associated the lateral spread of compression within the bottle shaped strut.
For service loads, the spread of compression may be modelled at a slope of 2 longitudinal:1
transverse (tan : = 1/2) to the axis of the strut, whereas for the ultimate condition the
dispersion angle narrows to tan : = 1/5 (Foster, 1998). When modelling the transverse
tension, the designer should be cognizant of tie reinforcing anchoring the strut, the strut angle,
and any reinforcing steel placed to confine the spread of compression; all of which influence
the dispersion of compression and hence the model (Baumann, 1998).
Several older structures lack the shear reinforcement to guarantee that plastic hinge will form
before shear failure of the member. Shear modelling in a frame analysis is dealt with at two
levels, the element level and the section level. At the element level, the Timoshenko beam
theory is typically used to model the shear behaviour. Displacementbased and forcebased
formulations for a Timoshenko beam are presented in Section 4.3.2.
At the section level, different approaches can be followed. The simplest relies on stress
resultant laws, where different nonlinear laws are given for the flexural, axial and shear
response. One such model has been proposed by Martino et al. (2000), where a fibre model is
used to model axial and bending responses, while a nonlinear law is used for the shear force
shear deformation response. While the shear response is decoupled from the other
deformations at the section level, the implementation of this section model in a forcebased
element allows coupling between axial and bending responses at the element level.
A fibre section model has been proposed for a Bernoulli beam by Petrangeli et al. (1999) that
includes shear deformations. In this approach, each fibre has basically three deformations,
axial strain, transverse strain (in the direction of the stirrups) and shear deformations, plus the
corresponding stresses. Given the section deformations e(x), the axial strain and shear
deformation of each fibre can be computed through compatibility. The third condition is that
the stress in the direction of the vertical stirrups is zero; that is, the sum of the forces in the
concrete and in the stirrups is zero. Given these three conditions (axial strain, shear
deformation and vertical zero net stress), the fibre state determination consists of finding the
corresponding axial stress, shear stress and vertical deformations and the fibre stiffnesses. The
fibre stresses and fibre stiffness are added to give the section forces and the section stiffness.
A concrete constitutive law based on the microplane theory is used to describe the concrete.
The fibre section model with shear deformations proposed by Petrangeli et al. (1999) is
implemented in a forcebased element and has given results that correlate well with available
experimental results.
Finally, the MCFT can also be used to describe the shear deformation of a reinforced concrete
member. The MCFT has proven to be successful in predicting the load deformation response
of reinforced concrete beams, with different amounts of longitudinal and transverse
reinforcement, using sectional analysis procedures (Vecchio and Collins, 1986). It has also
been implemented into FE programs for similar analysis (for example, Vecchio, 1989, 1990,
Foster and Gilbert, 1990, Foster, 1992, and many others). In the MCFT, constitutive
relationships are used to link average stresses and strains for both reinforcement and concrete
for analysis of reinforced concrete member behaviour. The stresses and strains used are
average values realizing that the maximum and minimum values can be different to the
average values. Conversely, local conditions may govern the failure but average conditions
are likely to be more relevant in representing the member behaviour. Figure 4.16 illustrates
the average stress and strain states employed in MCFT.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 139
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Strain hardening
h b
behaviour o the reinfforcement at
of a high straain levels ccan be inco
orporated
using th
he correspon
nding stresssstrain relattionship fro
om material tests.
The aveerage stresssstrain relaationship foor concrete is obtainedd from paneel tests and d can be
found elsewhere
e (V
Vecchio an nd Collins, 1982). Prin ncipal stresss and principal strain axes are
assumed d to be coin ncident as in rotating crack modeels. The beehaviour in panel tests showed
that craacked conccrete subjeccted to hig gh tensile strains in the directiion normall to the
compresssion is soffter and weeaker than in i standard cylinder teests. The cooncrete stresss in the
principaal direction can be relatted to the prrincipal con
ncrete strainn as follows:
e ) e ) 2 b b`
2
c 2 c
f c 2 ' f c 2 max c 2 7 c ` ` ` (4.29)
c )o d )o a `
d a
A number of models have been developed to determine the effect on concrete compressive
strength, and stiffness, for elements under the influence of transverse tensile strains (fc2max in
Eq. 4.29); for example, see Vecchio and Collins (1993), Belarbi and Hsu (1995), Kaufmann
(1998) and Vecchio (2000), amongst others. One such relationship that has proved robust with
time and can be used to compute fc2 is that of Vecchio and Collins (1986):
e 1 b
f c 2 max ' f c# cc `` (4.30)
d 0. 8 8 0. 34 ) )
1 oa
For concrete in tension, prior to cracking a linear elastic stressstrain relationship is used. The
principal tensile stress fc1 is given as follows:
where Ec is the tangent modulus of concrete, ) 1 is the principal tensile strain, fcr is the
tensile strength of concrete, and ) cr is the cracking strain of concrete. After cracking, tension
stiffening of concrete can be represented by:
f cr
f c1 ' (4.32)
1 8 500)1
A reinforced or prestressed concrete section can be discretized into a series of concrete strips
with superimposed longitudinal steel components (Figure 4.8). Each concrete strip is defined
by its width, b, depth, h, amount of transverse reinforcing steel, $t, and its position relative to
the top of the beam, yc. Similarly, the longitudinal steel components are defined by their
crosssectional area, As, yield strength, fy, and position relative to the top of the beam, ys. Pre
strain in reinforcement can also be considered in the analysis. Properties common to the entire
crosssection are usually the concrete strength, f c# , concrete strain at peak stress, "o, yield
strength of transverse steel, fy, and modulus of elasticity for reinforcing bars, Es. Sectional
analysis is carried out assuming plane section remains plane with equilibrium and
compatibility satisfied within each fibre. Hence, the estimates of the longitudinal strains and
the shear flow distribution through the depth of the section are required to compute the stress
strain conditions in a reinforced concrete member subjected to a given set of sectional loads.
First, an estimate of longitudinal strain distribution can be made. Strain in concrete for each
strip is assumed to be constant. The stress conditions in steel layers can be directly computed
from the assumed longitudinal strain distribution together with the uniaxial relationship of
steel from material tests. Second, a shear stress distribution is assumed for the section such
that the sum of shear stresses in each strip will be equal to the externally applied shear. Given
the longitudinal strain and shear stress for a strip, equilibrium and compatibility must be
satisfied to compute the compressive stress for the strip. The resultant of stresses for the strips
must balance the applied sectional forces, axial force, N, moment, M, and shear force, V.
For a section discretized with m concrete strips and n steel strips, equilibrium requires that
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 141
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
m n
Pi '1
f li bi hi 8 P f slj Asj ' N
j '1
m n
Pi '1
f li bi hi y ci 8 P f slj Asj y sj ' M (4.33)
Pv b h
i '1
i i i 'V
where fli is the compressive stress in the longitudinal direction in the ith fibre, and vi is the
shear stress in strip i. If these conditions are not satisfied, the longitudinal strains are adjusted
until equilibrium is satisfied. To determine the correct shear stress distribution, a second
section a small distance from the one considered can be analysed. Satisfying sectional
equilibrium for each case, the assumed stress distribution can be checked using the free body
diagram of the strip.
The procedure summarized above is shown in a flow chart in Figure 4.17 with the stress and
strain state of a section for a given set of loads computed using this procedure. Also, it is
possible to analyse the overall response of a section by analysing different increments of
longitudinal strains, shear stresses, and axial forces by using the procedure outlined above.
Bayrak and Sheikh (2001) proposed a plastic hinge analysis technique that can be used in a
sectional analysis platform in order to incorporate buckling of longitudinal bars in the
analysis. This technique employs slightly different displacement compatibility requirements
along with equilibrium considerations and constitutive relationships to evaluate the plastic
hinge response of tied columns. The following is a stepbystep description of the procedure
for the analysis of plastic hinges:
1) Standard sectional analysis procedure (Figure 4.17) is followed before the initiation of
buckling of the longitudinal bars.
2) At the initiation of bar buckling, the tie forces generated as a result of core concrete
bearing against the reinforcing bars are calculated.
3) By using proper boundary conditions and an assumed shape function for the forces
acting on longitudinal bars, their outward deflection at the midheight (e) between two
sets of ties can be calculated. Dividing this deflection by the longitudinal bar diameter
(d) yields an e/d ratio.
4) Using the e/d ratio calculated above and the ratio of the unsupported length of the
longitudinal bars (l) to the bar diameter (l/d), the relevant stressstrain curve for
reinforcing bars under compression can be selected and used as the constitutive
relationship for compressed bars in a sectional analysis. Figure 4.18 illustrates
experimentally determined stressstrain curves for reinforcing bars with initial
imperfections tested under compression (Bayrak and Sheikh, 2001).
SPEC IFY LO N G IT U D INAL ST EEL PRO PERT IES EST IMAT E "
SPECIFY SECT IO N LO AD S
D ET ERMINE
! t ,! dt , # lt , # m
D ET ERMINE
EST IM AT E LO NG IT U DIN AL ST RAIN D IST RIBUT IO N f st , f t , f d , f dt , f l , fd , "
C O M PU T E LO NG IT UD IN AL ST R ESSES
DIST R IBUT IO N FO R EAC H REBAR LAYER
NO
IS " = " ?
Symbols:
REPEAT CALC U LAT IO N S FOR $ l : Longitudinal strain
SECT IO N 2
v : Shear stress
$ d : Concrete principal compressive strain
% : Assumed angle of inclination of principal compressive strain
C O M PU T E RESU LT ING SHEAR
$ t : Transverse tensile strain
ST R ESS D IST R IBUT IO N
$ dt : Principal tensile strain
& lt : Normal shear strain
& m : Maximum shear strain
f st : Transverse tensile stress
NO SH EAR ST R ESS D IST . f t : Concrete transverse compressive stress
SAME AS ASSU M ED f d : Concrete principal compressive strain
? f dt : Concrete principal tensile stress
YES f l : Concrete longitudinal stress
f d : Principal compressive stress computed using assumed shear stress
% : Angle of inclination of principal compressive strain
O U T PUT
Figure 4.17: Sectional Analysis for Shear, Flexure and Axial Loads.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 143
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
4.3.5 M
Modelling Bond Slip in Beams
Bond plays a centtral role in the response of reinfo orced concrrete membeers by allow wing the
stress trransfer from
m the steel bars
b to the suurrounding concrete. PerfectP bondd is usually assumed
a
in the analysis off reinforcedd concrete structures and impliees full com mpatibility between
concretee and reinfoorcement strrains. This assumption n is valid onnly in regionns where neegligible
stress trransfer occuurs between the two com mponents (K Kauser and Mehlhorn, 1987) and can only
take plaace at early loading stagges and at loow strain leevels. As thee load is inccreased, craacking as
well as breaking of o bond unaavoidably occcurs and a certain am mount of bondslip takees place,
all of which
w will, innturn, affecct the stresss distributions in both the
t steel andd the concreete. Near
the craccks, high boondslips deevelop caussing relativee displacem ments betweeen concretee and the
reinforccement steeel. Due to this bonddslip, diffeerent strainns are obseerved in the steel
reinforccing bars thhan in the surrounding
s g concrete. At the cracks the loccalised straiin in the
reinforccing steel caan be many times greatter that the average
a straain and this is known as
a strain
localisaation.
Bondslip affects the overall response of reinforced concrete structural members. In particular,
two phenomena are worth discussing:
1. An increase in stiffness in the regions between two adjacent cracks. This effect,
called tension stiffening, can be included by either considering the tensile
resistance of the concrete or by increasing the tensile stiffness and strength of the
steel reinforcing bars. Tension stiffening mostly affects the member response
under serviceability conditions, since the effect of bond is completely lost when
member failure occurs.
2. An increase of flexibility at the member ends, due to the pullout of the reinforcing
bars at the interface either with beamcolumn joints or with the footings. Similar
drops in stiffness may also be caused by insufficient lap splice lengths.
These effects become particularly important and complex under seismic loading conditions,
when bond gradually deteriorates due to large strains and damage caused by multiple load
reversals.
RubianoBenavides (1998) proposed the use of rotational springs at the element ends to
account for the added flexibility due to bondslip. This approach is suitable for lumped
plasticity models but requires particular care in the selection of the rotational spring's
mechanical properties. A similar approach is followed by Filippou et al. (1999).
Another approach is followed by Monti and Spacone (2000), who modify the fibre section
model to add the effects of bond slip, modelled according to Monti et al. (1997a, 1997b). The
basic idea is fairly simple: the strain in the steel fibre is the sum of the actual strain in the
reinforcing bar and the strain equivalent to the bar slip. This model accounts not only for the
response of the reinforcing bar inside the beam but also for its anchorage outside the element,
in either a structural joint or a footing. The steel fibre strain is given by the sum of the effects
of the bar deformation and the anchorage slip. The response is still computed in terms of fibre
stress and stiffness, which are needed for the fibre section state determination. The steel fibre
strain is obtained from the section deformations using compatibility and is written as
1
) s8a ' ) s 8 ua ' ) s 8 ) a (4.34)
LIP
It is important to point out that the formulation proposed by Monti and Spacone (2000)
provides a solution to the bondslip problem within a single beam finite element. The element
can capture the base rotation in RC columns due to bar slips and can also describe slip and
bond failure at the bar splices. The fibre section model by Monti and Spacone (2000) is
inserted in the forcebased element by Spacone et al (1996).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 145
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
axial strain = )
curvature = 0
Figure 4.19: Slice response to axial deformation only (Monti and Spacone, 2000).
dN B $ x % n
8 P D bi $ x % ' 0
dx i '1
(4.35)
dN i $ x %
7 D bi $ x % ' 0, i ' 1, n
dx
where N B $ x % and N i $ x % are the axial forces in the beam and in bar i, respectively, and
D b i $ x % is the bond interface force between the beam and bar i.
dVB $ x %
7 py $ x % ' 0 (4.36)
dx
where VB $ x % is the beam section shear force and p y $ x % is the transverse distributed load.
dM B $ x % n
dx
7 VB $ x % 7 P y D $ x%
i '1
i bi '0 (4.37)
py (x)
dx
=
py (x)
concrete beam
Db1(x)
Mb (x) Mb (x) 8 dMb (x)
y1
NB (x) beam reference axis NB (x) 8 dNB (x)
yn
VB (x) VB (x) 8 dVB (x)
Dbn(x)
bars +
bar 1 N1(x) Db1(x) N1(x) 8 dN1(x)
y1
beam reference axis
yn Dbn(x)
bar n Nn (x) Nn (x) 8 dNn (x)
Figure 4.20: Slice of RC frame element with bond slip (Limkatanyu and Spacone, 2002).
where M B $ x % is the beam section bending moment and yi is the distance of bar i from the
element reference axis. The work by Limkatanyu and Spacone (2002) follows the Euler
Bernoulli beam theory and, thus, the shear deformations are neglected. The shear force VB $ x %
is removed by combining the above equations to obtain:
d 2M B $ x% n dD bi $ x %
7 p y $ x % 7 P yi '0 (4.38)
dx 2 i '1 dx
Compatibility
As for the element compatibility, the reinforced concrete element is treated as a Bernoulli
beam. On the other hand, the reinforcing bar slip are determined by the compatibility
relationship between the beam and the bar displacements:
dvB $ x %
ubi $ x % ' ui $ x % 7 uB $ x % 8 yi (4.39)
dx
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 147
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Starting from the above differential equations, Limkatanyu and Spacone (2002) proposed
three formulations: a displacementbased, a forcebased, and a twofield mixed formulation.
While the details of the three formulations may be involved, it is important to point out that
the displacementbased element with bondslip, while computationally robust, is quite
inaccurate and thus several elements per structural member must be used. On the other hand,
forcebased and mixed elements are much more precise and, though more complex to
implement in a general purpose FE program, they can be used in a coarser mesh leading to
computational cost savings.
All the tools needed for various steps described earlier are available at this point. Therefore,
the response of a section located in the plastic hinge region of a concrete column can be
predicted using the plastic hinge analysis procedure. At this point it should be appreciated that
the equations presented herein are derived for cases where buckling over one tie spacing takes
place. For cases where one or two tie sets rupture, the existence of ruptured tie sets can be
ignored and spacing can be modified accordingly so that the equations presented herein can
be used. In reality, according to the principal of minimum potential energy, the mechanism
that would require minimum energy to be stored in the system is the governing mechanism of
failure and, hence, that mechanism must be used in the analysis.
Figure 4.21 illustrates the sectional response of four specimens (AS3, AS17, AS18 and AS
19) tested by Sheikh and Khoury (1993). The moment curvature predictions, obtained using
the confined concrete stressstrain relationships suggested by Sheikh and Uzumeri (1982)
(SU), Kent and Park (1971) (MKP) and Mander (1988) (MAN) are also shown in these
figures. In the conventional sectional analyses performed to obtain the sectional responses, the
aforementioned confined concrete stressstrain relationships are used for the core concrete.
An unconfined concrete stressstrain relationship is used for the cover concrete and stress
strain relationships obtained from a tensile coupon test are used for longitudinal bars under
tension and compression. Predictions obtained using the plastic hinge analysis procedure, with
the Sheikh and Uzumeri model (SU+B) are also shown in Figure 4.21. The computer
program, SecRes99, (Bayrak, 1999) was used to obtain the predictions shown in Figure 4.21.
The use of the plastic hinge analysis procedure resulted in reasonably accurate predictions for
the behaviour of sections located in the plastic hinge region of the test specimens.
Conventional sectional analyses (fibreanalyses) could not provide a reasonable prediction for
the ultimate curvatures. With the use of the plastic hinge analysis technique, the reduction in
the load carrying capacity of longitudinal bars as a result of their buckling can be predicted.
The ultimate curvature and, hence, the failure of a column can therefore be determined with
reasonable accuracy.
Reinforced concrete frame elements, similarly to concrete solid elements, can lead to
numerical inconsistencies that derive from localization problems. When the crosssection
response starts softening (as may be the case of reinforced concrete column sections), the
element response becomes non unique and depends on the number of integration points per
element and/or the number of elements per structural member. Localization issues, though
similar in nature, tend to be different in displacementbased and in forcebased frame
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 149
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
elements. Localization in displacementbased solid finite elements has been studied by Baant
and Oh (1983), Baant and Planas (1998) de Borst et al. (1994), among others and the theory
extends to frame elements. Coleman and Spacone (2001), Scott and Fenves (2006) and
Valipour and Foster (2007) discuss localization in forcebased elements and propose various
solution strategies.
To illustrate the numerical problems encountered in forcebased frame elements when plastic
hinges form, we will consider the case of a steel cantilever beam under an imposed transverse
tip displacement. A single forcebased element is used for the entire member. As the applied
tip displacement increases, a plastic hinge forms at the base where the maximum moment
occurs. Figure 4.22 illustrates the response of the forcebased element for the cantilever beam
with an elasticstrain hardening section behaviour. Unloading is prescribed in the final steps
to clarify the peak displacement and curvature demands. The base shear is plotted against the
tip displacement on the right and the base curvature (i.e. the curvature of the first integration
point) is shown on the left. The response is objective at both the element and the section
levels for models with four or more integration points. Three integration points do not
accurately integrate the element integrals leading to overprediction of the stiffness in the
strainhardening region.
Base Shear
3 IP
3 IP
4 IP
4, 5, 6, 7, 8 IP
5, 6, 7, 8 IP
O
3, 4, 5, 6, 7, 8
Integration
Points (IP)
Elasticstrain hardening
momentcurvature
Curvature Displacement, O
Figure 4.22: Cantilever beam with elasticstrain hardening section response (Colemanand Spacone, 2001).
Figure 4.23 shows the response of the same forcebased element to an imposed tip
displacement with an elasticperfectly plastic momentcurvature behaviour. Here, the
prediction of the element forcedisplacement response remains objective while the peak
curvature demand varies with the number of integration points. The loss of objective
curvature prediction is due to the localization of the inelastic curvature at the base integration
point. When this bottom section reaches the plastic moment, the column reaches its load
carrying capacity. As the tip displacement increases, the curvature of the base integration
point increases with constant (plastic) moment, while all the other integration points remain
linear elastic and do not see any change in either curvature or moment. The length of the base
integration point and, thus, the plastic hinge length, becomes a function of the number of
integration points used. As the number of integration points increases, the plastic hinge length
decreases and the curvature demand in the base integration point must increase to give the
same prescribed tip displacement. From here the nonobjective section prediction of Figure
4.23 is obtained.
Base Shear
3, 4, 5, 6, 7, 8 IP
8 IP 7 IP 6 IP 5 IP 4 IP 3 IP O
3, 4, 5, 6, 7, 8
Integration
Points (IP)
Elasticperfectly plastic
momentcurvature
Curvature Displacement, O
Figure 4.23: Cantilever beam with elasticperfectly plastic section response (Coleman and Spacone, 2001).
P (constant)
Base Shear
O 3, 4, 5
Integration
Points (IP)
3 IP
R/C beamcolumn
3 IP
4 IP
4 IP
5 IP
5 IP
Curvature Displacement, O
Figure 4.24: RC beamcolumn modelled with strain softening section response (Coleman and Spacone, 2001).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 151
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 4.25 shows the schematic bending moment and curvature distributions for three, four
and five integration points along the column height. This case refers to the elasticperfectly
plastic section response of Figure 4.23 and helps clarify the localization process in force
based elements. The plastic curvature Cp occurs when the plastic moment Mp is first reached.
Since equilibrium is strictly satisfied in the forcebased element, the bending moment remains
linear. When the plastic moment Mp is reached at the first (base) integration point, the applied
force cannot increase and the tip displacement increases under constant applied load (and
constant base moment). Because the bending moment cannot increase beyond Mp, adjacent
integration points remain elastic and the inelastic curvature localizes at the first integration
point. The tip displacement is computed as the weighted sum of the curvatures at the
integration points. The first integration point has a finite length LIP=1 = w1L proportional to its
integration weight w1. The larger the number of integration points, the shorter the length of
the first integration point and the larger the curvature at the first integration point to obtain the
same tip displacement, as shown in Figure 4.25.
GaussLobatto
integration point
Moment Curvature
Mp C p Mp Cp Mp Cp
A) B) C)
3 Integration Points 4 Integration Points 5 Integration Points
Figure 4.25: Moment and curvature profiles for an elasticperfectly plastic cantilever modelled with a single
forcebased element (Coleman and Spacone, 2001).
In summary, it is the ability to capture a jump from elastic to inelastic behaviour that makes
the forcebased formulation both attractive and prone to unique numerical problems. One can
observe that as the distance between the first (plastic) and second (elastic) integration point
varies, the response also varies. This implies that the number and placement of the integration
points not only influences the accuracy of the integration but also the postpeak response. For
hardening materials, plasticity usually spreads beyond a single integration point and
numerical problems are limited to a nonsmooth response if too few integration points are
used (Figure 4.22). For perfectly plastic and softening crosssection responses, the curvature
tends to localize at a particular integration point and problems with objectivity arise (Figure
4.23 and Figure 4.24).
For further reading on the state of research into the effects of localisation in fibre based
elements and on methods for obtaining objective results, the reader is referred to Scott and
Fenves (2006) and Valipour and Foster (2007).
discontinuity induced by the crack opening basically prevents stresses normal to the crack.
The failure is termed localized because material outside the fracture process zone remains
practically undamaged.
+
O
+ A region B
)
L +
B
OlL
A
overall response
region A
)
Figure 4.26: Uniaxial compression test under displacement control (Coleman and Spacone, 2001).
The concrete cylinder can be idealized as a series system, where one element (region B in
Figure 4.26) represents the weak link. When region B reaches the peak strength and starts
unloading, the concrete in regions A must unload in order to maintain equilibrium. From these
observations it is concluded that concrete failure in compression, as with tension, occurs in a
localized manner and requires special attention in a numerical model.
The concept of constant fracture energy in tension is widely used to regularize meshsensitive
smeared crack displacementbased elements in continuum FE analyses (Baant and Oh, 1983,
Baant and Planas, 1998, among others). The concept is applied here to forcebased beam
elements that soften in compression. While the constant fracture energy concept is not as
widely accepted for compression as it is for tension, experimental research (Lee and Willam,
1997, Jansen and Shah, 1997) and analytical investigations (Markeset and Hillerborg, 1995)
show that the theory also holds true also for localization in compression. The main idea of the
regularization process is to assume that the uniaxial stressstrain relationship for concrete is
supplemented by an additional material parameter, the fracture energy in compression Gcf ,
defined as:
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 153
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
where )i indicates inelastic strain and h is a length scale. For smeared crack elements, h
represents the size of a single element in the crush band. For forcebased frame elements, h
becomes the length of the softening integration point LIP.
Even though the proposed procedure is general, the regularization is applied here to the Kent
and Park (1971) law used for the concrete fibres of the fibre section. The prepeak behaviour
is given by a parabola, followed by a linear postpeak softening branch until a stress of 0.2 f c#
is reached at a prescribed strain labelled ) 20 . The residual strength is assumed to remain
constant for strains larger than ) 20 . The fracture energy is defined here from the peak stress
fc# until the end of the softening branch (Figure 4.27). This definition is similar to that of
Jansen and Shah (1997), where the fracture energy is defined from the peak stress until
0.33 f c# . It follows that ) 20 must be calibrated to maintain a constant energy release.
Assuming that G cf is known from experimental tests (Jansen and Shah, 1997), Eq. 4.41,
together with the definition of G cf from Figure 4.27, leads to:
G cf 0.8 f c'
) 20 ' 7 8 )0 (4.42)
0.6 f c' LIP E
where E is Youngs modulus and ) 0 is the strain corresponding to peak stress as indicated in
Figure 4.27.
f c' G cf
h
E
0.2 f c'
)o )20 )
Figure 4.27: Kent and Park (1971) stressstrain law and compression fracture energy (Coleman and Spacone,
2001).
Theoretically, Eq. 4.42 implies that the constitutive law must be calibrated for each separate
integration point in a structural model. In practice, plastic hinges generally form at the
element ends where the extreme integration points lie. If all elements in a model are
integrated with the same scheme and with a number of integration points, ) 20 only varies for
elements of different length, L. Linking the constitutive law to the element length is a
straightforward process.
As the number of integration points increases, the extreme integration point must furnish
larger inelastic strains to satisfy the constant fracture energy criteria. This is equivalent to
assuming a constant stressdisplacement relationship (Figure 4.28b), rather than a constant
stressstrain law as shown in Figure 4.28a.
+ Given: L, f c, )p, G cf +
wIP varies
4 IP 4, 5, 6 IP
fc
5 IP
6 IP
)o ) O
(a) (b)
Figure 4.28: Stressstrain curves with constant fracture energy: a) stressstrain curves for 4, 5, and 6 IP
schemes; b) stressdisplacement curve (Coleman and Spacone, 2001).
Curvature postprocessing
Once the global forcedisplacement response is regularized, in some cases there is still a need
to postprocess the results to obtain an objective prediction of the curvature demand in the
plastic hinge region. This is due to the fact that the plastic zone length is the length of the first
integration point and does not necessarily correspond to the physical length of the plastic
hinge. As with the inelastic strains in a smeared crack model, the constant fracture energy
criterion is insufficient for regularization of the internal element deformations. Because
different mesh sizes in a smeared crack model must produce the same crack opening
displacements, the magnitude of inelastic strains must vary. While the inelastic strains in a
smeared crack solid finite element do not have any physical meaning and can remain non
objective, inelastic curvature demands may be of particular interest in frame elements. A
simple procedure is presented below to obtain an objective prediction of the true curvature
demand on the member.
Figure 4.29 shows a deformed interior beam with plastic hinges forming at both ends. The
plastic rotation  i and the relevant displacements O i are indicated at the near (N) and far (F)
ends of the element. These are essentially fictitious quantities used to formulate the curvature
scaling law. The bending moment diagram and the curvature profile corresponding to elastic
perfectly plastic behaviour are also shown in Figure 4.29. Inelastic curvatures, indicated by Ci,
are concentrated at the extreme integration points and spread over a length LIP.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 155
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
O iF  iF
 iN O iN
L
Mp
Moment diagram:
LIP
Mp
!i
Curvature profile:
M
Ce '
EI
LIP
Figure 4.29: Beam with plastic hinge at each end (Coleman and Spacone, 2001).
The total curvature in the plastic hinge region can be separated into elastic and inelastic
curvature components as C ' Ce 8 Ci . Considering the geometry of Figure 4.29, the inelastic
hinge rotation is  i ' Ci LIP . Neglecting the elastic curvature in the other integration points and
using a small angle approximation for  i , the inelastic curvature can be approximated as:
Oi
CiMODEL m (4.43)
ZL L W
LIP X 7 IP U
Y2 2 V
MODEL
where Ci indicates the inelastic curvature that results from the analysis. Substituting the
actual length of the plastic hinge Lp for LIP yields a similar approximation for the true
curvature demand:
Oi
CiPREDICT m (4.44)
Z L Lp W
Lp X 7 U
Y2 2 V
PREDICT
where Ci indicates the inelastic curvature demand based on the assumed plastic hinge
length Lp. Finally the model output can be scaled according to:
The scale factor is computed by solving for O i in Eq. 4.43 and substituting into Eq. 4.44 to
obtain:
wIP L2 (1 7 wIP )
scalefactor ' (4.46)
Lp (L 7 Lp )
The doublecurvature case shown in Figure 4.29 prevails in the analyses of buildings under
lateral loads. On the other hand, some structural members, such as cantilevers, experience
single curvature and the plastic hinge forms at one end only. Such is the case of bridge piers
subjected to seismic loads in the transverse direction. The doublecurvature derivation is
readily recast in terms of single curvature by replacing the L 2 term in the denominator of
Eqs. 4.43 and 4.44 with the full length L. Taking this singlecurvature approach results in:
wIP L2 (2 7 wIP )
scalefactor ' (4.47)
L p 2( L 7 L p )
From the previous discussion, it appears that if the length of the first integration point
corresponds to the length of the plastic hinge, that is if LIP ' Lp , no postprocessing of the
curvature is needed because the curvature is objectively predicted by the element. One might
then suggest to select the number of GaussLobatto integration points in such a way that
LIP [ L p . This may cause two problems:
1) the number of integration points may be too small for short elements (causing
undesirable reduced integration) or too large for long elements (increasing the
computational cost); and
2) in most cases, the length of the element would also have to be adjusted, thus
introducing an additional element in each member and greatly increasing the
computational cost of the analyses.
Different integration schemes in which the user can define the length of the integration points
would solve the issue but may compromise the accuracy of the numerical integration. The
regularization approach described above is general and does not affect either the element
formulation or the integration scheme.
Some practical considerations are presented here to help analysts when using nonlinear
elements for static and dynamic analyses of frames. When a fibre section model is used, one
of the important question is how many fibres or layers are needed to discretize the section.
Spacone (1994) studied the sensitivity of a simple circular section cantilever beam response to
the number of section fibres (Figure 4.30). Their results, and those of similar parametric
studies, indicate that using too many fibres overkills the problem and increases the
computational cost of the analyses. While no precise indication on the number of fibres to use
exists, it appears that for a rectangular section fifteen layers are a good compromise between
accuracy and computational cost in a simple bending analysis. On the other hand, users
should always test their section discretization by comparing the results of different, more and
less refined, fibre meshes to assess the optimal fibre section discretization. Such initial tests
should be performed on simple structures such as a cantilever beam with or without axial
load.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 157
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
200
P,d 1000 kips COMPRESSIVE AXIAL FORCE
GAUSSLOBATTO INTEGRATION
18 #24
55.78 ft
100
SHEAR P (kips)
4 integration points
0
100
200
2 1 0 1 2
TIP DISPLACEMENT d (ft)
Figure 4.30: Response of a cantilever beam with circular cross section for three different section mesh
refinements.
A similar problem applies to the number of integration points to use in an analysis. Two
integration schemes are typically used for line elements, Gauss and GaussLobatto. The two
schemes, applied to a natural [1,1] domain, are given by
1 m
I ' G f $R % dR ' P wh f $Rh % (4.48)
71 h '1
1 m71
I ' G f $R % dR ' w1 f $R1 ' 71% 8 P wh f $Rh % 8 wm f $Rm ' 1% (4.49)
71 h'2
where Eq. 4.48 is the Gauss scheme, and Eq. 4.49 is the GaussLobatto scheme (Stroud and
Secrest, 1966). In Eq. 4.49, h denotes the monitored section and %& is the corresponding
weight factor. The Gauss scheme with m integration points permits the exact integration of
polynomials of degree up to (2m1), while the GaussLobatto scheme with m integration
points integrates exactly polynomials of degree up to (2m3). The Gauss scheme is preferred
in displacementbased elements, while the GaussLobatto scheme is preferred in forcebased
elements where monitoring the end sections is important (Spacone et al., 1996).
Similar to the issue of the number of fibres, the selection of the number of integration points
is important, because too few points may lead to inaccurate results, while too many
integration points in computationally inefficient and can cause severe strain localization
problems. Figure 4.31 and Figure 4.32 show a study on the sensitivity of the response of a
circular cantilever beam to the number of integration points. Two to five GaussLobatto and
Gauss integration points are used in a single forcebased element. From the figures, it appears
that four or five GaussLobatto or Gauss integration points are sufficient. Although these
results have not been generalized, experience shows that more than five integration points are
200
P,d 1000 kips COMPRESSIVE AXIAL FORCE
GAUSSLOBATTO INTEGRATION
18 #24
100 55.78 ft
SHEAR P (kips)
nGL = 2
100
nGL = 3
nGL = 4
nGL = 5
200
2 1 0 1 2
TIP DISPLACEMENT d (ft)
Figure 4.31: Response of a cantilever beam with circular crosssection for different numbers of GaussLobatto
integration points.
200
P,d
18 #24
55.78 ft
100
SHEAR P (kips)
0
nG = 2
nG = 3
nG = 4
100
nG = 5
Figure 4.32: Response of a cantilever beam with circular crosssection for different numbers of Gauss
integration points.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 159
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
not needed for accuracy. Analysts are, however, strongly encouraged to run preliminary
parametric studies on a simple structure (a cantilever beam, for example) to assess the number
of integration points necessary to reach a converged solution.
As discussed in the previous section on strain localization, the selection of the number of
integration points is closely related to the length of the plastic hinge that may form in the
structural member being considered. Without the inclusion nonlocalisation modelling (see
Valipour and Foster, 2007), plastic hinges tend to localize at a single integration point in
forcebased elements, while plastic hinges tend to extend to entire elements in displacement
based elements. The details of the localization characteristics of forcebased elements have
been previously discussed and should guide analysts in selecting the number of integration
points in an element that is likely to develop plastic hinges. Furthermore, it is important to
note that the plastic hinge length Lp is a quantity that is not easily computed. The following
expression suggested by Paulay and Priestley (1992) suggested that the plastic hinge length be
taken as:
where L is the distance from the plastic hinge to the nearest zero moment point, fy is the steel
yield stress and db is the bar diameter. There exist other formulas in the published literature
for estimating the plastic hinge length, mostly obtained from experimental results combined
with analytical studies.
4.5 References
Anagnostopoulos, S. (1981), Inelastic Beams for Seismic Analysis of Structures, Journal of
Structural Engineering, ASCE, 107(ST7), pp. 12971311.
Bannon, H., Briggs, J., and Irvine, M. (1981), Seismic Damage in Reinforced Concrete
Frames, Journal of Structural Engineering, ASCE, 107(ST9), pp. 17131729.
Bayrak, O. and Sheikh, S.A. (2001), Plastic Hinge Analysis, Journal of Structural
Engineering, ASCE, Vol. 127, No. 9, September, pp. 10921100.
Baant, Z.P. and Oh, B.H. (1983), Crack band theory for fracture of concrete. Materials and
Structures, Vol. 16, pp. 155177.
Baant, Z.P., and Planas, J. (1998), Fracture and size effect in concrete and other quasibrittle
materials, CRC Press, Boca Raton.
Belarbi, A., and Hsu, T.T C. (1995). Constitutive laws of softened concrete in biaxial tension
compression, ACI Struct. J., Vol. 92, No. 5, 562573.
Bertero, V.V., Aktan, A., Charney, F, and Sause, R. (1984), Earthquake Simulator Tests and
Associated Experimental, Analytical and Correlation Studies of OneFifth Scale Model, in
Earthquake Effects on Reinforced Concrete Structures, American Concrete Institute,
SP8413, Detroit, pp. 375424.
Brancaleoni, F., Ciampi, V., and Di Antonio, R. (1983), RateType Models for Non Linear
Hysteretic Structural Behavior, EUROMECH Colloquium, Palermo, Italy.
Charney, F., and Bertero, V.V. (1982), An Evaluation of the Design and Analytical Seismic
Response of a Seven Story Reinforced Concrete FrameWall Structure, Report EERC 8208,
Earthquake Engineering. Research Center, Berkeley.
Clough, R., and Benuska, L. (1967), Nonlinear Earthquake Behavior of Tall Buildings,
Journal of Mechanical Engineering, ASCE, 93(EM3), pp. 129146.
de Borst, R., Feenstra, P.H., Pamin, J., and Sluys, L.J. (1994), Some current issues in
computational mechanics of concrete structures. Computer modelling of concrete structures.
Proceedings of EUROC, Eds. H. Mang, N. Bi!ani!, R. de Borst, pp. 283302.
ElTawil, S. and Deierlein, G.G. (2001), Nonlinear analyses of mixed steelconcrete moment
frames. Part I  beamcolumn element formulation. Part II  implementation and verification,
ASCE, Journal of Structural Engineering, Vol. 127, No. 6, pp. 647665.
Filippou, F.C., and Issa, A. (1988). Nonlinear Analysis of Reinforced Concrete Frames under
Cyclic Load Reversals, EERC Report 8812, Earthquake Engineering Research Center,
Berkley.
Filippou, F.C., D Ambrisi, A., and Issa, A. (1999), Effects of reinforcement slip on
hysteretic behavior of reinforced concrete frame members, ACI Structural Journal, Vol. 96,
No. 3, pp. 327335.
Filippou, F.C., and Issa, A. (1988), Nonlinear Analysis of Reinforced Concrete Frames under
Cyclic Load Reversals, EERC Report 8812, Earthquake Engineering. Research Center,
Berkeley.
FIP (1998), Commission 3 on FIP 1996 Recommendation for Practical Design of Structural
Concrete, Fdration Internationale de la Precontrainte, May.
Foster S.J. (1992), An Application of the Arc Length Method involving Concrete Cracking,
International Journal for Numerical Methods in Engineering, Vol. 33, No. 2, January, pp. 269
285.
Foster S.J. (1998), Design of Non Flexural Members for Shear, Journal of Cement and
Concrete Composites, Vol. 20, No. 6, pp. 465475.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 161
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Foster S.J., and Gilbert R.I. (1990), Nonlinear Finite Element Model for Reinforced
Concrete Deep Beams and Panels, UNICIV Report No.R275, School of Civil Engineering,
University of New South Wales, Kensington, December, 113 pp.
Iwan, W. (1978), Application of Nonlinear Analysis Techniques, in, Iwan W. ed., Applied
Mechanics in Earthquake Engineering, ASME, AMD, 8, New York, pp. 135161.
Jansen, D.C. and Shah, S.P. (1997), Effect of length on compressive strainsoftening of
concrete. J. Eng. Mech., ASCE, Vol. 123, No. 1., pp. 2535.
Kauser, M., and Mehlhorn, G. (1987), Finite elements models for bond problems. Journal
of Structural Engineering, ASCE, Vol. 113, No. 10, pp. 21602173.
Kent, D.C. and Park, R. (1971), Flexural Members with Confined Concrete, Journal of the
Structural Division, ASCE, Vol. 97, ST7, July 1971, pp. 19691990.
Lai, S., Will, G. and Otani, S. (1984), Model for Inelastic Biaxial Bending of Concrete
Members, Journal of Structural Engineering, ASCE, 110(ST11), pp. 25632584.
Limkatanyu, S. and Spacone, E. (2002), R/C Frame Element with Bond Interfaces. Part 1:
DisplacementBased, ForceBased and Mixed Formulations. Part 2: Element State
Determination and Numerical Validation. ASCE Journal of Structural Engineering, Vol. 128,
No. 3, pp. 346364.
Mander, J.B., Priestley, M.J.N., and Park, R. (1988), Observed StressStrain Behaviour of
Confined Concrete, Journal of Structural Division, ASCE, Vol. 114, No. 8, August 1988, pp.
18271849.
Marti, P. (1985a), Basic Tools of Reinforced Concrete Beam Design, ACI Journal
Proceedings, Vol. 7, No. 1, Jan.Feb., pp. 4656.
Martino, R., Spacone, E., and Kingsley, G. (2000), Nonlinear Pushover Analysis of RC
Structures. ASCE Structures Congress, Advanced Technology in Structural Engineering, M.
Elgaaly editor, Philadelphia, PA, May. 8 pp (CDROM).
Monti, G., Filippou, F.C., and Spacone, E. (1997a), Finite element for anchored bars under
cyclic load reversals. Journal of Structural Engineering, ASCE, Vol. 129, No. 5, May
Monti, G., Filippou, F.C., Spacone, E. (1997b), Analysis of hysteretic behavior of anchored
reinforcing bars. ACI Structural Journal, Vol. 94, No. 2, MayJune
Monti, G., and Spacone, E. (2000), Reinforced Concrete Fiber Beam Element with Bond
Slip. ASCE Journal of Structural Engineering, Vol. 126, No. 6, pp. 654661.
Navier, CL. (1826), Resume des lecons de la resistance des corps solides, Paris.
Otani, S. (1974), Inelastic Analysis of R/C Frame Structures, Journal of the Structural
Division, ASCE, 100(ST7).
Paulay, T., and Priestley, M.J.N. (1992), Seismic design of reinforced concrete and masonry
buildings, Wiley, New York
Petrangeli, M., Pinto, P.E. and Ciampi, V. (1999), Fiber element for cyclic bending and shear
of RC structures. I: theory ASCE Journal of Engineering Mechanics, Vol. 125, No. 9, pp.
9941001.
Prager, W. and Hodge, P. (1951), Theory of Perfectly Plastic Solids, John Wiley and Sons,
New York.
Schlaich, J., Schfer, K., and Jennewein, M. (1987), Toward a Consistent Design of
Structural Concrete, Journal of the Prestressed Concrete Institute, Vol. 32, No. 3, MayJune,
pp. 74150.
Scott, M.H., and Fenves, G.L. (2006), Plastic hinge integration methods for forcebased
beamcolumn elements. Journal of Structural Engineering, Vol. 132, No. 2, pp. 244252.
Sheikh, S.A. and Khoury, S.S. (1993), Confined Concrete Columns with Stubs, ACI
Structural Journal, V. 90, No. 4, JulyAugust 1993, pp. 414431.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 163
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Sheikh, S.A. and Uzumeri, S.M. (1982), Analytical Model for Concrete Confinement in Tied
Columns, Journal of the Structural Division, ASCE, Vol. 108, ST12, December 1982, pp.
27032722.
Spacone, E. (1994), FlexibilityBased Finite Element Models for the Nonlinear Static and
Dynamic Analysis of Concrete Frame Structures Ph.D. Dissertation, Department of Civil
Engineering, University of California, Berkeley.
Spacone, E., Filippou, F.C., and Taucer, F.F. (1996), Fiber BeamColumn Model for
Nonlinear Analysis of R/C Frames. I: Formulation. II: Applications. Earthquake Engineering
and Structural Dynamics, Vol. 25, No. 7, pp. 711742.
Takayanagi, T. and Schnobrich, W. (1979), Non Linear Analysis of Coupled Wall Systems,
Earthquake Engineering and Structural Dynamics, Vol. 7, pp. 122.
Takeda, T., Sozen, M.A., and Nielsen, N. (1970), Reinforced Concrete Response to
Simulated Earthquakes, Journal of Structural Engineering, ASCE, 96(ST12), pp. 25572573.
Valipour H.R., and Foster, S.J. (2007), A Novel Flexibility Based BeamColumn Element
for Nonlinear Analysis of Reinforced Concrete Frames, UNICIV Report No. R447, School
of Civil and Environmental Engineering, The University of New South Wales, September.
Vecchio, F.J. (1989), Nonlinear Finite Element Analysis of Reinforced Concrete Membrane
Elements, ACI Journal, Proceedings Vol. 86, No.1, JanFeb., pp. 2635.
Vecchio, F.J. (2000), Disturbed stress field model for reinforced concrete: formulation, J.
Struct. Eng., Vol. 126, No.9, 10701077.
Vecchio, F., and Collins, M.P. (1982), The Response of Reinforced Concrete to InPlane
Shear and Normal Stresses, Publication No. 8203, Department of Civil Engineering,
University of Toronto, March, 332 pp.
Vecchio, F.J., and Collins, M.P. (1986), The Modified Compression Field Theory for
Reinforced Concrete Elements Subjected to Shear, ACI Journal, Vol. 83, No. 2, Mar.Apr.,
pp. 219231.
Vecchio, F.J., and Collins, M.P. (1993), Compression response of cracked reinforced
concrete. J. Struct. Eng., Vol. 119, No. 12, 35903610.
Depending on the task to be performed, the user has to choose the right method of
computation in order to avoid oversophistication on the one hand, yet including the decisive
nonlinear effects of a given problem on the other. This chapter deals only with the
displacement based nonlinear finite element method. Some approaches that combine the
lower bound theorem of the theory of plasticity with linear optimisation algorithms and with
other numerical solution techniques are described in Chapter 9.
5.2 Notation
Ec initial elastic modulus of concrete
Complementary energy
Eh hardening modulus of steel
)c concrete strain when concrete strength is reached
)cu ultimate compressive strain of concrete
)tu cracking strain of concrete
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 165
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
)t tensile strain
Gf fracture energy of concrete
ft direct tensile strength of concrete
fcm mean value of concrete compressive strength
fc concrete compressive strength
R vector of point loads
( vector of nodal force parameters
H deflection matrix
e
C, C matrix containing coefficients of a linearized yield criterion
e
Cs, Cs vector of yield moments
nx, ny shear forces per unit of length
mxx, myy, mxy bending and twisting moment per unit of length
mp, mpx, mpy plastic moment per unit of length
e
f vector of generalized nodal forces
e
u vector of nodal displacements
e
K stiffness matrix
d vector of design variables
W matrix of coefficients corresponding to d of a weight function
q, p distributed loads
Rd design value of resistance
Sd design value of action
SdL design value of crosssectional action assuming linear elastic material
SdNL design value of crosssectional action assuming material nonlinearity
aS area of reinforcement per unit of length
&R partial safety factor for structural resistance
^ load factor
wd total deformation of compressive softening zone
n Poissons ratio of concrete
To demonstrate the practicalities of using nonlinear FE analysis, we shall review the results
of modelling of a two span deep beam tested by Cervenka and Gerstle (1971). The
dimensions of the beam are shown in Figure 5.1a and the reinforcement arrangements for the
beam are shown in Figure 5.1b with the reinforcement consisting of #3 (USA) deformed bars
placed as shown (note that only one half of the specimen is shown with symmetry).
The properties of the concrete and the reinforcing steel as used in the analysis are shown in
Table 5.1.
166 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Table 5.1: Material properties for the Cervenka and Gerstle (1971) deep beam.
Concrete Reinforcement
Ec 20 250 MPa Es 190 000 MPa
' 0.2 fy 353 MPa
fcm 26.75 MPa Eh 700 MPa
"c 0.00264
wd 0.5 mm
ft 3.1 MPa
Gf * 58 N/m
Note: * Fracture energy (Gf ) calculated using the CEBFIP Model Code 1990 (1993).
A plane stress analysis was performed using 4node quadrilateral isoparametric finite
elements, using 4 integration points per element and with the constitutive relationship for the
concrete taken as the fractureplastic model of Cervenka (2002). In this approach the tensile
behaviour is described by the fracture energybased smeared crack model and the
compressive behaviour by a nonassociated plastic flow rule. Softening of concrete behaviour
in both tension and compression are considered. It is worthy of note that the tension stiffening
effect is modelled indirectly through localization of strains and no specific tension stiffening
parameter appears in the model. Perfect bond between concrete and reinforcement is assumed.
The deep beam was modelled using two different FE meshes; a rectangular mesh (Figure
5.2a) and a triangular mesh (Figure 5.1b). The sensitivity of the results to the element size and
shape was compared using three rectangular meshes with 5, 10 and 20 elements across the
span and with one mesh of 10 triangular elements across the span. The results of the analyses
are shown in the form of load versus displacement plots in Figures 5.3 and 5.4 and cracking
patterns in Figures 5.5 and 5.6.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 167
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
140
120
100
Force P [kN]
80
Experiment
60
Mesh 5
40
Mesh 10
20 Mesh 20
0
0 1 2 3 4 5
Displacement [mm]
Figure 5.3: Load versus displacement: quadrilateral mesh results compared with the experimental data.
120
100
Force P [kN]
80
60
Experiment
40
Mesh 10 Quadrilateral
20 Mesh 10 Triangular
0
0 1 2 3 4 5
Displacement [mm]
168 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 5.5: Comparison of experimental and FE crack patterns at failure (Mesh 10).
(a) (b)
Figure 5.6: a) Cracks and compressive stresses (Mesh 20); b) plastic strains in bars (Mesh 10).
The behaviour of the deep beam includes the propagation of cracks from early stages of
loading up to failure. At the maximum load, the horizontal bars at the bottom yield and
deform plastically and the concrete at the top right corner fails in compression. The analyses
show that the analytical models can reflect the real behaviour with accuracy for this failure
mode of this case type. They, thus, provide for acceptable engineering solutions for this
failure model for this case. To fully validate the model, other benchmark analyses need to be
considered for other failure modes, as outlined in Chapter 7 of this report.
In addition to strength results, the analyses can show the trends of various effects, which can
be utilized in practical cases. For example, we observe the response during the crack
propagation before maximum load is reached (that is, the ascending branch of load
displacement diagram). In this stage the response is dominantly affected by interaction of
partially cracked concrete and reinforcement, the so called tension stiffening effect. This
behaviour can be observed on the load displacement diagrams of Figures 5.2 and 5.3 and in
the crack patterns, Figures 5.4 and 5.5. The study shows that the larger elements give lower
stiffness in this stage and that smaller elements reflect better individual discrete cracks due to
better modelling of strain localization. If we continue to increase the number of elements, so
that the element size is several times smaller then the crack spacing, the crack pattern would
be close to the experimentally observed discrete cracks. However, in practical cases extremely
large numerical models are computationally expensive and the tension stiffening effect may
be underestimated.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 169
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
If we use triangular elements of a similar mesh size, the response is stiffer in the post
cracking, serviceability, range (see Figure 5.4). The reason for the stiffer response of the
triangular mesh is due to the effect of the shape function of the triangular element that
produces a constant strain field, while the quadrilateral element has a semilinear strain field.
Next we examine the effect of the mesh size on the maximum load at failure. The load
carrying capacity is exhausted by the attainment of the yield stress of reinforcement in the
bottom portion of the beam and the surpassing of the compressive strength of the concrete in
the top right corner. This is a typical bending mode of failure and larger elements give a
higher load response. The higher response is also reached by triangular (more stiff) elements.
In this particular case, the quadrilateral meshes 10 and 20 give near identical responses for the
maximum load and indicates that mesh 10 is sufficiently fine, while mesh 5 and the triangular
mesh 10 give slightly higher failure loads.
The descending shape of the load displacement diagrams after the peak indicates a limited
ductility, which is influenced by the softening model adopted for the concrete in compression.
This limited ductility is not observed in experiments, where concrete behaviour in the panel
corner may be influenced by the confining effect of the thickened central rib forming the
central support. This difference can be reduced by assuming a perfectly plastic behaviour for
the concrete, in this region, but such assumptions are not rationally founded since it applies
only to a small region of a panel and must be used with care.
This case study illustrates some of the general features valid for numerical simulation of
reinforced concrete structures. In concluding, we note that:
1. Two distinct constitutive models for concrete (smeared cracks, fracture energybased
crack band) and for reinforcement (bilinear) are capable of simulating reasonably
well the behaviour of our reinforced concrete panel. Other models, however, can be
adopted and it is for the user of such software to ensure that the models being used for
their problem are capable of predicting the response with accuracy. Benchmarking is
discussed in more detail in Chapter 7.
2. Finite element size and type can significantly influence the structural response. A
mesh sensitivity study is required to show the objectivity of results.
3. A coarse mesh can underestimate the structural resistance in postcracking stage of the
analysis and overestimate the ultimate load capacity.
4. Low order elements overestimate the load response in the whole range.
These conclusions are true for reinforced concrete structures with more or less uniformly
distributed reinforcement, such as for the deep beam example illustrated. Structures with
discrete reinforcement (for example, beams without web reinforcement) are less sensitive to
the underestimation of tension stiffening but can be more dependent on the fracture model
adopted for the concrete (see Cervenka, 1998).
One possibility for modelling of 2D Structures is the application of volume elements. The
reason why one should refrain from doing so is twofold; firstly, it is known from the nature of
the 2D problem, that stresses and strains normal to the middle surface can be neglected. Using
volume elements requires significant computational effort in calculating something that is
known a priori. Secondly, it is necessary to choose the volume elements with lateral
dimensions of the same magnitude otherwise the stiffness matrix becomes ill conditioned
170 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
and resuults in an ennormous nuumber of volume elemeents as theirr characterisstic length has
h to be
of the order of the plate
p or sheell thicknesss.
With a linear
l strainn variation over
o the element thickn
ness, the strrain state "xxx(k), "yy(k), #xy(k) for
layer k is
i given by:
) xx (k ) ' ) xx 8 S xxx T z (k )
) yy (k ) ' ) yy 8 S yyy T z (k ) (5.1)
where "xx, "yy, #xy, (xx, (yy annd (xy (see Figure
F 5.8) are the strrains and thhe curvaturees in the
middle plane and z(k)
z designaates the disttance of lay yer k from the
t referencce surface. WithW the
strains given
g assumming a planee stress statte, the distriibuted internnal forces oof each layeer can be
calculatted via material modeels for the reinforcem ment and concrete. Seectional forrces and
momentts are attainned by inteegrating thee layer conttributions thhrough the element th hickness.
The elemment stiffneess matrix can
c be evaluuated in a similar way by b adding uup the layer stiffness
matrices.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 171
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
To verify the approach described above, the orthogonally reinforced plate element M7, tested
by Marti et al. (1987), is evaluated. The specimen had dimensions of 200 x 1700 x 1700 mm
and was loaded with pure torsional moments in the reinforcement directions. Torsion was
applied by couples of equal and opposing forces, P, applied at the corners B and D of the
specimen (Figure 5.9a) with the specimen supported at the corners A and C. The
reinforcement data is shown in Figure 5.9b (the reinforcement ratio is *x = *y = 0.25%) and
the material properties are shown in Figure 5.9c.
A FE analysis of plate ML7 was performed for a single, layered, shell element. The element
was divided into concrete layers and reinforcement layers and the numerical computations
were performed both with and without including tension stiffening in the formulation. Due to
the low reinforcement ratios of the plate, tension stiffening is shown to have little influence on
the results, as shown in the moment versus curvature relationships for the plate plotted in
Figure 5.10.
Concrete:
Ec = 35500 MPa
n = 0.20
fc = 44.4 MPa
ft = 4.0 MPa
)c = 0.0025
)cu/)c = 1.5
)t/)ucr = 10
(a) Reinforcement:
fy = 479 MPa
(b) ES = 200 000 MPa
Eh = 0.0 MPa
(c)
Figure 5.9 Loading and deformation of test specimen ML7 (Marti, 1987): a) applied load; b)
Reinforcement; and c) material properties.
Moment  M [kNm/m]
50
40
30
20 Experiment
without TSE
10
with TSE
0
0 5 10 15 20 25 30 35
Curvature 7 f [1/mm]*1E6
172 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
There are many ways that 3D solid structures can be modelled and, as for any nonlinear
model, it is important that the model be calibrated to the problem being considered and
verified using appropriate procedures. In this section, three of the more common methods for
the development of models for the analysis of 3D structures are discussed: namely, models
developed based on nonlinear elasticity, models developed on the theories of plasticity and
fracture and microplane models. Some examples of the implementation of such models are
presented in the sections following.
Although local stress/strain fields of a reinforced concrete plate or solid element subjected to
uniform external forces are irregularly distributed, the load carrying mechanism can be
reduced to a combination of space averaged onedimensional stress flows developing in the
cracked concrete and steel. The spaceaveraged constitutive models are the basis of 3D
modelling of reinforced concrete solid elements. The compressiontension field modelling of
1D has been successfully applied to 2D modelling with the same concept extended to 3D
problems.
Shear stresses on a crack plane of a reinforced concrete element cause rotation in the direction
of principal stresses from those at the initial cracked state. It has been shown that rotation of
the principal stress may lead to formation of cracks in a new direction with almost full closure
of the initial cracks (Vecchio and Collins 1982, Stevens et al. 1987). By taking the mechanical
condition of the latest developed cracks, the concept of a rotating crack approach was
implemented into FEM codes. In this approach, the local axes of cracked concrete in three
directions (1,2,3) are considered to coincide with those of principal stress axes and
information on previously developed cracks is not necessarily kept in the analysis. The
compression field theory (CFT) (Mitchell and Collins, 1974) and subsequent modified
compression field theory (MCFT) (Vecchio and Collins, 1986) provided a valuable
contribution in the implementation of twodimensional FE analysis of reinforced concrete.
Vecchio and Selby (1991) extended the 2D rotatingcrackbased constitutive models of the
MCFT to the 3D case. In their approach, prior to cracking the reinforced concrete was taken
to be a linear elastic, isotropic, material. After cracking, the concrete and reinforcement are
considered to contribute separately to the structural element stiffness and its response to
boundary tractions. Cracked concrete is modelled as a nonlinear orthotropic, path
independent, material with the principal axes corresponding to the direction of the principal
average strains. For concrete in the direction of the largest principal compressive strain ()3),
the average stressstrain relationship is based on the MCFT (Figure 5.11a), in consideration of
the effect of the coexisting tensile strain in the 1direction ()1) and its cause of degradation of
the concrete compressive stiffness and strength in the 3direction. For concrete in tension and
before cracking, in the direction of principal tensile strain ()1), a linear relationship is used.
Whereas, after cracking, a decay function is adopted to represent tension stiffening effect
(Figure 5.11b). The intermediate principal stress (in the 2direction) is evaluated according to
the twodimensional MCFT, depending on its strain. If the strain is tensile, the tension
stiffening model is used and if compressive the compression model of MCFT is used.
Reinforcing bars are modelled using a trilinear relationship with strain hardening.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 173
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
2: intermediate
fc3 principle direction
fc 1: direction of normal
fc3max Softening to crack
3: direction of largest
principle compressive
strain
)0 )3
fc1 (a)
Local equilibrium
)cr )c1
(b)
Figure 5.11: Concrete material models adopted in the MCFT and applied to 3D. (Vecchio and Selby, 1991).
To validate their approach, Vecchio and Selby (1991) conducted series of numerical tests
using the constitutive relationships described above. One of these tests is shown in Figure
5.12 for the analysis of hollow reinforced concrete box beam sections tested by Onsongo
(1978). The test beams, measuring 410 mm by 508 mm in section and with a span of
2290 mm, were subjected to a constant moment and torque along their length. For the
analysis, 750 mm lengths of the beams were modelled with a mesh of 1200 elements and
1760 nodes and resulting 5280 degrees of freedom. The torque was applied as nodal forces at
the end of the bars. The longitudinal reinforcement was uniformly smeared over the
respective areas. The comparison of analytical results in terms of capacity prediction and
loaddeformation is shown in Figure 5.12b and Figure 5.12c. It is seen that reasonably
accurate predictions were obtained for the strength of both underreinforced and over
reinforced sections.
An alternative approach to rotating crack approach of Vecchio and Selby (1991) is the multi
directional smeared crack model. This model allows for consideration of several cracks within
a single element that need not necessarily be orthogonal to existing cracks. Here, information
of all cracks assigned for each crack direction is stored in the analysis. The number of
allowable cracks is usually limited, however, to reduce the complexity of computation (de
Borst and Nauta, 1985, Barzegar 1989, Hofstetter and Mang, 1995, Fokuura and Maekawa,
1998).
174 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 5.12: Comparison of threedimensional analysis using a MCFT formulation with experimental results:
a) experimental layout; b) and c) comparison of results (Vecchio and Selby, 1991).
composition of inplane
decompoded constitutive
planebase +// modelling
partial stresses +..
+/. (1,2)
(1,3) decomposed
plane
3 2
1 (2,3) decomposed plane
+// +..
+22 +22
+/2 (1,3) +.2 (2,3)
Figure 5.13: Breakdown and recomposition of load carrying mechanism of 3D cracked solids of concrete
The total stress is computed on the basis of the partial stress on the three subspaces. Hauke
and Maekawa (1998) took into account the difference of concrete mechanics near and far
from the reinforcing bars through consideration of two different zones in the reinforced
concrete space (Figure 5.14a). One for the concrete near the steel bars and subjected to
tensionstiffening behaviour, the so called reinforced concrete zone (due to influence of bond
on the concrete stiffness), and a second for the concrete far from reinforcement the so called
plain concrete zone which may have a localized crack in the concrete control volume (An
et al., 2000, Hauke and Maekawa, 1998). The post cracking tension model in this approach is
expressed as
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 175
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
+ t ' f t ( ) tu / ) t ) c (5.2)
where c is a parameter representing the steepness of the descending branch and )tu is the
cracking strain. In order to avoid mesh dependency of the results when dealing with
localization of cracks in plain concrete zone, the softening parameter in Eq. (5.2) was defined
according to fracture energy and the crack band width (Baant and Oh, 1983). On the other
hand, for the reinforced concrete zone, the softening parameter, c, is taken as constant and
equal to 0.4, irrespective of element sizes. If such a crack is located in the plain concrete zone,
the softening behaviour varies with the crack orientation. Also, for a control volume
containing smeared reinforcing bars in one or several directions, the average softening
stiffening behaviour depends on the crack inclination. Hauke and Maekawa (1998) used a 2nd
order interpolation function for considering the influence of crack inclination on the tension
softeningstiffening parameter. The normalized interpolated fracture energy (area under the
tension stiffening model as shown in Figure 5.14) in the desired direction, n, is computed as
(a)
(b) Softening Mixed behaviour
Gf*(2) {Gf, lr,2, ft} Gf*(n) {Gf*(1), Gf*(2), n }
2 +t / ft => c(2) => c(n%5or c()
+t / ft
)t
1
(2)
)t
3 Crack n
normal
n2
Stiffening
2 Gf*(1) {c(1) = 0.4}
Directional
angle
3 REbar +t / ft

n1 (1)
)t
3
2 imaginary orthogonal
1 reinforcement system
1
Figure 5.14 a) Three dimensional zoning approach; b) interpolation scheme for tension stiffening model
(Hauke and Maekawa 1998).
For the reinforcing steel, the average yield stress of the bars is taken as a function of an
average uniaxial strainstress relationship. In developing this relationship, the influence of
crack inclination with respect to direction of reinforcing bar is considered (see Hauke and
Maekawa, 1998, and Maekawa et al. 2003).
As an example, a 1/5 scale model tested by Ichikawa et al. (1998) of a pier under constant
axial load of 147 kN, applied at an eccentricity of 700 mm from the centre of cross section is
considered (Tsuchia et al., 2001). The outline of the experiment is shown in Figure 5.15 with
cyclic torsional loading applied to the specimen as shown in Figure 5.15d. In such a coupled
action, cracks developed in three directions, i.e. twoway shear cracks due to reversed cyclic
176 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
torsion and shear as well as horizontal cracks due to flexural loading. The comparison of the
horizontal load versus displacement relationships recorded from the experiment and
calculated in the 3D FE analysis and, also, the torsion versus displacement relationships and
longitudinal displacement are shown in Figure 5.16. A good correlation is observed for the
numerical modelling when compared with the test results.
Crack modelling based on the theory of plasticity permits a unified treatment of the
compressive and tensile behaviour of concrete. This type of concrete modelling is usually
developed from the rate constitutive equations (Hofstetter and Mang, 1995). Several
constitutive models in 3D strain space exist on the fractureplasticity formulation of plain
concrete. For example, Pramono and Willam (1989) developed a fracture energybased
plasticity constitutive model that accommodates strain hardening and softening. They adopted
a fracture energy release approach to describe the degradation of triaxial strength in tension
and in compression in concrete with low levels of confinement.
A unified treatment of plasticity and fracture can effectively make the return mapping
algorithm possible (Wilkins, 1964) which guarantees the solution for all magnitudes of strain
increments (Hofstetter and Mang, 1995, Cervenka and Cervenka, 1999). The main difficulty
in the development of a unified model for both the compressive and tensile regions is the
algorithm for the combination of the two models.
On the basis of plasticity theory, fracture can be modelled by using a Rankine yield criterion
(maximum tensile stress criterion) as:
where +i (i =1, 2, 3) is the principal stress and fT is the tensile strength of the concrete. An
appropriate softening law should be considered and can be taken as a function of the
accumulated damage of the concrete (Feenstra, 1993) such that fT in Eq. 5.4 is given as a
function of a damage variable. One such model is the crack band model of Baant and Oh
(1983), which can be used to describe mesh sizedependent softening laws.
For hardening/softening plasticity of concrete there exists several models to describe concrete
crushing. For example, the failure surface of Drucker and Prager (1952), the fiveparameter
ultimate strength surface of Willam and Warnke (1975), the four parameter model of Ottosen
(1977) and threeparameter failure surface of Menetrey and Willam (1995).
Cervenka et al. (1998) developed an algorithm for the combination of plasticity with fracture.
The model is able to handle physical changes like crack closure and is not restricted to any
particular shape of hardening or softening laws. The model can be used to simulate concrete
cracking, crushing under high confinement and crack closure due to crushing in other material
directions (Cervenka and Cervenka, 1999). The fracture is modelled by an orthotropic
smeared crack model based on the Rankine tension criterion (Eq. 5.4). Further, it is assumed
that strains, and stresses, are converted into the material directions that, in case of a rotating
crack model, correspond to the principal directions and, in the case of fixed crack model, are
given by the principal directions at the onset of cracking. The Menetrey and Willam (1995)
threeparameter failure surface was used to model concrete crushing.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 177
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
(a) (b)
(c)
178 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In the Cervenka et al. model, the strain is decomposed into elastic, plastic and fracture
components, as introduced by de Borst (1986). The new stress state is then computed from the
increment of each strain component based on the material used. The combined algorithm
determines the separation of strains into plastic and fracturing components in an iterative
scheme as the stress equivalence in each model must be preserved. Details of algorithm can
be found in Cervenka et al. (1998).
As an example of the verification of the model, an experiment test on prestress cable anchors
is simulated (Cervenka and Cervenka, 2007). The prestressing force is transferred from the
tendon to the concrete through special cylindrical anchors embedded into concrete. The
anchor is surrounded by spiral reinforcement and loaded by compressive forces to simulate
the action of the prestressing. A comparison of the load versus displacement obtained from
the FE model and the capacity of the anchorage, as obtained from the experiment, is shown in
Figure 5.17.
A second example, presented in Figure 5.18, shows plastic strains at failure for a bridge over
the Berounka River near Prague in the Czech Republic (Cervenka and Cervenka, 2007). The
model includes the curved box girder, the piers and the foundation. The box girder and piers
were modeled using higher order shell elements; the foundation by brick and tetrahedral
volume solid elements and the prestressing cables by string elements. Various stages at
service conditions were analysed as well as verification of the strength limit state. The
example shows how nonlinear analysis can be applied to complex large structures for
optimization of design and for verification at various limit conditions.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 179
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
180 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
In dealing with plasticity of polycrystalline metals, Taylor (1938) proposed that the stress
strain relationship be specified independently on various planes in the material. The
assumption was that the stresses on such a plane are the resolved components of the
macroscopic stress tensor, called the static constraint, or that the strain components on such a
plane are the resolved components of the macroscopic strain tensor, known as the kinematic
constraint. This idea developed for plasticity (Batdorf and Budianski, 1949), is known as slip
theory. Later, the model was extended by Baant and his co workers to concrete and
geomaterials and is referred to as the microplane model. This model is for analysis of quasi
brittle materials that exhibit softening behaviour (Baant and Gambarova, 1984, Baant and
Prat, 1988a, 1988b, Baant and Obolt, 1990, Carol et al. 1992 and Baant et al., 1996a,
1996b). Baant and Prat (1988a) developed an advanced kinematic constraint version of
microplane model which was extended by Obolt and Baant (1992) to a more general cyclic
form with rate sensitivity.
The 3D version of the microplane model is constructed on the basis of uniaxial relationships
between stress and strain components on planes of various orientations within a material
(Figure 5.19). These planes may be imagined to represent the weak planes in the micro
structure such as the inter particle contact planes, interfaces, planes of microcracks, etc.
!
!
(a) (b)
!
!
(c)
!
Figure 5.19: Microplane model: a) integration sphere; b) microplane stressstrain components; and c) stress
transfer through a number of contact planes (microplanes).
The strain components on the microplane are taken to be projections of the macroscopic strain
tensor (kinematic constraint approach) from which the normal and shear strain component are
computed. The normal microplane strain component is then decomposed into volumetric and
deviatoric part to realistically model concrete dominated by compressive load and to control
the initial elastic value of the Poissons ratio (Baant and Prat 1988a). The simplicity of the
model is due to the fact that constitutive properties are characterized entirely by only
relationships between the stress and strain components of microplanes. Knowing the stress
strain law for each microplane component, the macroscopic stiffness and stress tensors are
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 181
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
calculated by the stress integration over all microplanes. Baant and Prat produced the strain
stress microplane laws for the first loading and unloading part, while Obolt and Baant
(1992) introduced more general rules at the microplane level, including rate sensitivity
(Figure 5.20).
(a) (b)
Figure 5.20: Cyclic stressstrain relationship; a) volumetric microplane component; b) "Deviatoric microplane
component (Obolt et al. 2001).
In 3D analysis of reinforced concrete structures, the bond between reinforcement and concrete
can be modelled by a layer of 3D elements around the reinforcing bars. Hoehler and Obolt
(2001) adopted a microplane model for these bond elements such that the shear stressstrain
relationship provides a realistic shear stressslip response. Figure 5.21 shows the application
of the microplane based FEM program in simulation of an experiment on a beam column
connection that was tested by Ma et al. (1976). The analysis was performed by a displacement
controlled method as set out in Hoehler and Obolt (2001). The applied displacement history
is shown in Figure 5.21b (note that in initial stage, the experiment was loadcontrolled). The
test and numerically simulated results are compared in Figure 5.22 and a reasonable
correlation is observed.
LB16
LB13
Displacement (inch)
(b)
LB15
(a)
Figure 5.21: Details of by Ma et al. (1976) test: a) layout of experiment; b) time history of "applied displacement
in the Hoehler and Obolt (2001) analysis.
182 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Experiment
MASA
(c)
Displacement
(inches)
(a) (b)
(d)
Figure 5.22: 3D microplane FEM simulation and comparison with experimental data: a) section of FEM mesh;
b) comparison of loaddisplacements; c) distribution of principal strains from analysis; and (d)
axial strain in reinforcing bars from the analysis (Hoehler and Obolt, 2001).
A second example of the application of microplane models is that of Liu and Foster (2000)
where 3D modelling was used to examine the effects of cover spalling in the design of high
strength concrete columns under high axial loading. Figures 5.23a and 5.23b show the mesh
used for the modelling of column CS19 tested by Razvi and Saatcioglu (1996). One eighth of
the column modelled due to symmetries about XY, YZ and XZ planes. The concrete was
modelled using 20node isoparametric solid elements with numerical integration on a
2 o 2 o 2 Gaussian quadrature. The insitu mean concrete strength was taken as fc = 83 MPa,
being 90% of the mean cylinder strength tested under standard conditions. The longitudinal
and transverse reinforcement were modelled using 3node truss elements with numerical
integration on a 2 point Gauss quadrature. A perfect elastoplastic stressstrain relationship
was used for both the longitudinal and transverse reinforcement with a modulus of elasticity
of 200 GPa. The yield strengths of the longitudinal and tie reinforcement was 450 MPa and
400 MPa, respectively.
To include the effect of cover spalling in the analysis, a phenomenon commonly observed in
laboratory tests of axially loaded columns, the cover was taken to be spalled from the cross
section when a threshold transverse tensile strain was reached at the interface between the
cover concrete and the core. Such tensile strains occur due to the effects of the restraint of the
Poissons effect provided to the core by the tie reinforcement (see Foster et al., 1998). In this
analysis, the cover was taken as spalled if this transverse tensile strain between the cover and
the core elements exceeded 750 B). The spalling was included in the analysis by setting the
cover elements to a low stiffness once the threshold strain had been reached and, thus, the
cover elements do not contribute to the axial capacity of the section once spalling of the
element was deemed to have occurred. The failure of the column was by an axial localization
of the concrete in compression around the XZ axis of symmetry, as shown in Figure 5.23c.
In the Razvi and Saatcioglu experiment, the axial strain was measured over a gauge length of
300 mm (or 150 mm from the centreline with symmetry). The load versus axial strain results
from the FE model is compared to the laboratory measurements in Figure 5.24, with and
without cover spalling included in the model. A good correlation is observed between the FE
results and the experimental data only when cover spalling is included. If the cover spalling
phenomena, as observed in the laboratory tests, was not accounted for in the numerical
analysis, the strength would be over predicted by some 20%. This example is used here to
demonstrate the care needed when undertaking nonlinear FE modelling that all influences
significant to the failure condition can be captured in the analysis.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 183
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
P = 3840 kN
(descending)
Note:
Cover elements
not shown for
clarity
Figure 5.23: Finite element mesh for Razvi and Saatcioglu (1996) column CS19: (a) details and dimensions;
(b) 3D view; (c) xz displacements at P = 3840 kN on the descending curve (dimensions in mm).
8000
7000
6000
5000
Load (kN)
4000
3000
Experimental
2000 FEM  without cover spalling
FEM  cover spalling at 750 microstrain
1000
0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
Axial strain
Figure 5.24: Finite element analysis of the Razvi and Saatcioglu (1996) column CS19 with and without cover
spalling.
184 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
(b)
(a)
Figure 5.25: a) Wall with tendon duct; and b) details of anchorage.
The model of the wall represents a typical anchoring detail in bridge structures, where the pre
stressing tendon is anchored in a relatively thin wall. The reinforcement arrangement,
schematically shown in Figure 5.25b, must provide sufficient resistance to a splitting action of
the anchor as well as confinement effect to concrete and ensure safe tendon anchoring. The
goal was to determine the horizontal force to be resisted by the reinforcement. The study
included variations of the amount of tie and spiral reinforcement and was used as a basis to
improve design of anchoring regions with optimal arrangements of reinforcement.
To meet the design goals, a 3D solid analysis was employed. The model was reduced to using
advantage of two symmetry axes as shown in Figure 5.26. It includes concrete, steel anchors
and all reinforcing bars that are embedded in the concrete solid elements. A fractureplastic
constitutive model was used for the concrete and a VonMises plasticity model for steel of the
anchoring body. The reinforcing bars were modelled using a bilinear elasticplastic law with
hardening.
The loading, representing the anchoring force of the prestressing tendon, was applied by
prescribed forces at the top surface of the anchor. The force was applied in load steps up to
the value corresponding to full prestressing and then progressively increased to failure.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 185
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Examples of results of the analyses for the anchoring system are shown in Figure 5.27. The
stress state of the anchoring region was evaluated in detail and the concrete, compressive
stresses and plastic deformations were examined and crack widths were evaluated. The
stresses in the reinforcement were also examined.
The response of the anchor with spiral reinforcement (300mm in diameter) can be seen from
the loaddisplacement diagram in Figure 5.28a and the stresses in the spiral reinforcement,
Figure 5.28b, for different concrete quality (0.6 and 0.8 fc) and for different amounts of tie
reinforcement. The response can be compared with the level of service load that is applied at
the prestressing stage.
(a) (b)
Figure 5.27: Results of FE analysis: a) cracks and compressive stress contours; b) stress in reinforcing bars.
186 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
3.5
2.0
3.5
3
Anchoring force [MN]
Service load
2.5
1.5
SP300 C0.8
1 SP300 C0.8 ties 1x
0
0 100 200 300 400
Stress in spiral reinforcement [MPa]
(b)
Figure 5.28: Results from FE analyses: a) Anchoring force verses displacement diagram; b) stress in spiral
reinforcement due to anchoring force.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 187
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
vibration
protector
generator
machine
base
23.0m
casing
pile
For design, the high pressure of running water inside of the casing is sustained by both the
casing pipe, made of steel, and the surrounding reinforced concrete structure. If the
hydrodynamic pressure was to be carried by the steel casing alone, its thickness becomes so
large that welding quality could not be well assured. Thus, the reinforced concrete is expected
to partially bear the pressure. At the same time, concrete cracking caused by the high pressure
must be controlled so that the performance of powerhouse is maintained and able to support
the weight of the plant and to act as a foil against underground water entering through the
188 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
rock joints. A conventional design, based on 3D linear analysis, shows that the quantities of
hoop reinforcement needed is 3.3% around the casing pipes, as shown in Figure 5.29. The
arrangement of the reinforcing bars in 3D is laborious and the high quantity makes high
quality concreting works problematic as compaction is difficult.
In reality, however, the thermal stress component declines significantly due to stiffness
reduction by cracking and a linear analysis based design (strength based design) tends to bring
about more reinforcement than is needed to meet serviceability and safety requirements. On
the other hand, however, the steel stresses in the composite structures tend to be
underestimated. Thus, a nonlinear analysis for the performance assessment is required in
view of both safety and cost benefits.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 189
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
the steel casing. Localized damage may cause unexpected vibrations that would make the
power generator unstable. An approximately 36% reduction of costs related to the casing
runner was achieved by using nonlinear 3D FE modelling.
While this example shows some of the advantages of moving to higher methods of modelling,
before adopting any nonlinear analysis results great care is needed to ensure that the
modelling accurately simulates the 3D crack propagation and size effects. Thus, tension
softening of concrete around the plain concrete zone and tension stiffening of concrete close
to the casing runner were taken into account and the accuracy of the 3D nonlinear analysis
was verified by experiments (see Kato, 2000, and Kato, 2001).
5.6 References
An, X., Maekawa, K., Okamura, H. (1997), Numerical Simulation of Size Effect in Shear
Strength of RC Beams, Journal of Material, Concrete structures and Pavements, 35(564), pp.
297316.
Baant, Z.P., and Gambarova, P. (1984), Crack Shear in Concrete: Crack Band Microplane
Model, Journal of Engineering Mechanics, ASCE, 110, pp. 20152035.
Baant, Z.P., and Prat, P.C. (1988a), Microplane Model for BrittlePlastic Material: I
Theory, Journal of Engineering Mechanics, ASCE, 114(10), pp. 16721688.
Baant, Z.P., and Prat, P.C. (1988b), Microplane Model for BrittlePlastic Material: II
Verification, Journal of Engineering Mechanics, ASCE, 114(10), pp. 16891702.
Baant, Z.P., and Oh, B.H. (1983), Crack Band Theory for fracture of concrete, Material and
Structures, RILEM, Paris, 16, pp. 155177.
Baant, Z.P., and Obolt, J. (1990), Nonlocal Microplane Model for Fracture, Damage and
Size effect in Structures, Journal of Engineering Mechanics, ASCE, 116(11), pp. 24852504.
Baant, Z.P., Xiang Y., and Prat, P.C. (1996a), Microplane Model for Concrete I. Stress
Strain Boundaries and Finite Strain, Journal of Engineering Mechanics, ASCE, 122(3), pp.
245262.
Baant, Z.P., Xiang, Y., Adley, M., Prat P.C., and Akers, S. (1996b), Microplane Model for
Concrete II. Data Delocalization and Verification, Journal of Engineering Mechanics,
ASCE, 122(3), pp. 263268.
Carol, I., Prat, P., and Baant Z.P. (1992), New Explicit Microplane Model for Concrete:
Theoretical Aspects and Numerical Implementation, International Journal of Solids and
Structures, 29(9), pp. 11731191.
190 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
CEBFIP Model Code 1990 (1993), Comit EuroInternational du Bton, Thomas Telford,
London.
Cervenka, V. (2002), Computer simulation of failure of concrete structures for practice, 1st
fib Congress, Concrete Structures in 21 Century, Osaka, Japan, Keynote lecture in Session 13,
289304
Cervenka, J., Cervenka, V., and Eligehausen, R. (1998), FracturePlastic Material Model for
Concrete, Application to Analysis of Power Actuated Anchors, Proc. FRAMCOS, 3, pp.
11071116.
Cervenka, V., and Gerstle, K. (1971), Inelastic Analysis of Reinforced Concrete Panels: (1)
Theory, (2) Experimental Verification and application, Publications IABSE, Zrich, V.31
00, 1971, pp.3245, and V.32II,1972, pp.2639.
De Borst, R., and Nauta, P. (1985), NonOrthogonal Cracks in a Smeared Finite Element
Model, Engineering Computations, 2, pp. 3446.
Drucker D.C., and Prager W. (1952), Soil Mechanics and Plastic Analysis or Limit Design
Quarterly Journal of Applied Mathematics, 10(2), pp/ 157165.
Feenstra, P.H. (1993), Computational Aspects of Biaxial Stress in Plain and Reinforced
Concrete, PhD Thesis, Delft. University of Technology, The Netherlands.
Foster, S.J., Liu, J. and Sheikh, S.A. (1998), Cover spalling in HSC columns loaded in
concentric compression, Journal of Struct. Engrg., ASCE, 124(12): 14311437.
Fukuura, N., and Maekawa, K. (1998), MultiDirectional Crack Model for InPlane
Reinforced Concrete Under Reversed Cyclic Actions4 Way Fixed Crack Formulation and
Verification, in de Borst et al (eds.), Computational Modeling of Concrete Structures,
Balkema, Rotterdam and Brookfield, pp. 143152.
Hauke B., and Maekawa, K. (1998), ThreeDimensional Reinforced Concrete Model with
MultiDirectional Cracking in de Borst et al (eds.), Computational Modeling of Concrete
Structures, Balkema, Rotterdam and Brookfield, pp. 93102.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 191
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Hofstetter, G., and Mang, H.A. (1995), Computational Mechanics of Reinforced Concrete
Structures, Friedr. Vieweg & Sohn Verlagsesellschaft mbHg, Braunschweig/Wiesbaden.
Ichikawa, H., Ogasawara, M., and Maekawa K. (1998), Experimental Study of Seismic
Performance of RC Piers Subjected to Eccentric Axial Force and Torsional Moment, Proc. of
JCI, 20(1), pp. 149154.
Kato, S., Iizuka, K., and Nanbu, S. (2000), Nonlinear Behavior of Reinforced Concrete
Structures Subjected to High Water Pressure, Proc. of JCI, Vol.22, No.3, pp.6772.
Kato, S., Yamaya, A., Zhao, Y., and Ozawa, H. (2001), Study on Structural Performance of
Inner concrete using Ringgage, Proc. of JCI, Vol.23, No.1, pp.631636.
Liu, J., and Foster, S.J. (2000), A Three Dimensional Finite Element Model for Confined
Concrete Structures, Computers and Structures, Vol. 77, No. 5, October, pp 441451.
Ma, S.Y.M., Bertero, V.V., and Popov E.P. (1976), Experimental and Analytical Studies on
the Hysteretic Behavior of Reinforced Concrete Rectangular and TBeams, Earthquake
Engineering Research Center (EERC), Report No. UBC/EERC 762, University of California,
Berkeley.
Maekawa, K., Irawan, P., and Okamura, H. (1997), Path dependent three dimensional
constitutive laws of reinforced concreteformulation and experimental verification, Struct.
Eng. Mech., 5(6), pp. 743754.
Maekawa, K., Pimanmas, A., and Okamura, H. (2003), Nonlinear Mechanics of Reinforced
Concrete, Spon Press, London.
Marti, P., Leesti, P., and Khalifa, W.U. (1987), Torsion Tests on Reinforced Slab Elements,
Journal of the Structural Division, ASCE.
Menetrey, P., and Willam K.J., (1995), Triaxial Failure Criterion for Concrete and its
Generalization, ACI Structural Journal, 92(3), pp. 311318.
Mitchell, D., and Collins, M.P. (1974), Diagonal Compression Field TheoryA Rational
Model for Structural Concrete, ACI Journal Proceeding, 71(8), pp.396408.
Okamura, H., and Maekawa, K. (1991), Nonlinear Analysis and Constitutive Models of
Reinforced Concrete, GihodoShuppan Co., Tokyo.
Ottosen, N.S. (1977), A failure criterion for concrete., Journal of Engrg. Mech. Div., ASCE,
103(EM4): 527535.
Onsongo, W. (1978), The Diagonal Compression Field Theory for Reinforced Concrete
Beams Subjected to Combined Torsion, Flexural and Axial Load, Thesis presented to the
University of Toronto, Toronto, Canada.
Obolt, J., and Baant, Z.P. (1992), Microplane Model for Cyclic Triaxial Behavior of
Concrete, Journal of Engineering Mechanics, ASCE, 118(7), pp. 13651386.
Obolt, J., Li, Y., and Kozar I. (2001), Microplane Model for Concrete with Relaxed
Kinematic Constraint, International Journal of Solids and Structures, 38, pp. 26832711.
192 5 Analysis and desing! of surface and solid structures using nonlinear models
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Pramnono, E., and Willam, K.J. (1989), Fracture EnergyBased Plasticity Formulation of
Plain concrete, Journal of the Engineering Mechanics, ASCE, 115, pp. 11831204.
Razvi, S.R, and Saatcioglu, M. (1996), Tests of high strength concrete columns under
concentric loading. Report OCEERC 9603, Department of Civil Engineering, University of
Ottawa, Canada.
Stevens, N.J., Uzumeri, S.M., and Collins, M.P. (1987), Analytical Modeling of Reinforced
Concrete Subjected to Monotonic and Reversed Loadings, Publication No. 8701,
Department of Civil Engineering, University of Toronto, Toronto.
Taylor, G. I. (1938), Plastic Strain in Metals, J. Inst. of Metals, London, 62, pp. 307324.
Tsuchia, S., Tsuno, K., and Maekawa, K. (2001), Nonlinear ThreeDimensional FE Solid
Response Analysis of RC Columns Subjected to Combined Permanent Eccentric Axial Force
and Reversed Cyclic Torsion and Bending/Shear, Journal of Material, Concrete structures and
Pavements, 683(52), pp. 131143.
Vecchio, F.J., and Collins, M.P. (1982), The Response of Reinforced Concrete to InPlane
Shear and Normal Stresses, Publication No. 8203, Department of Civil Engineering,
University of Toronto, Toronto.
Vecchio, F.J., and Collins, M.P. (1986), The Modified CompressionField Theory for
Reinforced Concrete Elements Subjected to Shear, ACI Journal, V. 83, No. 2, MarchApril,
pp. 219231.
Vecchio, F.J., and Selby, R.G. (1991), Toward CompressionField Analysis of Reinforced
Concrete Solids, Journal of Structural Engineering, ASCE, 117(6), pp. 17401758.
Willam, K.J., and Warnke, E.P. (1975), Constitutive Model for the Triaxial Behavior of
Concrete, Proceeding of the International Association for Bridge and Structural Engineering,
19.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 193
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
[Blank Page]
!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Arguably, nonlinear elasticity is the most general and most popular of the approaches used to
model reinforced or prestressed concrete structures, and most commercial FE packages have
the ability for the user to use nonlinear elastic models. Methods for analysis of reinforced
concrete elements, based on concepts of linear or nonlinear elasticity, encompass a wide range
of approaches, many of which have met with good success.
Presented below is the basic formulation for a smeared rotating crack model using a secant
stiffness approach. An example of the formulation of a fixed crack model based on a tangent
stiffness approach is provided by Maekawa et al. (1999). A hybrid formulation, combining
elements of both fixed and rotating cracks, is described by Vecchio (1990, 2000, 2001). To
simplify the discussion, the formulation presented will be limited to the case of two
dimensional plane stress; all formulations are easily expandable to the general three
dimensional case.
Consider a reinforced concrete element, of unit dimensions, subjected to the plane stress
condition +. The element contains n components of reinforcement, where the ith
reinforcement component is oriented at angle :i to the reference xaxis (0 ? i ? n). The
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 195
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
amount of reinforcement in the ith direction is denoted by the reinforcement ratio *i. The
constitutive response of the element is described as
+=D) (6.1)
where + and ) are the total stress and total strain vectors for the element, and D is the
composite secant stiffness of the material.
At this point, it is being assumed that the strains are common for all components, meaning:
perfect bond between concrete and reinforcement; no initial strains in the materials (e.g., no
concrete shrinkage); and no subsequent volume changes due to thermal expansion, creep, or
other types of elastic or plastic mechanisms (these are treated later). The composite material
stiffness matrix D is comprised of contributions from the concrete and each reinforcement
component; thus:
n
D ' Dc 8 P D s i (6.3)
i '1
For each reinforcement component, the material stiffness must first be formulated in the
direction of the reinforcement, then transformed back to the reference axes (x, y). Thus, for
reinforcement component i,
Z * i E si 0 0W
X U
Ds i ' X 0 0 0U (6.4)
X 0 0 0UV
Y
The reinforcement stress f si is found from a suitable stressstrain model for reinforcing steel
(e.g. elasticplastic with strain hardening), evaluated for the steel strain ) si . To transform
back to the reference axes, the following transformation is used:
where
and r = :i.
Prior to cracking, in low confinement states, concrete can be sufficiently well represented as
an isotropic material; thus
Z1 n 0 W
Ec X U
Dc ' n 1 0 (6.8)
1 7 n 2 XX U
Y0 0 (1 7 n ) / 2 UV
After cracking, concrete behaves as an orthotropic material, with the directions of orthotropy
aligned with the crack direction. Thus, the concrete material stiffness matrix must be
evaluated accordingly; that is, first defined with respect to the principal stress directions
(defined by angle , normal to the crack direction) and then transformed back to the reference
axes. Thus,
Dc ' T T Dc T (6.9)
where T is the transformation rotation matrix as before except that r = . The concrete
stiffness D is evaluated as
Z Ec1 0 0W
X U
D 'X 0 Ec 2 0U (6.10)
X 0 0 Gc UV
Y
where Ec1 and E c 2 are the effective secant moduli for the concrete in the principal tensile
stress and principal compressive stress directions, respectively, and Gc is the effective secant
shear modulus. Note that the absence of offdiagonal terms in D implies that postcracking
Poissons ratios are ignored; these can be included in a slightly more rigorous formulation
described later. The effective secant moduli are evaluated as:
f f Ec1 T Ec2
Ecl ' c1 ; Ec 2 ' c 2 ; Gc ' (6.11)
)1 )2 Ec1 8 Ec2
where fc1 and fc2 are the postcracking principal tensile stress and principal compressive stress
in the concrete, respectively, evaluated based on the concrete principal tensile and
compressive strains, )1 and )2, and using appropriate stressstrain relationships. Models for
calculating fc1 can include the effects of tension softening, tension stiffening, and other
influencing mechanisms (e.g. local stress conditions at crack locations). Models for
calculating fc2 will typically consider compression softening and/or confinement effects.
Note: The nature of the constitutive models for calculating fc1 and fc2 will depend on whether
a fixed or rotating crack approach is being used, and some caution is needed in selecting
these.
The accuracy and range of application of this basic formulation can be considerably enhanced
by introducing provisions for modelling elastic and plastic offset strains. With respect to the
concrete, this will allow the consideration of such effects as dilation (postcracking Poissons
effects), thermal strains, shrinkage, and plastic residual strains (due to yielding and
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 197
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
unloading). For the reinforcement, it will permit the inclusion of prestressing strains, thermal
strains, and plastic offsets (yielding).
The general approach used is to first decompose, for each material, the total strains into elastic
or plastic offset strains and net elastic strains. Thus, for concrete,
where ) are the total element strains (as before), )c are the concrete net elastic strains, )co are
the concrete elastic offset strains (e.g. due to shrinkage, thermal, or Poissons effects), and )cp
are the concrete plastic offset strains (due to unloading, damage). Similarly, for the
reinforcement
where )si are the net elastic strains for the ith component reinforcement, )soi are the
reinforcement elastic offset (due to prestressing or thermal effects), and )spi are the plastic
offsets (due to yielding, unloading). It is essential to note that it is the net elastic strains, )c
and )si , that must be used within constitutive relations when determining concrete and
reinforcement stresses, and when defining the effective secant moduli.
n
" o ' Dc ( !co 8 !cp ) 8 P Ds i (! soi 8 ! spi ) (6.15)
i '1
As an example, consider the formulation required to represent dilation effects in the concrete.
Let n21 represent the strain in the 2direction due to a stress in the 1direction; similarly, n12
represents a strain in the 1direction due to a stress in the 2direction, where 1, 2 are the
principal stress directions. Thus, define
p
!co ' ) cx
o o
) cy o
& cxy qT
(6.16)
where
o 1 1
) cx ' ) co1 (1 8 cos 2 ) 8 ) co2 (1 7 cos 2 ) (6.17a)
2 2
o 1 1
) cy ' ) co1 (1 7 cos 2 ) 8 ) co2 (1 8 cos 2 ) (6.17b)
2 2
o
& cxy ' ) co1 T sin 2 7 ) co2 T sin 2 (6.17c)
and where, in this case, ) co1 ' n12 T ) c 2 , ) co2 ' n 21 T ) c1 and  is the inclination of the principal
stress direction.
Note that +si represents the contribution of the ith reinforcement component to the total
element stresses. The actual stress in the reinforcement must be evaluated as
f si ' + si * i . (6.19)
In this basic formulation, an iterative solution procedure is required. Some estimate of the
material secant moduli is first made, and the element composite stiffness matrix is assembled.
For a given element stress condition, the element total and net strains are then calculated.
From the net elastic strains thus determined, the material secant moduli can be recalculated.
The process is repeated until the secant moduli have converged to stable values. This
approach is applicable to postpeak strain/strain conditions with equal numerical stability as
for prepeak conditions.
Additional details pertaining to this basic formulation can be found in Vecchio (1990).
Equally viable formulations for a fixed crack approach, or for tangentstiffness based models
(either fixed or rotating crack based), can be found in the literature. More esoteric
formulations based on multiple nonorthogonal cracks, microplane models, or hybrid crack
models can also be found, and these generally provide equal capacity for accurate modelling
of reinforced concrete based on elasticity principles.
6.2.2 Plasticity
The flow theory of plasticity is a natural generalization of the classical work of Tresca (1868),
SaintVenant (1870), Lvy (1870) and von Mises (1913). Its principal ingredients are the
yield condition, the flow rule and the hardening law. The total strain is additively decomposed
into the elastic strain and the plastic strain. It is assumed that the current state of an
elementary material volume is fully described by the total strain, plastic strain, and by some
additional hardening (or softening) variables that characterize the changes of microstructure
and are collected in a vector k. The basic equations include the elasticplastic split,
) ' )e 8 ) p
(6.20)
and hardening law, which consists of two parts  the definition of the hardening variables, in
general in the form of rate equations
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 199
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
and the dependence of the parameters q appearing in Eqs. 6.22 and 6.23 on the hardening
variables,
q ' h( f ) (6.25)
In the above, )={)11, )22, )33, &23, &13, &12}T is the column matrix of engineering strain
components; )e and )p represent elastic and plastic strain, respectively; +={+11, +22, +33, +23,
+13, +12}T is the column matrix of stress components; De is a square matrix of elastic moduli
(the elastic material stiffness matrix); f is the yield function; g is a function specifying the
direction of plastic flow; and ^ is the rate of the plastic multiplier. A dot over a symbol
denotes differentiation with respect to time. However, recall that the theory is rate
independent. Time plays here only the role of a formal parameter that controls the loading
process, and it does not need to have the meaning of real physical time. The rate equations,
Eqs. 6.23 and 6.24 could also be presented in an incremental form, with rates replaced by
infinitesimal increments.
The yield function defines the elastic domain in the stress space, which is bounded by the
yield surface. States for which f<0 are elastic, states corresponding to f=0 are plastic, and f>0
is not possible. Of course, as the variables q grow, the yield surface evolves  it can expand,
shrink, translate and even change its shape. Plastic flow can take place only if the current state
is plastic; this is expressed by the condition.
Indeed, if the material is in an elastic state (f<0), Eq. 6.26 implies ^! ' 0 and according to
Eqs. 6.23 and 6.24, the plastic strain and the hardening variables remain constant. On the
other hand, in a plastic state we have f=0 and Eq. 6.26 does not restrict the rate of the plastic
multiplier. This rate should never be negative, because g in Eq. 6.23 specifies the oriented
direction of evolution (e.g., the plastic strain rate in a uniaxial tensile test can be zero or
positive). This is described by an additional restriction, ^! < 0 . Combining it with Eqs. 6.22
and 6.23 we obtain the loading/unloading conditions in the socalled KuhnTucker form as,
During plastic flow, the yield function must remain equal to zero, and so the rate of its change
is also zero. This consideration leads to the consistency condition,
^! f! ' 0 (6.28)
Suppose that the current values of all variables as well as the rate of the total strain are given.
The basic equations make it possible to compute the rates of all variables. If the current state
is elastic (f < 0), the plastic multiplier, plastic strain and hardening variables remain constant,
and the stress evolution is governed by the elastic law. If the current state is plastic (f = 0), the
plastic flow can continue or the material can unload elastically. The former case is
characterized by f! ' 0 and ^! @ 0 . In the latter case, we have ^! ' 0 and, as the subsequent
stress state must be elastic, the yield function must decrease, which corresponds to f! 9 0 . The
special case when f! ' 0 and ^! ' 0 is called neutral loading.
If plastic loading takes place, the rate of plastic multiplier can be calculated from the
condition f! ' 0 . Applying the chain rule, we can write this condition as,
T T
e sf b e sf b
f! ' c ` "! 8 cc `` q! (6.29)
d s" a d sq a
Combining the rate form of the elastic stressstrain law with the additive split and the flow
rule we get
The rate form of Eq. 6.25 combined with Eq. 6.24 yields
sh sh
q! ' f! ' ^! k (6.31)
sf sf
By substituting Eqs. 6.30 and 6.31 into Eq. 29 we have a linear equation for ^! , from which
f T+ De )!
^! ' (6.32)
f T" De g 7 f Tq Hk
e De gf T+ Deb
+! ' c De 7 ` )! (6.33)
c f + De g 7 f Hk `a
T T
d q
sf (+ ,q )
)! p ' ^! (6.34)
s+
As the gradient sf ls+ is normal to the yield surface, Eq. 6.34 is sometimes called the
normality rule. Yielding of metals is usually insensitive to the volumetric part of the stress
tensor, and it can be conveniently described by the von Mises yield condition
f (+ ) t J 2 (+ ) 7, 0 ? 0 (6.35)
in which , 0 is the yield stress in shear and J2 is the second deviatoric invariant of the stress
tensor. The J2invariant can be computed from the principal stresses +1, +2 and +3 according
to the following formula:
1
p
J 2 ' (+ 1 7 + 2 ) 2 8 (+ 2 7 + 3 ) 2 8 (+ 3 7 + 1 ) 2
6
q (6.36)
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 201
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
It is customary to present yield conditions graphically in the principal stress space with axes
+1, +2 and +3. The line on which all three principal stresses are equal is the hydrostatic axis,
and planes perpendicular to this line are deviatoric planes. Intersections of the yield surface
with deviatoric planes are deviatoric sections, and intersections with planes that contain the
hydrostatic axis are called meridians.
The yield surface corresponding to the Mises criterion is a cylinder rotationally symmetric
about the hydrostatic axis and is generally not acceptable for concrete structures. All
deviatoric sections are identical circles, and all meridians are identical straight lines parallel to
the hydrostatic axis. Any normal to the Mises cylinder lies in a deviatoric plane. This means
that the flow rule associated with the Mises condition gives purely deviatoric plastic flow,
which is for metals in a good agreement with reality. Volume changes are in this case purely
elastic.
In Eq. 6.35, the shear yield stress , 0 is a constant material parameter, and so the yield surface
does not evolve during plastic flow. This is the case of perfect plasticity (i.e., plasticity
without hardening or softening), which does not use any hardening parameters. More
advanced models take into account the changes in the microstructure induced by plastic flow
and the resulting evolution of the yield surface. The simplest hardening model only expands
or shrinks the yield surface radially. Such isotropic hardening can be described by a single
scalar hardening variable, k. To characterize the total amount of plastic flow, the strain
hardening model uses the accumulated plastic strain, defined by the rate equation as,
2
f! ' )! p (6.37)
3
where  .  denotes the Euclidian norm of a tensor. The role of the scaling factor 2 3 is to
make k coincident with the usual plastic strain under uniaxial loading, provided that the
plastic flow is purely deviatoric. Alternatively, the workhardening model defines k as the
total plastic work, the rate of which is
f! ' + T )! p (6.38)
By substituting from Eqs. 6.23 and 6.25 we can convert either of these definitions into the
general form of Eq. 6.24. The dependence of the current shear yield stress on the hardening
variable, , 0 ' h(f ) , can easily be identified from a uniaxial test. The yield function is then
defined as
f (+ ,, 0 ) t J 2 (+ ) 7, 0 (6.39)
f (+ , : ) t J 2 (+ 7 : ) 7 , 0 (6.40)
A classical evolution law for the back stress was proposed by Prager (1955) and improved by
Ziegler (1959). Realistic modelling of metals under cyclic loading requires even more
complex models (Mroz 1967, Dafalias and Popov 1975, Krieg 1975).
For pressuresensitive materials such as concrete, rock, gravel or soil, it is important to take
into account the effect of volumetric part of the stress tensor by the first stress invariant,
I1 ' + 1 8 + 2 8 + 3 (6.41)
By introducing a dependence on the volumetric stress into the Mises condition, we obtain the
DruckerPrager criterion as
f (+ ,, 0 ) t : I1 (+ ) 8 J 2 (+ ) 7, 0 ? 0 (6.43)
If the friction coefficient : is constant, the yield surface is a rotationally symmetric cone with
straight meridians and circular deviatoric sections. Experimental measurements on concrete
specimens indicate that the meridians are curved and the deviatoric sections have the form of
a rounded triangle, whose shape changes from almost triangular for tensile and low
compressive hydrostatic pressures to almost circular for high compressive hydrostatic
pressures. As the deviatoric section is not circular, the yield function must depend on the third
deviatoric invariant as,
It is more convenient to transform J3 into the socalled Lode angle, which has a direct
geometrical meaning. This angle, , is defined by the relationship
3 3 J3
cos 3 ' (6.45)
2 J 23 / 2
and its geometrical interpretation in the deviatoric plane is indicated in Figure 6.1a. A fairly
general description of the strength envelope can be presented in the form:
where c1, c2 and c3 are material parameters and &()) is a suitable function of the Lode angle
related to the shape of the deviatoric section. A failure criterion of the form given by Eq. 6.46
was first proposed by Ottosen (1977) as
g Z1 W
hh cos X 3 arccos( K cos 3 )U if cos 3 < 0
r ( ) ' i Y V (6.47)
Z L
hcos 7 arccos(7 K cos 3 )W
1
if cos 3 ? 0
hj XY 3 3 UV
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 203
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
where K is a shape factor affecting the outofroundedness of the deviatoric section; see
Figure 6.1a. Another function &()) with the desired properties was constructed already by
Willam and Warnke (1974) but they exploited it in a somewhat different manner. Their
function,
is derived from an ellipse in the deviatoric plane. Parameter e is the eccentricity, again related
to the shape of the deviatoric section, as illustrated in Figure 6.1b. Mentrey and Willam
(1995) presented a failure criterion of the form of with function &()) given by Eq. 6.48.
Both definitions given by Eqs. 6.47 and 6.48 lead to a convex failure surface if the parameters
K or e are chosen within a certain admissible range. The surface is smooth everywhere except
for the intercept with the hydrostatic axis. If the parameters are determined from the same set
of experimental data, there is usually only slight difference between the surfaces generated by
the Ottosen criterion and the MentreyWillam criterion. As an example, Figure 6.2 shows the
deviatoric sections of the Ottosens surface constructed with parameters c1 = 0.16 MPa1,
c2 = 0.5 MPa1, c3 = 0.005 MPa2 and K = 1. These values correspond to the recommendation
of the CEB Model Code 1990 (1993) for concrete of grade C20.
Finally, let us point out that associated flow rules usually mispredict plastic changes of
volume in frictional materials such as concrete, and they have to be replaced by non
associated ones, often written in the form
sg
)! p ' ^! (6.49)
s+
where g(),q) is a new function called the plastic potential. We might say that the flow rule is
associated if the yield function is used at the same time as the plastic potential. A general
plastic potential defines a set of equipotential surfaces such that the plastic strain always
grows in the direction normal to the surface on which the current stress state is located. It is
often possible to assume that the flow is associated in the projection onto the deviatoric plane
and only its volumetric part is nonassociated. A suitable plastic potential can derive from the
yield function by modifying only the term that reflects the pressure sensitivity of the material.
From the DruckerPrager criterion (Eq. 6.43), we could derive a plastic potential as
g ( I1 , J 2 ) ' :r I1 8 J 2 (6.50)
where :r is the dilatancy coefficient, in general different from the friction coefficient : that
appears in the yield function. The most difficult component of a plasticity model for concrete
is a realistic description of hardening and softening. This issue will be addressed later in the
sections dedicated to particular models.
K=1.0 e=0.5
K=0.9 e=0.6
K=0.8 e=0.7
(a) (b)
Figure 6.1: Influence of the shape parameter on the deviatoric section correspondingto zero mean pressure
according to a) Ottosen, b) Mentrey and Willam.
80 250
0 MPa 30 MPa
10 MPa 50 MPa
20 MPa 200 100 MPa
60 30 MPa 200 MPa
150
40
100
20
50
0 0
50
20
100
40
150
60
200
80 250
(a) (b)
Figure 6.2: Deviatoric sections of Ottosens surface at a) low and b) high levels of mean pressure.
The structure of a typical model based on damage mechanics will be explained using the
simplest example  the isotropic damage model with a single scalar parameter. The stress
strain equations are written in the total form of
where N is the damage parameter. Initially, N is equal to zero, and the response of the material
is linear elastic. As the material deforms, the initiation and propagation of microdefects such
as voids or microcracks decreases the stiffness, which is reflected by the growth of the
damage parameter. The model described by Eq. 6.51 is based on the simplified assumption
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 205
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
that the stiffness degradation is isotropic, i.e., stiffness moduli corresponding to different
directions decrease proportionally, independently of the direction of loading.
Similar to plasticity, we introduce a loading function f specifying the elastic domain and the
states at which damage grows. In damage theory it is natural to work in the strain space. The
loading function is therefore chosen to depend on the strain, ), and on an additional parameter
k that describes the evolution of the elastic domain. States for which f(), k) < 0 are supposed
to be below the current damage threshold. Damage can grow only if the state reaches the
boundary of the elastic domain. This is described by the loading/unloading conditions that
again have the KuhnTucker form of,
It remains to link the damage threshold k to the damage parameter N. It would be possible to
use a damage evolution law in a rate form. However, as both variables grow monotonically, it
is more convenient to postulate an explicit relation between their total values as
N ' g (f ) (6.53)
The function g affects the shape of the stressstrain diagram and can be directly identified
from a uniaxial test.
f () , f ) ' )# () ) 7 f (6.54)
where, )~ is the equivalent strain. The formula defining the equivalent strain plays a similar
role to the yield function in plasticity, because it directly affects the shape of the elastic
domain. The simplest choice is to define the equivalent strain as,
)# ' ) or (6.55)
)# ' ) (6.57)
or,
T
)# ' ) De ) / E (6.58)
where McAuley
M brackets < > denote thee positive part of. For F scalars, < x > = maax(0 , x),
i.e., < x > = x for x positive annd < x > = 0 for x neg gative. For symmetric
s ttensors, succh as the
strain teensor ), the positive paart is a tensoor having th
he same prinncipal axes as the origiinal one,
with priincipal valuues replacedd by their poositive partss. Consequeently Eq. 6.57 can be rewritten
r
as
3
)~ ' P)
2
i (6.59)
i '1
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 207
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
where )i, i = 1, 2, 3 are the principal strains. The elastic domains corresponding to Eqs. 6.57
and 6.58 are shown in Figure 6.3b. If a model corresponding to the Rankine criterion of
maximum principal stress is desired, one may use the definitions
1
)# ' max De) i
(6.60)
E i '1,2,3
or,
3
1 1
P
2
)# ' De) ' De) i
(6.61)
E E i '1
where < De) >i, i = 1, 2, 3, are the positive parts of principal values of the effective stress
tensor De). The former definition exactly corresponds to the Rankine criterion while the latter
rounds off the corners in the octants with more than one positive principal stress; see Figure
6.3c.
An important advantage of damage models is that the stress evaluation algorithm is usually
explicit, without the need for an iterative solution. For a loading function in the form of
Eq. 6.54, the variable f has the meaning of the largest value of equivalent strain that has ever
occurred in the previous deformation history up to its current state. For a prescribed strain
increment, the corresponding stress is computed simply by evaluating the current value of
equivalent strain, updating the maximum previously reached equivalent strain and the damage
parameter, and reducing the effective stress according to Eq 6.51. Depending on the definition
of equivalent strain, we may have to extract the principal strains or principal stresses. This can
be done relatively easily, since closedform formulas for the eigenvalues of symmetric
matrices of size 2x2 or 3x3 are available.
Finally, let us emphasize that the purpose of the present brief introduction was to illustrate
some of the basic aspects of the approach based on damage, using the simplest possible
model. Damage mechanics literature contains a large number of models with various levels of
complexity and sophistication, e.g., twoparameter scalar damage models in the compliance
form (Ladevze 1983, Mazars 1985, Borderie 1991) or in the stiffness form (Mazars 1985,
Mazars 1986), models with two secondorder damage tensors and one scalar parameter
(Ramtani 1990, Papa and Taliercio 1996), a model characterizing damage by a fourthorder
damage tensor (Chaboche 1979), or models characterizing damage directly by the compliance
tensor or the stiffness tensor (Ortiz 1985, Simo and Wu 1987, Yazdani and Schreyer 1988) to
name only a few.
Smeared crack models present the total strain as a sum of two parts  one corresponds to the
deformation of the uncracked material, and the other is the contribution of cracking. The
response of the uncracked material can be governed by a general nonlinear material law but
usually is assumed to be linear elastic. The strain decomposition is written as
) ' )e 8 )c (6.62)
The crack strain, )c, represents in a smeared manner the additional deformation due to the
opening of cracks. A crack is initiated when the stress state reaches a certain failure surface.
Models aiming at the description of fracture under shear and compression exploit more
general criteria (Weihe 1995, Peekel Instruments 1994). The initiation criterion specifies also
the orientation of the crack. According to the Rankine criterion the crack plane is normal to
the direction of maximum principal stress. In a local coordinate system with axis n normal to
c
the crack and axes m and l lying in the crack plane, the only nonzero crack strains are ) nn,
c c
& nm and & nl. In other words, opening of the crack contributes to the normal strain in the
direction normal to the crack plane, and sliding of the crack contributes to the shear strains in
planes normal to the crack plane. The corresponding strain components in global coordinates
are obtained by the transformation as
where ni, mi and li, i=1,2,3 are the components of unit vectors in the direction of axes n, m and
l. In a compact notation, transformation rule (Eq. 6.64) can be written as
The components of ec are assumed to be directly related to the traction transmitted by the
crack, s = {snn, snm, snl}T which is a projection of the stress on the crack plane. As s and ec
form a workconjugate pair, the same as + and )c, the stress transformation formula
s ' T T+ (6.66)
contains the transpose of the strain transformation matrix T from Eq. 6.65.
s nn ' f () nn
c
) (6.67)
The cracking law (Eq. 6.67) for normal components is sufficient for the rotating crack model,
which assumes that the crack normal rotates and remains aligned with the current direction of
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 209
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
maximum principal strain. In contrast, the fixed crack model freezes the crack direction at the
moment of crack initiation. The crack in general transmits shear that produces relative sliding
of the crack faces, represented by shear components of crack strain. In simplistic versions of
the fixed crack model, the shear traction is taken as proportional to the shear crack strain, with
a proportionality factor (G, where G is the shear modulus of elasticity and ( < 1 is the so
called shear retention factor (Suidan and Schnobrich 1973). This is not very realistic because
such a model crack can transmit large shear tractions even when it is widely open. If the shear
retention factor is constant, a very small value has to be set, e.g. ( = 0.01, to limit the spurious
stress transfer that may lead to the stress locking 2. A better remedy is to make ( variable and
decrease it to zero as the crack opening grows (Cedolin and Poli 1977). It is also possible to
formulate the relation between s and ec in the spirit of damage theory, with a scalar damage
parameter depending on an equivalent crack strain that is computed from ec. Whatever choice
is made, the relations can always be transformed into the incremental form with tangential
stiffness matrix Dc as+
$
s! ' Dc e! c (6.68)
The rate forms of Eqs. 6.62, 6.63 and 6.65, can be combined to yield
By substituting this into the rate form of Eq. 6.66 and comparing with Eq. 6.68 we obtain a set
of equations for e!c from
Z1 7 n 0 0 W
E X 0 U (6.71)
D e ' (1 7 2n ) / 2 0
(1 8 n )(1 7 2n ) X U
XY 0 0 (1 7 2n ) / 2UV
is the same in any coordinate system, and no matrix multiplication is needed. Moreover, if the
cracking laws for the normal components and for the shear components are decoupled, then
the crack stiffness matrix Dc is diagonal, and the inversion of the diagonal matrix D e 8 D c is
straightforward. Substituting Eq. 6.70 back into Eq. 6.69 we finally obtain the relation
between the increments of stress and total strain in the form
D ' De 7 De T ( D e 8 D c ) 71 T T De (6.73)
D ' (C c 8 TC c T T ) 71 (6.74)
with C e ' De71 = elastic compliance matrix, and C c ' D c71 = tangent crack compliance matrix.
Eq. 6.74 is formally simpler than Eq. 6.73 but it requires the inversion of a 6x6 matrix while
in Eq. 6.73 we have to invert only a 3x3 matrix (sometimes even diagonal).
So far we have tacitly assumed that the crack opening is monotonically increasing. However,
it can also happen that the crack starts closing (unloading), and the tangent stiffness has to be
replaced by the stiffness valid for unloading. Usually it is assumed that unloading takes place
to the origin (Figure 6.4a), so that the crack strain completely disappears when the applied
stress is removed. A model with some permanent strain (Figure 6.4b) is more realistic but in
most practical simulations there is little difference between the two unloading models. When
the applied normal stress changes sign to compression, the crack closes completely and
becomes locked. The crack strain should not assume negative values because the crack
faces cannot overlap. This is reflected by the vertical part of the diagram in Figure 6.4a. Upon
complete crack closure, the material restores its original stiffness, and the overall response is
purely elastic as long as the normal traction remains negative. If this traction changes sign
again, the crack starts reopening at constant secant stiffness, and when the crack opening
reaches its previous maximum, the basic curve described by Eq 6.67 is followed.
! !
"c "c
(a) (b)
Figure 6.4: Cracking law a) with unloading to the origin, b) with some permanent strain.
The approach outlined above can be generalized to the case of m cracks with different
orientations. The basic equations remain formally the same but the column matrices s and ec
now consist of blocks that correspond to individual cracks, the transformation matrix T is
augmented accordingly, and the tangent crack stiffness matrix becomes blockdiagonal:
Z D c1 x % x W
X U
Xx D c2 x U (6.76)
T ' pT1 T2 % Tm q ,
Dc ' X
X & & UU
Xx x % D cm UV
Y
However, for m cracks the number of possible combinations of loading and unloading is 2m,
and it may be difficult to find the correct one. Another difficulty is a proper generalization of
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 211
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
the crack initiation criterion. If the standard criterion of maximum principal stress was kept,
the normals of individual cracks could be arbitrarily close to each other. The number of
generated cracks could be theoretically infinite. Then, it was proposed to impose additional
restrictions such as the minimum threshold angle between a new crack and the already
existing ones (Borst and Nauta 1985).+
As an alternative to the multiple fixed crack model, the rotating crack model takes into
account the possible existence of cracks with various orientations by a continuous adjustment
of the orientation of a single crack. The crack normal is assumed to always coincide with the
direction of maximum principal strain. Here, the directions of principal stress and strain
remain aligned and no shear tractions or shear crack strains appear in the formulation. This
simplifies the cracking law because a relation between the normal components of type
Eq. 6.67 is needed. On the other side, it is necessary to take into account that the
transformation matrix T does not remain constant because the crack normal can rotate. For
example, when differentiating Eq. 6.65 we have to write
)!c ' Te
! 8 Te!
c c (6.77)
This complicates the derivation of the tangent stiffness matrix but the resulting formula is
remarkably simple (Baant 1983, Willam et al. 1987). In principal coordinates, we have:
Z(C# + C# c ) 1 xW
D'X e U (6.78)
Y x D# s V
This formula covers the general case, in which cracking can take place in all three principal
planes. Matrix
Z 1 7n 7n W
~ 1X
C e ' X7 n 1 7 n UU (6.79)
E
XY7 n 7n 1 UV
~
is the block of the elastic compliance matrix that corresponds to normal components, and Cc
is the diagonal crack compliance matrix with current compliances of three mutually
orthogonal cracks on the diagonal. If some of the principal directions is not cracking, the
corresponding crack compliance is set to zero. The block of the stiffness matrix (Eq. 6.78)
that corresponds to shear is
Z s2 7 s3 W
X 0 0 U
X 2$e2 7 e3 % U
~ s3 7 s1
Ds ' X 0 0 U (6.80)
X 2$e3 7 e1 % U
X s1 7 s2 UU
X 0 0
XY 2$e1 7 e2 % UV
where the symbols si and ei, i = 1, 2, 3, stand for principal stresses and principal strains. The
tangent stiffness matrix D can be rotated into the global coordinates using the standard
transformation rule.
The shear coefficients in Eq. 6.80 are properly defined only if the principal strains ei,
i = 1, 2, 3, are mutually different. The case of two or three equal principal strains is a weak
point of this particular form of the rotating crack model (Jirsek and Zimmermann 1997). An
alternative formulation that avoids this deficiency is embodied within the Disturbed Stress
Field Model, described by Vecchio (2000, 2001).+
Unlike conventional tensorial models that relate the components of the stress tensor directly to
the components of the strain tensor, microplane models work with stress and strain vectors on
a set of planes of various orientations (socalled microplanes). The basic constitutive laws are
defined on the level of the microplane and must be transformed to the level of the material
point using certain relations between tensorial and vectorial components. The most natural
choice is to construct the stress and strain vector on each microplane by projecting the
corresponding tensors, i.e., by contracting the tensors with the vector normal to the plane.
However, it is impossible to use this procedure for both the stress and the strain and still
satisfy a general law relating the vectorial components on every microplane. The original slip
theory for metals worked with stress vectors as projections of the stress tensor; this is now
called the static constraint. Most versions of the microplane model for concrete and soils have
been based on the kinematic constraint, which defines the strain vector e = (e1, e2, e3) on an
arbitrary microplane with unit normal n = (n1, n2, n3,) as
ei ' ) ij n j (6.81)
where, )ij, i, j=1, 2, 3 are the components of the strain tensor. When dealing with tensorial
components, we use the Einstein summation convention implying summation over twice
repeated subscripts in productlike expressions. For example, subscript j, on the righthand
side of Eq. 6.81 appears both in )ij and in nj, and so a sum over j running from 1 to 3 is
implied.
The microplane stress vector, s = (s1, s2, s3,), is defined as the workconjugate variable of the
microplane strain vector, e. The relationship between e and s is postulated as a microplane
constitutive equation. A formula linking the microplane stress vector to the macroscopic stress
tensor follows from the principle of virtual work, written here as
3
+ ij O) ij '
2L Gy
s i Oei d y (6.82)
are components of the corresponding virtual microplane strain vector. Note that the
summation convention implies summation over i and j on the lefthand side of Eq. 6.82,
summation over i on the righthand side of Eq. 6.82, and summation over j on the righthand
side of Eq. 6.83. Integration in Eq. 6.82 is performed over all microplanes, characterized by
the components of their unit normal vectors, ni, i = 1, 2, 3. Because of symmetry, the
integration domain y is taken as one half of the unit sphere, and the integral is normalized by
the area of the unit hemisphere, 2L/3.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 213
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Substituting Eq. 6.83 into Eq. 6.82 and taking into account the independence of variations O),
we obtain (after certain manipulations restoring symmetry) the following formula for the
evaluation of macroscopic stress components:
3
+ ij '
4L Gy
( si n j 8 s j ni ) dy (6.84)
s'~
s (e, n) (6.85)
Realistic models for concrete that take into account the complex interplay between the
volumetric and deviatoric components of stress and strain (Baant and Prat 1988, Obolt
1995, Baant et al. 1996) usually lead to more general microplane constitutive laws of the
type
that are affected by some components of the macroscopic stress, +, for example by its
volumetric part. Instead of a direct evaluation of the explicit formula Eq. 6.86, the
macroscopic stress is then computed as the solution of an implicit equation, and the stress
evaluation algorithm involves some iteration.
In Section 3.4, we started with a point on the equilibrium path (qo , & p) , as shown in
Figure 3.18, where qo is the current displacement vector, & is a load magnification parameter
and p is a reference vector of applied loads. Equilibrium of the discretized structure was then
written as
or
where f is a vector of internal forces and is a function of the current displacement state. The
solution of the equation system is then obtained from
Kq f = 0 (6.90)
and can be calculated directly for a linear system. For a nonlinear system, however, an
iterative approach is needed with successive solutions of the linear system until convergence
is obtained.
A number of algorithms have been introduced into finite element programs to trace
equilibrium paths. The majority of these use either the NewtonRaphson (NR) or modified
Newton (mNR) techniques as their basis. The NR and mNR algorithms work well when linear
or bilinear material relationships are being used, however, become more and more inefficient
as higher degrees of nonlinearity are introduced. A further disadvantage of the NR and mMR
method is that without the addition of special techniques a falling load path cannot be
handled. An early method of obtaining limit points was to use displacement control as given
by Batoz and Dhatt (1979). This method has been used successfully on many occasions;
however, it has a number of drawbacks, many of which are highlighted in Crisfield (1981). A
further problem when trying to apply this method to finite elements which have a high degree
of material nonlinearity (such as those modelling concrete) is that the initial solution, which
forms the basis of further iterations, may be well away from the final equilibrium state and
divergence and/or algorithm failure may occur.
A major improvement to the standard NR mNR techniques was introduced by Riks (1972)
and Wempner (1971) and later modified by Crisfield (1981) and Ramm (1981) and involves
control of the load/displacement path. The constant arc length method was later modified to
include line searches and accelerations (Crisfield, 1983).
The most frequently used iteration scheme for the solution of nonlinear equations is some
form of the NewtonRaphson procedure. In the case where Eq. 6.88 cannot be solved exactly
the residual forces at the ith iteration can be written as
d df (qi )
(r (qi )) ' 7 ' 7 K (qi ) (6.92)
dq dq
If an approximate solution q ' qi is obtained, then the truncated Taylor expression may be
written as
dr
r (qi 81 ) ' r (qi ) 8 (qi ) O qi ' 0 (6.93)
dq
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 215
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Denoting K $qi % , the tangent stiffness matrix for the current displacement state, as K Ti , then
71
Oqi ' K Ti ( _^ p 7 f ( qi )) (6.96)
The NewtonRaphson solution process is illustrated in Figure 3.18, noting that for every step
the current stiffness matrix is formed and the linearized equations solved for O qi .
From the above calculations it is seen that the formulation of a new tangent stiffness for each
iterative cycle and the solution of a new system of equations must be undertaken. In
computing terms, this can be time consuming and costly. To overcome this difficulty the
approximation K Ti ' K To is often made. This modifies Eq. 6.96 to
71
O qi ' K To ( _^ p 7 f ( qi )) (6.97)
and resolution of the same equation set is repeatedly used. The solution at each iteration is
sped up; however, more iterations to convergence are required, as can be seen in Figure 3.19.
The overall economy of the solution procedure is dependent on the problem size and non
linear behaviour. An updated mNR approach may be adopted where the tangent stiffness
matrix, K Ti , is updated if convergence is not obtained after a predetermined number of
iterations, n, and continues to be updated every n iterations following.
The incremental displacement method was developed to overcome problems with load control
solution schemes and their inability to deal with structural stability problems or problems
where the loading decreases with increasing displacements.
A discrete representation of a linear problem is given by Eq. 6.90 where the internal force
vector is defined by Eq. 6.88. In a usual linear analysis the load vector, ^ p , is given and
Eqs. 6. 88 and 6.90 can then be solved directly for the displacements, q. If, however, the jth
component of q is chosen as unknown, the standard form of Eq. 6.90 needs to be modified to
solve for & and (n1) unknown displacements of q.
Batoz and Dhatt (1979) developed an algorithm such that the structures stiffness matrix, K, is
not modified, instead a solution of two vectors q a and q b is sought such that
K q a ' r uh
v (6.98)
K q b ' p hw
where r is the residual vector defined in Eq. 6.91. By using this method the structural stiffness
matrix remains banded and symmetric.
If the stiffness matrix, K, of Eq. 6.90 is nonsingular, then Eq. 6.98 maybe solved. The
solution for & and q is given by
q ' q a 8 ^ qb (6.99)
q j 7 q aj
^' (6.100b)
q bj
qj
^' (6.101b)
q bj
The algorithm of Batoz and Dhatt can be applied to the NewtonRaphson solution procedure.
Letting the quantities (qo , ^o p) denote the displacement and load state at a point on the
equilibrium path, as shown in Figure 6.5, then instead of increasing the load parameter, the jth
component of q is incremented by _q j . The initial displacement vector, q, is modified so that
qi ' qo where qij ' qij 8 _q j .
The residual force vector due to the modified initial displacements is computed and the
displacement vectors due to the residual force, r, and unit load, p, are computed
simultaneously from Eq. 6.98, that is,
where K $qi % is the nonlinear tangent stiffness matrix corresponding to the current
displacement state, q i .
j
It is desired that the jth element of O qi is equal to zero (that is, O qi ' 0) . This leads to
e b b
O ^i ' 7 c _ q aj q 8 _ q j ` (6.104)
d j ai
The structure displacement vector can now be computed together with the loading parameter
using
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 217
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Figure 6.5: Schematic representation of incremental displacement method using mNR iteration.
_ qi 81 ' _ qi 8 O qi
qi 81 ' qi 8 O qi
If the algorithm is used with the modified NewtonRaphson method, then the structure
stiffness matrix, K $qi % is updated only at selected iterations. A summary of the solution
method is depicted in the schematic outlined in Figure 6.5.
The constant arc length method was developed to overcome some of the problems associated
with relying on merely load or displacement control. The principle behind the arc length
method is to increment the loaddisplacement path in such a manner as to derive the benefits
of both load and displacement control. This was first introduced by Riks (1979) using a
normal to the tangent stiffness and later modified by Crisfield (1981) to use a circular path;
the latter method being slightly less likely to fail.
Let the quantities $qo , ^o p % denote the displacement and current load vector at a given point
on the equilibrium path, as shown in Figure 6.6. The equation governing equilibrium is then
given by Eq. 6.88 and the residual force vector, r, given by Eq. 6.91.
Since the arc length method treats both & and q as variables, an extra equation is necessary to
enable solution of Eqs. 6.88 and 6.91. This extra equation comes from the constraint
relationship
where _' is the prescribed incremental length and _ qi is the incremental displacements after
the (i1)th iteration (see Figure 6.6).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 219
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
The solution procedure starts by incrementing the load parameter by _^ , and forming the
current tangent stiffness pKa q . The initial change in displacement vector, O qo can be
calculated by
H
Oqo ' _ ^1 K a71 . p I (6.107)
Letting HK 71
a . I
p ' # T gives
By now applying the constraint equation 6.106 the arc length can be set as
and remains constant for the remainder of the iterations within this load step.
From Eq. 6.91 the residual force vector, r i , may now be calculated. The solution continues
with the calculation of O q i from
H
O qi ' K a71 . p ri 8 O ^i . p q I (6.110)
Note that the load parameter & must be modified by O ^i in order not to violate the constraint
equation (that is, _^i 81 ' _^i 8 O^i ). Letting HK 71
a . ri I ' # , Eq. 6.110 may be rewritten in
i
the form of
O qi ' # i 8 O ^i #T (6.111)
_ qi 81 ' _ qi 8 Qi O qi (6.112)
where Qi is an acceleration parameter obtained by line searches (if line searches are not used
then Qi ' 1.0) and substituting Eqs. 6.112 and 6. 111 into Eq. 6.106 gives
where
a1 ' Qi #TT #T u
h
h
p
a2 ' 2 #TT _qi 8 Q1#TT #i q h
v (6.115)
h
T T h
a3 ' 2 #i _ qi 8 Qi # i # i h
w
The two roots of Eq. 6.114 correspond to the two points of intersection between the
equilibrium path and the surface described by Eq. 6.113. The appropriate root is the one that
gives forward progression and may be obtained by ensuring an acute angle ) exists between
_ qi and _ qi 81 where
Qi Z T
cos  ' 1 8 # _qi 8 O^i #TT _qi W (6.116)
2 XY i UV
_'
In most cases Eq. 6.116 will yield one positive value of cos ) (corresponding to an acute value
for )) and one negative value and hence there is no problem finding the root that gives
forward progression. In the event that both values for cos ) are positive, the appropriate root
is the one that is closest to the linear solution a3 a2 . Once the value for O^ i has been
obtained, substitution into Eq. 6.111 yields Oqi and subsequently into Eq. 6.112 for O^ i 81 .
As yet application of the arc length method to the modified NewtonRaphson technique has
only been discussed. The constant arc length method can be applied equally well using the
NewtonRaphson technique or any updated mNR method by recalculating the #T vector
whenever the tangent stiffness matrix [Ka] is reformed. The arc length, _' , however, is not
recalculated until the end of the current load/displacement increment.
Irons and Elsawaf (1970) and Elsawaf (1979) first applied line search concepts to nonlinear
finite element problems. Since then many researchers have adopted the line search technique
as a method of accelerating towards a solution. Crisfield (1983), as a method to avoid
numerical difficulties caused by concrete cracking, introduced line search techniques into the
arc length method.
The line search concept seeks a scalar Qi such that the energy z at qi 81 is stationary in the
direction of Qi , that is;
sz sz sq
' ' 7 riT81 #i ' S j ' 0 (6.117)
sQi i 81 sqi i 81 sQi
S j 9 r So (6.118)
where
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 221
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
S o ' 7 #Ti . r i u
h
v (6.119)
S j ' 7 # i . r i 81 hw
T
and * is the required solution tolerance. The work of Crisfield (1982, 1983), Wolfe (1975),
Dixon (1972), Shanno (1978) and Foster (1992) indicate that a slack tolerance of * = 0.8 is
optimum. For a complete mathematical formulation of the line search methods see Crisfield
(1982, 1983).
r i 9 ) max r j
( j ' 1,2,3... i ) (6.120)
T
where r i ' O q i . O q i or r i ' r (q i )T . r (q i ) , as appropriate.
When modelling concrete structures, due to the changing of constitutive relationships from
the uncracked to cracked states and the change in the material strain relationship as a function
of the principal stress ratio, the displacement criteria to set the solution tolerance is often used.
Ideally when using one of the Newton techniques to solve nonlinear problems, the increase in
load or displacement, as the case may be, from a converged state to a new state, should reflect
the current degree of nonlinearity. If the load step is too large then the problem will be slow
to converge. If the load step is too small, then a larger number of steps is required to define
the loaddisplacement state than needed. Crisfield (1983) remarked that, for modelling of
reinforced concrete it may not be wise to relate the load increment size to the number of
iterations required to achieve results at the previous load step. This is due to unpredictable
nature of concrete and the large redistribution of forces when cracking occurs. Substantial
change in the model occurs frequently and the number of iterations required to give
convergence may increase dramatically. This, however, does not necessarily affect subsequent
load steps and no apparent advantage is obtained by reducing the initial load increment (+&o)
for the following load steps. In fact, substantial computer time may be taken to map a small
aberration in a few elements that ultimately do not affect the results of the model. A more
reliable method of automatic load incrementation is to monitor the performance of a key
displacement component within the model and modify the load increments (for load control
methods or the constant arc length method) within a set maximum and minimum in order to
obtain optimum displacement increments.
There are
a many othher compliccating issuees related to specific orr special struuctures not covered
in this report andd it is for the designner to preaassess theirr importancce before adopting
a
computeed results. In
I the case of o columns,, for examp ple, if inform
mation on a postpeak response
r
is desirred issues such
s as spaalling of coover concreete and bucckling of thhe reinforceement in
compresssion membbers needs consideratio
c on. Figure 6.7
6 illustrates typical ffailure in th
he plastic
hinge reegion of a cyclically
c looaded colummn specimen n tested by Bayrak (1999). Since buckling
b
of the longitudinal
l l bars is geenerally noot considereed in conveentional anaalytical pro ocedures,
behavioour at large inelastic cuurvatures is generally overpredict
o ed with resppect to stren
ngth and
ductilityy.
6.4.2 E
Effects of ageing
a and distress in concrete
The vollume changee of the matterial at an early age iss mainly cauused by tem mperature risse due to
hydratioon reaction and autogeenous and drying
d shrin
nkage rootedd in the mooisture behaaviour in
micro pores.
p Here,, the internaal stress gennerated by the volumee change is strongly deependent
on creeep behaviouurs. In addition, for a PC design n, considerration of crreep is esseential to
evaluatee the loss of
o prestress. Here, a coouple of evaaluation meethods for sshrinkage an nd creep
are briefly reviewed.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 223
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
where, )(t) is the total strain, )e(t) the instantaneous strain, )c(t)the creep strain (which is
divided into basic and drying creep), )s(t) is the shrinkage strain and )T(t) is the thermal
dilatation. For stresses up to approximately 40% of the strength of concrete, creep can be
assumed to be proportional to the stress (known as DavisGranville relationship). Based on
this assumption under constant stress +, the above equation can be transformed as,
where, J(t, t) is defined as a compliance function or creep function, which represents the
strain at time t produced by a unit constant stress acting since time t. It can be written as,
1 1 8 C$t , t #%
J $t , t #% ' 8 C $t , t #% ' (6.123)
E $t #% E $t #%
where, E(t)is the elastic modulus of concrete at time t # , C $t , t # % is the specific creep that
represents the creep strain at time t produced by a unit constant stress acting since time t # and
C(t, t # ) is a creep coefficient which represents the ratio of the creep strain to the elastic strain
at time t. These creep functions have been formulated based on concrete properties, structural
dimensions, and atmospheric conditions by many past researches and design codes.
+ c $t #% + $t #%
) c $t , t #% ' T C$t , t #% ' c pC 0 T ( c $t 7 t #%q (6.124)
Ec Ec
where, +c: constant stress applied at time t, Ec: modulus of elasticity at the age of 28 days, C0
(t, t # ): notional creep coefficient and it depends on the ambient relative humidify, size of the
member, and the mean compressive strength at the age of 28 days, and (c (t, t # ): coefficient
describing the development of creep with time after loading and it depends on the ambient
relative humidity and size of member.
) s $t , t s % ' ) so T ( s $t 7 t s % (6.125)
where, )so is the notional shrinkage coefficient and it depends on the ambient relative
humidity, the mean compressive strength at the age of 28 days, and the type of cement, (s is
the coefficient to describe the development of shrinkage with time, which depends on the size
of member, t is the age of concrete and ts is the age of concrete at the beginning of shrinkage
or swelling.
All national codes of practice include simplified creep and shrinkage models in their contents
that can be used in modelling. For example, the Japan Society of Civil Engineers (JSCE)
Code 2002 has the following model for creep:
p H
) #c $t , t #, t 0 % +#cp ' 1 7 exp 7 0.09$t 7 t #%
0.6
IqT )#
cr
(6.127)
) #cr ' ) #bc 8 ) #dc
# is the ultimate value of specific basic creep strain, which is a function of unit
where, ) bc
weight of water and cement and watertocement ratio and ) dc # is the ultimate value of specific
drying creep strain, which is a function of ambient relative humidity, unit weight of water and
cement, watertocement ratio, exposure surface area, and structural dimension.
The above models are generally used to predict a mean value of shrinkage and/or creep over
the cross section of a certain member. For practical purposes, such simple treatments are easy
to use, but their applicability should be clearly kept in mind. That is to say, they dont give
local properties within the cross section of a concrete member, such as the variation of
internal stresses, moisture states, and generation of micro cracking. In addition, these models
adopt the conventional separation between the shrinkage, drying creep, and basic creep in
their formulations. Strictly speaking, however, these behaviours should not be treated
separately but they are rather several aspects of one physical phenomenon.
Other models
Recently, several models based on the microphysical phenomena in concrete, such as cement
hydration, moisture transport/equilibrium, and microstructure of cement paste, have been
proposed (Baant and Prasannan, 1989, Baant et al., 1997, Lokhorst and Breugel, 1997,
Maekawa and Ishida, 2001). Since they try to simulate the actual rheological phenomena in
the material and structure from the mesoscopic viewpoint, it is expected that they are a
breakthrough giving a solution to classic research topics in the concrete engineering, i.e.,
creep and shrinkage. Figure 6.8 shows an example of such an approach for a creep test by
Ross (1958) undergoing a complex loading and unloading history and modelled by Chong
(2004) using the solidification formulation of Baant and Prasannan (1989).
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 225
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Steel corrosion
Acceleration
Spalling
Structural performance
Where second ordder effects are a importaant in assesssing the sttrength or sservice limiit states,
these neeed to be inncluded in thhe modellinng. Many, iff not most, commerciaal packages used for
modelling of concrrete structurres will havve some cap pability for modelling geometricaally non
linear problems
p toogether with the nonlinear mateerial modellling descriibed in the various
chapterss of this repport. An exxample of a common sttructural eleement wherre designerss will be
familiarr with the issue
i of seccond order effects in their
t treatm
ment of slennder columnns where
theeffecct of geometric nonlinnearity is to magnify beending mom ments. A seccond, less common,
c
examplee is in creeep buckling of a beam mcolumn, su uch as thatt shown in Figure 6.10 0, where
geometrric nonlineearity combbines with time effeccts and othher material nonlineaarities to
engendeer failure.
Steel ch
hannel section
eT Eccentric loadiing
Strong wall Ten
nsioning cable
1250
Dial gaugee
Stirrups 10 at 15
50
Clearr cover 15 mm
2N12
1250
Dial gaugee
5000
2N12
1250
A A 150
Dial gaugee
Test co
olumn Section AA
1250
Hydraaulic jack
on loading arm
Isectio
eB
Load cell
6.5 R
Reference
es
Batoz J.,
J and Dhattt G. (1979)), Incremenntal displaceement algorrithms for nnonlinear prroblems,
Int. J. Num.
N Meth. Engng., Vool 14, pp. 122621266.
Baant Z. P. (19883), Comm ment on orthotropic models forr concrete and geomaaterials,
Journal of Engineeering Mechaanics, ASCE
E, 109, pp.8
849865.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 227
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Baant Z. P. and Prat P. (1988), Microplane model for brittle plastic materials. I: Theory, II:
Verification, Journal of Engineering Mechanics, ASCE, 114, pp.16721702.
Baant Z.P. and Prasannan S. (1989), Solidification theory for concrete creep. I.
Formulation, II. Verification and application, Journal of Engineering Mechanics, 115(8),
pp.16911725.
Baant Z. P., Xiang Y., and Prat P. C. (1996), Microplane model for concrete. I: Stressstrain
boundaries and finite strain, Journal of Engineering Mechanics, ASCE, 122, pp.245254.
Baant, Z. P., Hauggaard, A. B., Baweja, S., and Ulm, F.J. (1997). Microprestress
solidification theory for concrete creep. (I: Aging and drying effects, II: Algorithm and
verification). Journal of Engineering Mechanics, ASCE, Vol. 123, No. 11, pp. 11881201.
Borst R. and Nauta P. (1985), Nonorthogonal cracks in a smeared finite element model,
Engineering Computations, 2, pp.3546.
Carol I., Prat P., and Baant Z. P. (1992), New explicit microplane model for concrete:
Theoretical aspects and numerical implementation, International Journal of Solid and
Structures, 29, pp.11731191.
CEBFIP Model Code 1990 (1993), Thomas Telford Services, Ltd., London, for Comit
EuroInternational du Bton, Bulletin dInnformation No. 213214, Lausanne, pp.437.
Cedolin L. and Poli S. Dei (1977), Finite element studies of shearcritical R/C beams,
Journal of Engineering Mechanics, ASCE, 103(3), pp.395410.
Chong, K.T. (2004). Numerical Modelling and Time Dependant Cracking and Deformation of
Reinforced Concrete Structures, PhD Thesis, The University of New South Wales, School of
Civil and Environmental Engineering, Australia,
Chong, K.T., Foster, S.J., and Gilbert, R.I. (2008). Timedependent modelling of RC
structures using the cracked membrane model and solidification theory, Computers &
Structures, 86, pp.13051317.
Crisfield, M.A. (1981), A fast incremental/iterative solution procedure that handles snap
through, Comput. Struct., Vol 13, pp.5563.
Crisfield, M.A. (1982), Accelerated solution techniques and concrete cracking, Comp. Meth.
Appli. Mech. Engng. Vol 33, pp.585607.
Crisfield, M.A. (1983), An arc length method including line searches and accelerations, Int. J.
Num Meth. Engng. Vol 19, pp.12691289.
Dixon L.C.W. (1972), Nonlinear optimization, The English University Press Ltd, London.
Elsawaf, A. (1979), The conjugateNewton method for nonlinear finite element problems.
PhD. Thesis, University of Calgary, Canada.
Foster S.J. (1992), An application of the arc length method involving concrete cracking, Int.
J. Num. Methods Eng., Vol. 33, No. 2, pp. 269285.
Irons, B.M., and Elsawaf, A. (1970), The conjugateNewton algorithm for solving finite
element equations, in formulations and algorithms in finite element analysis, (Eds. K.J.
Bathe, J.T. Oden and W. Wunderlich) MIT Cambridge, MA, pp. 656672.
Jirsek M. and Zimmermann T. (1997), Rotating crack model with transition to scalar
damage: I. Local formulation, II. Nonlocal formulation and adaptivity, LSC Internal Report
97/01, Swiss Federal Institute of Technology, Lausanne, Switzerland.
Lvy M. (1870), Mmoire sur des quations gnrales des mouvements intrieures des corps
solids ductiles au del des limites ou llasticit pourrait les ramener leur premier tat, C.
R. Acad. Arts Sci., 70, pp.1323.
Lokhorst SJ and Breugel K. (1997), Simulation of the effect of geometrical changes of the
microstructure on the deformational behaviour of hardening concrete, Cement and Concrete
Research, 27(10), pp.14651479.
Mazars J. (1985), A model of a unilateral elastic damageable material and its application to+
concrete , In Fracture toughness and fracture energy of concrete, Elsevier, New York.
Mentrey Ph. and Willam K. J. (1995), A triaxial failure criterion for concrete and its
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 229
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Ortiz M. (1985), A constitutive theory for the inelastic behavior of concrete, Mechanics of
Materials, 4(1), pp.6793.
Papa E. and Taliercio A. (1996), Anisotropic damage model for the multiaxial static and
fatigue behaviour of plain concrete, Engineering Fracture Mechanics, 55(2), pp.163179.
Ramm E. (1981), Strategies for tracing nonlinear response near limit points, in
formulations and algorithms in finite element analysis, (Eds. K.J. Bathe, J.T. Oden and W.
Wunderlich) MIT Cambridge, MA, pp. 656672.
Riks E. (1972), The application of Newtons method to the problem of elastic stability, J.
Appl. Mech., Vol. 39, pp.10601066.
Ross, A.D. (1958), Creep of Concrete under Variable Stress, ACI Proceedings, 54(9),
pp. 739758.
Shanno D.F. (1978) , Conjugate gradient methods with inexact searches, Math O.R., Vol. 13,
pp. 71113.
Simo J. C. and Wu J. W. (1987), Strain and stress based continuum damage models: Part I
Formulation, International Journal of Solids and Structures, 23(7), pp.821840.
Tresca. (1868), Mmoire sur lcoulement des corps solides, Mmoires prsents par divers
savants, 18, pp.733799.
Vecchio, F.J. (2000), Disturbed Stress Field Model for Reinforced Concrete: Formulation,
ASCE J. Structural Engineering., Vol. 126, No. 9, pp. 10701077.
Vecchio, F.J. (2001), Disturbed Stress Field Model for Reinforced Concrete:
Implementation, ASCE J. Structural Engineering., Vol. 127, No. 1, pp. 1220.
Yazdani S. and Schreyer H. L. (1988), An anisotropic damage model with dilation for
concrete, Mechanics of Materials, 7(3), pp.231244.
Weihe S. (1995), Modelle der fktiven Ribildung zur Berechnung der Initiierung und
Ausbreitung von Rissen, PhD thesis, Universitt Stuttgart.
Wempner, G.A. (1971), Discreet approximations related to nonlinear theories of solids, Int.
J. Solids Struct., Vol. 7, pp. 15811599
Willam K. J. and Warnke E. P. (1974), Constitutive model for the triaxial behavior of
concrete, Concrete Structures Subjected to Triaxial Stresses, volume 19 of IABSE Report,
pp.130, International Association of Bridge and Structural Engineers, Zurich.
Willam K., Pramono E., and Sture S. (1987), Fundamental issues of smeared crack models,
In S. P. Shah and S. E. Schwartz, editors, Fracture of Concrete and Rock, pp.192207, Society
for Experimental Mechanics, Bethel, Connecticut.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 231
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
[Blank Page]
!
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
NLFEA tools can be used for the design of a structure as well as for conducting a detailed
investigation of a particular aspect of its behaviour such as the maximum stress in
reinforcement, crack orientation and width, or energy absorbing capacity. For the purpose of
general design and analysis, it is necessary that the practitioner evaluate the programs ability
to predict the strength and behaviour of a wellselected group of experiments (benchmark
tests) in order to determine a global safety (or strength reduction) factor to use in his or her
design. The predicted capacity of a structure is then multiplied by this global safety factor to
obtain a safe design capacity. Thus, a higher global safety factor is appropriate when the user
has reliable material data for setting material characteristic values and when the program
provides accurate predictions for the behaviour of physical experiments that capture the range
in behaviours that are expected in the structure being modelled. The more comprehensive the
validation activity, the greater confidence the designer can have in any calculated global
safety factor. In the absence of less reliable material data, or when the NLFEA program does
not provide a good estimate of the capacity of representative physical test specimens, or when
an extensive validation procedure is not conducted, the results of the analysis should be
penalized by the use of a lower global safety factor.
Chapter 7 provides guidance for a designer who is trying to assess the appropriate global
safety factor to apply to the results from a particular finite element prediction as well as for
the user who is trying to investigate a particular aspect of the behaviour of a concrete
structure. The most important segment of this chapter is in Section 7.2 which describes the
steps in the calibration and validation of NLFEA programs. Section 7.3 discusses the factors
to consider in the determination of a global safety factor. Section 7.4 addresses several
important issues in the selection, use, and validation of NLFEA programs. The remainder of
the document is devoted to the presenting case study examples. Section 7.5 presents an
example of the design of a squat concrete shear wall with openings in which the global safety
factor is first determined through a comparison of experimental test data with the predictions
of a specific NLFEA program. In the second case study (Section 7.6), a comparison is made
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 233
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
between two different NLFEA programs for investigating the detailed behaviour of a deep
beam. One of these programs is a large commercial program in which several behavioural
models are available for the user to select between. The use of such a program requires
additional care by the user as is illustrated in this example. This chapter concludes in
Section 7.7 with a discussion of future trends in model validation.
The steps and flow of information in the calibration and validation of a NLFEA program is
shown in Figure 7.1. The case of the design of a shear wall using a 2Dimensional continuum
finite element analysis program is used in this example in order to provide more specific
details for the types of information that need be exchanged at each step.
Modelling
*Any Calibration based on Element or Structural Level Experiments Must be Completed with Caution to Avoid
Tuning a Model to Fit One Specific Class of Experiments; Importance of Validating Using a
Comprehensive Series of Test Data that Captures the Range of Critical Behavior Characteristics cannot be
Overemphasized
Figure 7.1: Overview of procedure for model calibration and validation (after Okamura and Maekawa, 1991).
In the use of all NLFEA programs, it is necessary to specify the mechanical characteristics of
materials. This is likely to include initial elastic stiffness as well as peak stress and strain
values, concrete shrinkage strain, thermal expansion coefficients and other standard material
properties. The more accurately the user can specify the required information, the more
reliable will be the results of any analysis. In selecting material values, the user should use
mean (average) material properties and recognize that the material properties in structural
concrete may differ from those in material test specimens. For example, due to additional
shrinkage in thin members and the restraint due to the presence of reinforcement it is common
for structural concrete to crack under stresses that are considerably below what is obtained
from material test data. It is also important to recognize that there can be a wide variation in
some material properties and the reliability of analysis results is reduced if assumed values are
used. To illustrate this latter point, Figure 7.2 plots the tensile cracking stress (fct) as a
function of cylinder compressive strength (fcm). This wide variation illustrates the importance
of using test data and not simply relying on typically assumed relationships, such as the one
shown in Eq. 7.1, if the results of the analysis are expected to be sensitive to this value.
Figure 7.2: Variation in tensile splitting strength as a function of cylinder compressive strength.
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 235
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
Since a NLFEA program does not provide the exact capacity of a structure, it is necessary to
begin the use of any investigation with an evaluation of the program through a comparison of
predicted and measured response of test structures that have been loadtested to failure. By
examining the ratio of measured to predicted capacity of a well selected set of experiments,
the user can determine a global safety factor to apply to the results of an analysis. This global
safety factor, which will be labelled ', will be a fraction less than or equal to 1.0. The safe
capacity predicted by the program for the design or analysis being conducted will then be the
capacity predicted by the NLFEA program multiplied by this global safety factor, '. The
global safety factor may be considered to be the end product of the model validation process.
While Level 1 was purely a calibration activity in which material parameter values were set,
Level 2 it not always a pure validation exercise. This is because the experimental test data that
is most suitable for model evaluation is often the most suitable test data for setting structural
level parameters that account for such effects as compression softening, tension stiffening,
shear retention, loss in bond, crack spacing, etc. The tests that can serve this dual purpose are
simple tests where there is little chance for misinterpretation of the test data. These tests may
be referred to as fundamental or elementlevel tests. When evaluating a tool, or for that matter
calibrating structurallevel parameters, the user needs to predict the behaviour of a
systematically arranged set of elementlevel benchmark tests that encompass the range in
possible behaviours that may be experienced by the structure being designed or studied.
Consider the following example to illustrate the evaluation of NLFEA programs at the
element level. The construction of the Panel Element Tester at the University of Toronto
(UofT) in 1979 allowed uniform tests on reinforced concrete shear panels to be performed for
the first time. Based on the results of an IABSE colloquium in June of 1981, it was decided to
hold an international competition (Collins, 1985) to test the quality of various numerical
predictions for a collection of four panel elements (AD). Panel D is shown in Figure 7.3.
The tests were performed in August of 1981 and while the measured material properties were
made available, the actual experimental results were kept confidential to ensure that the
results of the contest would be true predictions. A total of 27 entries were received from 13
countries. The tests had been selected such that they would be difficult to predict; none of the
elements failing in biaxial shear. It was found that the predictions varied substantially from
the observed behaviour. Figure 7.4 presents the details of the panels. The best entry was able
to predict all 4 panels to within 17% of the measured capacity. However, other predictions
were in error by as much as a factor of 3 in the measured strength, as illustrated in Figure 7.5.
What made the UofT competition so valuable is that each of the four selected panels evaluates
an important aspect of the biaxial response of structural concrete. Panel A was selected so that
the accuracy of the shear transfer relationship and the compression softening model would
have little effect on the shear response of the test specimen. In Panel B, the softening of the
compression model dominated the response. In Panel C, it was most important to accurately
model both tension stiffening and shear transfer across cracks. In Panel D, it was important to
accurately model all aspects of behaviour in order to make a good prediction of behaviour.
When reviewing the accuracy of the 27 entries it was clear that using only 2 or 3 of the panel
tests would have led to an inaccurate assessment of the capabilities of the numerical tools;
thereby illustrating the importance of the selection of a systematically arranged set of
validation experiments (benchmark tests).
Since 1982 there has been significant advancements in the development of NLFEA programs
for structural concrete. Figure 7.6 provides a comparison of the predicted versus measured
shear behaviour of the four panels using a more recent NLFEA program and labelled as
NLFEAP1.
Y
t t
PANEL A PANEL B
t t + x ' 70.7,
sx
+ y ' 70.7,
0 9 ( 9( u 0 9 ( 9( u
X sy
f c' ' 20.5 MPa $ 0 ' 0.00190 f c' ' 20.5 MPa $ 0 ' 0.00190
f xy ' 442 MPa * x ' 0.01785 f xy ' 442 MPa * x ' 0.01785
f yy ' 442 MPa * y ' 0.01785 f yy ' 442 MPa * y ' 0.01785
Y
t t
PANEL C PANEL D
t t + x ' 7(, 7 3.9) 9 0
sx
+ y ' 7(, 7 3.9) 9 0
0 9 ( 9( u 0 9 ( 9( u
X sy
f c' ' 19.0 MPa $ 0 ' 0.00215 f c' ' 21.7 MPa $ 0 ' 0.00180
f xy ' 458 MPa * x ' 0.01785 f xy ' 441 MPa * x ' 0.01785
f yy ' 299 MPa * y ' 0.00713 f yy ' 324 MPa * y ' 0.00885
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 237
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
12 , 8
Number of Predictions
Number of Predictions
7 ,
10 ,
6 ,
8 5 +x
6 *x5'5*y 4
3
4 +y
2
2 *x5'5*y
1
0 0
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7
Predicted/Observed Predicted/Observed
6
9
,
Number of Predicitons
Number of Predictions
8 5 ,
7 ,
, +x
6 4
5 3
4 *x5'52.55*y
3 2 +y
2 *x5'5 25*y
1
1
0 0
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8
Predicted/Observed Predicted/Observed
10
ObservedA
AnalysisA
ObservedB
8
AnalysisB
ObservedC
Shear Stress(MPa)
AnalysisC
6 ObservedD
AnalysisD
In order to make a thorough evaluation of a particular NLFEA program and thereby determine
the global safety factor, it is necessary to make detailed comparisons with carefully selected
test data of both element level tests (discussed in Section 7.2.4) and tests of complex
structures. These complex or structurallevel tests can account for the effects of complex
geometries, loadings, and edge effects. In evaluating a model with structurallevel tests, many
types of comparisons can be made. The most common and often important comparisons are
initial stiffness, capacity, deformation at peak load, and deformation at failure. The overall
loaddisplacement response captures all of these measures. It is also possible to compare a
wide range in other performance measures including the development, propagation, width and
orientation of cracks, the strains in reinforcement, and the distribution of compressive strain in
concrete. These other comparisons are determined by the specific needs of a particular
investigation.
An example is now used to illustrate how global factors of safety are selected in a codeof
practice; the equivalent approach can be used for determining ' for any particular NLFEA
program. Figure 7.7 plots the ratio of the experimentally measured versus ACI31802 code
calculated shear strength of reinforced concrete beams with shear reinforcement, where the
code calculated nominal shear capacity (Vn) is determined by
fc# Av f y d
Vn ' Vc 8 Vs ' bwd 8 (7.2)
6 s
where bw is the width of the web, d is the member depth, Av is the area of shear reinforcement,
and s is the spacing of the transverse reinforcement (units are in N, mm, and MPa).
3.0
2.5
2.0
Vtest / VACI
1.5
1.0
0.5
0.0
0 20 40 60 80 100 120 140
f'c (MPa)
!
fib Bulletin 45: Practitioners guide to finite element modelling of reinforced concrete structures 239
Copyright fib, all rights reserved. This PDF copy of fib Bulletin 45 is intended for use and/or distribution only within National Member Groups of fib.
As shown in Figure 7.7, Eq. 7.2 does not provide a very accurate estimate of the measured test
strength nor does it always produce a conservative result. However, with the use of a suitable
value of C, then this somewhat inaccurate equation can be safely used for design with the safe
design strength being taken as CVn.
In the development of codes of practice, the global safety factor is selected so that there is a
small probability that the actual capacity is less than the design strength (C o nominal
capacity). Based on an assumed normal distribution, there is a direct relationship between the
mean (,), the standard deviation of the strength ratio (s), the strength reduction factor C and
the percentage of the design strengths (or the Fractile level, F) that are expected to be less
than C o nominal capacity. This relationship between these values is given by
eC 7u b
F ' zc ` (7.3)
d s a
C ' z 71 ( F ) T s 8 u (7.4)
To illustrate this relationship, consider the data described in Figure 7.7 for which the mean is
, = 1.35 and the standard deviation is s = 0.37. The statistical values in Table 7.1 can be
calculated using Eqs. 7.3 and 7.4. If no strength reduction factor is used (that is C = 1.0), then
in 17.2% of the cases it would be expected that the actual strength is less than the design
strength. When a strength reduction of C = 0.75 is applied, which is the current value for shear
used in ACI318, then for only 5.24% of the cases is the actual strength expected to be less
than the design strength (see Figure 7.8).
Table 7.1: Example of relationships between global safety factor and fractile level.
B = 1.35
s = 0.37
COV = 0.28
Fractile level
F
In many cases, the practitioner has limited choice in the NLFEA program that is available to
apply to a particular problem due to financial, time, or technical constraints. In other cases,
the practitioner has some time to evaluate and select the most suitable program for their short
or longterm needs. There are many factors that influence which model is most appropriate to
use in an analysis, ranging from technical to practical issues, as discussed below.
Capability: Are the analytical models (constitutive models and failure theories) incorporated
in the NLFEA program capable of capturing the types of behaviour most relevant to the
design challenges (or problems) to be studied? Some of the behaviours that are difficult and
infrequently modelled well in structural concrete include the benefit of passive confinement,
damage accumulation due to cyclic loading, dowel action across crack interfaces, fracture in
plain concrete regions, interface shear transfer and compression softening.