Professional Documents
Culture Documents
Doctor of Philosophy
by
December 1990
To my parents,
for their care and support.
ABSTRACT
This work is concerned with the development of efficient and accurate methods for the
nonlinear analysis of steel frames, where the effects of geometric and material
nonlinearities are considered.
Distinction is made between the requirements of large displacement analysis in plane and
space frames, and an incremental approach is adopted for 3D analysis to avoid problems
associated with finite rotations.
To address the efficient nonlinear analysis of elastic frames, a new finite element quartic
formulation is derived in a convected local system. Verification examples demonstrate the
ability of the quartic formulation to model elastic beam-columns with initial imperfections
using only one element per member.
Two approaches are employed for the elasto-plastic analysis of steel frames. The first is
an approximate approach based on plastic-hinge idealization, while the second accounts
for the spread of plasticity across the section depth and along the member length through
the use of elasto-plastic cubic elements. Both approaches are considered in the context of
an automatic mesh refinement process, where the analysis is always started with one
quartic element per member, and elasto-plastic elements are inserted where and when
necessary during the analysis.
Non-structural mass and damping formulations are developed and used in conjunction
with Newmark's time integration scheme to provide an efficient nonlinear dynamic
analysis capability. Consideration is given to the modelling of distributed mass as well as
to damping representation based on Rayleigh's approach.
The above formulations and procedures are implemented in a new nonlinear static and
dynamic analysis program named ADAPTIC. The main features and analysis capabilities
of ADAP1'IC are discussed, and verification examples are performed to demonstrate the
extreme efficiency of the developed code in comparison with industry-standard packages.
3
ACKNOWLEDGEMENTS
I would like to thank my supervisor Dr. Amr Elnashai for his careful guidance and
continuous encouragement throughout the duration of this study. A great deal of what has
been achieved is due to his invaluable suggestions and clear sense of direction.
This work was initiated by an industrial research project addressing the behaviour of
offshore structures. I owe the opportunity for working on this project to Professor
Patrick Dowling, the Head of Department, whose global supervision and unfailing
support have been a tremendous asset.
I wish to thank the Steel Construction Institute for funding part of this research, and for
providing me with office space and computing facilities.
Thanks is also due to the University of London for partially funding this work through
the Edmund Davis Scholarship. The extension of the initial one-year duration of the
scholarship to two years is also gratefully acknowledged.
Finally, I would like to thank Mr. Nick James in the Computer Centre for providing
access to powerful computing hardware, which was used for running the last verification
example of this thesis.
4
TABLE OF CONTENTS
Page No.
Abstract 3
Acknowledgements 4
Table of Contents 5
List of Figures 11
Notation 12
Chapter 1: Introduction 19
1.1 Preamble 19
1.1.1 Sources of nonlinearity in structural behaviour 19
1.1.2 Role of experimentation 20
1.1.3 Implications on the design process 21
1.1.4 Importance of nonlinear analysis 21
1.2 Nonlinear Analysis Methods 22
1.2.1 Solution of differential equations 22
1.2.2 Finite element method 23
1.2.3 Finite segment method 24
1.2.4 Physical models 25
1.2.5 Phenomenological models 25
1.3 Scope and Organization of the Present Work 26
5
2.3 Quartic Beam-Column Formulation 46
2.3.1 Kinematics 46
2.3.2 Generalized strains and stresses 48
2.3.3 Chord forces 51
2.3.4 Chord tangent stiffness 53
2.3.5 Global analysis 54
2.4 Verification 55
2.4.1 Quartic formulation beam-column properties 55
2.4.2 Square diamond frame 57
2.4.3 Buckling of imperfect truss 57
2.4.4 Lateral torsional buckling 58
2.4.5 Circular bend 58
Chapter 3: Material Nonlinearities 77
3.1 Analysis Methods 77
3.2 Plastic Hinge Quartic Formulation 80
3.2.1 Plastic hinge properties 80
3.2.2 Chord forces 82
3.2.3 Chord tangent stiffness 90
3.2.4 Global analysis 91
3.3 Elasto-Plastic Cubic Formulation 93
3.3.1 Kinematics 93
3.3.2 Generalized strains 94
3.3.3 Gauss integration 95
3.3.4 Material stress-strain law 96
3.3.5 Generalized stresses 97
3.3.6 Chord forces 97
3.3.7 Chord tangent stiffness 99
3.3.8 Global analysis 100
3.4 Automatic Mesh Refinement 100
3.4.1 Plastic hinge approach 101
3.4.2 Distributed plasticity approach 103
3.5 Verification 104
3.5.1 Modelling of beam-columns 105
3.5.2 Two-bay frame 106
3.5.3 Jacket structure 107
6
Chapter 4: Material Models for Mild Steel 122
4.1 Material Behaviour 122
4.2 Material Models 123
4.2.1 Bilinear model 125
4.2.2 Multisurface model 126
4.3 Verification 130
4.3.1 Properties of the multisurface model 130
4.3.2 Fixed ended beam-column 131
4.3.3 Sway frame 131
Chapter 5: Compliant Supports and Joint Flexibilities 145
5.1 Plane Frames 145
5.1.1 Localdisplacements 145
5.1.2 Globalforces 146
5.1.3 Global tangent stiffness 147
5.2 Space Frames 148
5.2.1 Local displacements 148
5.2.2 Global forces 151
5.2.3 Global tangent stiffness 151
5.3 Formulation Characteristics 151
5.3.1 Force-displacement relationships 152
5.3.2 Global analysis 152
5.4 Verification 153
5.4.1 Frame with flexible joints 153
5.4.2 Space truss 154
Chapter 6: Nonlinear Dynamic Analysis 165
6.1 Analysis Methods 165
6.2 Approach Used in the Current Work 168
6.2.1 Contribution of inertia 169
6.2.2 Contribution of damping 170
6.3 Dynamic Elements 172
6.3.1 Lumped mass element 172
6.3.2 Cubic mass element 172
6.3.3 Dashpot damping element 176
6.3.4 Rayleigh damping element 176
7
6.4 General Considerations 179
6.4.1 Global analysis 180
6.4.2 Distributed mass modelling 180
6.5 Verification 181
6.5.1 Elastic cantilever 181
6.5.2 Clamped beam 182
6.5.3 Elastic frame with Rayleigh damping 182
6.5.4 Jacket structure 183
Chapter 7: 'ADAPTIC' - A Nonlinear Analysis Program 196
7.1 Proportional Loading (Static Analysis) 196
7.1.1 Loadconirol 197
7.1.2 Displacement control 198
7.1.3 Automatic displacement control 199
7.2 Time History Analysis (Static/Dynamic) 201
7.2.1 Static vs dynamic analysis 202
7.3 Solution Procedure 205
7.3.1 Iterative strategy 205
7.3.2 Assembly and reduction (Frontal method) 209
7.4 Loading and Support Conditions 212
7.5 Program Structure 213
7.5.1 Datafile 213
752 'READ' 217
7.5.3 'ANALYSE' 217
7.5.4 Output files 218
7.6 Examples 218
7.6.1 Lee's frame 218
7.6.2 Elastic arch 220
7.6.3 Four-storey frame 220
7.6.4 K-frame 220
7.6.5 Elastic dome 221
7.6.6 Triangular frame 222
7.6.7 Frame with rigid/flexible joints 223
7.6.8 Three-dimensional jacket structure 224
8
Chapter 8: Summary and Conclusions 249
8.1 The Developed Formulations 249
8.1.1 Quartic formulation 250
8.1.2 Cubic formulation 250
8.1.3 Joint formulation 250
8.1.4 Mass and damping formulations 250
8.2 Automatic Sub-division 251
8.3 ADAPTIC 252
8.4 Future Research 253
References 255
Appendix B: Data and Output Files 275
B.1 Data Structure 275
B.1.1 Typeofanalysis 275
B. 1.2 Material properties 275
B.1.3 Sectionproperties 275
B. 1.4 Element groups (2D) 276
B. 1.5 Element groups (3D) 277
B.1.6 Nodal coordinates 279
B. 1.7 Element connectivity (2D) 279
B. 1.8 Element connectivity (3D) 280
B.1.9 Imperfections (2D) 281
B.1.10 Imperfections (3D) 282
B.l.11 Restrained freedoms 283
B.1.12 Loading 283
B. 1.13 Solution procedure and output 286
B.2 Output Files 286
B.2.1 Pilename.OUT' 286
B.2.2 Filename.NUM' 287
9
B.3 Lee's Frame 288
B.3. 1 Static (3 elements) 288
B.3.2 Static (5 elements) 293
B.3.3 Dynamic (Force incrementation) 295
B. 3.4 Dynamic (Displacement incrementation) 298
B.4 Elastic Arch 299
B.4. 1 Two quartic elements 299
B.4.2 Four quartic elements 301
B.5 K-Frame 302
B.5.1 Plastic hinge approach 302
B.5.2 Distributed plasticity approach 307
B .6 Elastic Dome 309
B.6. 1 Dynamic analysis (=5%) 309
B.7 Frame with Rigid/Flexible Joints 312
B.7.1 Flexible case 312
10
LIST OF FIGURES
Page No(s).
Chapter 2: Figures 2.1-2.19 60-76
Chapter 3: Figures 3.1-3.18 109-121
Chapter 4: Figures 4.1-4.12 132-144
Chapter 5: Figures 5.1-5.8 155-164
Chapter 6: Figures 6.1-6.7 185-195
Chapter 7: Figures 7.1-7.18 226-248
11
NOTATION
General Rules
- All notations are at least defined where they first appear, and a summary of operators
and symbols is given in the following sections. Notations from other work (Chapter 4:
section 4.2.2) do not necessarily obey the rules hereafter, and are hence defined only
- Generic symbols of matrices and vectors are represented by bold font-type with left
side subscripts or superscripts (e.g. G , a U). This rule also applies to three-
dimensional matrices.
- Subscripts and superscripts to the right side of the generic symbol indicate the term of
- Equations are identified by numbers located on the right-most margin and composed of
up to three entries (e.g. 3.22.a) with the first entry indicating the Chapter in which the
equation appears.
Operators
a : partial differentiation.
12
Symbols
a vectorin 3D space.
A cross-sectional area.
13
2c vector of direction cosines of local z-axis at end(2) of 3D element
z
14
Fp plastic axial force capacity.
I identity matrix.
condensation.
M reduced plastic moment capacity in the local y direction due to axial force.
reduced plastic moment capacity in the local z direction due to axial force.
15
H local mass matrix of cubic mass element.
C
T section torque.
I : matrix for transformation of global to local displacements.
matrix for transformation of generalized stresses to chord forces of cubic
element.
16
U element global displacements
g
(u 1 ,v 1 . a 1 , u 2 ,v 2 , a 2). cumulative for2D analysis
(u 1 . v 1 ,w 1 , a 1 , u2,v2, w2,a2, 2' 2)' incremental for 3D
analysis.
17
Newmark's time integration parameters.
current time-step.
£ material strain.
rate of twist.
icy curvature in the local y direction.
curvature in the local z direction.
1cz
load factor.
p. : strain-hardening parameter.
function of (cb) used for plastic interaction between the biaxial moments.
18
CHAPTER 1
INTRODUCTION
1.1 Preamble
Over the past two decades, the study of the nonlinear behaviour of structures has been the
focus of numerous research programmes, primarily due to its implications on design
economy and structural reliability. The rapid advancement of computers, in terms of
accessibility and performance, provided a much needed environment for the
implementation and application of nonlinear structural analysis tools. This also motivated
the development of advanced analysis methods capable of modeffing the various aspects
of nonlinear structural behaviour to a high level of accuracy. However, despite the
availability of such methods, their considerable demand on computing resources has
discouraged their wide application in preference to other simplified methods, in which
accuracy is compromised for the sake of efficiency.
The present work was motivated by the need for efficient computer-based tools for
nonlinear dynamic analysis of steel frames, which also maintain a high level of accuracy
in the response prediction. As a result, new formulations have been developed, and
traditional nonlinear analysis methods have been adapted to new procedures designed to
make a more efficient use of computing resources.
Nonlinear behaviour refers to the condition of a structure where the response to realistic
loading cannot be linearly extrapolated from the response to small loads, without
incurring significant inaccuracies. In general, engineering structures exhibit nonlinear
behaviour, though to varying degrees, mainly due to material and geometric effects.
Material nonlinearities can form a major contribution to the global structural nonlinear
behaviour, depending on the level of loading and the type of material under consideration.
Mild steel, for example, exhibits an almost linear response up to the point of yield, and
hence its nonlinear effects are detected under relatively high loads. Whereas materials
such as concrete and high strength steel are inherently nonlinear even under small strains,
and therefore induce nonlinearities on the structural level at low loads.
19
On the other hand, geometric nonlinearities are associated with the effect of a change in
geometry due to load application on the overall structural response. In general, geometric
nonlinearities are caused either by large displacements, or by structural instability due to
factors such as the beam-column action and flexure-torsion coupling.
In the context of steel frames, other sources of nonlinearities in the structural behaviour
can be identified. These include the effects of semi-rigid connections, residual stresses,
and initial imperfections (Birnstiel & Iffland, 1980; Chen, 1980; Jones et al, 1982;
Ackroyd & Gerstie, 1983).
Experimental work has always been essential for the understanding of the nonlinear
behaviour of structures, as well as the verification of developed nonlinear analysis tools
and new design concepts. Over the past few years, significant research efforts were
devoted to the study of nonlinearities in steel frames through experimental work on the
levels of the overall structure, the individual components, and the constituent material.
On the structural level, full scale experiments were restricted to simple structural
configurations. Wakabayashi et al (1974) performed four full scale experiments on braced
and unbraced one bay steel frames, where the effects of cyclic horizontal loading were
studied. Popov et al (1980) investigated the inelastic cyclic behaviour of two one-sixth
planar models of tubular towers, and compared the behaviour of braces as part of the
structure with experiments on individual braces.
On the level of individual components, considerable efforts were dedicated to the study of
the hysteretic behaviour of bracing members (Higginbotham & Hanson, 1976; Jam et al,
1978; Popov & Black, 1981). Nakashima et al (1983) studied the behaviour of beam-
columns, where the post-buckling instability in bending and lateral torsion were
considered.
On the level of the constituent material, experimental results were used to develop uniaxial
and multiaxial material models in which the effects of cyclic loading were included.
Popov and Petersson (1978) performed cyclic experiments on mild steel to verify a
multisurface model accounting for triaxial stress states. Cofie and Krawinkler (1985)
described uniaxial experimental results which showed the phenomena exhibited in the
cyclic response of mild steel, namely hardening, softening, and mean-stress relaxation.
20
1.13 Implications on the design process
Most current codes of practice for static design adopt a philosophy by which the limit
state is checked on the level of the individual components instead of that of the overall
structure. This is often achieved through determining the distribution of forces within the
structure using simple analysis methods, such as linear elastic analysis or plastic hinge
analysis, with the latter mainly employed for steel frames. The limitation of such a
philosophy is that the effect of nonlinear behaviour on the distribution of forces within the
structure is not accurately accounted for, especially if significant geometric and material
nonhinearities are involved prior to the ultimate state. Furthermore, such an approach does
not allow for redistribution of forces after the failure of individual members, and hence
does not consider the structure reserve strength.
In the case of seismic design, the effects of nonlinearities are accounted for through
scaling down the elastic design forces by a behaviour factor reflecting the ability of the
system to undergo post-yield deformation without significant loss of strength. Such a
behaviour factor depends to a large extent on the shape and layout of the structure, and
hence its accurate determination requires the understanding of the nonlinear dynamic
behaviour of various structural configurations.
The advent of powerful computers in recent years has paved the way for the development
and implementation of several nonlinear analysis methods, which vary in terms of
applicability, efficiency, and accuracy. The main motivation behind such developments,
however, has been the urgent need for such tools in the field of structural design as well
as in experimental and theoretical research.
The use of nonlinear analysis in the design of structures has not so far been utilized to its
full potential. Despite its considerable demand on computing resources, accurate
nonlinear analysis allows a better assessment of structural safety and reliability than can
21
be achieved by other simplified methods. This realization represents a major drive behind
the efforts to improve the efficiency of nonlinear analysis methods, and will combine with
the rapid advancements in the field of computing to lead to the inevitable inclusion of
nonlinear analysis as part of the design process.
Hereafter, the different nonlinear analysis methods are reviewed, and their relative merits
in terms of applicability, efficiency and accuracy are outlined. This is followed by a brief
description of the scope of the present work as well as the organization of this thesis.
22
in bending, the differential equations can be simplified further by neglecting axial and
shear strains (Holden, 1972; Mattiasson ,1981).
Despite its accuracy, this method involves a considerable computational effort in the
numerical integration, and becomes much more complicated in the context of dynamic
analysis. Such considerations have restricted its use to static nonlinear analysis of single
members or simple structural configurations.
This method of analysis involves the discretization of individual structural members into a
mesh of finite elements, where the type of elements used is determined according to the
requirements of the particular problem. Whilst two main formulation procedures have
been applied within the finite element method, namely stress and displacement-based
(Robinson, 1973), the latter procedure was adopted to a greater extent for nonlinear
structural analysis. The displacement-based finite element formulation is derived from the
principle of stationaiy total potential energy, which requires the description of the element
strain energy in terms of the element nodal displacements. This is achieved through the
use of interpolation functions for the assumed element displacement field, as well as
strain-displacement relationships in conjunction with the material constitutive law. In
most cases, the degree of interpolation within the finite element is such that one element
cannot represent the displacement field of a whole member, especially if a high level of
nonlinearity is envisaged. For such situations, the use of a number of elements per
member is necessary to achieve accurate results, with the accuracy increasing
asymptotically with the number of elements.
Another approach employs engineering assumptions regarding the section response, with
the element strain energy obtained on the level of stress and strain resultants. The
relationship between the resultants of stress and strain is based on section properties such
23
as the cross-sectional area and the second moment of inertia (Wen & Rahimzadeh, 1983;
Meek & Loganathan, 1989), with the concept of shear area used to describe shear effects
(Chugh, 1977). This approach is computationally superior to the continuum approach,
but its inherent assumptions represent a source of inaccuracy especially for elasto-plastic
analysis, where considerable difficulties arise in the calculation of equivalent section
properties reflecting accurately the effect of realistic material laws.
The advantage of the finite element method is its versatility in modelling the nonlinear
behaviour of complex structural configurations with a systematic and computer-oriented
approach. Moreover, its extension into the domain of dynamic analysis is straight-
forward, where the effects of inertia and damping can be represented by equivalent nodal
loads via the use of lumped or consistent mass and damping idealizations. However, its
main disadvantage is the need for a fine mesh of elements to obtain an accurate
representation of the nonlinear member response, especially if material plasticity is
involved. Apart from imposing difficulties on the problem definition, this consideration
leads to a significant computational effort, and can render the analysis of some structural
configurations prohibitively expensive.
This is similar to the finite element method in the sense that a number of elements is
needed to model the nonlinear member response. However, the main difference is that the
element tangent stiffness matrix is directly extracted from elastic beam-column theory,
except that the elastic axial and bending rigidities are replaced by instantaneous values in
order to perform elasto-plastic analysis (Saleeb & Chen, 1981; Sugimoto & Chen, 1985).
Moment-axial force-curvature and axial force-moment-axial strain relationships are
employed to determine the values of the instantaneous section rigidities, and can be
modified to model such effects as local buckling (Sohal & Chen, 1988).
This method was mainly applied to the 2D analysis of beam columns, since its extension
into the 3D domain would involve significant complications in the generalized stress-
strain relationships. While its use of explicit relationships and matrices makes it an
efficient tool for nonlinear analysis, the explicit derivation of such relationships becomes
a formidable task if complex cross-sections and/or realistic material laws are considered.
24
1.2.4 Physical models
This method is mainly concerned with the pre- and post-buckling behaviour of beam-
columns, where a simplified idealization is adopted to provide an approximate but
efficient analysis tool.
One approach is to use the same function to represent the deflected shapes in the pre- and
post-buckling ranges (Toma & Chen, 1983; Papadrakakis, 1983), with plasticity effects
considered at midspan, as well as at the ends for beam-columns with fixed ends. For this
type of physical models, effects of material plasticity are usually established using
moment-axial force-curvature and axial force-moment-axial strain relationships dependent
on the cross-sectional shape.
Another approach divides the beam-column in the post-buckling range into two distinct
parts separated by a zone of plasticity. A plastic hinge without axial interaction (Toma &
Chen, 1982; Smith, 1984) or including axial interaction (Papadrakakis & Chrysos, 1985;
Papadrakakis & Loukakis, 1988) is often used to model the plasticity zone. However,
procedures allowing for the spread of yield have also been employed (Shibata, 1982;
Ikeda & Mahin, 1986).
This method was developed alongside the physical models approach (Higginbotham &
Hanson, 1976), where the main objective is to represent the hysteretic axial behaviour of
beam-columns. Phenomenological models are empirical formulations which follow
hysteretic loops available a priori, and whose main parameters are entirely determined by
the user. These parameters include physical properties such as cross-sectional area and
effective length factor, as well as other properties relating to the required shape of
hysteresis loops (Maison & Popov, 1980; Jam et al, 1980; Popov et al, 1985).
25
Whilst phenomenological models are only intended to represent bracing members subject
to inelastic buckling, their main advantage is that they employ only one element degree of
freedom (axial), and they are more efficient than other methods. However, considerations
such as the complexity of model logic, and the unavailability of experimental or analytical
solutions to determine the model parameters represent serious shortcomings if an accurate
prediction of hysteretic behaviour is required. Moreover, such modelling cannot be used
as a predictive analysis tool, since adequate information on the expected behaviour is
needed a priori.
The purpose of the present work is to develop a nonlinear static and dynamic analysis
capability for steel frames, where the basic phenomena involved in the nonlinear
behaviour, namely geometric and material nonlinearities, are considered in the context of
a computationally efficient approach. Amongst the available nonlinear analysis methods,
the finite element approach is considered to be the most suitable in terms of accuracy and
versatility, hence it is chosen as the basis of formulation throughout this work. However,
the essence of this research is to develop new formulations and analysis techniques by
which the finite element method retains its accuracy while attaining a much higher level of
efficiency than has been hitherto possible.
26
Material nonlinearities are considered in Chapter 3, and two approaches for modelling the
elasto-plastic behaviour of steel frames are presented. The first is an approximate
approach, in which the elastic quartic formulation is extended to allow idealized plastic
hinges at the element ends. These hinges switch from rigid to plastic states and vice versa
according to an interaction surface in the stress-resultant space, which accounts for the
interaction between the axial force and the biaxial bending moments. The second
approach represents more accurately the spread of yield along the member length and
across the section depth, and is based on a cubic formulation derived on the level of the
constituent material, where only direct stresses are considered to contribute to the
condition of material plasticity. The versatility of the cubic formulation is such that any
material constitutive law can be applied.
It is realized in this work that, whilst one quartic plastic hinge element per member, or at
most two if member global buckling is involved, provides an approximate prediction of
the elasto-plastic response, accurate modelling can only be achieved if a fme mesh of
cubic elements is employed. Since this represents a waste of computational effort for the
members which remain elastic for a large part of the analysis, and since one quartic
element is shown to represent accurately the elastic response of such members, a process
of automatic sub-division is suggested. This implies that the analysis is always started
with one quartic element per member, and new elements are created automatically during
analysis when and where needed, as discussed in Chapter 3 for both the approximate and
accurate approaches. Verification examples illustrate the relative accuracy of both
approaches, and demonstrate the improvement in efficiency due to the process of
automatic sub-division.
Two material models for mild steel are described in Chapter 4, with both models
employed by the cubic formulation. The first is the bilinear model with kinematic strain-
hardening, while the second is the multisurface model presented by Petersson and Popov
(1977). Sample analyses indicate that whilst the bilinear model is more efficient than the
multisurface model, the latter captures more realistically the behaviour of mild steel under
cyclic loading.
In Chapter 5, a formulation accounting for cyclic joint behaviour is presented, and related
geometric nonlinearity effects are discussed for the 2D and 3D cases. The formulation is
shown to cover other aspects of structural analysis, such as the effects of inclined
supports and structural gaps.
27
Nonlinear dynamic analysis is discussed in Chapter 6, where the use of non-structural
mass and damping elements is suggested in the context of a step-by-step time integration
procedure based on Newmark's scheme (Newmark, 1959). Two mass elements are
presented; the first is based on a lumped mass idealization, while the second employs an
Updated Lagrangian (UL) system to model distributed mass. Whereas hysteretic damping
is accounted for by the structural elements, dashpot and Rayleigh damping elements are
developed to represent equivalent viscous damping in linear dynamic analysis.
Finally, a nonlinear frame analysis program (ADAPTIC), containing all the developed
formulations, is discussed in Chapter 7. Description of the analysis capabilities of
ADAPTIC, including static analysis under proportional loading as well as time history
static and dynamic analysis, is first given. This is followed by discussion of the
incremental iterative strategy employed, as well as the automatic step reduction process
adopted in the case of convergence failure. For the solution of the stiffness equations, the
frontal method is found to be particularly advantageous, and its merits regarding the
process of automatic sub-division are pointed out. The structure of ADAPTIC is then
outlined, and the input data and results output files are briefly described. To demonstrate
the accuracy and efficiency of the developed code, example analyses are performed on
real and theoretical structures, and comparisons are made with solutions selected from the
literature, and with the results of a commercial nonlinear analysis package using
comparable element types.
28
CHAPTER 2
GEOMETRIC NONLINEARITIES
Owing to the complexity of geometrically nonlinear space frame formulation, mainly due
to the non-commutativity propeity of large rotations in 3D, there is a case for dealing with
2D and 3D analyses separately. Whereas this leads to larger source files in the context of
computer implementation, it reduces the computational cost considerably for 2D analysis.
Hereafter, a rigorous treatment of large displacements in plane and space frames is given,
followed by verification tests to substantiate the derived matrices and the assumptions
made.
The two main sources of geometric nonlinearity in plane frames are the beam-column
action and the effect of large displacements. Reisnour and Batla (1984) demonstrated with
a simple frame example the beam-column effect, and showed that erroneous results can
be obtained if it is neglected. Moreover, modelling of large displacements in realistic
structures becomes essential if the post-ultimate response is required.
Various approaches have been adopted for geometrically nonlinear finite element analysis
of plane frames. These approaches identify mainly with classifications according to the
formulation reference system (Eulerian, Lagrangian, etc..) and the degree of the
interpolation function used for element displacements, as discussed in subsequent
sections.
29
2.1.1.1 Total Lagrangian system
This is a system in which the local element freedoms and interpolated displacements (u,v)
are always referred to the initial undeformed configuration of the element, as shown in
figure 2.1. Its main disadvantage is the complexity of the strain-displacement
relationships when considering large displacements. A brief review of previous work
related to this system is given below, based on selected representative research papers.
Mallet and Marcal (1968) also used the it system to develop a finite element beam-
column formulation for moderate displacements. Cubic and linear shape functions were
used for the transverse and axial displacements, respectively, and three matrices
appearing in expressions for element forces and tangent stiffness were derived
In this system, local freedoms are referred to the last known equilibrium configuration, as
shown in figure 2.2. This, with the assumption of small displacement increments
between consecutive configurations, allows the use of simplified incremental strain-
displacement relationships, and reduces the five nodal freedoms required by large
displacement it formulations to three (u 1 ,v 1 , a.). where (a) is a rotational freedom.
30
Yang (1973) employed the Updated Lagrangian (UL) system in a finite element beam-
column formulation assuming cubic and linear shape functions for the transverse and
axial displacements, respectively. However, a straight element was assumed in the strain
energy expression, thus resulting in a tangent stiffness matrix that does not account for
bowing in the last known configuration.
Wen and Rahimzadeh (1983) discussed the need for finite element based derivations as
opposed to formulations adopting the theoretical stability functions approach. The main
reason is that the latter representation is complicated by the dependence of parameters on
the unknown axial force, thus requiring an exhaustive iterative procedure. Wen and
Rahimzadeh also presented an UL formulation based on cubic and linear distributions for
the transverse and axial displacements respectively, and pointed out that a limit of 15°
should be set on the increment of rotation of the element chord. Moreover, they showed
that a more accurate (less stiff) solution is obtained if the bowing strains are averaged
along the length. However, they did not observe that such an assumption does not require
the definition of a shape function for the axial displacements, but rather can be used to
obtain it, as will be shown in section (2.3.2.1). Similar to Yang (1973), Wen and
Rahimzadeh did not account for the bowing in the last known configuration. In the
representation of flexible members, this could lead to significant inaccuracies which can
only be avoided if a fine mesh of elements is employed.
This is a moving system with the x-axis joining the two ends of the element in its current
unknown configuration, as shown in figure 2.3. The assumption of small displacements
relative to the element chord is utilized to simplify the strain-displacement relationships,
while exact transformations to the global system allow for very large global
displacements.
31
Eulerian and global systems, an intermediate system coinciding with the local axes of the
undeformed element was used, a process resulting in additional computational steps.
In the present work, the Eulerian system is chosen for the formulation of plane frame
structural elements accounting for geometric nonlinearities and large displacements. The
choice is made in the light of the following advantages:
ii. The assumption of small displacements relative to the deformed element chord
is justifiable for realistic structures, even when large nodal displacements arc
involved. This favours the Eulerian system, as it avoids the complicated strain-
displacement relationships required by large displacement TL formulations.
iii. Eulerian plane frame formulations require three freedoms per node
(u., v 1 , a 1 ) , while large displacement TL formulations employ five
freedoms. Therefore, the former represents significant savings on the size of the
stiffness matrix and hence computational time.
v. The Eulerian system is ideally suited to performing yield checks and creating
new elements in the process of automatic mesh refinement, as discussed in
Chapter 3.
On the other hand, formulation in the Eulerian system also suffers from shortcomings.
Oran and Kassimali (1976) discussed the complications associated with distributed mass
when using Eulerian formulations in dynamic analysis. This is a direct consequence of
the difficulty of accurately modelling distributed loads if a moving system is employed,
since the transformation of distributed loads to equivalent nodal loads becomes a complex
function of the unknown deformed configuration of the element.
32
For simplicity and uniformity of terminology with 3D formulations, the Eulerian system
is hereafter referred to as the chord system, as it is associated with the displaced element
chord.
Three transformations are required between the chord and global systems for a complete
nonlinear analysis capability, namely, transformation of global to chord displacements,
transformation of chord to global forces, and transformation of chord to global tangent
stiffness. A relationship between chord forces and displacements, particular to the type of
formulation, is used with the first two transformations to establish global forces
corresponding to a set of global displacements. The third transformation defines the
global tangent stiffness matrix required for the iterative procedure.
The three basic chord displacements (u = 0 2')I) can be determined from the
global displacements ( gu = (u 1 . v 1 , a 1, u 2 , v2, a2)T) according to geometric
X'E=XE+ (u2_u1)
'E'1'E+ (v2_v1)
I 2 2
L'=4(X'p) +(Y'E)
(Y' 'i
= arctanj -4 I + nit
XE)
01=a1+p—p
02=a2+p— p1
(2.1)
33
2.1.2.2 Transformation of forces from chord to global system
The principle of virtual work can be used to establish the global element forces
1' i = 1F ,F 1 ,F 1 ,F 2 ,F 2F
g \ xl y 2)") from the three basic chord forces
(c = (M 1 ,M 2 ,F)1) (figures 2.5.a & 2.5.b):
3
g 1 f.Bu.=
Cj
g 1
au.
-
g 1.= Ug U
'j Cf.=T.. 1.
C j
j= J j=1
3
f.= T. . f.
g 1 j=1 1,3C3
(2.2.a)
where,
s, S
L'
C,
—s
L'
T=4 O 0
1' (2.2.b)
with,
XI
,_ E
C- &
(2.2.c)
34
The transformation of the chord tangent stiffness Jc to the global element tangent
stiffness g k can be establish by differentiation of (2.2):
at. ar.
g 1 = 3 ( air
C k _____
g i,j = aguj k= i i,k aguj + aguj
k
3 ( 1 3 ar acumi i,k
k =ZIT
g u. C'kI
k=1! i,ktmIacUm agujf+ agj
But,
a Ci k
___ ___
acUm ai. i,k a2 ck
a c U m_ ck k,m ' au =Tjm & ii agigj
u U.S6i,j,k
Hence,
3 II' 3
g k. .= H i. kCkk,mTj,m}+SGi,j,kC1k)
L,j k=1Um=1 1, (2.3.a)
where, g k and Jc are 6x6 and 3x3 matrices, respectively, and is a 6x6x3 matrix of
second derivatives of chord with respect to global displacements, given by:
2 c's'
1=—G14 1 S62,2, 1=5G251
L'2
,2 ,2
c —s
S6 1,2, 1 S61,5, 1S62, 1,1 S62,4,1 =
2c' s'
s64,1,1=s64,4,I=s65,2,1=s65,5,1=
L'2
,2 ,2
s —c
S64,2,1S64,5,1S65,1,1S65,4,1
L'2
All otherS 6. . are zero
(2.3.b)
35
s:1,1.3__s:1,4,3_S64,1,3=S:4,4,3_L
S:2,2,3_S:2,5,3_SG5,2,3=S:5,5,3
Geometric nonlinearities in space frames include the previous aspects discussed for plane
frames, i.e. the beam-column action and the effect of large displacements. However, an
additional aspect particular to space frames is the flexure-torsion coupling responsible for
lateral torsional instability.
The problem becomes even more involved when attempting to resolve the question of
flexure-torsion coupling in the presence of section warping. As it is necessary in many
situations to model a member using a number of elements, additional degrees of freedom
relating the waiping of adjacent elements become mandatoiy.
36
Chu and Rampetsreiter (1972) included the effect of warping on torsional rigidity in an
elastic beam-column formulation. An explicit stiffness matrix neglecting the torsion-
flexure coupling was derived on the basis that a member is represented by a single
element, and that joints are stiff enough to provide warping restraint at the element ends.
The authors discussed the need to use a number of elements per member in problems
involving flexure-torsion coupling, and pointed out the necessity of employing extra
freedoms relating the warping of adjacent elements. Large displacements, however, were
not dealt with in a consistent manner. Besides using secant instead of tangent stiffness for
the iterative procedure and buckling analysis, the secant stiffness applied only to small
displacements.
Oran (1973-b) extended a 2D elastic Eulerian formulation (Oran, 1973-a) into 3D. The
formulation assumed small element relative displacements but large global displacements,
with the beam-column action modelled using stability functions. A nodal triad of vectors
defming the deformed orientation of the joints was determined from increments of nodal
rotations, with the increments assumed small enough to allow vector manipulation. Oran
used the joint orientation vectors to establish the deformed element principal axes of
bending, which were then employed to obtain the displacements relative to the deformed
element chord.
37
conservative, the authors adopted a method of moment generation using forces of fixed
direction. This lead to the definition of two types of conservative moments: quasi-
tangential and semi-tangential. The authors pointed out that semi-tangential moments
correspond in the virtual work expression to sequence-independent rotations, referred to
as semi-tangential rotations. Observing the commutativity property of such rotations, the
authors derived a geometric tangent stiffness characterized by symmetry, and applied it to
initial buckling problems. However, the validity of the geometric matrix for analyses
involving very large rotations is questionable, since the derivation was performed
assuming small rotations in order to avoid complex expressions for work done by
moments.
Chan and Kitipornchai (1987) derived a geometric stiffness matrix including warping
effects for thin-walled beam-columns, and applied it to initial buckling problems. A
different formulation was employed for geometrically nonlinear analysis, in which the
effects of warping and large rotations were ignored.
It is also realized that including warping effects within formulations based on engineering
assumptions represents a formidable task. Apart from the complexities associated with
extra warping freedoms and their representation within a global system, it is quite difficult
to quantify warping effects in the presence of section plasticity without resorting to
continuum modelling. Therefore, in this work section warping will be neglected, an
assumption with serious implications only to members mainly subjected to torques or
lateral torsional instability.
38
As for modelling large global rotations, definitions referring to the undeformed
configuration have several shortcomings. Euler angles in the virtual work expression
correspond to moments which are neither fixed in direction nor orthogonal, particularly at
large rotations. While virtual work expressions involving semi-tangential rotations and
moments achieve the required simplicity only for small rotations. An incremental
approach similar to that of Oran (1973-b) is thus adopted, only that significant
modifications have been necessaiy to allow for large chord displacements encountered in
plastic hinge analysis. Oran's approach employs joint orientation vectors for determining
the element principal axes after deformation, and obtains the total chord displacements,
assumed to have small values, from the orientation of these axes relative to the element
chord. In the current approach, the calculation of chord displacements is performed
incrementally using element-based rather than joint orientation vectors, as discussed in the
following sections.
Within an incremental approach, the assumption is made that increments of global rotation
are small enough to allow vector manipulation. According to Oran (1973-b), the
transformation of a vector 0 due to an increment of global rotations is given by:
3 1 — 7 131
E T. .o. with, T=I 1 -(XI
J
a 1]
However, this relationship applies only to very small increments, since the orthogonality
property of transformed unit vectors is satisfied to the first order in rotation.
3
6 = 1. .0.
j=1 (2.4.a)
where,
39
(2 2
'j +Y af3 ay
2 ---r- f3+-r.
a13 (a2+yZ)
rT
2 -a+
ay (a +p)
2
- 13 --r- (2.4.b)
Using such a transformation reduces the amount of spurious lengthening of vectors upon
rotation, and preserves to a greater extent the orthogonality property of unit vectors, as
may be shown by the product of the various rows and columns of rT in (2.4.b).
In the virtual work expression, the moments corresponding to the global rotation
increments of the previous section are not fixed in direction. This can be demonstrated by
equating the virtual work of these moments to that of moments about fixed axes:
3
3 fm.6r.=
f
Zm.8r.
! 1 1
1=1 j=1 (2.5)
where, &r and 8r are infinitesimal increments of rotation about fixed axes and as
defined in the previous section, respectively.
3 3
&rT j .0. 8T1 k0k=
j=1 k=1
3 3 3 3 3
&T. .0.E
. i1k1 rTk,j
.o.= Z&T
j j=lk=1 r i,k r Tk ,j
j=1 k=1
& rT ij 6T jk rTk,j
(2.6)
This relationship represents a set of 9 equations with only 3 unknowns (&z,33,&y), and,
thus, a unique solution exists only for a transformation matrix rT preserving exactly the
40
orthogonality property of unit vectors. Since 1T in (2.4) is accurate to the second order
in rotation, the solution of (2.6) can only be obtained to the first order.
while the infinitesimal transformation due to rotations about fixed axes is given by:
o —&y 813
6T= &y 0
_613f
& 0
1 7
8a 2
813 =- 1
87 a
2 2
or,
3
8r.= mTj .ör.
i=1 (2.7.a)
where,
41
'1 13
1 2
2
a
m- 1
2
13 a
1
2 2 (2.7.b)
The combination of (2.5) and (2.7) demonstrates that the global moments corresponding
to the current definition of rotations are related to moments about fixed axes by the
expression:
3
m T j rn.
j=1 '
(2.8)
It is thus evident that the current global moments are not moments about fixed axes, but
rather defined in a system rotated by half the increment of rotations (a,13,y), as shown in
figure 2.7. This corresponds to the semi-tangential moment definition (Argyris et a!,
1978), only it is applied in the context of an incremental approach.
Global displacements are applied incrementally from the last equilibrium configuration,
while element deformation is established in a convected chord system following the
unknown configuration, as shown in figures 2.8.a & 2.8.b.
To allow for large chord displacements, as may be the case for plastic hinge elements, an
incremental approach is also adopted, where the increment of chord rotations is obtained
from the incremental rotations of element unit vectors relative to the element chord axis
42
(figure 2.9). At the last equilibrium configuration, the two vector sets, ('cO 0) and
0),
(2c , c representing the principal axes at the element ends, are normal to the chord
vector C , but not necessarily identical; the latter consideration depending on the
The vectors of principal axes are modified according to the increment of global rotations
at the two ends, while the chord vector is only affected by translational displacements.
Hence,
yC ZC
/X'
C( E E E
x
\i' LCLC
'c''
i,jYj
3
Lt
i j=l r iJ
2CC..... 2T
1 • r .2Cc
y
2T .c?
1 1,J J
(2.9.a)
where,
C 0
X E =X E + U 2 — U1
Z=Z E + w2w1
I 2 2 2
L'
= J (x) + () + ( z) (2.9.b)
and,
T rT(ai,131,Ti)
2T
(2.9.c)
43
With rT ( U' f , ' y) defined in (2.4.b).
The inclination of the current principal axes relative to the chord axis determines the
increment of relative element rotations. However, for the calculation of the increment of
twist, a fictious vector 21 C is defined as the transformation of the principal y-axis at
end (1) due to rotation increments of end (2). Assuming small incremental values, the
increment of chord displacements 6 CU can be established according to the following
relations:
68 =-X c C . IcC
lz Z
2c
60 =-x c. y
=-X c'. 2c
682z Z
6A =LC_LO
where,
and,
y 1
2T
1c
j=1' i,jY •j
(2. 1O.c)
The current chord displacements are obtained by updating the displacements of the
last configuration:
u =u°+ 8u (2.11)
44
Once equilibrium is achieved, the principal axes should be reset to a position normal to
the current chord axis, so that the above equations can be applied for the next incremental
step. This is done according to the following procedure:
f o_ 1..c
x'• x'
1co_
y - Rccxxci
2O...
y
= x c° ( c° after resetting)
(2.12)
As for the 2D case, the transformation of chord forces Y to global element forces is
obtained from the principle of virtual work (Figures 2.1O.a & 2.lO.b):
T
ct=(M
ly
,M
lz
,M
2y ,M
2z FM T) (2.13.a)
'Fxl
g 1= \ ,F yl 1,F
(2.1 3.b)
6
f.= 1.= T. f
g i j1 U g U j C J j=i i,jC j
(2.1 3.c)
The 12x6 terms of matrix T are first derivatives of chord with respect to global
displacements, and can be obtained explicitly from (2.9) and (2.10), as shown in
Appendix (A.1.1).
45
2.2.2.5 Transformation of tangent stiffness from chord to global system
A transformation similar to that obtained in (2.3) for the 2D case can also be derived for
3D formulations:
6 f( 6
gkjj = k=ltJm= Tjk 'km T m} + S G jjk C1k)
(2.14)
Here, the chord tangent stiffness k is a 6x6 matrix. While the geometric term G is a
12x12x6 matrix established in Appendix (A.1.2).
A new formulation is developed herein for modelling elastic beam-columns with initial
imperfections using one element per member. The distribution of the transverse
displacements in the chord system is based on a quartic shape function, hence the name
quartic formulation. The motivation for such a development is the poor performance of
the existing cubic beam-column formulation for cases of high nodal restraints, mainly due
to the absence of an intermediate degree of freedom within the element length.
As part of the adopted strategy, the derivation is performed in the chord system, where
the 2D representation can be directly extracted from the 3D case. A relationship between
chord forces and displacements is needed to obtain global element forces from global
displacements, whereas a chord tangent stiffness matrix is required to establish the global
element tangent stiffness; both requirements are addressed hereafter.
2.3.1 Kinematics
Eight chord degrees of freedom are employed for the 3D quartic formulation, as shown in
figure 2.11. These are referred to the imperfect configuration, and include the six basic
chord freedoms plus two mid-side transverse displacements.
46
a - 0a a a a t a t a\T
q -( ly' 0 lz' 0 2y'°2z' y ' z/ = displacements including imperfections =
T
((e
1y
(2.15)
The movement of a reference point along the element length due to deformation can be
described by four displacement values (a,u,v,w), as shown in figure 2.12. Assuming
small deformations, the interpolation functions for the reference point displacements and
the imperfection shape are given by:
a(x)= OT(t)
(2.16.a)
v'(x) = (2L(e1
2y - + 16t,)(-) + (L(e + Oi))() _
2
i(e1 oi )+84)() +t,
( 2 2y ly (2.1 6.d)
47
where, the constants of interpolation are the solution of a system of simultaneous
equations representing the element boundary conditions.
No interpolation function is needed for the axial displacement (u), since the constant axial
strain criterion is employed, as discussed in the following section.
This allows the strain state within a cross-section to be solely determined by a set of four
generalized strains: centroidal axial strain, rate of twist, and curvature strains about the
two principal axes.
Ifsmall chord displacements are assumed, the expression of the centroidal axial strain can
be approximated to:
1 2
8(dx) du 1l(dv dv'
dx +(...+ dx, dx) dx)J
(2.17)
48
An interpolation function for the axial displacement (u) can only be linear in the absence
of additional axial freedoms, and, hence, its employment in (2.17) would result in a
variable centroidal axial strain (ce) along the length of the element. This, however, does
not comply with axial equilibrium, and leads to over-stiff elements as remarked by
Wen and Rahimzadeh (1983).
The alternative is to assume a constant (c,, ), and to employ this condition to obtain the
distribution of (u). Hence,
Jcth
i ui?)
11 1( .2 .2 2
1 du 1 'civ dv' (dw dw' (dv" dw' '1)
dx —)
L
2
.2 2 2
c=
A 1 II1I('dv dv"
tt J1id?) L
(dw dw'
ai)
2 (2.18)
x
+ 1)2 2 2
u(x)= e(x+
-k)- (t + dw'
dx)
(dv1'
dx)
(dw1
dx
(2.19)
For the quartic formulation, the centroidal axial strain of (2.18) can be expressed
explicitly in terms of the chord freedoms as:
49
'a 2 a 2 a 2 .2 .2 .21
A - 410 +0lz +0 +0 ai __O
0 ly
L 1 Y
hlz_ 012y_ 012z ]+
1{
105 2y 2z
r 2 •2
If t a \ t az_
2
\ (t' ' ( j\ 2
Y _
256 + t)
0 a 0a 0 a 0i 9 i 0i e 1+
2.5 0 ly 2y lz 2z ly 2y lz 2zJ
[a
' fta\
a I I (0 a 0 a fzI
0 v 2z
8Eea
( 2y
/ (41(oi _0i)()]}
2z 1 (2.20)
The generalized stress associated with (e) is the centroidal axial force given by:
F=EA c (2.21)
With the flexure-torsion coupling neglected, the expressions for the rate of twist and
curvatures about the y and z axes can be established for small chord deformations
according to:
,. dct 0T
D
dx L (2.22.a)
—0
dx2 LIL I 2y ly)
50
Assuming that the y and z axes are the principal axes of bending, the generalized stresses
corresponding to the above generalized strains are a torque and two moments given by:
T=GJ
M y EI y iCy
M=EI1c (2.23)
where, GJ is the St. Venant's torsion constant, El and El are the bending rigidities
about the two principal axes, and , ic and ic represent the rate of twist and the
curvatures about the two principal axes respectively.
The chord forces q1 corresponding to the chord displacements q U defined in (2.15) can
be explicitly obtained using the principle of virtual work:
2
q 1.6 U.
1 q1 f(Fc&c+T6+My6Ky+Mz6Kz)dx
2
ae ____ ____ aK ••'
q11
J( u +GJCp
qj qj
+EIyKyp u
qj
+EIz1CzpuZ
qjj I'
L
- (2.24)
Combining (2.20) and (2.22) with (2.24), the following expressions can be derived for
the chord forces of the quartic formulation:
T
,Mlz ,M ,M2z , F, M T, T y, T z)
ql=(M
ly 2y (2.25.a)
El (
M =-7.2O _25.6--)+--1 80 a
ly L ly–1.20 1o5 ly
+z5e_8-i-J (2.25.b)
51
El I 6!) FL( 80 a 150a —8--
Mlz = — 7. 20
lz - 1. 20
2z - 25. Iz 2z L)
L L + 105 (2.25.c)
El I t
M2y =
L
1.20
ly +7. 20 + 25. 6-_
L
i ) + -1!j- ( 50 + 80 + 8-f.)
(2.25.d)
a
EI (
M= 1. 20 +7.20 + 25. 6 .)+
L -- ( 50 + 80 a2 + 8-)
105 (2.25.e)
F=EAC (2.25.0
GJ
1T L "T (2.25.g)
El (
f
T= — —25.60ly +25.60 2y
t
L
(2.25.h)
El 1 tz) F( 89 a 89a
T= -(- 25. 60 iz + 25. 602z 204. 8 L + 105 lz 2z +512—L,)
(2.25.1)
where, the terms of u are defined in (2.15), and CC is obtained from (2.20). The two
forces (T i , T) are transverse mid-side forces corresponding to the two additional
chord freedoms (tt) as shown in figure 2.14.
The beam-column coupling between the axial force, the bending moments, and the mid-
side transverse forces is exhibited in the above equations. However, in the context of a
global analysis procedure, only the basic chord freedoms are available, which leaves
the two remaining freedoms (t , t ) belonging to q U to be determined from the
conditions (T = T = 0). Due to the coupling between the mid-side forces and the
axial force, an iterative procedure is required of the form:
52
1ty, _ 1t y1 I8tyl
t t zl,- itzj 8tJ
-1
aT aT
J &t y at a t JT1 qk77
where, qk is the chord tangent stiffness before static condensation, defined in the next
section.
The chord tangent stiffness before static condensation is an 8x8 matrix defined as:
a i.
k
q iiaqcj
(2.27)
q
k. .=k. .+EALIJ.V.
1 1 (2.28.a)
where,
El El
1k22=1k44=7.2-+-j--FL
El El
1k 13 =1 k 31 =- 1.2-j -+ -FL, 1k24=1k4 1.2-+-FL
2=
El
L2
1
El
L2 10
El El
GJ 512F 512F
L
All other terms are zero
(2.28.b)
53
and,
ta \
1 = i.-(80 , +2. 50a2Y — 8--J ij2 +2.50-8.y)
= 1
7 105L 105 801 z802z 5121.)
(2.28.c)
The two main requirements for a complete nonlinear global analysis capability are:
54
2.3.5.1 Plane frames
For a set of global displacements g U the chord displacements are first obtained
according to (2.1). A simplified version of (2.25) and (2.26) is then applied to determine
the corresponding chord forces J predicted by the 2D quartic formulation. These are
finally used in (2.2) to establish the element global force contribution gt
For the determination of the global tangent stiffness gk (2.3) is employed with a 2D
version of the chord tangent stiffness Jc, obtained in (2.29) after static condensation of
the mid-side transverse displacement
In 3D, global displacements g11 represent incremental rather than total displacements.
The increment of chord displacements 8,u is first established according to (2.10), and
used to determine the cumulative chord displacements . The chord forces f are then
obtained from (2.25) and (2.26), and substituted in (2.13) for the element global forces
gt•
The 3D global tangent stiffness contribution gk is determined from (2.14) using the
statically condensed chord tangent stiffness 1c of (2.29).
2.4 Verification
The quartic formulation has been implemented in a new nonlinear analysis program,
(ADAPTIC), described in some detail in Chapter 7. This program is employed in the
examples below to establish the accuracy of the quartic formulation, and to demonstrate
its ability to model elastic beam-columns using one element per member.
For an objective assessment of beam-column formulations, the extent to which the axial
force affects the bending stiffness must be demonstrated. Figure 2.15.a shows curves for
the non-dimensional bending stiffness parameters (s) and (cs) of a perfect element
subjected to a compressive axial force (F). For the quartic formulation, the bending
parameters (s , cs) and the axial load parameter (p) are defined according to:
55
______
S= (EI' ' (E1,,\ '
—F
El
(2_Y)
L) L) '¼ L)
where Jc is the chord tangent stiffness alter static condensation as established in (2.29).
The comparison in figure 2.15.a with the theoretical stability functions approach
demonstrates the significant improvement in accuracy attained by the quartic formulation
over the existing cubic formulation (Jennings, 1968). This is only due to the additional
mid-side freedoms (t >, , t which enable a better description of the deformed shape in
the chord system.
For a more practical assessment, the elastic buckling loads given by the cubic and quartic
formulations are compared with the theoretical values. For a simply supported strut
modelled by one element, the prediction of the lowest buckling load can be obtained
directly from the previously established values of(s) and (Cs):
Similarly, if the strut has identical joint rotational stiffnesses at both ends, then the
percentage error in the estimate of the buckling load can be plotted against an effective
length factor (k) defmed as:
k=J(1)
The remarkable accuracy of the quartic formulation exhibited in figure 2.15.b shows its
superiority over the cubic formulation, and reinforces its ability to model beam-columns
using one element per member. Although this accuracy is slightly diminished for effective
56
length factors (kz0.7), the range of accurate modelling using one quartic element per
member is significantly increased if plasticity is considered, as discussed in section 3.5.1.
However, for the few cases of members with effective length factors (k<0.7) and
undergoing elastic buckling, two quartic elements may be necessary to achieve very
accurate results.
The large displacement response of the diamond structure depicted in figure 2.16.a was
first obtained by Jenkins et a! (1966), and later presented in tabular form by
Mattiasson (1981). Both authors, adopted the elliptic integrals approach in the analysis of
the structure, and neglected the axial deformations.
The diamond frame is analysed under tension and compression by the 2D quartic
formulation, and comparison with the results of Mattiasson is shown in figure 2.16.b.
Only one quartic element per member is used, and excellent agreement is obtained over
the large displacement range. The slight disagreement at very large compressive
displacements can be attributed to the severe deformation in the chord system, at which
point the approximation of the strain-displacement relationships in (2.20) and (2.22) is no
longer accurate. This can be remedied by the use of two quartic elements per member,
which reduces significantly the chord deformations. However, such a situation would not
arise for realistic structures without involving plasticity, the treatment of which is
discussed in Chapter 3.
The simple plane truss shown in figure 2.17.a has a parabolic imperfection of (11500) in
the inclined compression member, which induces buckling under a horizontal force (P).
The truss is first analysed without imperfections using LUSAS and the quartic
formulation. The LUSAS element is based on an Updated Lagrangian approach
employing cubic interpolation for the transverse displacements. When the truss is
modelled with 2 elements, LUSAS overestimates the buckling load by 55%. While the
results given by 2 quartic elements compare favourably with the theoretical solution and
that given by 6 LUSAS elements.
For the imperfect case, LUSAS is unable to account for imperfections with 2 elements.
However, 2 quartic elements give identical results to those of 6 LUSAS elements where
57
the imperfection is accounted for by modifying the nodal coordinates along the inclined
member.
As discussed earlier, lateral torsional instability is not given a high priority in this work
due to the complexities associated with extra warping freedoms and the modelling of
warping in the presence of section plasticity. This permitted the flexure-torsion coupling
to be neglected when deriving the quartic formulation in the chord system. However,
such coupling can still be accounted for if a number of elements is used per member,
since the effect of nodal displacements on geometry is accurately modelled according to
section (2.2).
If the rectangular strip was supported and restricted against twist at both ends, 1 quartic
element would not show any buckling characteristics, and a number of elements would be
required. Also, for cross-sections that warp under torsion, the predicted results would not
be as good since non-uniform torsion is involved, thus, inducing warping strains and
stresses.
The 450 circular bend depicted in figure 2. 19.a has been extensively used for the
verification of finite element formulations. Bathe and Bolourchi (1979) performed the
large displacement analysis using 8 straight beam elements based on continuum
mechanics and employing the Updated Lagrangian approach. While Surana and Sorem
(1989) used 8 three-node parabolic elements derived from the isoparametric family. To
account for large rotations, Surana and Sorem chose the rotational transformation as the
average of the six available sequence dependent transformations, thus, eliminating the
bias towards a specific sequence of application.
58
The bend is analysed here using 1 and 2 imperfect quartic elements, and applying the load
in 15 equal increments. With 2 elements, the quartic formulation gives identical results to
those of Surana and Sorem, as shown in figure 2.19.b. While 1 quartic element still
provides a very good comparison even though the bend curvature conesponds to high
imperfection levels, thus, compromising the assumption of small deformation relative to
the element chord.
The bend is re-analysed using 2 imperfect quartic elements, but applying the load in
5 increments only. The comparison in figure 2.19.c demonstrates that the adopted 3D
incremental approach is almost insensitive to the size of the load increment applied.
59
2
Y,v
2
I - - -
x,U
2
Y,V Current unknown configuration
y .
x,u
60
Y,v
x,U
61
Y
L
2
p
1
XE
F'
x
Y
2
Parallel lines)
__
1
1
I
U2
62
Y
4VJM1
-'N.
— — — — 2
•NI
— —
1
F1 1
FZs*tFy1i..'N.
63
x
F,
L72
r>. ,-
'I 13 X
F1 #1 4q
ZF2
64
Y,v, x
Current unknown
configuration
(Convected system) P244272
Yç,
TT:::.4'
11/ , __ 11
.1/ /I .2
z
Last equilibrium
configuration
- -
"1' " I IY
_-jjz__j, X, u, a
Initial undeformed
configuration
Z,w,y
1*:
L+ 2y>I
65
Y C
2ccu.,1xc
2o
Current unknown
configuration
1 CC
y
_iiiii072co
y
4ic
(Convected system)
2
1c'
z
Last equilibrium _____ Y
configuration
cTi
x
E
VXOE
Figure 2.9 Element vectors for the current and previous configurations
66
y
X
__) M ,, M
MT lz MT
Y F
-. F2
- t
r>' F2F2
F,
F 72 1
F 1 Current unknown
zi configuration
2
SN.--.,
--
1u' x
Initial undeformed
configuration
67
(L
2 X
lii
Imperfect
configuration
LI
(L (L
Z(¼ (0) z-
0 lz
Ltz
1
0' lz I tiz
Imperfect 02;
configuration
L/2 112
y (0)
L 0
68
y (0)
y(x),v
x)
a(x)
w(x)
z(0jk #< ' - '
z(x), w
dv + dv'
I I
x I I
I I
DIr-
I
L I
::::;:;;iy --fl-
L—_—)rs dw' dw + dw'
z z
dx + du
Before deformation After deformation
69
F
hperfect
MT M configuration MT
ly
MT
70
5
2 (s)
.SS - I •••• S S IS
Cubic formulation
1 - - -. Quartic fomiulation
Beam-column theoiy
0
0.0 0.5 1.0 1.5 2.0
Axial force parameter (p)
100
90
80
70
60
30
20
10
0
0.5 0.6 0.7 0.8 0.9 1.0
Effective length factor (k)
71
L=2m
[ L J2P JL 2P
U
EA=2.lxlO8N
EI=1.75x 10 N. m2
10
8
6
c -2
-6
-8
-10
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Displacements u & v (m)
72
P
5m
>1
0.0 Q-
0.00 0.01 0.02 0.03 0.04 0.05
Horizontal displacement (m)
73
Z,w
Semi-tangential
moment (M)
1.2
1.0
0.8
0
U
0.6
0.2
0.0
-0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25
Lateral and vertical displacements (m)
74
Fully fixed end
P /
'R=lOOin
450
x
un
Z,w
R=lOOm
X,u
75
16
14
12
1
0
8
2
0
-40 -30 -20 -10 0 10 20 30 40 50 60 70
Tip displacements (in)
16
14
12
'I
0
8
2
0
-40 -30 -20 -10 0 10 20 30 40 50 60 70
Tip displacements (in)
76
CHAPTER 3
MATERIAL NONLINEARITIES
The accurate modelling of material nonlinearity effects in steel frame members requires
the use of appropriate material models, and the representation of the spread of plasticity
within the section and along the member length. In the context of a finite element
approach, this often demands a fine mesh and a detailed consideration of the section
shape, thus, involving significant computational effort. To avoid this drawback, a plastic
hinge approach can be employed, but at the cost of accuracy.
Wen and Farhoomand (1970) presented a 3D plastic hinge formulation neglecting the
effect of geometric nonlinearities. In order to account for the spread of plasticity along the
member length, the authors assumed the hinges to be lumped at a distance from the
member ends. However, such an assumption would lead to an overestimate of member
strength, a fact readily demonstrated for a cantilever.
Hobbs and Jowharzadeh (1978) used the dynamic relaxation method to solve the
differential equation of 2D beam-columns, and employed incremental moment-thrust-
curvature relationships to account for the spread of plasticity within the section. The
77
authors derived these relationships for a rectangular solid section, and discussed the
difficulty of explicit derivation when unloading is involved.
Anagnostopoulos (1981) pointed out the complexities associated with plastic hinge
modelling when attempting to represent the spread of yield and strain-hardening in beam
members. The author discussed the implications on such modeling of the loading
sequence as well as the continuous change in the moment gradient and the relative
magnitude of end rotations. As a result, it was concluded that simplifying assumptions
need to be made, thus, restricting the applicability of such approach to specific loading
configurations.
Yang and Saigal (1984) distinguished between the 'section' and 'fibre' types of elements
in modelling material nonlinearity effects. In the 'section' type, inelastic behaviour is
defined for the cross-section as a whole, not for individual points. This approach can be
used to account for the spread of plasticity along the member if the length is divided into a
number of sections. Otherwise, it can be employed in the context of a plastic hinge
representation which requires the definition of multi-dimensional action-deformation
relationships relating moments and axial force to hinge rotations and axial deformations.
On the other hand, the 'fibre' type includes division of the cross-section into a number of
areas where stresses and strains are monitored, and the contribution to the overall member
stiffness is determined from the governing stress-strain law. This approach provides a
better simulation of the spread of plasticity, whereas the 'section' type is computationally
more efficient. The authors presented a 2D beam formulation based on the 'fibre' type for
modelling material nonlinearities, and applied it to the analysis of metal and reinforced
concrete beams.
Corradi and Poggi (1984) derived an elasto-plastic 2D formulation neglecting the effect of
geometric nonlinearities. In an effort to reduce the amount of computation involved in
establishing the elasto-plastic section response, the authors used a moderate number of
sampling points within the cross-section, and employed polynomial fitting to determine
the overall section behaviour.
Sugimoto and Chen (1985) applied the finite segment method to solve the elasto-plastic
post-buckling problem of 2D tubular beam-columns. In this method, a member is divided
into a number of segments across which the curvature and axial strains are assumed
constant. Moment-thrust-curvature and thrust-moment-axial strain relationships were
employed to establish the tangent stiffness matrix of each segment, and an incremental
approach without iteration was applied to establish the global member behaviour. The
78
disadvantage of this approach is the difficulty of establishing explicitly the required
relationships between forces and deformations for various cross-sections, and the
complexity of such relationships when unloading is involved.
Powell and Chen (1986) performed a numerical investigation into the feasibility of a
generalized plastic hinge with strain-hardening. The authors discussed the concept of
zero-length plastic hinge which implies infinite strains, thus, requiring relationships
between hinge forces and displacements instead of strains. This leads to a dependence of
the hinge properties on the length of the element as well as the loading configuration if
strain-hardening is to be modelled. The authors pointed out the computational advantage
of this approach, but advised care to be taken in specifying the hinge strain-hardening
properties.
In the current investigation, two approaches are considered for modelling material
nonlinearities. The first is an extension of the quartic formulation presented in Chapter 2,
where a plastic hinge idealization is employed. No effort is made in this formulation to
include parameters for modelling strain-hardening or spread of plasticity, since the added
complexity does not guarantee sufficient accuracy. The second approach is based on an
elasto-plastic cubic formulation, where the spread of yield within the section and along
the length is accounted for by using a number of monitoring points across sections
corresponding to a second order Gaussian quadrature rule. In addition to its accuracy, the
advantage of this formulation is its ability to accommodate any stress-strain law, thus,
accounting explicitly for strain-hardening effects. Both approaches are considered in the
context of an automatic member subdivision process, where adaptive mesh refinement is
performed during analysis when and where necessary.
79
3.2 Plastic Hinge Quartic Formulation
Rigid-perfectly plastic hinges are added to the quartic formulation (section 2.3) to
provide a simple yet effective method for analysis involving material plasticity. The
resulting formulation is intended for preliminary investigations, since the effects of spread
of plasticity and strain-hardening are not accounted for. The element forces and stiffness
are considered in the chord system, in line with the derivation of the elastic element. This
permits the effects of geometric nonlinearity to be addressed in exactly the same manner
as discussed in sections 2.1 and 2.2.
Hinges of the rigid-plastic type are allowed at either end of the element, as shown in
figure 3.1. It is assumed that the contribution of shear stresses to plasticity is negligible,
and, hence, the formation of a plastic hinge is governed by the interaction of the two
principal moments and the axial force:
Plastic displacement increments are allowed at the plastic hinges, and are assumed to obey
the associated flow rule:
8u=
6u= Njhbh
aM aM Iz 0 0 0
NT = lY
0
2z
aF
(3.2)
N represents the components of the normals to the interaction surface, while b contains
positive scalars for the two hinges. Also, the summation range variable "h" indicates only
the hinges which are plastic:
80
Only hinge (1) plastic = h = 1
The need to establish the interaction surface of a general section shape by analytical
functions prompted the use of polynomial fitting to a selected number of interaction
points. Three curves determine the interaction surface, as shown in figure 3.2:
(M M
2
3 I M =0: biaxial moment interaction at zero axial force
MZJ))
YP (3.4)
It is assumed that the biaxial moment interaction in the presence of the axial force is
identical to that at zero axial force, but with reduced plastic moments. Hence, the equation
of the interaction surface can be expressed as:
I M M)
1=1
(3.5)
Each of the interaction functions "f1" and "f2" is composed of three polynomial functions
established over three adjacent intervals, as shown in figure 3.3. Conditions of continuity
of values and slopes at the two intermediate interaction points are used to establish the
constants of the polynomial functions, with the slopes chosen to satisfy the curve
convexity. On the other hand, function "f3" defines the non-dimensional biaxial moment
interaction, and is assumed to have the form:
81
(Y
3M',' M',)
(M
I
MY )2 M 2
m=J(M) +(M') & Cb= 2 2
/(M)
(3.6)
where "si" is a function defined by three polynomials established over three adjacent
intervals of bending direction, as shown in figure 3.3. A constant function "V"
(v c b = - i) corresponds to a circular interaction curve between the biaxial moments.
Since plasticity is lumped at the element ends, the chord forces can be directly
obtained from the elastic chord displacements u, only that an incremental approach is
necessary due to the path-dependence of the problem. To ensure that the chord forces
remain within the boundaries of the interaction surface, the plastic hinges exert
incremental plastic deformation 6u so that only part of the chord displacements
increment 6 u is elastic:
= - 8u (3.7)
Thus, for an increment of chord displacements 8 j, .. r can be obtained using the elastic
element properties once 81 u is determined. If both hinges are rigid at the start of the
current increment, 6u is taken as zero. If at least one hinge is plastic at the start of the
current increment, &u is determined in accordance with the following section.
The calculation of the plastic deformation must ensure that the forces at the plastic hinges
do not exceed the interaction surface. This condition can be expressed infinitesimally
using the tangency equation:
82
6
N. 8f.=0
i,g C
j=
(3.8)
6
8I=X k16u
J=1 (3.10)
Hence, the combination of the flow rule in (3.2) with (3.7), (3.8) and (3.10), results in
the following system of equations with the scaling factors b as unknowns:
6 6
D b = N. k. .8 u.
h g,h h
i=lj=l i,g i,j C j
(3.1l.a)
where,
66
U = N. e k . .N.
g,h i,gC i,j j,h
(3.1l.b)
This represents one or two simultaneous equations, depending on the number of plastic
hinges which is reflected in the range variables "g" and 'h'. The solution to (3.1 1.a)
yields an estimate for the scaling factors which can be expressed as:
66..
ek 8u.
b h = Dh'
,g N i,g C j,j C j
g i=lj=1 (3.12)
1
in which U is the inverse of the lxi or 2x2 part of the D matrix.
83
If a scaling factor corresponding to a plastic hinge is negative, elastic unloading occurs.
In this case, the hinge is assumed rigid, and the scaling factor of the other plastic hinge, if
present, is re-calculated from (3.12) after re-establishing the range variables "g" and "h".
b=b+6b (3.13)
where the iterative correction 6b depends on the deviation of the stress states p from the
interaction surface, and can be estimated from:
(ap g
(3.14)
ap 6 6 1a g 1a cr i acu i Yacu i ) 6 6
N ek (-1)N j,h
h i=lj=laCf.JaCUJJaCu.JLabhJi=1j=l i,gC i,j
ap
abh g,h
(3.15)
1 (Pg_i)
bhbh+ D h,g
(3.16)
where again D 'is the inverse of the lxi or 2x2 part of the D matrix.
84
3.2.2.2 Scaling to the interaction surface
As previously mentioned, hinges which are rigid at the start of an incremental step are not
allowed to exert plastic deformation. It is therefore possible that stress states of rigid
hinges exceed the interaction surface after the application of an increment of chord
displacements 6 Cu. To remedy this violation of hinge strength, 6 is scaled down by a
reduction factor 'r' until convergence to the interaction surface is achieved.
Because of geometric and material nonlinearities within the element formulation, the
relation between the interaction values p of rigid hinges and the reduction factor 'r' is
nonlinear. Therefore, the scaling procedure must be iterative, and proper allowance must
be made for the case when both element hinges are rigid and exceeding the interaction
surface simultaneously.
Once convergence is achieved, the corresponding hinge is taken as plastic, before the rest
of the increment "(1 - r)6u " is applied.
3.2.2.3 Sub-incrementation
In this work, 6 ii is initially applied in one step, and the number of sub-increments is
then determined according to the relative position of the non-dimensional stress states of
the plastic hinges, as well as the relative orientation of the non-dimensional normals at the
start and end of the step. Mathematically, the number of subincrements "n" is expressed
as:
85
11J I
Ii
If hinge(1)rigid
ii (3.17.a)
([integer (looxDist.(2do,2dc))+11
( Angle( 0o oc)) J If hinge (2) plastic
t integer
1 1 1 I
If hinge (2) rigid j
1 1 f (3. 17.b)
n =Max.(n,n,n,n)
(3.l7.c)
where, superscripts (o) and (c) denote start and end of step respectively,
/I o 0
0 f1y M1 F°
1 \M' MZP'FP
I C
dC_('lY M 1 FC
1
(3.18.a)
(3.18.b)
/ M F°
2
M FC T
2d \M' M'
(3.18.c)
86
2O°=(MXN2,MZPXN:2,FPXN2)T ]
2 C ( M W X N 2 ,M,x N 2 Fx N2)
(3.18.d)
and,
13 2
Dist .( .d O ,. d C ) = I
dC d°
J i=lc i)
Angle(0,&)=cos_' .a1.o°
J 1)
¼.i=l (3.19)
kaH adI
3.2.2.4 Plasticity under pure axial action
Because the normal to the interaction surface is not uniquely defined at the point
corresponding to the full axial capacity (±F), numerical difficulties arise if a stress state
of a plastic hinge crosses this point. To avoid this problem, the interaction surface is
assumed to extend smoothly beyond (±F), and stress states are allowed to continue on
the extended branch, as demonstrated in figure 3.5 for bending in the x-y plane. Since
this implies a violation of the hinge strength requirement, an iterative scaling procedure,
similar to that discussed in section 3.2.2.2, is employed to establish the reduction
factor 'r' needed to bring the stress states back to the point of full plastic axial capacity
(±F).
Once at the point (±F), a further increment of chord displacements &cU will not cause
any change in the stress states if the components of plastic deformation lie within the
boundary normals. This is demonstrated in figure 3.6 for stress states at (F r), and
assuming positive plastic hinge rotations in the x-y plane:
N1 + N2
80 ]=> no change in stress states ] (3.20)
N N
p[ 1,1 3,2
When biaxial hinge rotations are involved, a simple mathematical representation becomes
more difficult, since the boundary normals are now represented by a conical surface
instead of two vectors. However, if the boundary normal with components proportional
87
to the hinge rotational increments is established, the check for the change of stress states
can be readily made.
It can be shown from (3.4)-(3.6) that the normals at the point (F 1,) are expressed in the
positive rotations quadrant as:
Cb I Cb2
N21— M2, N42_M
where,
The direction cosines "C 1" and "Cb' are chosen such that the components of the
corresponding normals are proportional to the increments of hinge rotations at both ends.
Hence,
N 1,1
ly
II = fl
''2,1 60r
lz
(f'2( F
f(Fp)]
2
2
f'1(F))
I (3.22.a)
Similarly,
88
(f'Fp)
I f' (F ,)
C b2 = 2
J(f'2(F)
_____ 2 60
I__2Y)
f(Fp)) I 3P I
\ 2z1 (3.22.b)
Once and " Cj are determined, N can be established from (3.21), and a check
similar to (3.20) can be performed:
However, since the condition of no change in stress states implies a zero increment of
elastic chord rotations and axial displacement, this condition can be expressed in terms of
the increment of chord displacements 8u as:
30 (60 30 '\
& +NJ+N52xIN+NII
1.•. 3,2 4,2))
T\
8u=( 6O
ly ,3O 6e23o2Mo)
lz
I
T I
I6Cu =(0,0,0,0,0,60 T) I
T
=(o,0,o,O,Fp,(oT)) J (3.24)
with similar expressions for different combinations of positive and negative increments of
chord rotations, and for the case of plasticity at (—Fr).
If the condition of (3.24) is not satisfied for an increment 3 CU, then the stress states at
the plastic hinges either undergo elastic unloading or follow a loading path on the
interaction surface. In the latter case, difficulties arise because the normals are not
89
uniquely defined at (F) and, hence, 6u cannot be estimated. To avoid this problem,
the element is partially unloaded from the condition of axial plasticity before applying
6 Cu. Upon reloading, the scaling to the interaction surface brings the stress states at the
plastic hinges to points different from (19, and 8u can then be determined as normal.
The chord tangent stiffness matrix Jc must reflect the state of hinges at the element ends,
whether rigid or plastic. If both hinges are rigid, then Jc is taken equal to the elastic
element chord tangent stiffness k. If at least one hinge is plastic, then k can be
derived as follows:
in which,
a C r.
ek i, k = elastic stiffness
a k
(3.26)
and,
u
aCk
abh - Nk,h
(3.28)
and,
90
6
D•1g Nm , g ckm,j
gm=i (3.29)
Thus, combining (3.25)-(3.29) results in the following expression for the chord tangent
stiffness matrix:
6 6
ck = Z N h D 1g N m, g
j, k {u k, j - ( g h m=1
k=1 m, j)}) (3.30)
For the special case of axial plasticity at (±F1,), the condition of no change in the hinges
stress states is assumed, hence, the chord tangent stiffness is taken as:
As for the elastic formulation, the two main requirements for a nonlinear global analysis
capability are:
However, one more consideration of equal importance need to be included for the plastic
hinge formulation, namely, the suppression of plastic hinges causing nodal singularity, as
discussed below.
For the plastic hinge analysis of space frames, the calculation of the element global forces
from the global end displacements g U is performed through the same sequence of
transformations as for the elastic formulation. The chord displacements
corresponding to g U are first established using (2.10), and then employed to determine
91
the element chord forces according to section 3.2.2. Once J is obtained, gf Can be
calculated according to the transformation of (2.13).
For plane frame analysis, a simplified 2D version of the plastic hinge quartic formulation
is employed. The transformations to the global system are performed in the same manner
as for the elastic formulation, where the appropriate relationships of section 2.1.2 are
used.
The transformation of the chord tangent stiffness (ic to an element global tangent
stiffness g k is again identical to that of the elastic formulation, with a simplified version
of Jc used for plane frame analysis.
The elastic chord tangent stiffness k required for the calculation of (ic in section 3.2.3
is that of the elastic formulation, where u is replaced in (2.15) to (2.29) by u. To
reflect cases of elastic unloading or formation of plasticity within an incremental step, the
state of the hinges, whether plastic or rigid, is considered at the current iteration.
Numerical difficulties often arise if two plastic hinges occur simultaneously at a node
joining two elements. This can be demonstrated for the beam example of figure 3.7.a,
where the applied load causes two adjacent plastic hinges to form at the loading point.
Since the stress states at both hinges are dominated by the bending component, the
rotational stiffness at the middle node becomes very small, therefore, resulting in an ill-
92
conditioned stiffness matrix, and leading to a diverging solution when the next load step
is applied.
To overcome this problem, nodes where two adjacent plastic hinges may occur are
identified, and one of the hinges is always suppressed by assuming it rigid for the current
load step. This is shown for the beam example in figure 3.7.b, where increments of
plastic deformation are not allowed at the suppressed hinge during load step (8P). The
equilibrium condition at the end of the load step requires that the stress states of the
adjacent hinges remain approximately identical, therefore, ensuring that the stress state of
the suppressed hinge does not exceed the interaction surface significantly. Any departure
from the interaction surface would be mainly due to out-of-balance forces and the
elongation of the plastic element which alters the symmetry of the problem. To correct
such departures, the suppressed hinge is activated while the other hinge is suppressed for
the next load step, so that both plastic hinges accommodate their share of plastic
deformation.
The computational advantage of the plastic hinge formulation justifies its use, but only to
represent members in which the constituent material is elastic-perfectly plastic, and for
cases where the spread of plasticity can be ignored. To overcome such restrictions, an
elasto-plastic formulation based on a cubic distribution for the transverse displacements
has been derived in the chord system. The material nonlinearity effects are accounted for
through a detailed consideration of the section shape and the material stress-strain law,
while the geometric nonlinearities are dealt with according to sections 2.1 and 2.2.
3.3.1 Kinematics
Six chord degrees of freedom are employed for the 3D cubic formulation, as shown in
figure 3.8:
cIj = (ol,el,e2,e2,,oT)T
(3.32)
93
a(x)= OT(t)
f O +0 \ (20 'Y+O2y'• 2
v(x)=I 1 2
I—I JX+(Oiy)X
Is. L ) I4 L
As for the quartic formulation, no interpolation function is needed for the axial
displacement (u), since the constant axial strain criterion is used.
The same assumptions of section 2.3.2 are made for the cubic formulation, so that the
strain state within a cross-section is uniquely determined by the four generalized strains:
centroidal axial strain, curvature strains about the two principal axes, and rate of twist.
The centroidal axial strain is assumed constant along the element length, hence, an
interpolation function for the axial displacement (u) is not required. Although in the
presence of material plasticity, this condition does not necessarily imply a constant axial
force along the element length, the variation of the axial force is insignificant if the
element is used to model short lengths of elasto-plastic members.
The derivation of the four generalized strains is similar to that of section 2.3.2, i.e:
L
+ (4)2})
1 hI-LJ(dv\2
LJ2Ldx)
0
d2v d2w
z= 2 '
(3.34)
which in combination with (3.33) leads to the relationship between the generalized strains
and the element freedoms:
112 2 2 2\ I
i2iO
s. ly
+02y +0Iz
+0 2zj
1-10ly02y +0 Iz
02z
A
30 (3.35.a)
94
I 2y +e1y)Jx [2(20
'.' ly+02Y)]
L2 - L (3.35.b)
0
(3.35.d)
Since the relationship between the generalized stresses and strains cannot be established
explicitly in the presence of material plasticity, the integration of the virtual work equation
to obtain the element forces is performed numerically. Only two Gauss integration
sections are employed along the length, again because the element can only represent
short lengths of members. The position of the Gauss sections, shown in figure 3.10, is
chosen such that exact integration of the virtual work equation is obtained for the elastic
element
L( '/
g 1 2. - 3
_L(
gH2_ 2. /5•
+3 (3.36)
The generalized strains at both Gauss sections are, hence, represented by a matrix U:
cc cc
- lcy(gH1) lcy(gx2)
S'
1Cz(gH)
C (3.37)
{ 2(e y 2 2
y lz 2z) .. ly 2y +0lz02z)J
- 30 (3.38.a)
95
( 0 \ 0
(fl) 2y)
Ky(gH1)=(l+V')I lYI
L) IL) (3.38.b)
( 0 1 \ 0 2y)
1C y ( gH) =(_/)I_ y 1 + (i+v')
L) IL) (3.3 8.c)
c( g) =- (1— (1+
L) L) (3.38.e)
0
(3.381)
Each Gauss section is divided into a number of areas across which strains and stresses
are monitored, as shown in figure 3.11 for certain cross-sectional configurations. If the
effect of shear strains on material plasticity is neglected, only direct strains need to be
established at the monitoring points:
4
e m, g = d •
SU.i,g
1=1 m,i (3.39)
where e m, g is the direct strain of monitoring point 'm' at Gauss section 'g',
d m,1 =1
d m,2 =_ !Jm
d =—z m
m,3
d =0
m,4 (3.40)
96
A uniaxial stress-strain relationship is employed to establish the direct stresses at the
monitoring points:
Sm g = ae m,g)
(3.41)
For the element tangent stiffness calculation, a material tangent modulus is required:
dSm,g(
E m,g_deg_cjeem,g
(3.42)
The bending and axial generalized stresses at a Gauss section are detennined from the
material direct stresses, while the relation between the torsion generalized strain and stress
is based on the elastic rigidity constant:
n
slF j,g = Arnd,jSm,g for i=l,3 ]
m=1
where m is the area of monitoring point 'm', d is defined in (3.40), and 'n' is the
number of monitoring points at a Gauss section.
Once the generalized stresses are obtained at the two Gauss sections, the element chord
forces can be established from numerical integration of the virtual work equation:
au. \
S j,g I
cfij1g1(4)( au Jshj,g
(3.44)
in which (is) is the integration weighting factor.
97
An alternative expression of (3.44) is:
4 2
Clj= cTj,j,gstj,g
j= 1 g = 1 (3.45.a)
where T is a 6x4x2 matrix which can be derived explicitly from (3.37) and (3.38):
L(401_02)
60
('+/) ('-v's)
CT 1,2,1 2 T122=— 2
L (4e - °23
T211=T212= 60
' T232=T122
(3.45.c)
I4 2y —e
L40
1)')
60
L(402—01)
T4,1, 1=T4 1,2 = 60
' T432=T322
(3.45.e)
98
1=T5 1,2 2
(3.45.0
T64 1 =T6, 4,2 =4 (3.45. g)
____ at.
= C1
4 2 1 (af. " (aT.
I C i,j,g
Sj,gI______
____
i,k aCuk j=1g1cTi,j,g
a Cu k ak
2 (1
4 sr i ] [ a su h l (acTijg
k = :i: cTi,jg( a aCuk )Sj,g]
C j,k au a cUk
j=1g=1[h=1 S h,g
But,
I at.
Sj,g
I
I au
S h,g
g
(aSuh g
au) J=()cTk,h,g
and,
2 (aT. 4
I C
aCuk )Sfj,g=i,k
=1g=1
j
Therefore, the chord tangent stiffness can be obtained from the expression,
4 2 4
i,kfk1,k+(L) Z Z T skj,h,g cTk,h,g
j=lg=1 h=l C i,j,g (3.46)
99
in which fk and 5k are determined according to the following,
fk 11=fk22=fk33=fk44= 15
2 FL
FL
fk 12 = fk 2 1fk 3,4 = fk 43 = --
'F' is the axial force, and all other terms are zero (3.47)
and,
n
X H
Sk j , h , g_m=l mdm,j
. Em,gd
t m,h
for j = 1, 3 & h = 1, 3
8k44 g = GJ
To establish the global response of the cubic element, transformations from the chord to
the global system are applied in accordance with sections 2.1 for plane frames and 2.2 for
space frames. As with the quartic formulation, a 2D simplified version of the cubic
formulation is employed in the chord system for plane frame analysis.
In the calculation of direct stresses at the monitoring points according to (3.41), strains
must be considered incrementally from the last equilibrium state, as discussed in
Chapter 4. This requires storage for material variables (stresses, strains, etc..) at the start
of the load increment and for the current iteration, with the variables of the former
updated only after global equilibrium is achieved.
It has been shown in Chapter 2 that one quartic element per member provides an accurate
representation of elastic beam-columns. For elasto-plastic analysis, however, more than
100
one element per member is often needed. With the plastic hinge approach, two quartic
elements are necessary to model the inelastic buckling of members where a plastic hinge
forms within the member length. While with the distributed plasticity approach, a number
of elasto-plastic cubic elements is required per member so that the spread of plasticity
along the member length is adequately represented.
In the elasto-plastic modelling of structures, the zones of plasticity are often unknown
a priori. Within a conventional finite element approach, this requires the use of a fine
mesh of elements for all members throughout the analysis, thus leading to excessive
computational effort.
In this work, a more efficient approach employing automatic mesh refinement is adopted.
The approach recognizes the accuracy of the quartic formulation in representing elastic
beam-columns, and, therefore, starts the analysis with one quartic element per member.
In the course of load incrementation, mesh refinement is introduced only in parts of the
structure which have developed plasticity. Through eliminating the need to represent
elastic members by more than one element, and employing a fine mesh for elasto-plastic
members only when and where plasticity is detected, this approach provides significant
computational savings. The process of automatic mesh refinement can be applied in the
context of plastic hinge or distributed plasticity modelling, as discussed in the following
sections.
Within this approach, a member is modelled at the start of the analysis by one plastic
hinge quartic element. In the course of analysis, plasticity is checked within the element
length, and if detected, as may be the case with brace buckling, the element is
automatically subdivided into two quartic elements before the analysis is continued.
The check for plasticity within the element length is performed at each load step after
global equilibrium has been achieved. To establish the stress state within the element
length, the calculation of the biaxial bending moments must allow for the effect of the
axial force in the presence of transverse displacements. This effect can readily be
accounted for if the chord system is employed (figure 3.12):
101
(M_—M_)
1)' 2' (Mly +M2\
M(x) -
Jx—F{v(x)+v'(x)}
= 2 L
IM +M \
MZ(x)=(;2)_I 'z
LiJx—F{w(x)+w'(x)} (3.49)
in which v(x), v'(x), w(x) and w'(x) are determined according to (2.16).
To determine the plasticity condition at a section, the plastic interaction formula is used:
CD(M(x). M(x), F) - 1 section at (x) is plastic )
(3.50)
The abscissa 'xd' along the element length with the highest interaction value "CD" is first
established. For the 2D formulation, this can be performed analytically, since the
maximum value of "CD" corresponds to the maximum value of bending moment which is a
polynomial function of 'x' according to (3.49). For the 3D formulation, biaxial bending
makes an analytical solution very difficult Therefore, only 9 abscissas along the element
length are considered, with 'xj' chosen as the abscissa having the highest "CD".
If the interaction value "CD" corresponding to 'x€j' satisfies the plasticity condition of
(3.50), element sub-division is performed in accordance with the following section.
If none of the elements requires sub-division for the current load step, the solution
proceeds to the next step. Otherwise, the current load step is re-applied, so that global
equilibrium according to the new element representation is established.
Plastic hinge elements which are the result of an earlier sub-division process are not
allowed to further sub-divide in the current load step, since the existence of more than one
plastic hinge within the member length leads to considerable numerical difficulties. Thus,
the spread of plasticity within the member length is neglected, and the buckling process is
represented by two plastic hinge elements only, where the location of the intermediate
hinge is determined by the first occurrence of plasticity.
102
3.4.1.2 Element sub-division
The process of sub-division of a plastic hinge quartic element involves the addition of a
new node and two new quartic elements, as shown in figure 3.13.
The only variables associated with the new node are global displacements. These are
determined for the last equilibiium configuration from the deflected shape of the original
element and the global displacements of its end nodes.
For each of the new elements, variables pertaining to the initial and last equilibrium
configurations must be established. These include initial direction cosines, initial
imperfections, orientation of the principal axes at both ends, chord displacements, plastic
hinge deformations, and chord forces. The determination of chord displacements, hence
chord forces, must allow for the nonlinear distribution of the axial displacement along the
length of the original element, as represented by (2.19). This proves to be an important
factor for convergence to be achieved when the current load step is re-applied.
While with the plastic hinge approach element sub-division is necessary only in the case
of member buckling, with this approach mesh refinement is required for modelling the
spread of yield regardless of whether or not buckling is involved.
At the start of analysis, a member is modelled by one elastic quartic element In the course
of load incrementation, yield is checked at the extreme fibres of selected sections along
the element length. if yield is detected, elasto-plastic cubic elements are inserted only in
the zones of plasticity, while the rest of the element is re-modelled with elastic quartic
elements, before the analysis is continued.
The check for yield along the length of the quartic element is performed for each load step
at selected sections. These sections correspond to the Gauss integration positions of the
cubic elements to be inserted if yield is detected, with the configuration of the refined
mesh specified before analysis, as shown in figure 3.14.
The biaxial moments at a section are calculated in the same manner as for the plastic hinge
approach using (3.49). if the combination of the axial force and bending moments at a
103
section results in yielding at the extreme fibres, then a cubic element is inserted at the
corresponding location. If none of the sections have yielded, then mesh refinement is not
performed.
The current load step must be re-applied if any of the structure elements has been
subjected to mesh refinement; otherwise, the analysis proceeds to the next load step.
Elastic quartic elements which are the result of an earlier mesh refmement process are
checked for yield at sections corresponding to the remaining elastic parts of the original
element. Thus, after a mesh refinement process, information must be passed to identify
for the new elastic quartic elements the sections at which yielding is to be checked as the
analysis proceeds.
Within the distributed plasticity approach, mesh refinement involves the creation of a
number of new nodes, new elasto-plastic cubic elements and new elastic quartic elements,
as shown in figure 3.15.
As with the plastic hinge approach, the global displacements of the new nodes are
established at the last equilibrium configuration from the deflected shape of the original
element and the global displacements of its end nodes.
For the new elements, variables corresponding to the initial and last equilibrium
configurations must be determined. For the cubic elements, these include initial direction
cosines, orientation of the principal axes at both ends, chord displacements, stresses and
strains at monitoring points, and chord forces. While for the quartic elements they include
initial direction cosines, initial imperfections, orientation of the principal axes at both
ends, chord displacements, chord forces, and mesh refinement configuration. Similar to
the plastic hinge approach, allowance must be made for the nonlinear variation of the axial
displacement along the original element length according to (2.19).
3.5 Verification
Three examples are presented hereafter to demonstrate the efficiency and accuracy of the
plastic hinge quartic formulation, the elasto-plastic cubic formulation, and the
corresponding automatic mesh refinement approaches. Although the cubic formulation is
104
capable of modelling any material stress-strain law, only the bilinear elastic-plastic model
is considered in the following examples, with further discussion on other material models
given in Chapter 4. All CPU times quoted are for a MicroVax II computer with s3Mb of
memoly.
It has been shown in section 2.4.1 that the quartic formulation can accurately represent
elastic beam-columns with effective length factors (k>O.7). However, the range of (k) is
significantly increased if plasticity is considered.
According to section 2.4.1, beam-column modelling using one quartic element remains
accurate as the long the compressive axial force (F) satisfies the condition:
2it2EI
(—F)^
L (3.51)
If the plastic hinge or distributed plasticity approach of section 3.4 is employed to model a
straight beam-column, the quartic element is subjected to a maximum axial force of:
El
(—F)=Minimum (aA. ()2J
(3.52)
where,
and,
Therefore when plasticity is considered, one quartic element can accurately represent
concentrically loaded straight beam-columns if the condition resulting from the
combination of (3.51) and (3.52) is satisfied:
105
in which (i/3i) is the slenderness ration, and 'r' is the radius of gyration.
As demonstrated in figure 3.16.a, this condition implies that for slenderness ratios below
(v'), the effective length factor (k) may have any value. While the restriction
(k> 0.7) applies only to very slender members which buckle elastically, although lower
values of (k) can still be allowed at the cost of a slight reduction in accuracy, as shown
also in figure 2.15.b.
The effect of initial imperfections, load eccentricity or end moments would be to reduce
the maximum axial force that the quartic element is subjected to. This allows lower
effective length factors for high slenderness ratios, with a rounding effect on the line
separating the zones of accuracy, as shown in figure 3. 16.b.
A two-bay frame is subjected to the action of two vertical mid-span forces and one
horizontal sway force, as shown in figure 3.17.a. The frame is analysed up to plastic
collapse using the plastic hinge and distributed plasticity approaches and employing the
automatic mesh refinement procedure.
The load deflection curves of figure 3.17.b demonstrate good agreement between the
plastic hinge and distributed plasticity approaches. A slight disagreement in the prediction
of the collapse load factor can be observed, which is mainly due to the polynomial
representation of the interaction curve within the plastic hinge formulation. Also, the
over-estimation of the collapse load factor by the rigid-plasticity theory is expected, since
it neglects the effect of an axial force on the plastic moment capacity.
With the plastic hinge approach, the analysis is started using 7 quartic elements, and no
mesh refinement is performed during the analysis since plastic hinges are not detected
within the element lengths. Figure 3.17.d shows the locations of plastic hinges after
collapse, and demonstrates agreement with the rigid-plastic collapse mechanism depicted
in figure 3.17.c.
With the distributed plasticity approach, the analysis is started using 7 elastic quartic
elements, and automatic mesh refinement is performed when and where yield is detected.
106
At the end of analysis, the resulting mesh consists of 7 elastic quartic elements and
24 elasto-plastic cubic elements, as shown in figure 3.17.e.
A comparison between the CPU times of the two approaches (1.5 miii for plastic hinge,
and 5.25 miii for distributed plasticity) demonstrates the efficiency of the plastic hinge
approach which still provides reasonable accuracy if the material response is essentially
elastic-plastic without strain-hardening.
The load-deflection curves of figure 3.1 8.b demonstrate good agreement between the two
adopted approaches up to the point of ultimate capacity. In the post-ultimate range, the
slight disagreement is mainly due to the inability of the plastic hinge approach to account
for the spread of plasticity to the mid-length of the top buckled brace.
With the plastic hinge approach, the analysis is started using 28 quartic elements, and
automatic subdivision of members into two quartic elements is performed when a plastic
hinge is detected within the member length. At the end of analysis, 32 quartic elements
are employed, as shown in figure 3.l8.c.
With the distributed plasticity approach, the analysis is started with 28 elastic quartic
elements, and automatic mesh refmement is performed allowing a maximum of 10 elasto-
plastic cubic elements per member. The resulting mesh at the end of analysis consists of
26 elastic quartic elements and 63 elasto-plastic cubic elements, as shown in figure
3.18.d.
In this example, the plastic hinge approach requires only a quarter of the CPU time of the
distributed plasticity approach (55 mm for plastic hinge, 3 hrs 45 mm for distributed
plasticity), which shows again the advantage of the plastic hinge approach for cases
where the material response is essentially elastic-plastic without strain-hardening.
107
Within the distributed plasticity approach, however, the process of automatic mesh
refinement provides significant computational savings. This is demonstrated by analysing
the jacket structure using a fine mesh of elasto-plastic cubic elements from the start of
analysis. A comparison between the results shows that an automatically refined mesh
provides the same accuracy as that of an initially fine mesh, while saving two thirds of the
required CPU time (3 hrs 45 mm with automatic refinement, 11 hrs 15 mm with initially
fine mesh).
108
'.-. Initial imperfection
Elastic deformed shape
[;J Rigid-plastic hinge
y
p
0 e
kJY J 0ly 2x 2x
I
L4AC
L L+i+iS
Ilic
k
L+AC+A p
My
i'M M \ M
Curve(3):f Y z 1=0
M)
Curve(1): M = f1(F)
/
/ /
0
Fp
F
MZ
Curve(2): M, = f2(F)
(M (M M
109
M
F F
F F
1 <ubic
'i' cubic
/
/
/
, pabolic
F
//
/
1 mz
( m
f3(mym)=Jm+m +'vI
.Jm+m
Figure 3.3 Idealization of interaction functions "f 1 ", "f2" and "f3"
110
Interaction
value p1 Iteration (3): Quadratic
Iteration (1): Linear
I
1 - -
I-
II I
(n)
Convergence: r,
,—Full increment
(1) C2)
0 r r 1 Reduction factor (r)
M
Original state at hinge(1)
M1,
F
:II:III::IIIIIIII:FC
111
+ +' + +
( w 5, i' N , (N 5,2' N 3,2)
Rigid Rigid
Plastic Plastic
Veiy small Active hinge
rotational stiffness Suppressed hinge
112
y (o) y (L) z(0) z(L)
kt O2Y2X
y(o) y(x),v
u ( x)
\ >)' \x) x,u
(:) Before deformation
• After deformation
w (x)
z(0) / a(x)
113
y
L L
,
LL
Gauss integration
cross-section
1 2
x
gHi
z gK2
yL
tz ______
Monitoringarea:R
yr _____
Monitoring area: II
I -section
1
Figure 3.11 Distribution of monitoring points for different cross-sections
114
y
Initial imperfection
v(x) M2
F
1 -
v'(x)
ly
Iz
Initial imperfection
w(x) M
1 Miz w'(x)
Figure 3.12 Variables for plasticity check within the element length
Subdivision Two new
point, elements
Current
configuration
1
New
I
Initial node
configuration
115
x Sections for yield check
I Potential elasto-plastic cubic elements
Original elastic quartic element
Yielding at exueme fibres lasto-plastic cubic elementsN.
Elastic \ Elastic quartic elements /
116
1.0
0.9
08
10.7
0.5
6 7
Slenderness ratio(4J)
1.0
O.9
08
0.7
..
0.6
0.5
6 7
Slenderness ratio
117
2500 2500 > 2500 >
1
- - - . 2 3
200
' 1< >1 In All dimensions in (mm) & (N)
C
CE=210x 10 N/mm2)
J
Cf
iEBeams Columns
Lay=300NI12
Material
1.2
1.0
•
. 0.8
Vertical at node(5) Horizontal at node(4)
0.6
0.01
-0.10 -0.05 0.00 0.05 0.10
Displacement (m)
118
____________ +
Plastic hinge)
Plastic hinges__)
(I_•
It
Figure 3.17.d Modelling of two-bay frame using the plastic hinge approach
Yielding regions)
-,
Figure 3.17.e Modelling of two-bay frame using the distributed plasticity approach
119
(11500) iinperfection
in horizontal plane
(11500) imperfection
in vertical plane
I
j
E=210x 10 N/mn
Vertical legs:
Tubular (1)270 x 6 liT
a ),=300 N/rrirn2
Other members:
Tubular c1)90x3 imn
= 350 N / 2
600
500
100
0
0.00 0.01 0.02 0.03 0.04 0.05
Displacement (m)
120
Plastic hinges I)
Figure 3.18.c Modelling of jacket structure using the plastic hinge approach
Yielding regions)
Figure 3.18.d Modelling of jacket structure using the distributed plasticity approach
121
CHAPTER 4
The uniaxial behaviour of mild steel in its virgin state is essentially linear elastic up to the
point of yield. Under monotonic loading, the characteristic yield plateau is followed by a
region of increased strength due to strain-hardening, as shown in figure 4.1.
When subjected to constant strain amplitude cycling, steel exhibits a response which
always converges to a stabilized saturation loop dependent only on the amplitude of
cycling, as confirmed by the experiments of Feitner and Laird (1967) carried out on high
stacking fault energy metals to which structural steel belongs. The curve joining the tips
of the stabilized cycles is known as the cyclic stress-strain curve, as shown in figure 4.2.
Due to the difference in shape between the cyclic and virgin curves, the transient response
under constant strain amplitude cycling is characterized by softening for small amplitudes
and hardening for large amplitudes. Relaxation to zero mean-stress accompanies the
process of cyclic softening or hardening if, during the transient response, the mean-stress
attains a non-zero value, as demonstrated in figure 4.3.
Therefore, a reasonably accurate cyclic model for steel must include the ability to
represent the virgin response, the steady state cyclic response, and the transient behaviour
involving softening, hardening, and mean-stress relaxation. Such a model must also be
capable of representing random cycling without spurious prediction of the real response.
122
4.2 Material Models
Several of the existing material models for structural steel were developed to represent
cyclic behaviour under generalized loading in the multiaxial stress-strain space. In the
current study, such models are considered only in their uniaxial context, and comparisons
are made with other models exclusively developed for the uniaxial case.
Dafalias and Popov (1976) discussed the relative advantages of kinematic and isotropic
hardening models. The authors pointed out that purely isotropic hardening is not suitable
for modelling stress reversals with Bauschinger effect (Timoshenko 1956), since such
type of hardening is not associated with any direction of loading. They also presented a
multiaxial model employing the concept of a bounding surface always enclosing the yield
surface, and assumed that the proximity of the two surfaces in the course of their coupled
translation and deformation during plastic loading determines the value of the plastic
modulus. A purely kinematic case of the model was suggested for uniaxial loading, and
the calculation of the model parameters from experimental results was demonstrated.
Despite the computational advantage of such a model, it does not represent accurately the
virgin response, and it predicts a sudden reduction in the elastic stress range once
plasticity is obtained. The latter disadvantage is mainly due to neglecting the isotropic
hardening, and, hence, the independence of the Bauschinger effect from the stress-strain
history.
Petersson and Popov (1977) presented an extension and a refinement of the Dafalias-
Popov model, motivated by the spurious prediction that the model exhibits if unloading
and reloading are involved without substantial plastic flow in the opposite sense. The
authors pointed out that this deficiency can be remedied by the use of a number of
intermediate surfaces between the yield and bounding surfaces, and presented a new
model based on this multisurface approach. The multisurface model is capable of
representing the virgin response of steel. Through the use of a weighting function applied
to the virgin and cyclic stress-strain curves, it also takes account of the gradual transition
from virgin to cyclic behaviour as a function of accumulated plasticity. The ability of the
multisurface model to represent cyclic softening, cyclic hardening and mean-stress
relaxation will be demonstrated in subsequent sections of this chapter.
Santhanam (1979) presented a uniaxial piecewise linear model employing two parameters
for cyclic stiffness degradation and yield stress growth. Despite its simplicity, this model
has shortcomings in that it is unable to represent the basic cyclic phenomena of softening
123
and mean-stress relaxation. Also, it has an inherent disadvantage of predicting spurious
behaviour if unloading and reloading are involved without substantial plastic flow in the
opposite sense, similar to the model of Dafalias and Popov.
Tseng and Lee (1983) employed a two surface definition similar to that used in the
Dafalias-Popov model, but assumed only isotropic hardening for the bounding surface
which expands upon contact with the enclosed yield surface. The authors suggested that,
for structural steel, the hardening of the bounding surface is obtained from the virgin
response, while the size of the yield surface is allowed to shrink with the accumulation of
plasticity to reflect the dependence of the Bauschinger effect on stress-strain history. They
also demonstrated the ability of the proposed model to represent the phenomena of cyclic
softening, cyclic hardening and mean-stress relaxation. However, an implication of the
use of isotropic hardening for the bounding surface is the dependence of the stabilized
constant strain amplitude cycles on the stress-strain history, which contradicts
experimental findings. Also, the proposed model can predict spurious results if unloading
and reloading are involved without substantial plastic flow in the opposite sense.
Cofie and Krawinlder (1985) proposed a uniaxial cyclic model employing the Dafalias-
Popov relationship for the determination of the plastic modulus. The authors suggested
that the bounding lines are updated upon stress reversal so that cyclic softening, cyclic
hardening and mean-stress relaxation can be accounted for. While experimental results
indicated that softening, hardening and mean-stress relaxation factors depend on the
number of excursions already performed and the cyclic strain amplitude, the authors
adopted constant values for these factors to avoid modelling complexities. In a discussion
by White and McGuire (1987), it was suggested that improvements to the model are
needed so that Bauschinger-type effects become dependent on the stress-strain history,
and a variable factor in the Dafalias-Popov equation is used such that small amplitude
cycles approach the bounding lines more abruptly.
124
Filippou et al of the Menegetto-Pinto model involved the addition of an isotropic
hardening parameter, such parameter does not relate directly to the gradual softening or
hardening behaviour observed under constant strain amplitude cycling.
In this work, the bilinear model with kinematic hardening and the Petersson-Popov
multisurface model are implemented within the elasto-plastic cubic formulation discussed
in Chapter 3. The versatility of the cubic formulation, however, allows the inclusion of
any other uniaxial stress-strain law.
Within this model, the elastic range remains constant throughout the various loading
stages, and the kinematic hardening rule for the yield surface is assumed to be a linear
function of the increment of plastic strain, as shown in figure 4.4.
Mathematically, the calculation of the current stress state can be expressed according to
this model as:
a = a° (a°-a -a° (a°+a -a°
)^(e_e0)4
E E
a = a° + E (e - €0) (4.1.a)
(a°+a -a°
E if(c—e°)>(
E
a=a+
(4.1.b)
a°- a - a°
a=ao+E(e_ c_ (a°-a -a°
E
E
a=a-
(4.1.c)
125
a: current centre of elastic range (centre of yield surface)
superscript (o): denotes values at the start of increment.
The tangent modulus, required for the element tangent stiffness calculation, is dependent
on the current stress state, and is given by:
da if(a_c) <a<(a+a)
dc
The uniaxial case of the multisurface model proposed by Petersson and Popov (1977) is
adopted to represent the cyclic phenomena of mild steel. A number of surfaces enclosing
the yield surface are each associated with a value for the increment of plastic strain, and
the expansion/contraction (isotropic hardening) and translation (kinematic hardening) of
these surfaces are governed by pre-specifled hardening rules.
At the start of a strain increment, the configuration of the loading surfaces may be as
shown in figure 4.6.a:
If the application of a strain increment results in plastic strain, the loading surfaces
expand/contract and translate in the direction of loading, as shown in figure 4.6.b:
L p') '
cz 'c . =a0 (c . +
'
{icc(1 +I)-4c)} (4.4)
where, the (±) depends on the sign of the plastic strain increment ba,, and:
126
Petersson and Popov suggested that isotropic hardening of the loading surfaces is
expressed as a function of accumulated plasticity. A weighting function dependent on the
cumulative plastic strain is hence applied to the virgin and cyclic stress-strain curves to
obtain the initial and current sizes of the loading surfaces (figure 4.7):
o (€ .= (c .+ w() { 1c a(c pi )_
pi) bpi) lc b(e Pi)}
+W( p) {Ka(Cpi)_Kb(Cpi)} ]
I pi) b'¼ pi) (4.6)
in which,
W( = weighting function
In the process of updating the centres of the loading surfaces according to (4.4), a
continuous representation of the surface sizes (ic°) as a function of (Cpi) is required so that
any increment of plastic strain (b.c) can be allowed. From (4.6), this can be guaranteed if
a continuous functional representation is adopted for the virgin and cyclic stress-strain
curves, as well as the weighting function.
In this work, each of these curves is represented by five cubic polynomials over five
adjacent intervals and a straight line at the end, as shown in figure 4.8. Conditions of
127
continuity of values and slopes are enforced at the common points to obtain the
parameters of the cubic polynomials and the straight line.
For an increment of strain (ie), the post-yield loading curve to be considered is given by:
(C .=a0 (e .'±ic0(c
aPpij vpij pi (4.8)
The linear elastic stress-strain relationship is employed if the cuirent stress state remains
within the yield surface:
However, if the condition of (4.9) is not satisfied, plastic strain is induced, and the post-
yield curve of (4.8) must be employed. Hereafter, the stress state is assumed to lie
initially on the the yield surface in order to simplify the following equations. When the
stress state is initially within the yield surface, the part of (b.c) required to bring the stress
state to the yield surface is first obtained, before the rest of the increment (b.c) is applied.
The total strain increment consists of elastic and plastic increments, and, hence, the
following relations hold:
b.E=
a=ar(AEp) (4.10)
This represents a system of three equations with three unknowns: (a), (.c) and (.cp).
However, since (ap) can be a cubic function of (p), Or(Cpi), according to the previous
section, an iterative procedure is required to obtain the exact solution of (4.10), which
may lead to an excessive computational effort. In this work, a cubic polynomial fit of (a)
as a function of (Ac) is employed in order to avoid iterations, as shown in figure 4.9,
and the parameters of this polynomial can be obtained from its values and slopes at the
interval extremities:
128
O p( L p2) — a(
1E=LE +
2 p2 E (4.11.a)
1
dE( i) — i 1
- do
dEp (z.c1)
da A - 1
1
da
:PP2) (4.11.c)
Since the actual virgin and cyclic stress-strain curves are modelled using a number of
polynomial functions according to the previous section, the exact solution of (4.10) does
not guarantee exact correlation with the actual response. For this reason, no effort is put
into obtaining the exact solution of (4.10), and the fitting procedure employing a cubic
polynomial whose parameters are given by (4.11) is found to give reasonably accurate
results while avoiding excessive iterative computations.
The tangent modulus is required in the calculation of the element tangent stiffness matrix.
If the current stress state is within the yield surface, the linear elastic modulus is used.
Otherwise, the tangent modulus can be determined from the elastic and plastic modulii.
This is established through considering that the infinitesimal increment of total strain is
composed of elastic and plastic parts resulting in the same infinitesimal increment of
stress when multiplied by their respective modulii:
129
da E
if lol <1a(0)1
da_ 1 if
1
E da
(4.12)
where the function "ar" represents the post-yield loading curve as defined in (4.8).
4.3 Verification
Three examples are presented hereafter to verify the implementation of the bilinear and
multisurface models in the new nonlinear analysis program 'ADAPTIC. Comparisons
are also performed between the two models in terms of realistic representation of actual
material behaviour and computational efficiency.
The predictions of the multisurface model under constant strain amplitude cycling are
studied for a material with virgin and cyclic stress-strain responses as in figures 4.10.a
and 4.10.b. The transition between the virgin and cyclic behaviour is governed by the
weighting function dependent on the accumulated plastic strain, as shown in figure
4.1O.c.
Under small amplitudes of cyclic strain, the multisurface model is capable of representing
the softening behaviour by virtue of the accumulated plastic strain and its effect on the
weighting function, as demonstrated in figure 4.10.d.
For large amplitudes of cyclic strain, the multisurface model can predict the hardening
behaviour, again through the use of the weighting function, as shown in figure 4.10.e.
130
4.3.2 Fixed ended beam-column
The cyclic axial response of the tubular beam-column depicted in figure 4.1 l.a is obtained
using the multisurface model. For the purpose of comparison with the results of
Mizuno et al (1987), the cyclic stress-strain curve (Kb) is employed from the start of
analysis. The load-displacement curves plotted in figure 4.1 l.b demonstrate reasonable
agreement with the prediction of Mizuno et al, while the slight disagreement at the load
peaks may be due to differences in the degree of accuracy in representing the cyclic curve.
The sway frame shown in figure 4. 12.a is assumed to have the material properties used in
the example of section 4.3.1. The frame is analysed under constant displacement
amplitude cycling, and the response is obtained using the bilinear model (ay=450N/mm2,
p=O.018) and the multisurface model.
The load-displacement curves of figure 4.12.b show that with the bilinear model, cyclic
stabilization is achieved after the first cycle. While with the multisuiface model, stiffness
degradation is observed as plasticity accumulates, and hardening is predicted in the cyclic
stress amplitudes. Although the multisurface model gives a better representation of the
actual material response, a comparison of CPU times demonstrates that the bilinear model
is computationally more efficient (5.5mm for bilinear model, 1 1.25mm for multisurface
model).
131
Figure 4.1 Virgin monotonic response of mild steel
Stress
Yield
Stabilized cycles
Strain
Cyclic stress-
Yield Strain CUTV
132
Softening under small strain amplitudes
Stress
Yield
Strain
Mean-stress at
each half cycle
Stabilized cycle
Mean-stress relaxation
133
Figure 4.4 Loading and unloading paths of bilinear kinematic model
a
(c,a)
at
ITI
(e°, aC)
-a I a
-
I
-I- f--
t(COGO)
-C C
__
134
Figure 4.6.a Possible configuration of loading surfaces before increment of plastic strain
135
C.
1
Virgin curve
C.
1
Cyclic curve
1.0
Ep
Weighting function
136
iCa
Cubic
//1
- a a
c., a
Virgin curve
Linear
Cubic
-
a a a a a
- - -
C.
1
Cyclic curve
1.0
Cubic
Linear
Weighting function
Figure 4.8 Polynomial representation of the basic curves of the multisurface model
137
da
Cubic polynomial ãi(p2)
a(.c1)
dap
ap (0)
Int.(i)
p1 p2
Post-yield curve of stress versus plastic swain increment
a
Cubic fit (exaggerated) AE
2)
a(Ac2)
ludo
-(Ac
dc
138
70(
- -
• - - -
----
500 • - - -
- -1 0.OlEf -
- -
•
j=210x N/mm
200
100. -- - - - _c . -
a a a a - -.'
a - - a
___•'..4.) -
I I I
0.00 0.01 0.02 0.03 0.04 0.05 0.06
C.1
70
- - -
6 •_
% 500
• __--10.OlEb-
E
E 400
300
200
100 -
aa --
a -..
a a______ a -
I' - - - - -
I I -
0.00 0.01 0.02 0.03 0.04 0.05 0.06
pi
139
1.2
_ 4_x_ 4J -4J- _4._
4 _4.;_
- ____
- -.-- -
1.0
0.8
0.6
0.4
0.2
---- -
0.0 • - - -_- - - - - -
140
600
400
' 200
'I
& -200
-400
-600 •
-0.02 -0.01 0.00 0.01 0.02
Strain
600
400
-400
-600
-0.02 -0.01 0.00 0.01 0.02
Strain
141
600
400
'I
-400
-600
-0.02 -0.01 0.00 0.01 0.02
Strain
142
1000N41 1O0ON4
r Section
tI)ll4x 23mm2J
- ..- --
300 _____ _____ _____ _____ I _____ I _____
I I I
I I I
E I I I
E II I I I
200 • 1 03
II I 210 x 1 N i mm
II I I I
II I I I
II I I I
100 •
JD-,
11 I I
II I I I
IIn I I I
114.3 I I 43 I
II I - I I
0. i. I i...i.... I.•.•T•
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040
C p1
Figure 4.11 .a Geometry and material properties of fixed ended beam-column
200
100
-200
-300
) -40 -30 -20 -10 0 10 20 30 40 50
A (mm)
Figure 4.11.b Cyclic axial response of fixed ended beam-column
143
5000 1r3
e'l 200
>1 I<>i
350
t}
Beam
Columns
1
z
-1
-2
-3
-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15
A(m)
144
CHAPTER 5
Consideration is given in this chapter to the effect of support conditions and joint
flexibilities on the nonlinear structural response. This is considered necessary since
modem design methods are shifting towards the inclusion of effects such as semi-rigid
connection behaviour and soil-structure interaction. Whereas the accurate modelling of
these effects requires an extensive study of the various types of connections and
boundary conditions, it is the aim of this work to establish a theoretical basis for such
modelling, verified herein by means of approximate force-displacement characteristics. A
formulation has therefore been developed which employs two coincident nodes, accounts
for geometric nonlinearities, and allows the use of nonlinear force-displacement
relationships. In line with member formulations, geometric nonlinearities of this 'joint'
formulation are dealt with separately for plane and space frames. It is shown that in
addition to its ability to model the effects of connection flexibility, the joint formulation
can also be used to represent support conditions including inclination effects and
structural gaps.
A local system equivalent to the Eulerian system of member formulation is employed for
the 2D joint development. This system follows the deflected position of the joint element,
and its orientation is always defined by that of node (1), as shown in figure 5.1.a.
Three local freedoms related to axial, shear and rotational actions are associated with the
joint element, as shown in figure 5.1.a, i.e:
= (u,v, 0)T (5.1)
145
u=(u 2 —u 1 )cos(p') +(v2—v1)sin(p')
0 =a2—a1
(5.2.a)
where,
p'=p+a1
p: initial orientation of joint element (5.2.b)
The principle of virtual work can be employed to establish the global element forces
( g 1 = IF ,F , F ,F ,F ,F
\X1 yl zzl x2 y2 zz2/j
from the three local element forces
3
1.6
g 1 g1
3au. 3
1.= Z-, f.= T. . 1.
g i j=1 O g U j c J
j=1
3
f .=
g 1
T. .1.
j=1
(5.3.a)
in which,
- C' SI 0
- - C' 0
—u —1
T=
C -SI 0
SI C, 0
0 0 1 (5.3.b)
with,
146
c=cos(p') & s'=sin(p') (5.3.c)
The transformation of the local tangent stiffness Jc to the global element tangent
stiffness gk can be establish by differentiation of (5.3):
at t
3 aT.
___ Ck i,k
gkjj=g=
kagu. + ggUj Ctk
j
g
3 (
k=1
=Ii(3
aCfkaCu m
ki1m=iacum U 1+
g jJ
aT.
i,k
p U c'kI
gj
But,
at Pu
ai. PCUk
k,m Tj,m & aU =6.
Sijk
acum Pu' PuPu
g j gigj
Hence,
3 f( 3
H
g k. .= k=1'hjm=lI i,k dCk,mlj,m}+SGi,j,kC1k)
(5.4.a)
where, and k are 6x6 and 3x3 matrices respectively, and J is a 6x6x3 matrix of
second derivatives of local with respect to global displacements, given by:
= =
2,3, 1 =G 3, 2,1 —5G 3, 5, 1 = S6 5, 3,1
5G 33 1=—u
147
S6 1,3,2 S6 3, 1,2 = S6 3,4,2 _$64, = C
3,2
2=5G
S6 2,3, 3,2,2 =S6 3,5,2 = —6 5,3,2 =
S6 3,3,2 = -
S
6. . =0
i,j,2
for all other (i,j)
(5.4.c)
As for the 3D member formulation, the treatment of geometric nonlinearities within the
3D joint formulation is performed in the context of an incremental approach. This means
that, despite the difference in nature between the local freedoms of member and joint
formulations, the main points discussed in section 2.2.2 apply equally for both
formulation types.
Global displacements are applied incrementally from the last equilibrium configuration,
while local displacements are established in a reference system following the unknown
configuration, as shown in figures 5.3.a & 5.3.b.
The element vectors at node (1) define the orientation of the joint, while the vectors at
node (2) define the axes of bending (figure 5.4). For the last equilibrium configuration,
the two vector sets (1 c °, c ° and (2c0, 2c°' are normal to the vector C° but not
) x
I
necessarily identical, the latter consideration depending on the cumulative twist of the
element.
148
The element vectors of the joint formulation are only affected by the increment of global
rotations:
3 0
r jj
'cc Z'T.
yl i T .1cc
Y j
j=
Z'T1 Zj
3
2c'= 2T . 2c
yl
j=i r Yj
(5.5. a)
in which,
T =T(z1,31,11)
21= T(cz '2''2
r 2 (5.5.b)
The local translational displacements of the joint element are a function of the cumulative
C
global translational displacements d and the joint orientation at end (1), whereas the
increment of local rotations are determined from the inclination of the element vectors at
end (2) relative to the local x-axis vector. However, for the calculation of the increment of
twist (80 ). a fictious vector 21c is defined as the transformation of the y-axis at
end (1) due to rotation increments of end (2). Assuming small incremental values, the
increment of chord displacements 8 can be established according to the following
relations:
149
6u= ( 8 u,8 v,8 w,8Ox,80y,807)T
6 U • cd U°
8 V =lycCcdC_v0
8w=c • d —w0
8O1tClC
3OycdIxcc
60=
(5.6.a)
where,
C 0
T
d =d +((u2_u1),(v2_v1),(w2_w1)) (5.6.b)
y 1
•1 2 T i,jY
1c°
j
(5.6.c)
The current local displacements Cu are obtained by updating the displacements of the last
configuration:
Cu Chj + 8u (5.7)
Once equilibrium is achieved, the bending axes at end (2) should be reset to a position
normal to the current x-axis at end (1). This is performed as follows:
l.o_
x'• x'•
Iy co_1
_yc
c
1co=icc
2o
IX XCc
150
5.2.2 Global forces
As for the 2D case, the transformation of local forces J to global element forces gt S
obtained from the principle of virtual work: (Figures 5.5.a &5.5.b)
cf= (FU,FV,FWMX,MYMZ)T (5.9.a)
I=(F F 1 F ,F
g xi zi xxi,F
yyi
(5.9.b)
6 JJ 6
f.=
g i j=1 O g U j C
1.= T.
J j=i ij Cj
(5.9.c)
The 12x6 terms of matrix I are first derivatives of local with respect to global
displacements, and can be obtained explicitly from (5.5) and (5.6), as shown in
Appendix (A.2.1).
A transformation similar to that obtained in (5.4) for the 2D case can also be derived for
the 3D joint formulation:
6 (1 6
8k1 = k=1 T ik k,m Ij m}
S6 i,J,k cfk)
tm (5.10)
Here, the local tangent stiffness (Jc is a 6x6 matrix. While the geometric term is a
12x12x6 matrix established in Appendix (A.2.2).
151
section 5.1 or 5.2 must be employed depending on whether plane or space frame analysis
is considered.
In this work, it is assumed that the force-displacement relationships in the local system do
not exhibit any coupling effects between the various freedoms. This implies that the local
tangent stiffness Jc is always a diagonal matrix, with each diagonal term obtained from
the force-displacement curve of the corresponding freedom.
The essential cyclic features of this curve are depicted in figure 5.6.b, where the second
line in each quadrant is assumed to be a bounding line for the cycling process. In addition
to its ability to represent approximately the characteristics of cyclic joint behaviour, this
curve can be employed to model the effects of structural gaps, as discussed in
section 5.4.
As for member formulations, the two main requirements for a complete nonlinear global
analysis capability involving the joint formulation are:
Although the joint formulation employed herein assumes decoupling between the actions
of the various local freedoms, the transformations of sections 5.1 and 5.2, derived for
plane and space frames, can be applied to a general formulation in which the coupling
effects are considered.
152
5.3.2.1 Plane frames
For a set of global displacements g U the local displacements are first obtained
according to (5.2). The local forces are then determined using the appropriate force-
displacement curve for each of the local freedoms, and fmally employed in (5.3) to
establish the element global force contribution gf.
To determine the global tangent stiffness gk (5.4) is used with a diagonal local tangent
stiffness Jc the terms of which are obtained from the force-displacement curves of the
local freedoms.
The 3D global tangent stiffness contribution g k is determined from (5.10) using the
diagonal local tangent stiffness dc obtained from the force-displacement curves of the
local freedoms.
5.4 Verification
Two examples are presented hereafter to demonstrate the applicability of the joint element
for modelling the effects of flexible joints, inclined supports and structural gaps.
The frame depicted in figure 5.7.a is composed of three elastic members joined by elasto-
plastic rotational joints (figure 5.7.b), and loaded up to collapse by a horizontal force 'P'.
The support at one end is inclined at a slope of 3:4, and allows a gap of 0.5 m before full
translational restraint is achieved.
Quartic elastic elements (Chapter 2) are used to model the frame members, while three
joint elements are employed to represent the two rotational joints and the inclined support,
153
as shown in figure 5.7.c. Large stiffness values are assumed for the shear and axial
freedoms of elements 'JEl' and '1E2', with the rotational properties as given in figure
5.7.b. While for element 'JE3', the shear properties of figure 5.7.d are employed to
model the 0.5 m gap, a large stiffness is employed for the axial freedom, and a zero
stiffness is used for the rotational freedom.
The load-deflection curve in figure 5.7.e exhibits two distinct parts. The first part
compares favourably with the theoretical collapse prediction assuming an infinite gap, and
represents the collapse behaviour before contact in the gap region is achieved. While the
second part represents the response beyond contact, and compares well with the
theoretical collapse load based on full translational restraint at the inclined support, with a
slight disagreement mainly due to change in geometry. The sequence of deflection of the
frame upon load application is shown in figure 5.7.f, where the rotational deformation at
the member joints and the support response in the gap region are apparent
The elastic space truss of figure 5.8.a comprises three members, of which one has pinned
ends about horizontal axes, and has a parabolic imperfection distribution in the vertical
plane of a very small amplitude (1/10,000). The truss is loaded by a vertical load P
beyond the buckling point of the pinned member.
To model the pin-ended condition of the imperfect member, two joint elements are
employed which have a zero stiffness for rotation about the local z-axis and large
stiffnesses for the remaining five local freedoms. While modelling of the truss members
is effected using three quartic elastic elements, as presented in Chapter 2.
The load-deflection curve in figure 5.8.b shows good agreement with the theoretical
buckling load, and demonstrates the applicability of joint elements to the representation of
pin-ended conditions of truss members.
154
x
y
Y
U2
2'
p4 '
I
S
'1
V2
U
V
1,2 *4
155
F\AU x
y
21FX2
F x1)1
12* tFyi
156
Y
Current unknown •2
I
configuration y
I x
I
z 1 2
2
Last equilibrium
configuration I3iA"1
II
Initial undeformed
configuration
y z
U
oy
'' 0 x
'.-
- 2
--- w
1 1
x
157
2c
Y
Current unknown I
configuration -" y " ,'
xcc
1
yco 2o
Last equilibrium
configuration .
o
Figure 5.4 Joint element vectors for current and previous configurations
158
y4 z
M (FV
- -)* 2 __
4 F ?vt -, -4'. -)*'
-- F M
1 1
x x
Y4j
F
Current unknown
configuration F
F 1 F1,'
hA
F
zl L .-)F
F 721
Initial undeforined
0
configuration -.-4' • -'
1,2 x
159
Figure 5.6.a Parameters of the asymmetric cyclic curve
160
4.0 2.25
4
Joint (1) Joint (2)
E=210x 10 N/m2
3J7'
(All dimensions in (ma)
Cross-section
30
20
10
z
E
J-io
- -20
-30
161
N2 N3
JE1
QE1
Ni
Figure 5.7.c Modelling of frame members, flexible joints and inclined support
4'
3(
2(
K
____2_
-2C
- - - - - - - - -- -
-U.O -U.) -U.4 -U.5 -U.2 -U.! U.0 U.! U.2 U.i U.4 U.) U.O
Displacement (m)
162
25
20
15
z
10
0
0.0 0.1 0.2 0.3 0.4 0.5
A(m)
163
z
P,A
E=210x N/rn2 Horizontal pin-axis
(1110,000) imperfection
in vertical plane
k>I
0.1
Cross-section
Horizontal pin-axis X
I dimensions in () z
4.0
x
2.6
Elevation
Figure 5.8.a Geometric configuration of space truss with pin-ended member
2.0
1.8
1.6
1.4
0.8
0.6
0.4
0.2
0.0
0.00 0.01 0.02 0.03 0.04 0.05
164
CHAPTER 6
In the previous chapters, consideration was given to the static nonlinear behaviour of
structures, where the duration of loading is orders of magnitudes longer than the structure
natural periods. When the rate of loading is high, or its frequency content is comparable
to or exceeding the range of natural frequencies of the structure, dynamic effects due to
inertia and damping become significant, and the structural mass and damping must be
properly modelled for an accurate prediction of the response. In this work, separate non-
structural dynamic elements are employed alongside the structural formulations of the
previous chapters, and numerical time-step integration of the equation of motion is
performed on the element level to obtain the nonlinear dynamic response of the overall
structure.
The available approaches for the nonlinear dynamic analysis of framed structures can be
classified according to the adopted method of integration of the equations of motion, and
the type of modelling used for the structural mass and damping.
165
involve the factorisation of the structure stiffness matrix, but are in general
unconditionally stable for linear systems. The latter property allows the use of
significantly larger time-steps than those permitted by the explicit schemes, only that a
limit on the time-step size is often required for accuracy.
Modelling of the structure mass within the fmite element method is preformed using
lumped or consistent idealizations. Whereas different approaches can be adopted in the
mass lumping procedure, the consistent mass idealization employs the same shape
functions used in the interpolation of the finite element displacements field. On the other
hand, equivalent viscous damping is often used to model the damping characteristics of
the structure, related to friction between sliding surfaces and hysteretic behaviour due to
material inelasticity. However, if the effects of material inelasticity are accounted for in
the static force-deformation response of the structure, the use of equivalent viscous
damping to model such effects can be avoided.
Wen and Farhoomand (1970) adopted an explicit time integration scheme for the dynamic
analysis of inelastic frames. The authors pointed out that "consistent" mass modelling,
while involving a computational effort significantly higher than lumped mass modelling,
does not necessarily lead to a commensurate increase in accuracy. For this reason, a
lumped mass idealization was employed, in which translational as well as rotational
inertia were considered.
Oran and Kassimali (1976) extended earlier elastic frame formulations (Oran, 1973-a,
1973-b) to include a dynamic analysis capability based on Newmark's implicit time
integration scheme (Newmark, 1959). The authors justified the use of a lumped mass
idealization by pointing out that the effects of rotary inertia lead to complications in the
adopted Eulerian formulation.
Adeli et al (1978) performed comparative studies between six explicit and implicit time
integration schemes applicable to nonlinear structural dynamic analysis. The authors
noted that the unconditional stability of Newmark (average acceleration) and Houbolt
implicit methods cannot be guaranteed in nonlinear problems. Using the results from the
nonlinear dynamic analysis of a plane stress plate, the authors concluded that among the
explicit schemes the central difference method is the best, while among the implicit
schemes the Park stiffly-stable method is somewhat better than the Newmark method
with Houbolt's procedure rated third. However, it was emphasised that any such
comparisons depend on the particular problem being investigated.
166
Remseth (1979) discussed three methods of nonlinear dynamic analysis based on the
Newmark time integration scheme. The first is the direct step-by-step method in which
the time integration operates on the whole structure freedoms without transformation to
another coordinate basis. The second is based on mode superposition, in which
eigenmodes and frequencies are obtained for each time step and employed to decouple the
equations of motion of the structure freedoms by virtue of the orthogonality property.
Time integration is then performed on a number of modes using generalized freedoms.
The third method applies transformation to a number of generalized freedoms on modal
basis using the same eigenmodes for a number of time steps. This results in a set of
coupled simultaneous equations corresponding to the number of chosen generalized
freedoms, thus reducing the size of the problem depending on the number of modes used.
Remseth pointed out that the mode superposition and transformation to generalized
freedoms methods are both capable of neglecting higher modes, but that the latter is more
computationally efficient since it does not involve an cigen-solution at each time step.
Using a number of comparison examples between the direct step-by-step and the
transformation to generalized freedoms methods applied to large displacement elastic
problems, the author concluded that a large number of modes may be needed for the
transformation in the latter method to obtain accurate results, which could lead to a
significant increase in computational time compared to the direct step-by-step method.
The applicability of the transformation to generalized freedoms method becomes
significantly more complex and inefficient when elasto-plastic analysis is considered,
since the eigenmodes continuously change during analysis, thus, making the
transformation process complicated, and requiring eigen-solutions to be obtained
frequently at various stages of the analysis.
167
6.2 Approach Used in the Current Work
In this work, direct step-by-step integration of the equations of motion is adopted, with
each equation representing the equilibrium condition at a structural freedom between the
externally applied load on the one hand, and contributions of inertia, damping and
deformation forces on the other. The contributions of inertia and damping are obtained
through the use of non-structural dynamic elements, which are treated during the solution
process in the same manner as the structural elements employed in determining the
deformation forces. This allows the same solution procedure used for static analysis to be
followed in dynamic analysis, provided that the contribution of the non-structural
dynamic elements to the global forces can be established from the increment of global
displacements, and that an equivalent static tangent stiffness matrix can be derived for the
dynamic elements to be used in the iterative solution procedure.
In the choice of the method of time integration, explicit schemes have not been favoured,
since their stability is only conditional. Moreover, their computational efficiency can only
be achieved with a diagonal mass matrix, and the case of zero diagonal terms in the mass
matrix requires special treatment involving the factorisation of part of the stiffness matrix.
Therefore, only implicit schemes have been considered, and the fact that the better
stability and accuracy of some of these schemes cannot be guaranteed in highly nonlinear
problems has influenced the choice of Newmark's method, which is the simplest and
most widely used for nonlinear analysis. This also has the advantage of enabling
comparisons with previous work without the influence of variations in the time-
integration scheme used.
t+t gU
t +{(1_y)gUt+?gUtt}it
gU (6.1)
or, expressing the accelerations and velocities in terms of the displacements at (t+At):
168
t+A I ____ —i) _____\
t+(_1
g U =_()gu -e;• 1g
where,
gut + ' : incremental displacements from the equilibrium configuration at time (t)
and,
The inertia forces contributed by a mass element at time (t+it) can be expressed as:
1t+At_. ..t+ At
g I g H i,j g u.
j
(6.3)
where,
The equivalent static tangent stiffness matrix required for the iterative procedure is defmed
as:
C
ktt_ agr
g j,j a gj
169
t+At au
air
g j
g j
ktt_ _t+Lx
g j,j -
u.
agj agj u'
But according to (6.3) and (6.2),
a ft+At
gj
a ..t+.At M
i,j
gj
and,
- t
a g j
1
agut At2
hence,
t+At 11 )
gkjj M..
lj3A t 2 g i,j
(6.4)
It is noted that in the above derivation of the equivalent static stiffness gk . the element
mass matrix is assumed to remain constant within the time step (At). Although the
effects of large displacements lead to a continuously varying gM especially for
distributed mass, such effects can be accounted for by updating gM at the end of each
time step, as discussed in section 6.3.2.
Hysteretic damping effects are accounted for within the structural elements of the
previous chapters, where the force-displacement relationships define the hysteresis loops.
However, other damping effects (e.g. due to friction) are represented by means of
equivalent non-structural viscous damping elements, whose contribution to the global
forces at time (t+At) can be expressed as:
170
.t+At
g1j - • gC.
i,j• gU.
•j
(6.5)
where,
gft +
: contribution of damping element to global forces at time (t+&)
g C : element damping matrix dependent on the element type as in section 6.3
The equivalent static tangent stiffness matrix required for the iterative procedure is defined
as:
t+A t
ktt_ agr
g j,j a gj
agj
I a gj
u.
gkjj = t+AtX
a gu. j agju'
; ft+t
a .t+t g Ci,j
gj
and,
• t+ L I
a gj__
Ut
agj
hence,
kt1_t'_Y ")gCjj
g j,j (6.6)
171
6.3 Dynamic Elements
In this work, four dynamic formulations are employed, namely, lumped mass, cubic
formulation for distributed mass, dashpot damping, and formulation representing
Rayleigh damping. The derivations are presented hereafter for the 3D case only, from
which the 2D case can be readily extracted.
This formulation is intended to represent concentrated mass with no rotational inertia. The
global mass matrix gM is, thus, a 6x6 matrix given by:
gM11=gM22=gM33=M
The cubic mass formulation has 12 local freedoms referred to the last equilibrium
configuration, as shown in figure 6.1.b:
T
u= 'd ,d ,d 1 ,c
\ xl yl
1ay
C
(6.8)
172
Assuming small values for the local freedoms J, the axial and transverse displacements
along the element length can be expressed using linear and cubic interpolation functions,
respectively:
u=dxl+(d2—d1)(t)
v=[L(a+a 2) —2(d2—d1)](.)3-
1L ( 2a +a
L V zi z2/ I4
y2 —d
yl )](f)2+[L(aJ](t)+d yl
The translational accelerations along the centroidal line can be obtained as a function of
the accelerations of the element freedoms by differentiation of (6.9):
ü= xl
Assuming that the mass is uniformly distributed along the element length and located on
the centroidal line, the local mass matrix M can be obtained using the principle of
virtual work:
'12 L
C1j8j H U &u = f mü8u+m&v+m&wcix
j,jC j) C
=
o
173
12 L
au av ____
M.
i,jcuj=J(mupcu.+mpj.+mfa')th
j1 Cj
0 (6.11)
Jv1 11 —M 7,7 — M
IA M
d ' i,i 6
13M
12,2_M313_fr18,8_M9,9_
35
11ill
M 26 - - M 35 - - M 8 l2 - J' 1 9,lI = 210
9M
M2,8_.M39_. 70
13ML
420
105
ML2
140
J'1.
1,3.M..
C
for all (i,j)
M.
1,3.=0 for all other (i,j)
(6.12.a)
where,
Since the local mass matrix M is derived in a system fixed during the time step (At),
direct transformations dependent on the direction cosines of the last equilibrium
configuration can be employed to obtain the global mass matrix g M (figure 6.2):
12 12
1,3 k=lm=1
T.M
1,k k,mT.j,m
(6.13.a)
in which,
174
Since the cubic mass formulation exhibits identical behaviour about the local 'y' and 'z'
axes, it can be shown that the global mass matrix gM obtained from (6.13) is dependent
O
only on the x-axis direction cosines
.= g M. .
g H.i,j for all (i,j)
j,i (6.14.b)
where,
I:3x3identity matrix
f(1,2)=—f(2, 1)=3
f(23)=—f(3,2)=l
f(3, 1)=—f(1,3)=2
f(1, 1)=f(2,2)=f(3,3)=0
xc= 0 (6.14.c)
175
6.3.3 Dashpot damping element
This is a one-node formulation representing a viscous dashpot damper. Hence, the global
damping stiffness matrix gC is a 6x6 matrix given by:
cx 0 o o 0 0
0 cy o 0 0 0
0 0 cz 0 0 0
gC
0 o 0 0
0 0 o 0 cyy 0
0 0 o o 0 c7-z (6.15)
where,
This formulation is derived in an Updated Lagrangian system, similar to the cubic mass
element (figure 6.2), where a Rayleigh assumption is employed to model the equivalent
viscous damping (Barbat & Miquel Canet, 1989). On the element level, this assumption
can be expressed as:
gC a 1 ( g M) +a2(gk)
(6.16)
where,
In this work, gM is taken as the global matrix of the cubic mass element given by
(6.14), while gk is assumed to be the elastic stiffness of an equivalent cubic Updated
Lagrangian formulation, as given by:
176
12 12
gk k=1 m=1T i,k
j,j = k,mTj,m
(6.17.a)
EA
1 k 17 =k7 = --
12 El
k22=ck28=ck88= L3
6 El),
k 26 =k 2 12 618_—Jc8,12
L2
12EI
Jc33=—k39=Jc99= L3
6E1
k 35 =k 3 ii=c.k5,9=19,ii
L2
ck44ck4i0cki0i0 L
4 El
L
2 El
=
L
4 El
L
2 El
L
k1 =k 1 for all (i,j)
It should be noted that in order for the global structure damping matrix to satisfy the
orthogonality conditions with respect to the vibration modes, Rayleigh damping elements
must be employed so that the resulting structure damping matrix is proportional to the
global structure mass and stiffness matrices. This becomes an essential requirement if the
critical damping ratios associated with each of the vibration modes are to be established.
177
6.3.4.1 critical damping ratios
For linear elastic analysis, use of the same Rayleigh constants for all damping elements
guarantees a global structure damping matrix proportional to the global structure mass and
stiffness matrices, provided that all 'structural' and mass elements are associated with
Rayleigh damping elements. For such case, the global structure damping matrix satisfies
the orthogonality conditions with respect to the initial mode shapes, and generalized
damping coefficients can be obtained for each mode:
c* = a m*+a k*
i ii 2i (6.18)
* *
where c.,m. andk.1 are generalized damping, mass and stiffness coefficients,
1 1
respectively, for mode (i).
The critical damping ratio for mode (i) can thus be obtained by comparing the generalized
damping coefficient in (6.18) to the generalized critical damping coefficient:
C" a m+a k*
i 1 1 21
(,=i= 2f
a1 ____
1 2co 2
(6.19)
where and are the critical damping ratio and natural frequency of mode (i),
respectively.
It is noted that if the damping matrix is proportional to the stiffness matrix (a 1 =0) the
higher modes are heavily damped, while if damping is proportional to the mass matrix
(a 2 =0) the higher modes are lightly damped, as demonstrated in figure 6.3 using plots
of equation (6.19). The case of (a 1 ^ 0 & a 2 ^ 0) guarantees a minimum value of the
critical damping ratio for all modes, and the trend of such ratios can be ascending and/or
descending depending on the Rayleigh constants and the distribution of natural
frequencies for the structure under consideration, as shown in figure 6.3.
With nonlinear analysis, specific modal critical damping ratios cannot be guaranteed
without continuous evaluation of the Rayleigh constants from (6.19), since the natural
178
frequencies are continuously changing. However, since such a process requires frequent
and expensive eigen-solutions, and since there is no guarantee that constant critical
damping ratios reflect accurately the actual structural damping effects, no effort is made in
this work to keep the critical damping ratios at pre-specified levels. Consequently, an
exact correlation of the Rayleigh damping with the structure stiffness matrix is assumed
unnecessary, a reason for not using an elasto-plastic stiffness matrix in the expression
of (6.16).
It is also noted that identical results can be obtained for a Rayleigh damping matrix
proportional to either the mass or stiffness matrix, provided that the critical damping ratio
for the predominant vibration mode is the same in both cases, and that no rigid body
motion is involved in the response. The latter condition is essentially because
proportionality to the mass matrix leads to prediction of damping forces for rigid body
motion, while proportionality to the stiffness matrix does not. Consequently, for the case
of support excitation, where rigid body motion is involved, the damped response of
structures can be sensitive to the nature of Rayleigh damping, as demonstrated in the
example of section 6.5.3.
The modelling of dynamic effects through the use of non-structural dynamic elements has
two main advantages. The first advantage is the ability to employ the frontal solution
procedure (Irons, 1970), which is an element-based procedure for the efficient assembly
and reduction of the global stiffness equations, without any modifications in the case of
dynamic analysis. The other advantage is allowing the freedom to use mass and damping
elements only where the structural mass and damping are significant, and with various
combination of mass, damping and structural elements (i.e. "lumped or cubic mass" with
"dashpot or Rayleigh damping" and "cubic or quartic structural elements").
179
6.4.1 Global analysis
As for the structural formulations of the previous chapters, the two essential requirements
for including the non-structural dynamic elements in a nonlinear analysis capability are:
For mass elements, the increment of global displacements gut + A 'is first employed in
(6.2) to obtain the current element accelerations g U which are in turn used in (6.3)
t+At
to determine the current global forces g f . The global tangent stiffness matrix 8k
needed in the iterative procedure is calculated according to (6.4). The nature of the mass
matrix g M required in (6.3) and (6.4) depends on the element type, and can be obtained
from (6.7) for the lumped mass element, or (6.14) for the cubic mass element.
+AI
For damping elements, the current element velocities 5ü' are first obtained from the
A
increment of global displacements gU + using (6.2), and then employed in (6.5) to
t+At
determine the current global forces g1 . The global tangent stiffness matrix g k
needed in the iterative procedure is calculated according to (6.6). The damping matrix gC
of the viscous dashpot damping element required in (6.5) and (6.6) depends on the
element type, and is obtained from (6.15) for the dashpot damping element, or (6.16) for
the Rayleigh element.
180
hence justified, since the 'consistent' approach is more complicated and not necessarily
more accurate.
In modelling the dynamic behaviour of structures with distributed mass, more than one
cubic mass element per member may be required for accurate results, depending on
whether the modes of vibration contain local member deformation shapes which cannot
be represented by a single cubic interpolation function. Also, the automatic sub-division
process described in Chapter 3 for elasto-plastic analysis is only applied to the structural
elements, since the sub-division of cubic mass elements leads to a sudden increase in the
number of modes and a change in the mode shapes, which results in dynamic
inconsistency between the analysis before and after sub-division. However, preventing
the automatic sub-division of cubic mass elements upon detection of plasticity has serious
implications only if member buckling is involved. In which case, the analysis should be
started with at least two cubic mass elements for each buckled member.
6.5 Verification
Four examples are chosen to demonstrate the applicability of the non-structural mass and
damping elements to nonlinear dynamic analysis, and to establish the relative merits of
different types of inertia and damping modelling in the context of the nonlinear analysis
program 'ADAPTIC.
The elastic cantilever depicted in figure 6.4.a supports a concentrated mass 'M' at its free
end, and is subjected to harmonic excitation, with the response studied for the cases of
zero and 5% critical damping. The results given by ADAPTIC are based on Newmark's
parameters (7= 0.5 & J3 = 0.25) and an integration time step (t = 0.009 sec).
For the undamped case, the non-dimensional response predicted by ADAPTIC compares
favourably at small excitation amplitude (P'=lkN) with the results of linear dynamic
theory based on a one degree of freedom idealization. At high amplitude (P'=lOOkN),
geometrically nonlinear effects become important, and the linear theory can no longer give
accurate results, as shown in figure 6.4.b.
For the damped case, 5% critical damping is introduced in the loading direction using the
dashpot damping element. A similar pattern is observed as for the undamped case, where
181
the comparison is favourable at small excitation amplitude (P'=lkN) between the linear
theory and ADAPTIC. At high amplitude (FlOOkN), the non-dimensional response is
reduced due to geometric nonlinearity effects, as shown in figure 6.4.c.
An elastic clamped beam is subjected to a dynamic step load of 640 lbs (2.85 kN) at its
mid-length, as shown in figure 6.5.a. The response of the beam was obtained by
Mondkar and Powell (1977) using five 8-noded plane stress elements in half the span
with lumped mass idealization, and employing Newmark's integration with y= 0.5,
= 0.25, and At = 50 psec.
An elastic frame, shown in figure 6.6.a, comprises a beam supporting distributed mass of
(1000 kWm), and two columns having relatively negligible mass. The frame is subjected
to two cases of dynamic step-loading, as shown in figure 6.6.b. The first consists of two
equal forces (each 15 kN) applied at the level of the distributed mass, while the second
represents the case of support excitation with a step-acceleration of (6 nzlsec 2). Both cases
are analysed using ADAPTIC and PAFEC, where a critical damping ratio = 5%) is
assumed for the first mode (w 1 = 15.47 rad/sec). This level of damping is achieved
through the use of two different Rayleigh idealizations; the first assumes proportionality
to mass only (i.e. a 1 = 2co & a 2 = 0), whereas the second employs stiffness
proportionality (i.e. a 1 =0 & a 2 = 2 1/a).
182
For the case of force excitation, the two aforementioned Rayleigh damping idealizations
predict identical frame responses. This is mainly because the first mode is the dominant
mode of vibration, while the higher modes, which are damped to different extents with
both idealizations, do not contribute significantly to the overall frame response. The
results in figure 6.6.c demonstrate excellent agreement between the predictions of PAFEC
and ADAPTIC for the two damping cases. Despite the fact that PAFEC deals only with
linear elastic analysis, the small magnitude of displacements for this particular example,
and the absence of the beam-column effect, render the comparison between ADA VflC
and PAFEC legitimate.
For the case of support excitation, the two damping idealizations predict significantly
different responses, as shown in figure 6.6.d. For damping proportional to stiffness, the
frame response is identical to that of the force excitation case, since the step-acceleration
at the level of supports is equivalent to the applied force divided by the mass. However
for damping proportional to mass, the frame drift does not stablilize about the static value,
since rigid body motion is involved, and additional damping corresponding to the rigid
body velocity is induced. Since for most structures damping is associated with relative
rather than rigid body movements, it is essential that damping is not taken proportional to
the mass for the case of support excitation, especially if significant rigid body velocities
are involved.
The tubular jacket structure depicted in figure 6.7.a supports a distributed mass of
40,000 kgfm, and is subjected to the Lefkas (Greece) earthquake of 4 November 1973,
with the acceleration amplitude scaled to 50% of the original time histoty (figure 6.7.b).
The structure is first analysed using a lumped mass idealization (figure 6.7.c), and the
time integration is performed using Newmark's parameters (y = 0.5 & 13=0.25) and an
integration time step (t = 0.02 sec). Comparisons in figure 6.7.d show excellent
agreement between LUSAS and ADAPTIC in the prediction of the elastic response, only
that LUSAS consumes 1 hr 24 mm CPU time while ADAPTIC requires only 6 mm.
ADAPTIC is also employed to establish the elasto-plastic response using the automatic
mesh refinement process discussed in Chapter 3, along with the bilinear material model
described in Chapter 4. Deviation from the elastic response can be observed from about
t = 3.9 sec. at which point buckling is initiated in the bottom part of the structure, as
shown in figure 6.7.c.
183
The effect of mass modelling on the elasto-plastic response is then investigated using two
additional configurations shown in figure 6.7.e, of which one assumes two cubic mass
elements per member in the bottom part of the structure so that effect of member buckling
can be considered. The comparison in figure 6.7.f between the results of the lumped
mass configuration and the two distributed mass configurations demonstrate 5 that, for this
particular problem, the nonlinear elasto-plastic response can be adequately predicted using
the lumped mass idealization. This is mainly due to the negligible mass of the structure
members compared to the super-imposed sway mass, which renders the effect of the
distribution of mass on member buckling insignificant.
Further dynamic analysis verifications are undertaken in Chapter 7. The dynamic facilities
of ADAPTIC have also been used extensively is the context of other research activities.
The extension of the program into the domain of composite and reinforced concrete
structures (Maclas & Elnashai, 1989; Izzuddin et al, 1990) has allowed its verification
against real collapse situations encountered during the Loma Piieta earthquake of 17
October 1989 (Elnashai et al, 1989). Also, ADAPTIC was employed in the design and
verification of a pseudo-dynamic testing facility which was newly developed at Imperial
College (Elnashai et al, 1990), as well as in the study of the effect of random material
variability on the seismic design parameters of steel frames (Maglaras, 1990; Elnashai &
Chryssanthopoulos, 1990). More recently, the program has been applied for investigating
the effect of nonlinear joint behaviour on the dynamic response of offshore jacket
structures (Elnashai & Izzuddin, 1991).
184
Y
2
Current configuration (t + itt)
Figure 6.1 .a Updated Lagrangian local system (x,y,z) of the cubic mass element
y,v d x2 z,w
dx
d
d xl
22d:u
y2
185
Y
2
Current configuration (t + itt)
,.0
x'..
Last equilibrium configuration (t)
1
Initial configuration (t0)
Figure 6.2 Unit vectors of the cubic mass and Rayleigh damping formulations
186
C3
C2
Cl
(01 3 (0.
2 0) 1
Ci
Cl
C2
C3
(01 3 0).
2 (0 1
Ci
I -
a2 ilc/a2
(01 (0 0) (0.
2 3 1
Figure 6.3 Effect of Rayleigh constants on the modal Critical damping ratios
187
M cx
M= 1000 kg
E=210x 10 N/rn2
C,= 1394 N. sec/rn
P(t)=P'sin(lOt) 10.1m
P'L3
3E1 0.lm
Cross-section
-2
-4
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
(sec)
188
0
-2
-4
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
t (sec)
189
—2" 1)1 psi
P(t)
io lb.
640 lb
>1
un 1/8in
oss-section
k >1< >1
Figure 6.5.a Geometric configuration and loading of clamped beam
1.0
0.8
0.6
.-
0.4
0.2
0.0
0 1 2 3 4 5
t (milhisec)
190
5.Om
1<
I 1000kg/rn (= 210x 1O3 N/mmi)
:100 x 100
)lUmflS
am : 75 x 150 mm 2 (RSSJ
E
!fl
15.47 rd/secJ
P(t)t2 P(t)t2
P(t)
30 kN
1L
Case (fl : Force excitation
a (t)
________________
6m/SeC2I
I
a._ a'—
191
0.08. ___________________
0.07. ADAPTIC (2C1co1. 0 )"
• PAFEC
0.06.
(a 1 ,a 2) = or
. 0.05.
( 0 ,tico)
0.04.
0.03.
Stadiift
I
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Time (sec)
Figure 6.6.c Response to force excitation for (mass or stiffness) proportional damping
0.08. ______________
,V°•O)
0.07. - - - ADAPTIC
006 : PAFEC
a = (2 0)
Time (sec)
Figure 6.6.d Response to support excitation for (mass or stiffness) proportional damping
192
5m
m=40,000 kgfm
>1
44
Gravity
g=9.81 rn/sec2
I I Vi
I:c1250x 10mm2
J1:(M50x 5
7800 kg/rn3
E=210x N/mm2 Vi
I I
a=300 N/mm2
Ground motion
0.6
0.4
0.2
0.0
-0.2
-0.4
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
193
m1g
mg
—:; Im
tmy
(m3
mg
m 2g
m2 m2
m
mi= 100, 200 k
I
3 m2=4O1.3kg
I
m 3 = 125.6 kj
0.12 _____
ADAVflC: Elasto-plastic
0.10 • ADAPflC: Elastic
008 - - - - - LUSAS: Elastic
0.06'
'0.04
0.02
0.00
-0.02'
-0.04 V
-0.06
-0.08 . •.i...i.i.,.
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
Figure 6.7.d Elastic and elasto-plastic response of jacket using lumped mass modelling
194
* 1' * *
0.12
0.10
0.08
0.06
002
0.00
-0.08
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
Figure 6.7.f Elasto-plastic response of jacket structure with different mass idealizations
195
CHAPTER 7
A nonlinear static and dynamic analysis program (ADAPTIC) has been newly developed,
based on the nonlinear formulations presented in the previous chapters. The program is
coded in Fortran 77 on a MicroVAX II computer with 13 Mb of physical memory. In this
chapter, the basic features pertaining to the analysis capabilities of ADAPTIC and the
adopted solution strategies are first outlined. This is followed by a brief description of the
program structure, as well as the format employed for data input and results output. A
number of nonlinear analysis examples are finally presented demonstrating the
applicability of the program to real and theoretical problems, and showing, where
possible, its accuracy and efficiency compared to existing software.
In this type of analysis, the applied loading is assumed to be always proportional to a set
of nominal loads, as represented by the equation:
P = A. (7.1)
where,
P : applied loading
nominal loading
A. :loadfactor
Hence, the incrementation of the load factor (A.) results in a proportional change in the
applied loading. Upon reaching the ultimate point due to buckling or plastic failure,
displacement control procedures can be employed to obtain the post-u1timate response, as
discussed in the following sections.
196
7.1.1 Load control
Load control refers to the case in which the load factor (X) is directly incremented, and the
global structure displacements U are determined at each load factor level, as shown in
figure 7.1. This is achieved through an iterative procedure aiming at reducing the out-of-
balance forces 6 to very small values, and employing the global structure tangent
stiffness matrix K:
GS1
1= Op_FS 1
where,
5 current load step
1 current iteration
n number of structure freedoms
6 : out-of-balance forces
F : global structure resistance forces
U : global structure displacements
K 1 : inverse of global structure tangent stiffness matrix
and the values of F and U at iteration (0) are the same as those of the last iteration of the
previous step.
When 6 becomes very small according to a specified tolerance criterion, the iterative
process represented by (7.2) is stopped, and the next load step is applied. If the next load
factor exceeds a limit point of the structure (e.g. X 4 in figure 7.1), convergence to a small
6 cannot be achieved. In such a case, displacement control at one of the structure
freedoms can be employed to obtain the post-ultimate response, as discussed in the
following section.
197
Displacement control is concerned with the determination of the load factor (.)
corresponding to a value for one of the global displacements (U ). This involves an
iterative procedure which can be derived as follows:
s,i fl _ 1 r s,i
OU. = K. k L 8? k+6k]
k=1 (7.3)
If ( u) is the required increment of displacement at the controlled freedom (f) for the
current step, the iterative procedure based on (7.3) can be represented by:
s,i-1 fl
,U J = IC' °F'k
k=1 j,k
s, O\
8u'=us_(u;u1_uf )
8U6gUhI
,j
=
s, I - 1
PU f
198
in which the values of F, U and ?. at iteration (0) are the same as those of the last iteration
of the previous step.
The iterative procedure of (7.4) is similar to that suggested by Batoz and Dhatt (1979), in
the sense that the iterative displacements & g U due to out-of-balance forces and the
displacements PU associated with the nominal load are obtained separately, and then
employed to obtain 3? and 6U.
i I s,i-1 s,i
,j 8U' ögU12 _8gUf11}
6A. =
s,i-1 I
{ pU12 P fi J ) (7.5)
The displacement control procedure, nodal or element based, fails if the load-deflection
curve of the controlled freedom exhibits a snap back phenomenon at any stage of the
analysis. This is demonstrated in figure 7.2, where convergence to the actual response
cannot be achieved when applying the displacement increment (zMJ 4). To remedy this
problem, a procedure is adopted within ADAPTIC in which the controlled freedom is
chosen automatically by the program during analysis, as discussed in the following
section.
s,Ol I s,O
U1
j Ip j (7.6)
199
For element control, a similar condition based on the maximum differential response is
employed to determine the control freedoms (fl,f2):
Trueb determined the magnitude of the increment of displacement (AUS) of the newly
selected freedom from the previous load factor increment (z) 1), and established the
sign of (Us) according to a stability criterion based on the determinant of the global
structure tangent stiffness matrix.
In ADAPTIC, the choice of the magnitude and sign of the increment of displacement
(U9 for the newly selected freedom (f) is based on maintaining tangential continuity in
the control parameter of the previous (failed) procedure. This is performed through the
use of a load factor increment (A)) equivalent to the increment of the previous control
parameter. Hence, for the cun-ent automatic control procedure,
I s,O
S
P f nodal displacement control at freedom (f)
(Xs_ 1 s_2)
load control
It is also recognised that the choice of a constant increment (AUS) for all the steps of the
current automatic control procedure may be numerically inefficient, especially if larger
increments can be applied at the later stages of the procedure without causing convergence
200
difficulties. For this reason, an approach applicable only to the automatic control
procedure is adopted, in which the ratio of the maximum allowable iterations to the actual
performed iterations in the current step is employed to adjust the next increment of
displacement (A1J5):
max VI
AUS + 1 .... [integer(4< 1 I AU $
3xi )J (7.9)
where,
The facility to perform time history analysis is included in ADAPTIC, where the applied
loads can vary independently in the time or pseudo-time domain. Mathematically, this is
expressed as:
J ii (7.10)
where,
°P :nominalloadatfreedom(j)
The solution at time (t) involves the determination of the global structure displacements U,
which can be performed through an iterative procedure of the form:
201
t,i-1 t i-1
6. =)..°P.-F.
.1 ii I
-1 t,i-1
8LJ. = K. kG
k=1 '
where,
t current time
1 current iteration
n number of structure freedoms
6 out-of-balance forces
F global structure resistance forces
U global structure displacements
K 1 : inverse of global structure tangent stiffness matrix
It is noted that, although the value of U at iteration (0) is the same as that of the last
iteration of the previous step at time (t-t), the value of F does not satisfy this condition
when dynamic analysis is considered, as discussed in the following section.
The iterative procedure of (7.11) is stopped when 6 becomes small according to a certain
tolerance criterion, and the next equilibrium configuration at time (t^At) is considered.
Unlike the case of proportional loading discussed in section 7.1, static time history
analysis does not allow for crossing limit points, since the load levels are governed by
load factors which are prescribed in the time domain, and thus, numerical failure occurs if
the applied load exceeds a limit point. However, such a numerical failure is unlikely in
dynamic analysis, since if the static limit point is exceeded, the excess load is balanced by
inertia and damping forces.
The only difference between static and dynamic time history analysis is that in the latter
non-structural mass and damping elements contribute to the global structure resistance
forces F and tangent stiffness matrix K. Hence, the assembly procedure for F and K in
static analysis can symbolically be represented by:
202
F=
C
K=k
C (7.12)
where,
F gf + gf+ Egf
C m d
K= k+ k+ k
d8 (7.13)
where,
This leads to another important difference between static and dynamic analysis, related to
I i3O\
the global resistance forces at iteration (0) of the current step F 1. Since the global
I t,O\
displacements at iteration (0) of the current step U / are the same as those at the last
I t-b.t,l\
iteration of the previous step ",,U I , the global forces of the structural elements are
the same, i.e:
1,0 = All
gf
(7.14)
203
Hence, for static analysis, according to (7.12), the global structure forces are the same:
t,O t—i1t,1
F =F (7.15)
However, in dynamic analysis, the contribution of inertia and damping forces depend,
according to (6.3) and (6.5), on accelerations and velocities rather than displacements,
i.e:
gf=('gMjkgUk) ]
m m
g1= ( g C j gUk)
d dk (7.16)
Noting that in (6.2) gU represents incremental global element displacements from the last
equilibrium configuration, the following relationships can be established between
accelerations and velocities at iteration (0) of the current step and those at the last iteration
of the previous step:
• t1,O .t-&,l
gU = gU +gU
1,0 t-&,1
g U = g U +AgU
] (7.17)
where,
(1
i g U_jjJ gu _()gu
(7.18)
204
Combining (7.13) and (7.16):
t,O 1)
F = Zgf + H jg U )+(gCjkguk
e m(k
+Z(ZgM. kAgUk)+(gCj,kghik)
t,O t-.&,l
F = F + gMj k A g U k )+ (gCjkAgUk)
(7.19)
Hence, for dynamic analysis, the global structure resistance forces must be updated at the
start of each time step by contributions from mass and damping elements, as given by
(7.19).
205
With the NR method, the stiffness matrix K is formed for all iterations, as demonstrated
in figure 7.3 for a single degree of freedom system. This method exhibits high
convergence characteristics, although the formation and reduction of the stiffness matrix
K can be rather computationally expensive.
While with the mNR method, the stiffness matrix K is only formed at iteration (0) of each
load step, as shown in figure 7.4. Although the computational effort relating to the
formation and reduction of K is significantly reduced in the mNR method compared to the
NR method, this advantage can be out-weighed in some cases by the increase in the
number of iterations required to achieve convergence.
A general strategy, of which the NR and mNR methods are special cases, is adopted in
ADAPTIC, where the number of stiffness reformations within a load step is pre-
specified, as shown in figure 7.5 for the case of one stiffness reformation. Best
convergence results are obtained for stiffness reformations at the initial iterations, where
the change in stiffness between two consecutive iterations in more significant than for
iterations close to the converged solution. Depending on the load step and the number of
initial stiffness reformations chosen, this iterative strategy combines the high convergence
characteristics of the NR method, as well as the computational savings of the mNR in the
formation and reduction of the stiffness matrix K.
Two alternative definitions are used in ADAPTIC for the description of convergence to
the equilibrium solution. The first is based on the norm of out-of-balance forces, while
the second employs the norm of the iterative increments of global displacements:
where,
convergence norm
206
S. i
convergence for step (s) at iteration(i) (7.21)
When convergence is achieved, iterations are stopped for the current step, and variables
are updated before the next load step is applied.
The current load step is, however, terminated if one of the following conditions is
satisfied:
or, mathematically,
.s .max
1. 1 >1
max
2. >r for any iteration (i) step (s) tenated]
si+1 s,i 1. .div
3. >r for i ^i } (7.22)
where,
is
: number of iterations before convergence at step (s)
i : specified maximum number of iterations
S. i
convergence norm after iteration (i)
207
max :
maximum allowable convergence norm
1div :
iteration after which divergence is checked
While condition (1) prevents excessive iterations within a load step, condition (2) is
employed so that excessive divergence from the true solution, which could lead to
arithmetic faults, is avoided. The check for divergence in condition (3) is performed for
all iterations after ('), thus disregarding temporary divergence which may occur in
the initial iterations.
When a load step is terminated due to one of the conditions of (7.22), the structure
variables are reset to their values at the beginning of the increment, and the load step is
reduced and re-applied using 'nR' sub-steps. This process is repeated if one of the sub-
steps is in turn terminated, only that the step reduction process is allowed to continue for
up to three reduction levels. Hence, if the original step is assumed to be on level (0), the
reduced steps at the various reduction levels are given by:
0 Original
1 Original/na
2 Originai/4
3 Originai/4
Once in a level other than level (0), the solution procedure always goes back to the
preceding level after the successful completion of the 'ER' sub-steps, and ultimately
returns to level (0) if convergence difficulties do not arise. This process thus provides a
means for avoiding local difficulties in predicting the nonlinear response, through local
refinement of the original step by up to a factor of (n).
Once in the vicinity of a limit point, the load step is terminated at level (3) due to
conditions of (7.22). For the case of time history analysis, the solution is hence stopped
since, as discussed in section 7.2, this analysis type does not deal with crossing limit
points. For the case of proportional loading, however, a different control phase
208
(displacement or automatic) can be started, thus allowing the post-ultimate response to be
obtained.
The global structure resistance forces F and tangent stiffness matrix K, required for the
iterative solution procedure, are assembled from element contributions. The assembly
process depends on the global freedom numbers associated with the end nodes of each
element, and is symbolically represented by:
F=Zf
C
K=Zk
C (7.23)
where,
The global element forces g f and tangent stiffness g k are obtained according to the
element type, as indicated in the appropriate sections of chapters 2, 3, 5 and 6, listed in
the following table:
Element type
fi ( gt & g k): Section
Quartic elastic 2.3.4
Quartic plastic hinge 3.2.4
Cubic elasto-plastic 3.3.8
Joint 5.3.2
Lumped mass 6.4.1
Cubic mass 6.4.1
Dashpot damping 6.4.1
Rayleigh damping 6.4.1
It is noted again that the element global displacements g IJ are absolute for 2D analysis
and incremental in 3D analysis, as discussed in chapter 2. Hence,
209
gU C U (2D analysis)
The reduction process of the assembled global stiffness mairix K, symbolically denoted
by K 1 in (7.2), (7.4) and (7.11), can be performed using Gaussian elimination operating
on the whole matrix after the assembly of all the structure elements. However, this proves
to be costly in terms of reserved storage for K, as well as the number of operations
involved in reducing K to an upper triangular matrix.
For this reason, the frontal method, due to Irons (1970), is adopted for the assembly and
reduction of the global stiffness equations. The method utilizes the feasibility of
eliminating the freedoms of a node just after the assembly of all the elements connected to
it, and makes use of the fact that eliminated freedomsslo not need to be considered in the
further stages of the elimination process. This means that a subset of the stiffness matrix
K, commonly referred to as the Gra,cLpa, can be focused upon, where the size and
contents of the Grandpa change continuously during the assembly process. As the
element stiffness matrices are assembled, the Grandpa accommodates the associated
active nodal freedoms. Once all the elements connected to an active node have been
assembled, the freedoms of this node can be eliminated, and the node is made inactive.
After elimination, the size of the Grandpa reduces, and increases only when the further
assembly of elements brings new nodes which do not already have allocated storage in
the Grandpa. The active nodes at various stages of the assembly process are referred to as
the nodal front, hence the name frontal method.
The size of the Grandpa, or the nodal front width, determines the efficiency of the frontal
method in terms of storage and number of operations, and is significantly dependent on
the assembly order of the elements. This is demonstrated for a two storey frame in
figure 7.6, where the use of assembly configuration (2) instead of configuration (1)
results in an increase of 4 nodes in the maximum nodal front.
The significant effect of the order of element assembly on the front width, and hence on
the efficiency of the frontal method, requires that an optimum assembly order is obtained
before the solution process is started. This can be a tedious task for the user, especially if
the structure under consideration is complex, or a large number of elements is involved.
210
For this reason, an automatic ordering procedure, resulting in optimum or close to
optimum assembly configurations, has been developed and included in ADAP'flC.
For elements with two nodes (e.g. member and joint formulations), the rank of an
unassembled element is taken as:
Whereas for one noded elements (e.g. lumped mass and dashpot damping formulations),
the rank of unassembled elements is taken as:
Priority is given in ranks (2, 3 & 4) to elements whose active node appeared first during
the assembly process. Any tie between elements having the same rank (4 or 5) is broken
by chosing the one with the least number of elements connected to its inactive nodes.
This simple procedure proves very effective as demonstrated in figure 7.7, where the
same front widths are obtained as the more involved method of Sloan and Randolph
(1983) employing concepts from graph theory.
211
7.3.2.2 Considerations related to element sub-division
The choice of the frontal method, as opposed to methods employing matrix bandwidth
optimization, has been significantly influenced by its versatility and efficiency when
considered in the context of the automatic element sub-division process of Chapter 3.
Since such a process involves the creation of new elements and nodes during analysis,
the bandwidth approach would require a major re-optimization process, as that developed
by Gibbs et a! (1976), whenever element sub-division is performed. This is due to the
fact that bandwidth optimization is achieved through re-numbering the nodes according to
a specific sequence, and that this sequence may need to be completely altered if an
optimum bandwidth is to be maintained after the creation of new nodes. The frontal
method, on the other hand, employs an element ordering procedure, as the one discussed
in the previous section, to achieve an optimum frontwidth. Such a procedure does not
need to be re-activated when element sub-division is performed during analysis, since the
new order of element assembly can be directly obtained from the previous one by
assuming that any new elements are assembled sequentially and in place of their original
element. This assumption can lead to a maximum increase in the frontwidth of only two
nodes, regardless of the number of original elements sub-divided or new elements
created. Also, the increase in the frontwidth is only local at the elimination stage of the
newly created nodes, as demonstrated in the example of figure 7.8.
In the iterative procedures presented in sections 7.1 and 7.2, the loading vector P is
assumed to consist of global forces applied at the nodal freedoms. To account for
restrained support conditions, the tangent stiffness matrix K as well as the out of balance
forces vector 6 are modified so that the iterative increments of displacement at restrained
freedoms are always zero. Supports can also have prescribed displacements which are
included in the loading vector P, hence, can vary proportionally to other applied forces
according to section 7.1, or be part of time history analysis according to section 7.2. The
method of accounting for supports in the iterative procedure is based on setting the rows
of K associated with restrained or prescribed freedoms to zero, except for the diagonal
terms which are set to one, and then setting the corresponding terms of 6 to the required
iterative increments of displacement, taken always as zero for restrained freedoms.
Distributed element loads are not considered in this work, mainly because the elastic
quartic formulation (Chapter 2), which is intended to represent whole members, loses its
212
accuracy in modelling beam-columns with only one element per member if significant
distributed loading is applied. Hence, for an accurate representation of distributed loads, a
number of elements per member is required, and since this allows the load to be
approximated by a number of concentrated forces applied at the intermediate nodes, the
individual elements can be assumed free from distributed loading. It should be
mentioned, however, that another factor influencing the decision to neglect distributed
element loads is the complexity of accounting for these loads within the Eulerian system.
As discussed in Chapter 2, the fact that the Eulerian system is a moving system implies
that the transformation of distributed loads to equivalent element end loads and then to
equivalent nodal loads is a function of the unknown element configuration, which can
lead to unjustifiably complicated derivations.
Two distinct but complementary sub-programs form ADAPTIC, as shown in figure 7.9.
The first sub-program (READ) is an interface which operates on the user data and stores
the resulting information in a set of temporary files. These files are then accessed by the
second sub-program (ANALYSE) which uses the stored information to perform the
required analysis.
A sequential file is employed for storing the data related to structural modelling and the
adopted solution procedure. The file format is free horizontally but fixed vertically. A
brief description of the data file is given hereafter with reference to Appendix B.1, while
more detailed information is given in (Izzuddin & Elnashai, 1989).
This module specifies the type of elements whether 2D or 3D, and the type of analysis
whether static, dynamic or eigenvalue, as shown in Appendix B.1.1. The eigenvalue
solution is based on the Lanczos algorithm, as described by Hughes (1987).
In this module (Appendix B.1.2), the different materials constituting the members of the
structure are defined. Material model no. (1) refers to the bilinear kinematic model, while
213
model no. (3) refers to the multisurface model, as presented in Chapter 4 (Models no.
(2), (4) and (5) are concrete models, not discussed in this work).
The various sections and their constituent materials are defined in this module (Appendix
B.1.3). Section type nos. (1), (2) and (3) refer respectively to the three available steel
sections; namely, rectangular solid, circular hollow, and I-sections. Composite and
reinforced concrete sections have also been added to ADAPTIC in the context of other
research programmes.
This module (Appendix B.1.4 & B.1.5) defines the element groups referred to in the
element connectivity module. Each group of elasto-plastic cubic elements (Type (1))
requires the definition of a section no. and the number of monitoring points.
Groups of elastic quartic elements (Type (2)), which automatically sub-divide into
elasto-plastic cubic elements (Type (1)), are identified according to the cubic group they
refer to, and the sub-division pattern they employ. Within each specified pattern, the
number of sub-elements (Type (1)) as well as their length ratios are pre-defined (e.g.
Ratios [1:1:1:1:1] imply five sub-elements of equal length).
Plastic hinge quartic elements (Type (4)), are grouped according to their section no. and
whether automatic sub-division into two elements is required when a plastic hinge is
detected along the element length (Entries '1' or '0' are used to indicate if sub-division is
required).
The groups of 3D elements (Appendix B.1.5) contain the same information as the groups
of 2D elements (Appendix B.1.4), except for the dashpot damping and joint
formulations. The 3D dashpot damping element requires the definition of three additional
damping constants corresponding to the three additional freedoms (One displacement and
two rotations). Also, the 3D joint element requires the definition of three additional curve
types for the additional shear and rotational freedoms. Curve type (1) refers to a linear
elastic force-displacement relationship, while curve type (3) refers to the asymmetric
relationship discussed in Chapter 5.
214
7.5.1.5 Nodal coordinates
In this module (Appendix B.1.6) the coordinates of structural and non-structural nodes
are defined, where the non-structural nodes are only employed to define orientation,
rather than connectivity, for some element types. In 2D analysis, the 2' ordinate, which
is the same for all nodes, is not specified.
Elements are defined in this module (Appendix B.1.7 & B.1.8) according to their type
and the previously established groups. Amongst the 2D formulations, only the joint
element (Type (11)) requires a third node to define the initial orientation of the local x-
axis, since its nodes (1) and (2) are initially coincident.
The 3D member formulations (Types (31) - (34) & (42)) require a third node to define
the plane in which the principal y-axis lies, and the 3D joint formulation (Type (41))
employs two additional nodes to define the local x-axis and the plane in which the local
y—axis lies. It is noted that for the cubic mass element (Type (40)), the orientation of the
principal axes is irrelevant, as discusses in Chapter 6, and hence a third node is not
required.
The initial imperfections of quartic elements are defined in this module (Appendix B.1.9
& B.1.1O) by specifying the out-of-straightness levels at three intermediate positions
along the element length. For 2D elements only one value is required at each position,
while for 3D elements two values are needed to define the out-of-straightness in the local
'y' and 'z' directions.
In this module (Appendix B.1.l 1), global freedoms where permanent restraints exist are
defined by specifying the conesponding node and global direction (Direction is given an
integer value [1-3] for 2D analysis, or [1-6] for 3D analysis).
215
7.5.1.9 Loading
For the case of proportional loading, the nominal loads, which can be applied forces or
support displacements, are defined and the control phases are established. For load
control, the load factor increment and the number of steps in which this increment is
applied are specified. While for displacement control, the control nodal or element
freedom, the displacement increment and the number of steps are defined. An option is
available in load and displacement control to establish the sign of the load factor or
displacement increment so that the loading or unloading path is followed. Alternatively,
the sign of the increment can be kept as specified. For automatic control, nodal or element
control employing translational or rotational freedoms is specified, and the conditions
used to terminate the control phase are established. These conditions can be based on the
range of interest for the load factor or the global displacements at a certain freedom.
For time history static analysis, the applied loads, which can be forces or support
displacements, are defined in the time domain by means of load curves which may be
different for each of the loads. Different time step sizes can be employed during analysis
through the use of the equilibrium stages module.
Time history dynamic analysis is identical to time history static analysis except that
Newmark's time integration parameters need to be defined, and support displacements are
replaced by accelerations in the module of variable loading.
In this module, the iterative strategy, the convergence criteria, and the type of output are
specified (Appendix B.1.13). This includes the maximum allowable iterations (i), the
number of initial stiffness reformations within an equilibrium step, the step reduction
factor (nR), and the iteration after which divergence is checked (1th1.1)• Also included are
the tolerance (e), the reference force or displacement values, and the maximum
convergence norm with a negative tolerance indicating convergence check based
on the norm of iterative displacements. In the output module, the frequency of output as
well as the requirement for output of stresses and strains are specified.
216
7.5.2 'READ'
The sub-program READ performs the preliminaiy tasks required before the start of the
incremental analysis. READ checks the data for clear errors such as reference to a non-
existent element group, inconsistency in element or node numbering, or loading at
restrained freedoms.
READ also obtains element properties from the user data, which includes the calculation
of bending and axial rigidities, the interaction surface parameters for plastic hinge
elements, and the monitoring points areas and centroidal distances for the elasto-plastic
cubic elements. Also included is the determination of the nodal locations where only two
plastic hinge elements meet, hence, locations where plastic hinge suppression may be
required.
After establishing the connectivity of all elements, READ performs the automatic ordering
procedure of section 7.3, and determines the order of element assembly for the frontal
solution. It also determines the storage necessary for the variable arrays employed during
analysis, and allows in its calculation for the possible introduction of new elements and
nodes as the analysis proceeds.
7.5.3 'ANALYSE'
This is the main part of ADAPTIC where the incremental analysis is performed.
ANALYSE uses the information provided by READ in the temporary files, and
determines the type of analysis and the solution strategy to be adopted accordingly.
217
7.5.4 Output files
The first file 'Filename.OUT echos the data file, and includes any warnings or error
messages if inconsistency is detected in the input data. The file then shows the adopted
order of element assembly for the frontal solution, and provides information on the
progress of analysis, as demonstrated in Appendix B.2.1. This includes the creation of
new elements and nodes upon element sub-division, as well as the formation and closure
of plastic hinges.
The second file Pilename.NUM' contains all the numerical results for the specified steps,
as demonstrated in Appendix B.2.2. In addition to global nodal displacements, nodal
velocities and accelerations are output when dynamic analysis is considered, and local
element forces and displacements are presented according to the element type.
7.6 Examples
In this section, the general capabilities of ADAPTIC are demonstrated through a number
of examples, and comparisons are made, where possible, with other work. All examples
are presented in diagrammatic form, with the data and output files of selected examples
included in Appendix B.
Lee et al (1968) performed the static analysis of the frame depicted in figure 7.1 l.a using
a procedure developed for elastic loop frames, and allowing for large bending but no axial
deformation. The frame was also analysed statically by Trueb (1983) using ten quadratic
isoparametric beam elements, and employing an automatic displacement control
procedure.
Two meshing configurations comprising three and five quartic elastic elements
respectively (Figure 7.11 .b) are employed with ADAPTIC to obtain the static response of
Lee's frame. For both cases, the proportional loading case is adopted, and the automatic
displacement control procedure is used to obtain the post-buckling response, as shown in
the respective data files given in Appendices B.3.1.1 and B.3.2.1. The first configuration
218
provides reasonable comparison with the results of Lee et al, while excellent agreement is
given by the second configuration, as shown in figure 7.11.c. The main reason two
quartic elements per member are needed for accurate results is that the frame undergoes
severe bending deformations, as shown in figure 7.11 .d, and the use of one element per
member implies very large rotations in the chord system, thus, not complying with the
assumptions made in Chapter 2. Output files, presented in detail for the first configuration
in Appendix B.3.1.2 and briefly for the second in Appendix B.3.2.2, show the step
reduction process (Levels 1, 2 & 3) employed when convergence difficulties arise, and
demonstrate the automatic selection of the control freedom upon the failure of the
previous control procedure.
In the context of static analysis, the increase of the applied load beyond the limit point
value leads to convergence failure. This implies that dynamic effects would be involved in
the actual post-buckling response if force incrementation is adopted, even if the rate of
force application is very low as to simulate static loading. To demonstrate such effects,
time history dynamic analysis employing force incrementation and assuming a low rate of
loading (1 ton/sec) is performed. In this analysis, the second meshing configuration
comprising five quartic elastic and five cubic mass elements is adopted, and a force of
(4 ton) is applied over an interval of (4 sec), as shown in the data file of
Appendix B.3.3.1. The time integration is performed using Newmark's parameters
(1 = 0.25 & y= 0.5), and assuming a time step (At = 0.004 sec). The response,
shown in figure 7.11 .e, demonstrates the agreement between dynamic and static analysis
in the pre-buckling range, and depicts the dynamic effects in the post-buckling range,
where oscillation about the stable equilibrium position is exhibited.
For structures subjected to a single load, prescribing the displacement associated with that
load often guarantees a static response even in the post-buckling range, on condition that
a low rate of displacement is employed. However, this is not the case with Lee's frame
because of the snap-back phenomenon, which represents a limit point in the displacement
domain. This is demonstrated by performing a time history dynamic analysis with
displacement incrementation, where the displacement is prescribed according to a
triangular acceleration impulse of magnitude (100 cm/sec2) and duration (0.5 sec), as
given in the data file of Appendix B.3.4.1. The response, shown in figure 7.1 1.f,
compares well with the static response before the snap-back, but exhibits high frequency
dynamic characteristics afterwards.
219
7.6.2 Elastic arch
The shallow arch depicted in figure 7.12.a was analysed by Oliver and Onate (1986)
using meshes of six and ten 3-noded isoparametric finite elements. The results from
ADAPTIC, on the other hand, are based on meshes of two and four elastic quartic
elements, where the arch curvature is modelled through specifying appropriate element
imperfections, as shown in Appendices B.4.l and B.4.2. In order to obtain the post-
buckling response, displacement control of the freedom associated with the applied load
is employed. The load-deflection curves shown in figure 7.12.b demonstrate that the
results given by four quartic elements are in excellent agreement with the prediction of
Oliver and Onate based on ten isoparametric elements, while the use of two quartic
elements still provides reasonable accuracy.
The frame depicted in figure 7.13.a is subjected to the static action of vertical and sway
forces, which are increased proportionally up to plastic collapse. Three cases of sway to
vertical load ratios (r=0.1, 0.24 & 0.5) were considered by Kassimali (1983), who
employed a 2D plastic hinge formulation neglecting plastic axial displacements, and
assuming a bilinear interaction curve independent of the section shape. The frame was
later analysed by Kam (1988), who accounted for the spread of plasticity across the
section depth and along the member length.
The results given by ADAPTIC are based on the plastic-hinge quartic formulation, where
favourable comparison is demonstrated in figure 7.1 3.b with the predictions of Kassimali
for the three load cases. The slight disagreement in the region of ultimate capacity is
mainly attributed to the difference in the interaction curves used, since that of Kassimali
does not allow any reduction in the plastic moment capacity until the axial force exceeds
15% of the plastic axial capacity.
7.6.4 K-frame
In this example, comparison is made between the prediction of ADAPTIC and the
analytical and experimental results reported by Soreide et al (1986) on the K-frame setup
shown in figure 7.14.a. Although Soreide et al did not include any information regarding
the level of imperfections, it was found that ADAPTIC provides best comparison with the
experimental ultimate capacity if parabolic imperfections of (111000) are assumed in the
diagonal members.
220
Soreide et al used the computer program USFOS, which employs a beam-column
formulation with sinusoidal or hyperbolic shape functions, and adopting the plastic-hinge
approach to model the effects of yield. The comparison with experimental results in
figure 7.14.b shows that the prediction of Soreide et al overestimates the ultimate
capacity by 17%, while underestimating the response in the post-buckling range.
With ADAPTIC, the plastic hinge approach is first used, where the corresponding data
file is given in Appendix B.5.1.1. The analysis is started with four plastic hinge quartic
elements, and upon detection of a plastic hinge within either of the diagonal elements,
automatic sub-division into two elements is performed, as reflected in the output file of
Appendix B.5.1.2, and as shown in figure 7.14.c. With this approach, the ultimate
capacity is overestimated by 9%, while the post-buckling response is comparable to that
predicted by Soreide et a! (figure 7.14.b).
The distributed plasticity approach is also used with ADAPTIC, where the analysis is
started with four quartic elastic elements, and automatic sub-division into elasto-plastic
cubic elements is performed upon detection of yield, as specified in the data file of
Appendix B.5.2.1. By the end of analysis, 32 cubic elements, employing the bilinear
kinematic material model (p.=0.l%), are inserted in the yielding zones, while the rest of
the frame is still modelled by elastic quartic elements, as shown in figure 7.14.d. This
approach provides more favourable comparison with the experimental results in terms of
ultimate capacity (0.2% overestimate) and post-buckling response prediction, as
demonstrated in figure 7.14.b.
The framed dome of figure 7.15.a was analysed by Remseth (1979) who employed a
Lagrangian beam-column formulation not accounting for large global rotations, and was
later considered by Shi and Atluri (1988) who applied transformations which become
erroneous when large nodal rotations are involved.
The static response to a concentrated load at the crown point is shown in figure 7.15.b,
where excellent agreement is demonstrated between the prediction of ADAPTIC
employing one elastic quartic element per member and the results based on the
formulation of Kondoh et a! (1986). The disagreement with the results of Shi and Atluri,
and more significantly with those of Remseth, can be attributed to the point mentioned
above regarding their consideration of large global rotations.
221
ADAPTIC is also employed to obtain the dynamic response of the dome to harmonic
excitation at the crown point, where one elastic quartic element is used per member, and
lumped mass idealization is assumed. Consideration is given to the damped case, where
damping is taken proportional to the mass, and is based on a critical damping ratio
( = 0.05) for the first mode (T = 0.155 sec), as shown in the data file of Appendix
B.6.1.1. The response is plotted in figure 7.15.c, where direct comparison with the
results of Shi and Atluri can only be qualitative due to the inconsistency of their
formulation regarding large rotations, and since they did not indicate the type of mass
modelling or the level of damping they employed.
The 3D frame of figure 7.1 6.a consists of identical horizontal members with the webs in
vertical planes, and identical columns whose webs lie in the vertical planes containing the
bisectors of the equilateral triangular planform. In addition to the initial static load (P5),
the frame is subjected to a dynamic step load (Pd) consisting of a horizontal force and a
couple with a fixed vertical direction, both applied at one corner node.
Elastic analysis of the frame was performed by Wen and Farhoomand (1970), where the
equivalent nodal mass matrix was obtained by assuming a rigid configuration for the
members at an incident node. The rigid length of a member considered at one node was
assumed to be a quarter of the total length, while the equivalent density was doubled, so
that translational inertia was not affected, and the equivalent rotational inertia could be
estimated. Wen and Farhoomand showed how the rigid length affected the fundamental
period of the structure: T = 0.110 sec for zero rigid length or lumped mass idealization,
T = 0.094 sec for rigid length 25% of the total, and T = 0.084 sec for rigid length 50%
of the total. This is reflected in figures 7.16.b and 7.16.c, where the results of Wen and
Farhoomand for the second case (rigid length 25% of total) are compared with the results
of ADAPTIC employing lumped and distributed mass idealization. The dynamic analysis
is performed with ADAPTIC using one quartic elastic element per member, and a lumped
mass element at each of the unrestrained nodes or one cubic mass element per member,
with the time integration based on Newmark's parameters (3 = 0.25 & = 0.5) and a
time step & of 0.004 sec. Consideration of the results demonstrate qualitative agreement
between ADAPTIC and the prediction of Wen and Farhoomand, with the lumped and
distributed mass idealizations exhibiting longer and shorter periods respectively. To
establish the accuracy of ADAPTIC results, the elastic response is also obtained from
LUSAS using one 3-noded beam element per member, where the results in figures
222
7.16.b and 7.16.c, corresponding to the lumped and distributed mass idealizations
respectively, demonstrate good agreement. It is worth noting, however, that LUSAS
consumed 30 mm 30 sec of CPU time for the lumped mass case and 37 mm 31 sec for
the distributed mass case, while ADAPTIC required only 4 mm ii sec and 5 mm 5 sec
for the two cases respectively.
Wen and Farhoomand also obtained the elasto-plastic dynamic response of the frame
using a plastic hinge approach, where interaction was considered between axial, bending
and torsional actions. On the other hand, the prediction of the elasto-plastic response with
ADAPTIC is based on one quartic plastic hinge element per member, with both cases of
mass modelling and the same time integration process as for the elastic case. The
comparison in figure 7.16.d shows good agreement between the results of ADAPTIC and
those of Wen and Farhoomand, and demonstrates that the total elasto-plastic response is
only slightly sensitive to the type of mass modelling compared to the elastic case.
In this example, the response of the frame of figure 7.17.a assuming elastic and elasto-
plastic member behaviour is obtained for two cases. The first case assumes rigid beam-to-
column connections, while the second employs the joint moment-rotation relationship of
figure 7.17.b. In all analysis cases, the frame is subjected to the San Salvador,
El Salvador, 1986 earthquake, where the acceleration input is scaled down to 50% of the
original time history record shown in figure 7.17.c.
With ADAPTIC, one quartic elastic element per member is employed, while the structure
mass is represented by two lumped mass elements, and the joint flexibilities are modelled
using two joint elements. The analysis is also performed with LUSAS, where each
member is represented by one 3-noded beam element, and appropriate joint elements are
used to model the beam to column connections and the lumped masses. Moreover, the
same time integration scheme (Newmark, 3 = 0.25 & 'y = 0.5) is adopted with both
ADAPTIC and LUSAS, and the same time step (At = 0.02 sec) is assumed.
Comparison of the results in figure 7.17.d shows reasonable agreement between the
results of ADAPTIC and LUSAS for both rigid and flexible cases. The slightly shorter
period predicted by LUSAS is due to the stiffer response of its beam element, which is
based on a cubic shape function for the transverse displacements, and hence cannot
223
predict the P-A effect of the vertical loads as accurately as the quartic element. In terms of
CPU time, LUSAS consumed 36 mm 7 sec for the rigid case and 40 mm 7 sec for the
flexible case, while ADAPTIC required only 1 mm 30 sec and 1 mm 48 sec for the two
cases respectively; a saving of about 95% in CPU time.
With ADAPTIC, the distributed plasticity approach is adopted, where the analysis is
started using three elastic quartic elements, and automatic sub-division into elasto-plastic
cubic elements employing the bilinear kinematic model is performed when and where
plasticity is detected. The structure mass is represented by two lumped mass elements,
while the joint flexibiities are modelled using two joint elements, as shown in the data ifie
of Appendix B.7.1.1.
The analysis is also performed using LUSAS, where a mesh of thirty 3-noded beam
elements, equivalent to the ADAPTIC mesh of potential elasto-plastic cubic elements, is
employed from the start of analysis, since the locations of plasticity are not known
a priori. Moreover, the same time integration scheme (Newmark, 13=0.25 & 7=0.5) is
adopted with both ADAPTIC and LUSAS, and the same time step (At = 0.02 sec) is
used.
Comparison of the results in figure 7.17.e shows excellent agreement between LUSAS
and ADAPTIC up to (t 2 see), at which point numerical failure occurs with LUSAS.
Also, in terms of CPU time, ADAPTIC performed the whole analysis (up to t =5 sec) of
both rigid and flexible cases in less than 5 mm, while LUSAS required in both cases
around 1 hr 30 mm for analysis up to (t - 2 sec). This huge difference in efficiency
(a factor of 45) can be attributed to differences in the type of elements and solution
procedure, but more significantly due to the requirement with LUSAS of a fine mesh of
elements from the start of analysis.
The 3D tubular jacket structure depicted in figure 7.18.a is subjected to the transient
signal of figure 7.1 8.b, and is analysed using ADAPTIC where the effects of geometric
and material nonlinearities are considered. Due to the large size of the problem, the
analysis is performed on a VAX 4000 computer with 64 Mb of physical memory, and
which has a central processing speed of about 9 times that of the MicroVAX II computer
used in earlier examples.
224
Two cases are considered in this analysis. The first assumes an initially refined mesh of
400 elasto-plastic cubic elements, while the second employs the automatic mesh
refinement process with 52 elastic quartic elements at the start of analysis. Identical
solution procedures are used for both cases, and the time integration is based on
Newmark' parameters ( = 0.25 & '1 =0.5) and a time step (At = 0.01 sec).
The comparison in figure 7.18.c shows excellent agreement between the results of the
two cases. It is to be noted, however, that the first case required 10 hrs 6 mm of CPU
time for the analysis up to (t = 5 sec), whereas the second case consumed only
4 hrs 24 mm for the whole analysis (i.e. t = 10 sec). The significant computational
savings (78%) resulting from the use of the automatic mesh refinement process are
attributed to two main reasons. The first is the requirement of only 104 elasto-plastic
cubic elements for modelling the spread of plasticity (figure 7.18.d), as opposed to 400
cubic elements needed with the initially refined mesh, since only part of the structure
undergoes plastic deformation. Whereas the second is due to the initiation of plasticity at
(t = 2 sec), and hence the ability to model the response of the jacket during the first
2 seconds of the transient signal using only 52 quartic elastic elements.
225
,1
Uf
226
P
P1
2,3>1 U
2,1 2,2 >
81J BU 6U
P1
I >1>1>1 U
2,1 2,2 2,3
6U 8U 6U
227
P1
I 2,1
>1 2,2 2,3
U
6U 8U 6U
228
Key- e (b, a)
e = order of assembly
b = active nodes before elimination
a = active nodes after elimination
Assembly key
4:(3,2) 8:(4,3)
j.
ci
)c
I3:(3,2) 12:(2,2)
- r' 8
1 L,.L ••(3
9:(7,6) 1O:(6,6)
00
I7:(8,7)
S)
I' )c
I V ••N
Assembly configtion (2): Maximum front =8 nodes
Figure 7.6 Comparison of two assembly configurations for a two storey frame
229
4
\2i\,AJAJA:32
5 9 13 17 21 Maximum front
I 3 Current procedure 4 nodes
2 Sloan & Randolph 4 nodes
2 6 12 Maximum front
1 5 11 20 Current procedure 7 nodes
4 10 19
31 3 Sloan & Randolph 7 nodes
3 9 18 30
8 17 29 41
40
71 283 48
142634
1 27
13 39 3 55
46 52
2 35 66 63
23 37 74 64
27
65
' L" . 2I 73 79 I 80 178
230
Key- c (b, a)
e=orderofassembly
b = active nodes before elimination
a - active nodes after elimination
Assembly key
4: (3,2)
231
Temporary files Output files
Iterative ________
procedure
No
Convergence?
Yes
Yes Element
sub-division?
Update variables
'1
Output results
232
O.8L
I(O•2t1(p >1
L=l2Ocm N
E= 720 ton/an2
Mass per unit length = o. x ton. sec2 /an2
L
F<3crn>1
Cross-section
(N: Node
QE: Quartic elastic element)
QEI N5
QE1
N1
Configuration(fl: 3 quartic elements Configuration(2: 5 quartic elements
233
2.0
1.6
1.2
I
0
0 20 40 60 80 100
Displacement (cm)
234
4
-1
0 20 40 60 80 100
Displacement (cm)
Figure 7.1 i.e Dynamic response of Lee's frame under force incrementation
15
10
oO
-5
-10
0 20 40 60 80 100
Displacement (cm)
235
4.7625 mm
27.69 mm
E= 72395 N/mm2 863.6mm
1<
Cross section
(N:Node
QE: Quartic elastic elemen)
QE2
NYYN3 NXYN3
Mesh (U: 2 quartic elastic elements Mesh (2: 4 quartic elastic elements
200
-
150
I,
I
z I
.100
0
/
2 Quartic elastic elements
50 4 Quartic elastic elements
0 Oliver & Onate (1986)
0
0 10 20 30 40 50
Displacement (mm)
236
P12 P Pf2
Columns:
W 12x79 (Bouom storey) QP15 • QP16
Top storey W10x60 (Other storeys) N12 N13 N14
?P7 QP8
Girders W16x40 QP13 • QP14
P12 P P12 N9 NiO Nil
P5 QP6
L= 12 ft
rP QPI1 • QP12
Other storeys Lg=3Oft
N6 N7 N8
E= 13000 tsi P3 QP4
- - IC
LJ.&.J,1
QP9 - QP1O
- -
1 " y N3 N4 N5
QP1
L ___ I I N:Node
Quac psü -
1< )rj< )'-I I_hinge eiementj
L /2 L /2 Meshing configuration
-g rn - g
25
20
.10
0
0
0 2 4 6 8 10
Horizontal deflection at top right joint (in)
237
2790 mm
Irr
Transverse beam:
b219x 4.37mm2
E=210x 10 N/mm2
a ) =414 N/rrnn2
Diagonal members:
D1O1.7x 3.30mm2
E=210x 10 N/mm2
a=335 N/mm2
700
600
--
500
I:
200
• Experiment
USFOS (Soreide et al, 1986)
- -. Plastic hinge approach
100 Distributed plasticity approach (0.1% hardening)
0
0 50 100 150 200 250 300
Displacement (mm)
238
(iNode
QE: Quanic plastic-hinge eleme&
Plastic hinge
I)
Initial modelling Final modelling
Dljy spread
(N:Node (
QE: Quartic elastic elemen) lasto-plastic cubic elements)
N3 N4 N5
QE3 QE4
QE1 QE2
N2
239
0.76
Y
Cross Section
(All members)
2.570 x
00
00
E=20690 d
I-
V =0.3 I-
P =2400 kgfm3
H
6.285
11h.IYU )rJ
240
0.8
0.6
004
0.2 SM &Atluri (1988)
• Kondoh eta! (1986)
Remseth (1979)
0.0
0.0 0.5 1.0 1.5 2.0 2.5
Vertical displacement at crown (m)
2. -
1.
I
\-/
\:'f /
/'
-2. P=68.8sin(41.88t) MN '-I 4,
-
- SM & Atluri (1988) I
ADAPTIC (;=5%)
-4. -
0.0 0.1 0.2 0.3 0.4 0.5
Time (sec)
241
Static load:
P 8(4,2)=P 8 (5, 2)= 2)=-30k
P 5(4,3)=lk ; P(4,5)=6k.in
Columns: W12x40
L1
C
Beams: W12x53
E = 29, 000 ksi
a ,, = 33 ksi
:1. y=o.3
3.2
2.8 Wen & Farhoomand (1970)
ADAPTEC (lumped mass)
__ 2.4 ""C LUSAS (lumped mass)
/ EI
I. p.
C'C1
I I
I,
•. 0.8
'QI
I.
'I
/ • '
'f/
I
/ I t : i
'S 7,
0.4 .1 t' ' ¼)
w
0.0 II I • I •
0.0 0.1 0.2 0.3 0.4 0.5
Time (sec)
Figure 7.16.b Elastic response assuming lumped mass with ADAPTIC and LUSAS
242
3.2
2.8
Wen & Farhoomand (1970) I
ADAPTIC (distributed mass) I
. 2.4 • LUSAS (disbuted mass)
a
I'
A
fl
"I
0.4
0.0 I
,
'J\kJi'1
\ / '".i IA'
Y1
I
I
'I
V
0.0 0.1 0.2 0.3 0.4 0.5
Time (sec)
Figure 7.16.c Elastic response assuming distributed mass with ADAPTIC and LUSAS
3.2
I
- 2.8
1.6
Wen & Farhoomand (1970)
E 12
"'' A1)AP'FIC (lumped mass)
0.8 ADAPTIC (distributed mass)
0.4
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Time (sec)
243
5000 [E=210x 10 N/m2
>1
200 ay=300x1O6N/mJ
mg4,
e, i-< >
m.
>1
1ok 15
rm=300x C
g=9.81 m/sec2J
Figure 7.17.a Geometric and loading configurations of frame with joint flexibility
1.5
1.0
0.5 7/ _ _
.I.0.0
_ _
E
o -0.5
-1.0
/
-1.5
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
Rotation (rd)
244
0.6
0.4
0.2
-0.2
-0.4
-0.6
0 1 2 3 4 5
Time (sec)
0.2
•
• .
• So5
•tS
• 9,
•• SI
S
e • .1
• •
0.1 ••
:. % :. ,'
a a $
AJ.)APJ1C (flex.) •. •• •
-0.1 _________
I. $ ••
S. •
LUSAS (rigid) a..
S
'" LUSAS (flex.)
-0.2 I • I • I •
0 1 2 3 4 5
Time (sec)
Figure 7. 17.d Rigid and flexible frame response assuming elastic members
245
0.14
ADAVfIC (rigid)
0.12
ADAPTIC (flex.)
0.10 LUSAS (rigid)
""' LUSAS (flex.)
0.08
0.06 LUSAS fails
I- numerically
0.O4
0.02
0.00
-0.02
0 2 3 4 5
Time (sec)
Figure 7.17.e Rigid and flexible frame response assuming elasto-plastic members
246
Sm
m3 fl0x10mm2
Ground motion
1.00
0.75
0.50
0.25
0.00
-0.25
-0.50
-0.75
-1.00
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
247
0.06
0.04
0.02
0.00
'I
-0.02
-0.04
-0.06
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
Figure 7.18.c Response of jacket structure with and without automatic mesh refmement
Initial 52 0
Final 44 104
__)tiOfl
Figure 7.18.d Deflected shape and modelling of jacket with automatic mesh refinement
248
CHAPTER 8
Advanced nonlinear analysis of framed structures has by and large been used for research
purposes, while its extensive application within the various industries has been
discouraged by the considerable computational effort involved. The purpose of this work
was to develop an analysis tool which predicts accurately the nonlinear response of steel
frames, while providing means for significant computational savings.
The main conclusion from this work is that prudent development and implementation of
advanced material and geometry characterizations, along with improved nonlinear
analysis procedures, can transfer the application of nonlinear dynamic analysis from the
research, computer-intensive domain to the practical, micro-computer domain. This was
adequately demonstrated in Chapters 2 to 7, where exceptionally accurate results for
various classes of nonlinear static and dynamic problems were obtained using a micro-
computer, and at a small fraction of the CPU time required by other finite element codes.
In the following sections, the main features of the developed algorithms and their
implementation within the new program (ADAPTIC) are highlighted. Also included are
comments regarding the impact of these developments on the nonlinear large displacement
dynamic analysis of framed structures.
The accuracy and versatility of the finite element method lead to the decision to its use as
the basis for the work presented in this thesis. A new quartic formulation for elastic as
well as plastic hinge analysis was developed, while a cubic formulation accounting for the
spread of yield along the member length and across the section depth was implemented. A
joint element was included to allow for joint flexibiities, inclined supports and structural
gaps, whilst non-structural mass and damping formulations were developed to model
nonlinear dynamic effects.
249
8.1.1 Quartic formulation
This is a new formulation developed in a convected system, where small chord but large
global displacements are allowed. Quartic and linear shape functions are employed for the
interpolation of transverse displacements and twist respectively, whereas the axial
displacement distribution is determined by the average axial strain criterion. The quartic
formulation, while neglecting warping effects and flexure-torsion coupling, accounts for
the bowing action due to bending, and is derived on the level of stress resultants.
Verification examples showed the ability of the quartic formulation to represent the elastic
behaviour of initially imperfect beam-columns with only one element per member. To
provide an approximate means for elasto-plastic analysis, idealized plastic hinges are
included, where the interaction of biaxial bending moments and axial force is considered.
This formulation is also derived in a convected system, where small chord but large
global displacements are allowed. Cubic and linear shape functions are used for the
interpolation of transverse displacements and twist, respectively, with the distribution of
axial displacement determined from the average axial strain criterion. Although the cubic
formulation is similar to the quartic formulation in the sense that it neglects warping
effects and flexure-torsion coupling, its derivation is obtained on the level of the
constituent material rather than stress resultants. This, with the use of two Gauss point
quadrature for the integration along the element length, allows the spread of yield across
the section depth and along the member length to be modelled.
This multi-purpose joint formulation is intended for modelling joint flexibilities, inclined
supports and structural gaps. Decoupled actions are assumed between the different
translational and rotational freedoms, and the effects of large displacements are accounted
for. A cyclic asymmetric force-displacement relationship is included with this
formulation, thus allowing cyclic joint behaviour to be modelled.
250
is developed in an Updated Lagrangian system, where cubic and linear shape functions
are assumed for the interpolation of the transverse and axial displacements, respectively.
Whereas hysteretic damping is accounted for within the elasto-plastic structural elements,
dashpot and Rayleigh damping elements are developed to represent equivalent viscous
damping due to other effects. In addition to its suitability to the frontal method of solution
of the stiffness equations, dynamic modelling with non-structural elements has the further
advantage of allowing the freedom to use mass and damping elements only where the
structural mass and damping are significant, which can lead to considerable computational
savings.
The process of automatic sub-division forms, in this work, the essence of computational
efficiency for elasto-plastic analysis. Whereas an accurate representation of elasto-plastic
members requires a fine mesh of cubic elements, it was realized that the use of such a
mesh from the start of analysis for members which remain elastic for a significant part of
the analysis results in an unnecessary computational effort. This is especially true since
one quartic element can model accurately the elastic behaviour of a whole member,
including the beam-column action and the effect of initial imperfections. Nevertheless, the
use of traditional finite element procedures would still require fine meshing for all the
structural members, since the zones of material plasticity are not usually known a priori.
The use of the automatic sub-division procedure overcomes such inefficiencies, as the
analysis is always started with one quartic elastic element per member, and elasto-plastic
cubic elements are inserted automatically during analysis in parts of the members which
exhibit material plasticity. Automatic sub-division is also applied with the plastic hinge
modelling procedure, where the analysis is started with one quartic element per member.
In the case of member buckling, a plastic hinge is formed within the member length, and
the original element is replaced by two quartic elements.
In this work, the efficiency of the automatic sub-division procedure was demonstrated in
the context of accurate elasto-plastic analysis, where savings amounting to two thirds of
the original computational time were achieved. Moreover, such savings can be
significantly increased for structures which exhibit material plasticity in local areas or
during the late stages of analysis.
251
8.3 ADAPTIC
The formulations and procedures developed in this work were implemented in a new
nonlinear analysis program named ADAPTIC. The computer code is structured in such a
manner that the inclusion of new formulations, section shapes, and material models is
straightforward. This is manifested by the fact that the program was extended by the
implementation of two uniaxial concrete models and a number of composite section
shapes, while on-going research is focussing on the effects of local buckling.
With static analysis, ADAPTIC was developed to allow for the cases of proportional and
time history loading. Also, dynamic analysis capabilities were included through the
implementation of the various mass and damping formulations, where allowance was
made for support excitation so that earthquake analysis can be performed.
To solve for the nonlinear structural response, an incremental iterative strategy was
employed, where the Newton-Raphson or more general procedures can be applied. In the
case of convergence failure, step reduction is performed according to a pre-specified
parameter. With dynamic analysis, such a measure represents an automatic time-stepping
procedure, which proves essential for the determination of the complete dynamic
response, especially for structures which undergo sudden deterioration in stiffness during
analysis.
The frontal method was adopted for the efficient solution of the linearized stiffness
equations, a choice mainly influenced by the suitability of this method to the automatic
sub-division procedure. A simple element ordering procedure was developed and shown
to give similar frontal widths to more sophisticated procedures based on graph theory.
252
other computational aspects related to pre-processing, storage, and data handling, it is
believed that the improvement is brought about mainly by the newly developed
formulations and analysis procedures.
The development of ADAPTIC has provided a framework for the implementation and
verification of new research ideas regarding various aspects of the nonlinear behaviour of
structural frames, including steel, concrete and composite structures.
Future research is needed in the light of the findings of this work to provide accurate
analysis tools and procedures, which address more aspects of the nonlinear structural
behaviour, while maintaining an adequate level of computational efficiency. Programs
should be undertaken to investigate a number of research ideas, prompted by the
requirement for new analysis capabilities and the desire to achieve even greater savings in
computing resources. These include:
1. Addressing the effects of section warping and flexure-torsion coupling in the context
of elastic and elasto-plastic analysis. The general aim will be to provide an extension
to the elastic quartic formulation for modelling such effects with one element per
member, and the development of an elasto-plastic formulation to account for section
warping in the presence of material plasticity. The two formulations can then be used
in the context of the automatic sub-division process to provide a versatile and
accurate analysis tool which models the elasto-plastic lateral torsional buckling of
frame members.
253
dynamic contributions, and may necessitate the use of an Updated Lagrangian
system to avoid the complexities associated with modelling distributed mass and
damping within an Eulerian system. A special treatment will also be essential for
eigenvalue analysis involving such formulation, since the mid-side freedom does not
appear in the global stiffness equations.
6. Accounting for thermal loading and the effect of elevated temperatures on frame
behaviour under fire conditions.
254
REFERENCES
Ackroyd M.H. and Gerstie K.H., 1983. "Elastic stability of flexibly connected frames",
J. Struct. Eng., ASCE, Vol. 109, No. 1, pp. 241-245.
Adeli H., Gere J.M. and Weaver W., 1978. "Algorithms for nonlinear structural
dynamics", I. Struct. Div., ASCE, Vol. 104, No. ST2, pp. 263-280.
Argyris J.H., Dunne P.C. and Scharpf D.W., 1978. "On large displacement small strain
analysis of structures with rotational degrees of freedom", Comp. Meth. App!. Mech.
Eng., (Vol. 14, pp. 401-451) continued on (Vol. 15, pp. 99-135).
Barbat A.H. and Mique! Canet J., 1989. "Structural response computations in earthquake
engineering", Pineridge Press, Swansea, U.K.
Bathe K.J. and Bolourchi S., 1979. "Large displacement analysis of three-dimensional
beam structures", mt. J. Num. Meth. Eng., Vol. 14, pp. 961-986.
Bathe KJ., Ekkehard R. and Wilson E.L., 1975. "Finite element formulations for large
deformation dynamic analysis", mt. J. Num. Meth. Eng., Vol. 9, pp. 353-386.
Batoz J.L. and Dhatt G., 1979. "Incremental displacement algorithms for nonlinear
problems", mt. J. Num. Meth. Eng., Vol. 14, pp. 1262-1267.
Belytschko T. and Engelmann B.E., 1988. "Explicit time integration with enhanced
stability for structural dynamics", Comp. Struct., Vol. 29, No. 4, pp. 587-590.
Besseling J.F., 1977. 'Derivatives of deformation parameters for bar elements and their
use in buckling and postbuckling analysis", Comp. Meth. Appi. Mech. Eng., Vol. 12,
pp. 97-124.
Birnstiel C. and Iffland J.S.B., 1980. "Factors influencing frame stability", J. Struct.
Div., ASCE, Vol. 106, No. ST2, pp. 491-504.
255
Castiglioni C.A., 1987. "Numerical simulation of steel shapes under cyclic bending:
effect of the constitutive law of the material", Costruzioni Metalliche No. 3, pp. 154-
175.
Chan SI. and Kitipornchai S., 1987. "Geometric nonlinear analysis of asymmetric thin-
walled beam-columns", Eng. Siruct., Vol.9, pp. 243-254.
Chen W.F., 1980. "End restraint and column stability", J. Siruct. Div., ASCE, Vol.
106, No. 5Th, pp. 2279-2295.
Chu K.H. and Rampetsreiter R.H., 1972. "Large deflection buckling of space frames",
J. Siruct. Div., ASCE, Vol. 98, No. ST12, pp. 2701-2722.
Chugh A.K., 1977. "Stiffness matrix for a beam element including transverse shear and
axial force effects", mt. J. Num. Meth. Eng., Vol. 11, pp. 1681-1697.
Cofie N.G. and Krawinkler H., 1985. "Uniaxial cyclic stress-strain behavior of
structural steel", J. Eng. Mech., ASCE, Vol. 111, No. 9, pp. 1105-1120.
Connor J.J., Logcher R.D. and Chan S.c., 1968. "Nonlinear analysis of elastic framed
structures", I. Struct. Div., ASCE, Vol. 94, No. ST6, pp. 1525-1547.
Corradi L. and Poggi C., 1984. "A refined finite element model for the analysis of
elastic-plastic frames", lim J. Num. Meth. Eng., Vol. 20, pp. 2155-2174.
Dafalias Y.F. and Popov E.P., 1976. "Plastic internal variables formalism of cyclic
plasticity", J. Appi. Mech., ASME, Vol. 43, pp. 645-651.
Elnashai A.S., Bommer J. and El-Ghazouli A.Y., 1989. "The Loma-Prieta earthquake of
17 October 1989", Report No. ESEE-89/1 1, Imperial College, London, U.K.
Elnashai A.S. and Chryssanthopoulos M., 1990. "Effect of random material variability
on seismic design parameters of steel frames", Accepted for publication in J.
Earthquake Eng. Struct. Dynamics, Vol. 19.
256
Elnashai A.S., El-Ghazouli A.Y. and Dowling P.J., 1990. "Verification of pseudo-
dynamic testing of steel members", Accepted for publication in J. Construct. Steel
Research, Vol. 16.
Elnashai A.S. and Izzuddin B.A., 1991. "Dynamic characteristics and analysis of steel
jacket structures with nonlinear flexible joints", Paper to be presented at the 10th
International conference on OMAE, Stavanger, Norway.
El-Zanaty M.H. and Murray D.W., 1983. "Nonlinear finite element analysis of steel
frames", J. Struct. Eng., ASCE, Vol. 109, No. 2, PP. 353-368.
Feitner C.E. and Laird C., 1967. "Cyclic stress-strain response of f.c.c. metals and
alloys", Acta Metallurgica, Vol. 15, pp. 1621-1653.
Fiippou F.C., Bertero E.P. and Popov E.P., 1983. "Effects of bound deterioration on
hysteretic behavior of reinforced concrete joints", Report No. UCB/EERC-83/19,
University of California, Berkley, California.
Gibbs N.E, Poole W.G. and Stockmeyer P.K., 1976. "An algorithm for reducing the
bandwidth and profile of a sparse matrix", Siam J. Num. Anal., Vol. 13, No. 2, pp.
236-250.
Higginbotham A.B. and Hanson R.D., 1976. "Axial hysteretic behavior of steel
members", J. Struct. Div., ASCE, Vol. 102, No. ST7, pp. 1365-1381.
Hobbs R.E. and Jowharzadeh A.M., 1978. "An incremental analysis of beam-columns
and frames including finite deformations and bilinear elasticity", Comp. Struct., Vol.
9, Pp. 323-330.
Holden J.T., 1972. "On the finite deflections of thin beams", mt. J. Solids Struct., Vol.
8, pp. 1051-1055.
Hughes T.J.R., 1987. "The finite element method: Linear static and dynamic finite
element analysis", Prentice-Hall Inc., pp. 582-630.
257
Hughes TJ.R., Pister K.S. and Taylor R.L., 1979."Implicit-explicit finite elements in
nonlinear transient analysis", Comp. Meth. App. Mech. Eng., Vol. 17/18, pp. 159-
182.
Ikeda K. and Mahin S.A., 1986. "Cyclic response of steel braces", J. Struct. Eng.,
ASCE, Vol. 112, No. 5, pp. 342-361.
Inoue K. and Ogawa K., 1974. "Nonlinear analysis of strain-hardening frames subjected
to variable repeated loading", Technology Report of the Osaka University, Vol. 24,
No. 1222, pp. 763-78 1.
Irons B.M., 1970. "A frontal solution program for finite element analysis", mt. J. Num.
Meth. Eng., Vol. 2, pp. 5-32.
Izzuddin B.A. and Elnashai A.S., 1989. "ADAPTIC: A program for the adaptive
dynamic analysis of space frames", Report No. ESEE-8917, Imperial College,
London, U.K.
Izzuddin B.A., Elnashai A.S. and Maths P., 1990. "Nonlinear dynamic analysis of
reinforced concrete structures on small computers", International Conference on
Education Practice and Promotion of Computational Methods in Engineering using
Small Computers, Macau.
Jam A.K., Goel S.C. and Hanson R.D., 1978. "Inelastic response of restrained steel
tubes", J. Struct. Div., ASCE, Vol. 104, No. ST6, pp. 897-910.
Jam A.K., God S.C. and Hanson R.D., 1980. "Hysteretic cycles of axially loaded steel
members", J. Struct. Div., ASCE, Vol. 106, No. STh, pp. 1777-1795.
Jenkins J.A., Seitz T.B. and Przemieniecki J.S., 1966. "Large deflection of diamond-
shaped frames", Int. J. Solids Struct., Vol. 2, pp. 591-603.
Jennings A., 1968. "Frame analysis including change of geometry", J. Struct. Div.,
ASCE, Vol. 94, No. ST3, pp. 627-644.
Jones S.W., Kirby P.A. and Nethercot D.A., 1982. "Columns with semirigid joints", J.
Struct. Div., ASCE, Vol. 108, No. ST2, pp. 36 1-372.
258
Kam T.Y., 1988. "Large deflection analysis of inelastic plane frames", J. Struct. Eng.,
AScE, Vol. 114, No. 1, pp. 184-197.
Kani I.M. and McConnel RE., 1987. "A simple and efficient beam element for the
combined nonlinear analysis of frameworks", Comp. Struct., Vol. 25, No. 3, pp.
457-462.
Kondoh K., Tanaka K. and Atluri S.N., 1986. "An explicit expression for the tangent-
stiffness of a finitely deformed 3-D beam and its use in the analysis of space frames",
Comp. Siruct., Vol. 24, No. 2, pp. 253-271.
Lee S.L., Manuel F.S. and Rossow E.C., 1968. "Large deflections and stability of
elastic frames", J. Eng. Mech. Div., ASCE, Vol. 94, No. EM2, pp. 521-547.
LUSAS, User manual. "Finite element analysis system", FEA Ltd. Surrey, U.K.
Ma S.Y.M., Bertero V.V. and Popov E.P., 1976. "Experimental and analytical studies
on the behavior of reinforced concrete rectangular and T-beams", Report No.
UCB/EERC-7612, University of California, Berkley, California.
Maths P. and Elnashai A.S., 1989. "A theoretical model for composite beam-columns
under cyclic loading", Report No. ESEE-89/10, Imperial College, London, U.K.
Maglaras D., 1990. "Seismic design parameters of steel frames by response surfaces",
MSc dissertation is Engineering Seismology and Earthquake Engineering, Imperial
College, London, U.K.
Maison B.F. and Popov E.P., 1980. "Cyclic response prediction for braced steel
frames", J. Struct. Div., ASCE, Vol. 106, No. Sri, pp. 1401-1416.
Mallet R.H. and Marcal P.V., 1968. "Finite element analysis of nonlinear structures", J.
Struct. Div., ASCE, Vol. 94, No. ST9, pp. 2081-2105.
259
Mattiasson K., 1981. "Numerical results from large deflection beam and frame problems
analysed by means of elliptic integrals", Tnt. J. Num. Meth. Eng., Vol.17, pp. 145-
153.
Meek J.L and Loganathan S., 1989. "Geometrically non-linear behaviour of space frame
structures", Comp. Struct., Vol. 31, No. 1, pp. 35-45.
Meek J.L. and Loganathan S., 1990. "Geometric and material non-linear behaviour of
beam-columns", Comp. Struct., Vol. 34, No. 1, pp. 87-100.
Mizuno E., Kato M. and Fukumoto Y., 1987. "Multi-surface model application to beam-
columns subjected to cyclic loads", J. Construct. Steel Research, Vol. 7, pp. 253-
277.
Mondkar D.P., and Powell G.H., 1977. "Finite element analysis of non-linear static and
dynamic response", Tnt. J. Num. Meth. Eng., Vol. 11, pp. 499-520.
Newmark N.M., 1959. "A method of computation for structural dynamics", J. Eng.
Mech. Div., ASCE, Vol. 85, No. EM3, pp. 67-94.
Oliver J. and Onate E., 1986. "A total lagrangian formulation for the geometrically
nonlinear analysis of structures using finite elements. Part II: arches, frames and
axisymmetric shells", mt. J. Num. Meth. Eng., Vol. 23, pp. 253-274.
Oran C., 1973-a. "Tangent stiffness in plane frames", J. Struct. Div., ASCE, Vol. 99,
No. ST6, pp. 973-985.
Oran C., 1973-b. "Tangent stiffness in space frames", J. Struct. Div., ASCE, Vol. 99,
No. ST6, pp. 987-1001.
Oran C. and Kassimali A., 1976. "Large deformation of framed structures under static
and dynamic loads", Comp. Struct., Vol. 6, pp. 539-547.
PAFEC, User manual. "Linear analysis program ",PAFEC Ltd., Nottingham, U.K.
260
Papadrakakis M., 1983. "Inelastic post-buckling analysis of trusses", J. Struct. Eng.,
ASCE, Vol. 109, No. 9, pp. 2129-2147.
Papadrakakis M. and Chrysos L., 1985. "Inelastic cyclic analysis of imperfect columns",
J. Struct. Eng., ASCE, Vol. 111, No. 6, pp. 1219-1234.
Petersson H. and Popov E.P., 1977. "Constitutive relations for generalized loadings", J.
Eng. Mech. Div., ASCE, Vol. 103, No. EM4, pp. 611-627.
Popov E.P. and Black R.G., 1981. "Steel struts under severe cyclic loadings", J. Struct.
Eng. Div., ASCE, Vol. 107, No. ST9, pp. 1857-188 1.
Popov E.P., Mahin S.A. and Zayas V.A., 1980. "Inelastic cyclic behavior of tubular
braced frames", J. Struct. Div., ASCE, Vol. 106, No. ST12, pp. 2375-2390.
Popov E.P. and Petersson H., 1978. "Cyclic metal plasticity: Experiments and theory",
J. Eng. Mech. Div., ASCE, Vol. 104, No. EM6, pp. 1371-1388.
Popov E.P., Mahin S.A. and Clough R.W., 1985. "Inelastic response of tubular steel
offshore towers", J. Struct. Eng., ASCE, Vol. 111, No. 10, pp. 2240-2258.
Powell 0. and Simons J., 1981. "Improved iteration strategy for nonlinear structures",
hit. J. Num. Meth. Eng., Vol. 17, pp. 1455-1467.
Powell G.H. and Chen P.F.S., 1986. "3D beam-column element with generalized plastic
hinges", J. Eng. Mech., ASCE, Vol. 112, No. 7, pp. 627-641.
Qashu R.K. and DaDeppo D.A., 1983. "Large deflection and stability of rigid frames",
J. Eng. Mech., ASCE, Vol. 109, No. 3, pp. 765-780.
Reisnour P.R. and Batla F.A., 1984. "Beam-column analysis of rigid frames",
Engineering Mechanics in Civil Engineering, A.P. Boresi et al, eds., pp. 1472-1475.
Remseth S.N., 1979. "Nonlinear static and dynamic analysis of framed structures",
Comp. Struct., Vol. 10, pp. 879-897.
261
Robinson J., 1973. "Integrated theory of finite element methods", John Wiley & Sons,
428 pp.
Saafan S.A., 1963. "Nonlinear behavior of structural plane frames", J. Struct. Div.,
ASCE, Vol. 89, No. ST4, pp. 557-579.
Saleeb A.F. and Chen W.F., 1981. "Elastic-plastic large displacement analysis of pipes",
J. Struct. Div., ASCE, Vol. 107, No. ST4, pp. 605-626.
Santhanam T.K., 1979. "Model for mild steel in inelastic frame analysis", J. Struct.
Div., ASCE, Vol. 105, No. ST1, pp. 199-220.
Shi G. and Atluri S.N., 1988. "Elasto-plastic large deformation analysis of space-frames:
A plastic-hinge and stress-based derivation of tangent stiffnesses", Tnt. J. Num. Meth.
Eng., Vol. 26, pp. 589-615.
Sloan S.W. and Randolph M.F., 1983. "Automatic element reordering for finite element
analysis with frontal solution schemes", mt. J. Num. Meth. Eng., Vol. 19, pp. 1153-
1181.
Smith E.A., 1984. "Space truss nonlinear analysis", I. Struct. Eng., ASCE, Vol. 110,
No. 4, pp. 688-705.
Sohal 1.5. and Chen W.F., 1988. "Local and post-buckling behavior of tubular beam-
columns", J. Struct. Eng., ASCE, Vol. 114, No. 5, pp. 1073-1091.
Soreide T.H., Amdahl J., Granli T. and Astrud O.C., 1986. "Collapse analysis of
framed offshore structures", 18th Annual Offshore Technology Conference, Houston,
Texas, pp. 95-102.
Stricklin J.A., Haisler W.E. and Riesemann W.A.V., 1971. "Geometrically nonlinear
structural analysis by direct stiffness method", J. Struct. Div., ASCE, Vol. 97, No.
ST9, pp. 2299-2314.
262
Sugimoto H. and Chen W.F., 1985. "Inelastic post-buckling behavior of tubular
members", J. Struct. Eng., ASCE, Vol. 111, No. 9, pp. 1965-1978.
Surana K.S. and Sorem R.M., 1989. "Geometrically non-linear formulation for three
dimensional curved beam elements with large rotations", mt. J. Num. Meth. Eng.,
Vol. 28, pp. 43-73.
Teixeira de Freitas J.A. and Smith D.L., 1983. "Fmite element elastic beam-column", J.
Eng. Mech., ASCE, Vol. 109, No. 5, pp. 1247-1269.
Timoshenko S.P., 1956. "Strength of materials", D. Van Nostrad Co., Princeton, N.J.,
Part II, 3rd edition, pp. 411-416.
Toma S. and Chen W.F., 1982. "Cyclic analysis of fixed-ended steel beam-columns", J.
Struct. Div., ASCE, Vol. 108, No. ST6, pp. 1385-1399.
Trueb U., 1983. "Stability problems of elasto-plastic plates and shells by finite
elements", A Thesis Submitted for the Degree of Doctor of Philosophy in the Faculty
of Engineering of the University of London, Department of Civil Engineering,
Imperial College, London, U.K.
Tseng N.T. and Lee G.C., 1983. "Simple plasticity model of two-surface type", J. Eng.
Mech., ASCE, Vol. 109, No. 3, pp. 795-8 10.
Vinnakota S., 1982. "Planar strength of restrained beam columns", J. Struct. Div.,
ASCE, Vol. 108, No. ST1 1, pp. 2496-2516.
Wakabayashi M., Matsui C., Minami K. and Mitani L, 1974. "Inelastic behavior of steel
frames subjected to constant vertical and alternating horizontal loads", Proc. Fifth
World Conference on Earthquake Engineering, Rome, Vol. 1, pp. 1194-1197.
Wen R.K. and Farhoomand F., 1970. "Dynamic analysis of inelastic space frames", J.
Eng. Mech. Div., ASCE, Vol. 96, No. EMS, pp. 667-686.
263
Wen R.K. and Rahimzadeh J., 1983. "Nonlinear elastic frame analysis by finite
element", J. Struct. Eng., ASCE, Vol. 109, No. 8, PP. 1952-197 1.
White D.W. and McGuire W., 1987. "Discussion on: Uniaxial cyclic stress-strain
behavior of structural steel, by Cofie & Krawinkler 1985", J. Eng. Mech., ASCE,
Vol. 113, No. 11, pp. 1803-1807.
Yang T.Y., 1973. "Matrix displacement solution to elastica problems of beams and
frames", hit. J. Solids Struct., Vol. 9, pp. 829-842.
Yang T.Y. and Saigal S., 1984. "A simple element for static and dynamic response of
beams with material and geometric nonlinearities", hit. J. Num. Meth. Eng., Vol. 20,
pp. 85 1-867.
Zayas V.A., Shing P.S.B., Mahin S.A. and Popov E.P., 1981. "Inelastic structural
modeling of braced offshore platforms for seismic loading", Report No. UCBIEERC-
81fl)4, University of California, Beridey, California.
264
APPENDIX A
The transformation matrix T referred to in (2.1 3.c) is a matrix of first derivatives of chord
with respect to global displacements. The 12x6 terms of T can be established explicitly
using (2.9) and (2.10), as shown below.
T
= (0 1 '°1'02y'02' &OT) (A.1.a)
T
gU(U1IV1,W1cZ1,31,Y1,U2,V2,W2,a2,f2,Y2)
(A.1.b)
acc •
ag U 1 Y
iE
ao1
(aT
T i,1 = agu i) iE fl
L_xaguj y
0 ieR (A.2.a)
265
acc
Ic E
- U .zc
I
gj
ao1
1 i2 = agU. c.Ia'rT o) iE'H
_c agU
)
0 I E (A.2.b)
( acc
2
. C iE
U. 1 Y
i,3
ao2 0 iER
g 1 \
a2T
r 2 01
j E II
gj U Y II
(A.2.c)
( acc •
1 E
I au.
= ae 2z[
T 0 i€
i,4 u.
c
(T 2o'
U ZC I iEH
1a
'¼ g j I (A.2.d)
aL
T- ____- a u. i E FI
a u. - g i
0 E (A.2.e)
0 I E
where,
266
C C C
aXck ( c Xtk Ij
i E fl
a g uJ - LC
____ ac
iEfl
au. -- u.
g i g i-6
I=3x3 identity matrix (A.3)
13 I a- —1 a
-Y 1 1 2
2 2
arT a T a '1 13
13
—a 1 —1
-1,_w=
—1 a 13
1 —a 2 '•'
2 -Y (A.4)
and,
-'I C
aI_,
i E fl
a.
aLC aLC
i E
a u. a u.
g g i-6
(A.5)
2
acc
•1 (i,j) E 1H)
au.au. Y
gigj
ac c a'T
a2o -
____ I r 1 ol
(i,j)
ly äu. Ipu.yC E (R,'R)
6 g 1 gj )
agujaguj
1 o•'
- auauyCl (i,j)e (H,A)
xcc.I _g i g j )
0 (ij)E (,aH)UJ(afl,1D)
(A.6.a)
267
2
a cC
• (i,j) E 1R)
au.au.
gigj
acc a'T \
a2o 1z ___ I not
- - - agU1II Z I (i,j) E (1R,R)
G
S i,j,2 u.a u. '¼IgU
j j
gigj
_a2i
- auauzdl ( I, j) E (R,fl)
xcc.I gigj
0 (i,j) E (1l, afl)U'
(A. 6.b)
a2 ,c
• (i,j) e
- agu1agu
0 (i,j)e ('rfl,aR)L(aA,)
6 - 2y -
S i,j,3 u.d U. - acc
gigj (ai 2co'1 (i,j) E (1A,R)
au.
g 1 gIau.y
j )
a22i
r 'I
201 (i,j) E (A,R)
- auauydl
xcc.Ig i g j )
(A.6.c)
- a2x cC 2c
u.a u. •
agigj (i,j)e(H,fI)
0 (i,j)e (,aR)U(aR,8)
a
S G.. = 2z =
u.a u. ac c (a2i 2
gigj (i,j) E (R,R)
g 1 . UZ I
gj I
a22i
r 2oI
- auauztl (i,j) E (R,fl)
xcc.Igigj
(A.6.d)
a2L
a2
________ I ___ (i,j)e(fl,fl)
SGiii5=gU.agUj gUjagUj
0 (i,j)e (nR,aR)U(aR,nR) ] (A.6.e)
268
0 (1, j) e (FI, a) '' (as' R)
( uT 10121
yC
u. u. zC
gigj
(i,j)E (.R,fl)
aeT
S61,j,6= pgigj
uPU- I a' ) 1 aT 1
aguj z )• agu )C ) (i, j) E (ft, R)
CC•
(pgujgtij yC] (i,j) E (A,R)
(A.6.f)
2rC
_______- __________
P gU j Pg U j - PgUj_6PgU_6 (i, j) E (1I, R)
(A.7)
1 1
0 0 01 a2T 0 0 0 0
a2T
aaacz - O jtaczat3 . 0 0 0
1
0 0 0 0
0 0 0
a2T a2T
—1 0 0
0 1
0 r - 0 0
aaa = aa 2
—1 1
0 0
2
r - r ' T
r a2 T 1-1 0 0
?yaczaa?y —1 0
0 0
(A.8)
and,
269
a2Lc
(i,j)E ('A,R)
au.au.
g1g3 LC
a2 L c a2LC
(1, j) E A)
agu.a
1 g gu.a
i g j-6
a2 L c a2LC
(i,j) E (R,fi)
aU.aU.aU.
g 1 g j g i-6au.
g j-6 (A.9)
T
= fuv,w,9,OyO) (A. 10.a)
T
= 132'r2) (A. 10.b)
270
adc
C
IE
agu
T11=
au
ag u
=1 xC0]1cd iEfl
I [ aguj
I E
(A.1 1.a)
C
ad
ie
aU
- av =(aT
T1,2—
agu1 1cO].dC iefl
aguj Y
0 I E fl
(A.11.b)
I cc.
ad
au
1E
- aw
T ii 3 =1 a ' T_1 O]•%JC iER
au1 [aguj zC
0 i E
(A.1 1.c)
0 iE
= ao
I
1Ip u iER
T gj
1.4 aU.
1
____ 1 01
IE
gj.Ia2T
au y I I (A.11.d)
1 0 IE
ao I r
iE'fl
Iau.x
T5aU
- g _lz
I(aT20
IE
ha g
1 I (A.1 1.e)
271
1 0 iE
T = ao zi 2cc.TT
y Iau.x
g
iefl
g u.
1,6
ar 2
IIp u Y I•X iEfl
\gj I (A.1l.f)
where,
a
i € un
a . -- i,k
a d ad
iEH
ag u. a g i-6
and the first derivatives of matrices and with respect to global displacements are
the same as for member formulations given by (A.4).
The geometric matrix 6 for the 3D joint formulation is a 12x12x6 matrix of second
derivatives of local with respect to global displacements. The explicit terms of J can
thus be obtained by differentiating once the terms of matrix T obtained in the previous
section.
(A.13.a)
272
0 (i,j)e U(aA,H)
____ o a
____ (i,j)E (1'n)
a 2 agu. y I • a g I
II._°'
S6 i,j,2 j )
g IagUj -
-
a2 T
co] • (i,j)E (R,'rR)
ag u.a U. Y
1 gj
(A.13.b)
( 0 (i,j)E (tA,iR)J(aA,I)
0 (i,j)E (tR,aR)U(aR,tR)'
a21 '
__________ ol
X I (i,j)e(fl,11.R)
a2o ,
.I a
gj g j u
I
1a2T " 'a'T
S61,j,5 agj
u agj
u= c0J. XJ (i,j)E (fl,R)
aguj agu1
( a2i 2o
I (i,j)E (R,A)
u
gj gj u zC
(A.13.e)
273
0 (i,j)E (tR,afl)U(aH,tH)'
S6j,j,I(S6j,j,k (i,j)E ( as ' a11) & k 1-46) (A. 13.g)
The second derivatives of matrices and with respect to global displacements are
the same as for member formulations given by (A.8).
274
APPENDIX B
The general structure of the data file employed by ADAPTIC is presented hereafter in its
sequential order.
IXXXZXXXXXX%X%XX1XX1ZXX1xZXXXZXzXxzxzxxxxxzx*xxxxxxxxxxxxxxzxzxzzx:xzzx
2D...EIe.ents/3D_EIe.ents (2/3)
3 (for 3D analysis)
ZXZXXXXXXZII*XXXXXXX%XXXXXZIXZZZXZXHXXXXZZXXZZXZXXXXXXXXXXXX%XZXXXZIX
ANALYSIS TYPE
EIGEN-UALUE STATICS DYNAMICS
0 1 0 (for static analysis)
IXZZXIIZXXXXXXX%IIXXXXXXXXXXZXZZZXZXXXZXZ:IXZXZXXZZ%XXXZX%XXIXIIXIIXXZ
ZZXZXZZIZZXXXXZZXXIXXXXZXXXXZXXXXZXZZXIZZXXXXXXXZXXXXZXZX%IZZXZIXIXZXX
MATER I ALS
++++++++++++++++++++++++++++++++++++++++++,+++,+++++++++++++++++++++++
NUMBER OF MATERIALS
XIII XIIXXIXXI XI XX XIXIXZXIIXXI XX XIII XI IIIXIIIIXXX I XXX XXXI 111111% XXX I XXX
SECTIONS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++,+,+++++++
275
NUI1BER OF SECTIONS
SECTION-NO. TYPE-NO.
MATERIAL IIO.(S)
DIMENSION(S)
ELEMENT GROUPS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
I ______________________________________________________________________
SUBDIUISION PATTERNS
NUMBER OF PATTERNS
+++++++++++++++++++++++++++++++++++++++++++++++++,++++++++++++++++++++
GROUP SECTION-NO.
+++++++++++++++++++++++++++++++++++++++++++++++.+.+++++,++,+++++++++++
276
ELEMENT TYPE (8) C Lumped t1.. Element )
NUMBER OF GROUPS
GROUP MASS
GROUP Cx Cy Czz
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++,+++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
I+++++++++++++++++++++++,+++++,++,+++++++++++++++++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
277
SUBDIUISION PATTERNS
NUMBER OF PATTERNS
GROUP SECTION-NO.
+++++++++++++++++++++++++++++++++++++++++++++++,++++++++++++,,++++++++
++.+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
GROUP MASS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
GROUP MASS-PER-UNIT-LENGTH
+++++++++++++++++++.++++++++++++++++++++++++++++.+++++++++.+++++++++.+
278
B.1.6 Nodal coordinates
IIIIIXIXXXXXXXXXXXxzXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXsxXXXXXXXXXXXIIXXXXXX
NODE X Y Z
zxxxxzxz;xxIXXXXXXIXX;IXXXXXXxXXXXXXXXXXXXXIXXXXUXXIUXXIXXXXZXXXxXXI
NODE X V 2
XXX XX XX UXX I XXX XXXIII XIX XXXXXXXXX XXIX XXXXXXXXXXXXXXX XXXI XXX XXXII XXXIII
XXXXXXXXXXIIXXXXXXXXIXIXXXXXXXXXXXXXXXXXXIXXXXIIXIXXXXXXXXXXIXXXXXXXXX
ELEMENT COHNECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++.+++++++++++++++++++++++++++++++++
279
ELEMENT TYPE (8) ( Lumped 110.. Element )
NUMBER OF ELEMENTS
+++++++++++++++++++++++++++++++++++++.+++++.++++++.+++++++++++++++++++
++++++4.++++++++++++++++++++++++4.++++++++++++++++++++++++++++++++++++++
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ +++++++++++
+++++++++++++++++++++++++++++++++++++++++++,++++++++++++++++++++++++++
l++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++++.+++++++++++++++++++++++++++++++++++++++,+++++++++
280
ELEMENT TYPE (33) C 3D Quartic Elastic Element )
NUMBER OF ELEMENTS
++++++++++++++++++++++++++++++++++++++++++++4+++++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++,+++++++++++,++++.++++++.+
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++.+++++++,+.++
++++++++++++++++++++++++++++++++++++++++++.+++++++++++++++++++++++++++
xx1xzxzxxzxxzzxxzzzxz111zxzzz1x*xxxxzxxxx1xxxxz1z1xsxzzxxx*1xzxxxxzzxz
IMPERFECTIONS ( Element Types ( 2, 3, 4 , 32, 33, 34 ] )
+++++++++++++++++++++++++++,++++++++++++++++++++++++++++++++++++++++++
281
ELEMENT TYPE (2) C Quartic Element sith Distributed PlastiCity )
IIUI1BER OF IMPERFECT ELEMENTS
+++++++++++++++++++++++++++++++++++++++++++++l.++++,,++++.+++++++++.+++
++++++++++++++++++++++++++++++++++++++++++++++++.+++.,,,+++.++++++++++
+++++++++++++++++++++++++++++++++++++++++++4.++++++++++++++++++++++++++
+++++++++++++++++++++++++++++++++++++++++++++++,+++++++++++ +++++++++++
282
B.1.11 Restrained freedoms
lx1zxxxxxxxxxzxxzxzzzxxxxzxxxxxxxxxx%xzxxzzxzxxxxzxxxxzxxxzxxxxxxxxuxx
RESTRA I HTS
NUMBER OF RESTRAINTS
B.1.12 Loading
The loading module depends on the type of analysis required; namely, static analysis
assuming proportional loading, time history static analysis, or time history dynamic
analysis.
LOADING
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++,+++++++
+++++++++++++++++++++++++++++++++++++++++++4.++++++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++,+++++++++++++++++++++++++
NUMBER OF PHASES
I ----------------------------------------------------------------------
PHASE Load/DIep./Auto.(1/2/3)
- 1 (Load control)
INCREMENT-FACTOR Load/Un I ood/Keep( 1/2/3) STEPS
283
PHASE Load/DIep./Ruto.(1/2/3)
- 2 (Displacement control)
Nodal/Eleaent(1/2) DIR. NODE/ELEIIENT(Type,Ho.)
PHASE Load/DIsp./Auto.(1/2/3)
- 3 (Automatic control)
Nodal/Element(1/2) Trone./Rot . (1/2)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
IXXXX XX%XX XX X*XX XXXXXXXXXXX XXXXXXXXXXXXX XX XXXXXXXXXX XX XXXXXX XXXZX XXX XXX
LOADING
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
I ______________________________________________________________________
LOAD-CURUE TIME-POINTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
284
EQUILIBRIUM STAGES
IIUMBER OF STAGES
+++++++++++++++++++++++++++++++,++++++++++++++.+++++++.++++.++++++++++
LORD I HG
HEIH1ARKS PARAMETERS
BETA GAMMA
O.250000E+OO O.500000E+OO (Average acceleration method)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER-OF-LOAD-CURUES STARTING-TIME
LOAD-CURUE TIME-POINTS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
EQUILIBRIUM STAGES
NUMBER OF STAGES
285
I++++++++++++++++++++++++++++++++++++++++++++++++,+++++++++,+4,++++++++
ITERATIUE STARTEGY
ITERATIONS INITIAL-REFORMS. STEP-REDUC. ITER.-DIU.
COHUERGENCE CRITERIA
TOLERANCE FORCE/TRANS. I1OI1ENT/ROT. MAX.-CONU.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
OUTPUT
FREQUENCY STRESSES
1111 XIIIXXXXXXX XXX XXIXXXXIX XXX XXX XXIX XX XXXIII XXXXXXX XXXI I IX I XXXI XXXIII
Two files are used for results output. The first Filename.OUT' contains general
information on the progress of analysis, including the creation of new nodes and elements
upon element subdivision, as well as the formation and closure of plastic hinges. While
the second Pilename.NUM' contains all the numerical results.
B.2.1 'Filename.OUT'
IXXXXXIXXXIXXXXXXXXXXXXXIXXIIXXXXXXXXXXIXXXXIXIXXXXIXXXIXXXIXXXXXXXXIXX
28 (34 , 9)
IXXXXXXXXIIXXIXXXXXXXXXXXIXXXIXXXXXXXXIIXXXIXXXXXXXXXIXXXXX*IIXXXIXXXXX
UARIRBLE LOADING
PHASE NUMBER - 1
TYPE • LOAD CONTROL
INCREMENT FACTOR •O.100000E+02
NUMBER OF STEPS - 50
286
URRIABLE
LOAD
OUTPUT FACTOR LEUEL CONU.-IIORM ITERATIONS
1 O.20000000E+O0 O O.801E-09 2
20 O.I0000000E+O1 O O.124E-08 2
B.2.2 'Filename.NUM'
3 3 0.59 18?IE+O5
287
DI SPLACEMENTS
MODE U U ALPHA
1 O.000000E+OO O.000000E+OO O.000000E.00
I0000000000000000000000000000000000000000000000000000000000000000000000
Ixx1XZ1Z%XXXXXX%XxXZXXZXZXZZZXXXZXZXXXXZXZXZXXXZZX1XXXXZXXXXXXXZXXZXXXZ
20_Elements/3D_Elements (2/3)
2
ZXXZ U XXXXXX U XX XXX XXXXXX XXX XXXX XIXXXXXXXXXXIXXXXX XXX IX XXX
ANALYSIS TYPE
EIGEN-URLUE STATICS DYNAMICS
0 1 0
XIXXIXIIXXIIXXXXXXXXXXIXIXXXXXXXXXIXIXIXXXXIXIXZXXXXXIXXXXIXXXXXXXXXXX
MATER I ALS
+++++++++++++++++++++++++++++++++++++++++++++++++,++++.+++++++++++++++
NUF1DER OF MATERIALS
SECTIONS
NUMBER OF SECTIONS
288
SECT I ON-HG. TYPE-NO.
1 1
MATERIAL NO.(S)
DIMENSION(S)
O.300000E+O1 O.200000E+O1
xxzxzxxxx%xx%xxxx*%xxzxxxzzxxxxz1xsxxxxxzzxzzxxzxxxxz1xzxxxxx1xxzx1zx1
ELEMENT GROUPS
+++++++++++,++++++++++++++++,+++++++++++++++++++++++++++++++,+.+++++++
GROUP SECTION-NO.
1 1
11z1zxz z xxs xxxn ix x xxx xx xx xixxxxxixx xxxxxx xxix xx xxxxxxxxx
ELEMENT CONNECTIUITY
AESTRA I NTS
NUMBER OF RESTRAINTS
4
RESTRAINT NODE DIRECTION
1 1 1
2 1 2
3 4 1
4 4 2
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxzxxxxxxxxxxxxxxxxxxxxxxxxxxxzxxxxxxxxx
LOAD I HG
+++++++++++++++++++,++++++++,++++++++++++++++,++++++++++++++++++++++++
+ + + + + + + + + + + + + + + + + + + + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +++
289
NUMBER OF PHASES
2
PHASE Load/OIsp./Auto.(1/2/3)
1 1
I HCREI1ENT-FACTOR Load/Un Ioad/Keep( 1/2/3) STEPS
O.200000E+O1 3 20
I ----------------------------------------------------------------------
PHASE Lood/DIep./Auto.(1/2/3)
2 3
Nodal/EIe.ent(1/2) Trans./Rot. (1/2)
I 1
Load/Un Ioad(1 /2) STOP-CONDITIONS
1 3
COND. L.F./DIsp.(1/2) 6.1. 1.1. (IIODE,DIR.)
1 1 O.200E+O1 -.200E+O1
2 2 O.120E+03 -.120E+03 4
3 2 0.120E+03 -.120E+03 4 2
+++++++++++++++.+++++++++++,++++++++++++++++++++++,+++++++++++++++++++
ITERATIUE STRRTEGY
ITERATIONS INITIAL-REFORMS. STEP-REOUC. ITER.-DIU.
6 6 5 5
CONUERGENCE CRITERIA
TOLERANCE FORCE/TRANS. MOMENT/ROT. I1RX.-CONU.
O.100000E-O5 O,200000E+O1 O.100000E+03 O.100000E+03
+++++++++++++++++++++++.++++++++++++++++++++++++++++.+++++++++++++++++
OUTPUT
FREQUENCY STRESSES
1 0
XXZIXIIIIXXXXI 111* II 11111 XX XXIIIXXXXXZ XXIXIIXII XXIXXX*X 1111111 III XIII
B.3.1.2 'LEE_STAT_3.OUT
IXxIzIIXIIxXIXxIIXXIxIxXXXIzIxIXXXIzXxXIXIzXIIuIXxzxxIIxIXIIIXIXXIIXI
XXXXIXIIIXIXXIIIXXXXXIXXXXIIXXXXIIIZXIXIIXIIIXXX %XIXXXXXIIXXXI
290
UARIR8LE LOADING
++ ++ ++ + ++ ++ + ++ +++ ++ +++ +, ++ + ++++ +
PHASE NUMBER - 1
TYPE - LOAD CONTROL
INCREMENT FACTOR -O.200000E+O1
NUMBER OF STEPS • 20
UARIABLE
LOAD
OUTPUT FACTOR 1 EU EL COHU.-NORI1 ITERATIONS
0. 1 0000000E+OO 0 O.f89E-09 4
O . 20000000E+OO 0 O.926E-08 4
3 0.3000000 OE+OO 0 0. 3OfE-07 4
4 O . I0000000E+O0 0 0.392E-07 4
5 O.50000000E+OO 0 O.357E-07 4
6 O . 60000000E+OO 0 O.314E-07 4
7 O.70000000E+O0 0 O.305E-07 4
8 O . 80000000E+0O 0 O . 343 E-O7 4
9 O . 90000000E+OO 0 O.459E-07 I
10 0. I0000000E+O1 0 O.71OE-07 I
11 0.1 1000000E+O1 0 0.1 18E-O6 I
12 0. 12000000E+01 0 O.187E-06 I
13 0. 13000000E+O1 0 0. 242E-06 I
14 0. 14O00000EO1 0 O . 244E-06 I
15 0. 1S000000E+O1 0 0.245E-06 4
16 0.1 6000000E+O1 0 O . 434E-06 I
17 0. 17000000E+O1 0 O.168E-11 5
18 0.1 8000000E+01 0 O.100E-08 5
0 0. 1 6200000E+O1 O.479E-12 I
0 0. 18400000E+O1 O.539E-1 1 I
0 0. 18600000E+O1 0. 197E-09 I
0 0. 18800000E+O1 O.803E-07 I
0 0. 18640000E+O1 2 O.736E-O7 3
0 0. 18880000E+01 2 O.499E-06 3
0 0. 18920000E+01 2 0. 340E-O9 4
0 0. 18928000E+01 3 0. 144E-08 3
0 0.1 8936000E+O1 3 O.929E-08 3
0 0. 18944000E+O1 3 0. 145E-O6 3
0 0. 18952000E+O1 3 O.462E-O7 I
PHASE( 1) TERMINATED
++++++++++++++++++++++
PHASE NUMBER - 2
NODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED '1
CONTROLLED NODE -
URRIABLE
DISPLACEMENT LOAD
OUTPUT INCREMENT FACTOR LEUEL CONU.-NORI1 ITER.
19 0. 12676187E+O1 O.18951791E+O1 0 0. 635E-09 3
20 0. 25352971E+O1 0.1 89O25O4E+O1 0 0. 121E-06 3
21 0. 50?05949E+01 0.1 8612923E+O1 0 0.1 IfE-lO 4
22 0.10141 190E+O2 0. 173O3899E+O1 0 0. 14 9E-O6 4
23 0.2O282379E+02 0. 10988723E+O1 0.22 1E-06 5
0 0. IO56f759E+01 O.77O76292E+O0 O.8O2E-1 1 6
0 0. 162259O4E+OO O.?45O8737E+0O 3 0. 9O8E-07 3
0 0. 16225904E+O0 0.71472775E.0O 3 0. 884E-06
291
0. 16225901 E.00 0. 6756239 1E+OO 3 0.5O6E-09
0. 1622590IE+O0 O.60712096E+OO 3 0. 258E-06
PHASE NUMBER • 2
NODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED - 2
CONTROLLED NODE - 3
URRIABLE
DISPLACEMENT LOAD
OUTPUT I HCREI1ENT FACTOR LEUEL CONU.-HORI1 ITER.
0 0.41 944479E+OO 0.581 67059E.00 0. 126E-O6 3
0 0. 4194 4479E.00 0.55552057E+OO O.317E-06 3
0 0.41 944479E+OO 0. 5281 5399E+OO 0. 939E-06 3
0 0.41 914479E.00 0.50019I81E+OO 0. 192E-1 1 I
21 0.11941479E+OO 0.47O37867E+OO O.618E-1O I
0 0.41 91 4479E+OO 0.43850362E+OO O.572E-O8 4
0 0.11 91ff 79E+OO O.fO383864E+OO 0.38 8E-1O 5
0 0. 83888959E-01 0. 396f8O75E+OO 2 0.47 1E-08 3
0 0. 83888959E-01 0. 388 95588E+OO 2 0. 795E-08 3
0 0. 83888959E-01 0.38 125O61E+OO 2 0. 135E-O7 3
0 0. 83888959E-01 O.37331996E+OO 2 0.225E-07 3
0 0. 83858959E-01 0. 36523680E+OO 2 0.36 1E-07 3
0 0.41 944479E+OO 0. 32069 129E.00 0.977E-07 I
0 0. 83888959E-01 O.3107697?E+OO 2 0. 796E-06 3
0 0. 83888959E-01 0. 3004 1029E+OO 2 0.300E-1 1 4
0 0. 8388 8959E-Ot 0.28955005E+00 2 0. 126E-1O f
0 0. 83888959E-01 0.2781 10??E+0O 2 0.215E-1O 4
25 0. 83888959E-Ol O . 26599233E+OO 2 0.321E-10 4
0 0. 83888959E-01 O . 25306309E+OO 2 O.383E-10 4
0 0. 83888959E-01 0.2391 1 390E+OO 2 0.489E-1O 4
0 0. 83888959E-01 O . 22397993E+OO 2 O.799E-1O I
0 0.83888959E-Ol O.20718640E+O0 2 0. 199E-09 4
0 0. 83688959E-01 0. 18813O88E+0O 2 0.976E-09 4
0 0. 83888g5gE-ol O.16562?89E+0O 2 0. 158E-07 I
0 0. 83888959E-01 0. 13685736E+O0 2 O.387E-1O 5
0 0. 16777792E-01 0. 12974569E+OO 3 0. 196E-07 3
0 0. 16777792E-01 0.12 187010E+O0 3 O.562E-07 3
0 0. 16777792E-01 0.1 1292542E+0O 3 0.215E-O6 3
0 0. 16777?92E-01 0. 1 0232483E.00 3 O.622E-1 1 4
0 0. 16777792E-01 O.88598328E-01 3 O . 337E-08 4
PHASE NUMBER - 2
NODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED - 2
CONTROLLED NODE - 2
U AR IA BLE
D I SPLACEI1EHT LOAD
OUTPUT INCREMENT FACTOR LEUEL CONU.-NORM ITER.
26 -. 10101251E+OO 0.72880333E-01 0 0.164E-O6 3
27 -.202085O1E.O0 0. 44 832164 E-01 0 0.731E-1 I 4
28 -.4041 ?003E+OO - . 260?9390E-02 0 0. 338E-08 I
-. 80831005E+0O -.78616533E-01 0 O.325E-06 4
292
B.3.2. 1 'LEE_STAT_5.DAT
%zxxxxxxxxxzxxx1xIxx%1x%z1xxxxx11xx1xxzzxxxz11xxxxxzxx1xxxxxxx:zxx*zxx
STRUCTURAL NODAL COORDINATES
NUMBER OF NODES
6
NODE X Y
1 0.0000 0.0000
2 0.0000 120.0000
3 24.0000 120.0000
4 120.0000 120.0000
5 0.0000 60.0000
6 12.0000 120.0000
xxxx1rxxzzxxzzxxzzxxxxxxxxxxzxxxxxxxxxxzxxxx11%z1%1xxxzxxzxxxxzxxzxxzz
ELEMENT CONNECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
293
B.3.2.2 'LEE_STAT_5.OUT
Iz1x1x1Xzzx%lXxxXXXXXZXZZXZZXlZXZXXZZzxxxxzxzxzxzxzzxzxxxxzxxxxxxxxzzzx
ELEMENT ASSEIIOLY ORDER
ORDER ELEI1ENT:(TYPE , IIUI1BER)
1 (3, 1)
2 (3, 2)
3 (3, 3)
4 (3, 4)
5 (3, 5)
IXXXIXZIZZIZZIZXXX%ZXXXXZ%ZXIXZZXIXIZXZXXZIXZXXXXX*XXZXZZXXXZXIZXXXXXXX
URRIABLE LOADING
++++++++++++++++++++++++++++++++
PHASE NUMBER - 1
TYPE - LORD CONTROL
INCREMENT FACTOR •O.200000E+O1
NUMBER OF STEPS - 20
LJARIABLE
LORD
OUTPUT FACTOR LEUEL
CONU -NORM
ITERATIONS
I O.10000000E+OO 0 O.325E-1O 4
0 O.16568000E+O1 3 0. 127E-06 5
PHASE( 1) TERMINATED
+ ++ + + ++++ +++ ++++++ ++++
PHASE NUMBER - 2
NODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED - 1
CONTROLLED NODE • 2
U AR IA BLE
DISPLACEMENT LOAD
OUTPUT INC RE liE MT FACTOR LEUEL CONU.-NORI1
hER.
19 O.95094226E+OO 0. 185?0528E+O1 0 O.505E-06 3
0 0.304301 52E-01 -.11205051E+OO 3 O.113E-11 6
294
PHASE NUI1BER - 2
MODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED • 1
CONTROLLED NODE - 5
UARIABLE
DISPLACEMENT LOAD
OUTPUT INCREMENT FACTOR LEUEL CONU.-NORM ITER.
34 0.7211 1015E+OO -. 15956601E+OO 0 0.720E-13 I
0 0.1 153?762E+OO -.67421f79E+0O 2 0.814E-07 5
PHASE NUMBER - 2
NODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED • 2
CONTROLLED NODE • 2
U AR IA BLE
DI SPLACEI1ENT LOAD
OUTPUT INCREMENT FACTOR LEUEL CONU.-NORM ITER.
41 -. 97055966E+O0 -.65154231E+00 0 0.641E-O7 3
50 -. 38822387E+O1 0.45439134E+O1 O O.157E-O7 5
B.3.3. 1 tEE_DYN_F_5.DAT
lxxxzxxxxzzxzzxxuxzxxzxzxxxxxxxxxxxzxxxzxxxzzxxxxxxzx%**zxz%11z%%1xzx1
20_EIe.ents/3D_EIe.ents (2/3)
2
xx11xxxxzxxxxxxxxxxxzxxxxx1xzxx1x1xxxz1xxxxxxzx1xxxxxxr1xxxxxz%zxxzxxx
ANALYSIS TYPE
EIGEN-UALUE STATICS DYNAMICS
0 0 1
ZIXIXXXIXIXXXXXXXZXZXXZZXZXXXIIIIXXIZI%XXXXXXJZIIXXIZIXZIXXXXXXX%XXXXI
MATER I ALS
+++++++++++,++,++,+,++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER OF MATERIALS
295
SECT tOIlS
++,++,++++,++++++++++++++++,+++++++++++++++++++++.++,+++++++++++,+++++
NuMBER OF SECTIONS
I1RTERIAL HO.(S)
OIMEHS ION(S)
O.300000E+O1 O.200000E+O1
zxx1zzxxxxzzxxx IXIZZIlIXIX I XIII XXXI IX IXIIIIXIIIXXXXIXXX 1111 1111111 1111
ELEMENT GROUPS
+++++++++++++++++++++++++++++++++++++++++++.+++++++++++++++ +++++++++++
GROUP MASS-PER-UNIT-LENGTH
1 O.210000E-O5
IIXXXXIXIIIIXIX1XXXXXXIXIIXXIIXXIXIXIIIIIIXXXXXIIIXXIXXXIIIIIIXIIIIIXX
ELEMENT CONHECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
296
ELEMENT TYPE (10) C Cubic floss Element )
NUMBER OF ELEMENTS
5
ELEMENT GROUP NODE1 NODE2
1 1 1 5
2 1 5 2
3 1 2 3
4 1 3 6
5 1 6 4
%XXXXXZXXXXIXXZXZXXXU%XXXXXXZXXSXZZZXZXXXXXXIZXXZXZZZZZZXZXXXXZXXXXXX
RESTRA I HIS
NUMBER OF RESTRAINTS
4
RESTRAINT NODE DIRECTION
1 1 1
2 1 2
3 4 1
4 4 2
x1x1xx11zzxxxxxzxxxzxxxz%xx1xzzzzxx%xzx1%xxxzxzxxxxxzxxxxxzxxxxzxxxx1z
LOAD I NO
+++++++++++++.++++++++++++++++++++++++++++++++++++++++++++++++++++..++
NEUFIARKS PARAMETERS
BETA GAMMA
0.250000E+OO O.500000E+OO
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER-OF-LOAD-CURVES STARTING-TIME
1 O.000000E+OO
LOAD-CURUE TIME-POINTS
1 1
TIME-POINT TIME LOAD-FACTOR
1 O.400000E+01 O.I00000E+01
++,+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
EQUILIBRIUM STAGES
NUMBER OF STAGES
ITERATIVE STARTEGY
ITERATIONS INITIAL-REFORMS. STEP-REDUC. ITER.-DIU.
6 6 5 5
CONVERGENCE CRITERIA
TOLERANCE FORCE/TRANS. MOMENT/ROT. MAX.-CONU.
0.100000E-05 0.200000E.01 0.100000E+03 0.100000E+03
++++++++++++++++++.++.+++++++++++++++++++++++++++++++++++..+,+++..+++++
297
OUTPUT
FREQUENCY STRESSES
5 0
B.3.4. 1 'LEE_DYN_D_5.DAT
+++++++++++++++++++++++++++++++++++++++++++++++++++++,+++++ ++++++++++.
HUM5ER OF URRYING LOADS
EQUILIBRIUM STAGES
NUMBER OF STAGES
298
B.4 Elastic Arch
IXXX XXX X XXXXXXXX XXXXX XXXXXXXXXXXXXXXXX XXX XXXXXXX XXX XXXXXXXXX XXX XX XXXXXX
20_Elements/3D_Elements (2/3)
2
XXXXX XXX ZXXXIX X XXX ZXXXXXXXXXXXXXZXXXXXXXXXXXX XXXXIXXXXX X XXX XXXXX ZXXXXX
ANALYSIS TYPE
EIGEN-UALUE STATICS DYNAMICS
0 1 0
XXXXXXXXXXXXXXX XXXXXZXXXXXXXXXXXX XXX X XX XXXXXXXXXXXXXXXXXXXX XX XXXXXXZXX
MATER I ALS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER OF MATERIALS
SECT I OHS
++.+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER OF SECTIONS
SECTION-NO. TYPE-NO.
1 1
MATERIAL HO.(S)
DIMENSION(S)
O.25f000E+02 O.4?6250E+O1
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXIXXXXXXX
ELEMENT GROUPS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
GROUP SECTION-NO.
1 1
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXZXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
299
ELEMENT COHHECTIUITY
RESTRA I HTS
NUMBER OF RESTRAINTS
6
RESTRAINT NODE DIRECTION
1 1 1
2 1 2
3 1 3
4 3 1
5 3 2
6 3 3
xZIXXIXXxIIIXxIXxXIXXXxxXXxxxxZxZZxZIXXXIXxZx*xXxIIxXxxXxxxxxIIXXZXxXX
LOAD I HG
NUMBER OF PHASES
300
PHASE Lood/DIsp./Auto.(1/2/3)
1 2
Hodol/EIe.ent(1/2) DIR. PtODE/ELEI1ENT(Type,Ho.)
1 2 2
DISP.-IIICREI1ENT Load/Unlood/Keep(1/2/3) STEPS
- . 500000E+02 3 25
+++++++++++++++++++++++++++++++++++++++++++++++++++++,+++++++++++++4++
I TERAT I UE STARTEGY
ITERATIONS ItIITIRL-REFORF1S. STEP-REDUC. ITER.-DIU.
4 4 5 3
OUTPUT
FREQUENCY STRESSES
0 0
This is the same as 'ARCH_2.DAT except for the number of nodes and elements.
IZZXXXX1XZXXXXX%Xx1ZXXxXX%Z%XXXXXXXXXZxXZXXXZZZXXXZZXXZXXXiXxlXXXxXXXZX
ELEMENT CONNECTIUITV
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
301
IMPERFECTIONS C Element Types [ 2, 3, 4, 32, 33, 34 ] )
+++++++++++++++++++++,+++++..+++++++++++++.++++++++++++++++++++++++++,
B.5 K-Frame
B.5.1.1 'K_PH.DAT
XIIXXXIIIXXXXIXXIXIIIIXIIXIIXIXXXXIXXXXIIXIXXIXIIIXXXXIXXXXXXXIXIXXIXX
20_Elements/3D_Elements (2/3)
2
IIIXXXIIIIXIIIIXXIIXXXIXIXXIIXXXIIXIXIIXIXIIXIXXXIXXIXIIXIIIIIXIXIIIXX
ANALYSIS TYPE
EIGEN-UALUE STATICS DYNAMICS
0 1 0
IIIXXXXXIXXIIXXXXIXXXXXIIXXIIIIIIXXIXXIXIXXXXIXXXXIXXXXXXXXIIIXIXXXIXX
MATER I ALS
NUMbER OF MATERIALS
2
MAT.-H0. MODEL-HO. PROPERTIES
SECT IOHS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER OF SECTIONS
2
I
302
SECT I OH-HO. TYPE-hO.
1 2
I1ATERIAL tIO.(S)
DIh1EHS ION(S)
0. 1O1?OOE+03 O.330000E+O1
SECTION-NO. TYPE-HO.
2 2
MATERIAL HO.(S)
2
DIMENSION(S)
O.219000E+03 O.f3?000E.01
1zx1xxzz zxxxx1zzx1z1z:zx xzxzzxzzx xxx xxx xxxxxxxxxx xxx xxx
ELEMENT GROUPS
ELEMENT CONHECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
303
RESTRA I HIS
NUMBER OF RESTRAINTS
10
RESTRAINT NODE DIRECTION
1 1 1
2 1 2
3 2 1
4 2 2
5 3 1
6 3 2
7 3 3
8 5 1
9 5 2
10 5 3
xxxx%xzzxxxxx1xzxxxsxxx1xxxx1zzxxxx1xxzxx1*xzzxxx1xzxxzxx1xxxz1zzxzxxx
LOAD ING
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Prop./TI.e-H I etory(1/2)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER OF PHASES
2
PHASE Load/Disp./Auto.(1/2/3)
2 3
Nodal/Element(1/2) Trane./Rot.(1/2)
1 1
Load/Un Ioad( 1/2) STOP-CONDITIONS
1 1
COND. L.F./OIep.(1/2) G.T. L.T. (HODE,DIR.)
1 2 0.000E+OO -.300E+03 4 2
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
ITERATIUE STARTEGY
ITERATIONS INITIAL-REFORMS. STEP-REDUC. ITER.-DIU.
10 10 10 6
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ++ + + + + + + + + + + + + + + + +_+ +
304
CONUERGENCE CRITERIA
TOLERANCE FORCE/TRANS. PIOMENT/ROT. I1AX.-CONU.
O.I00000E-05 O.500000E.06 O.100000E.08 O.100000E+05
OUTPUT
FREQUENCY STRESSES
0 0
111xxzxxzxzx1x1xxzxxzxxzxxxzzzxxzzzzzxzxxzzxxzxxxzx11zxzzzzzx*zzzxxxx1
B.5.1.2 'K_PH.OUT'
I%XXXXXX*XXXXZXXXXXXXXXXXXXXXZZXZXIZXXXXZX*XXIXIXZXXXXXXXXZXXXXIIXXXZXI
VARIABLE LOADING
+ +++ ++ + ++ +++ ++ ++ + ++++ +++ ++ + ++ +++
PHASE NUMBER - 1
TYPE - LOAD CONTROL
INCREMENT FACTOR -O.I00000E+OI
NUMBER OF STEPS • 25
VARIABLE
LOAD
OUTPUT FACTOR LEVEL
CONU. -NORM ITERATIONS
1 O.40000000E-01 0 0. 155E-07 2
2 O.80000000E-01 0 0. 242 E-O7 2
30 O.5?036000E+OO 3 O.153E-O?
PHASE( 1) TERMINATED
++ ++ + +++ + +++ + + ++++ ++++
PHASE NUMBER • 2
NODAL DISPLACEMENT CONTROL
FREEDOM CONTROLLED • 2
CONTROLLED NODE • 4
VARIABLE
DISPLACEMENT LOAD
OUTPUT INCREMENT FACTOR LEVEL CONU.-NORII ITER.
31 -. 736?4499E-03 O.5?038395E+OO 0 O.436E-08 2
P.H.F. ( 4, 2, 4)
305
32 -.44204699E-02 O.57029138E+OO 0 O.208E-08 2
33 -.26522320E-01 O.5683f792E+OO 0 O.162E-07 3
P.H.F. ( 4, 1, 1)
31 -. 10609128E+OO O.56249141E+OO 0 0.625E-O7 3
ssa*t**ssssssssss**s( ]s**ass**ssss*
SUBDIUISION OF ELEMENT( 4, 1)
306
B.5.2. 1 'K_DIS.DAT
This is identical to 'K_PH.DAT' except that elastic quartic elements are employed, and
automatic subdivision into elasto-plastic cubic elements is performed upon detection of
plasticity.
xzxxx%xx*xzzxzxzxxxxx%xxxuz111%x1xx1zxz111zx11x1x%xfl1%%z1x1xxxxxzzxx
hATER I ALS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
HUI1DER OF I1RTERIALS
2
I1AT.-HO. h100EL-NO. PROPERTIES
'xxxxxxxzxxxxizxzz11xxzxxxx1xxxxzxxx1xzxxxxx%1z1zzzxx11x1x111zsxzx1xx1
LEMEHT GROUPS
307
SUBDIUISION PATTERNS
NUMBER OF PATTERNS
2
PATTERN SUBELEMENTS RATIOS
1 10 1 2 3 2 1
1 2 3 2 1
2 10 1 2 3 4 5
5 4 3 2 1
XXXZXXXXZIXXXXXZIXZZXZXXZXXXXXIZZXXXXXXXXZXXZXZZXXXZXZZXXXZXZZXXZZXZZZ
ELEMENT CONNECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
308
B .6.1.1 DOM_DMP.DAT
xxxxzxrx1xx1xxzxxzzxxzx11xzxxx1xzzxz%zxzzx%xxx1zzxzzzzzxzxxxxzxxxzxxzx
2D_EIe.ents/30_EIeaente (2/3)
3
XXXIX XXIIIXXXXXXXXXXXX I XX XXXI XXXXXXXIX XXXIII I I I XIII XXXI
ANALYSIS TYPE
EIGEN-UALUE STATICS DYNAMICS
0 0 1
IXIXXIXXIXIXIXIIXXXIXXIIXIIIXIXIIXIXXXXIXXIXIXXXIIIXXIXXIIXZXIIIIIXXXI
MATER I ALS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER OF MATERIALS
SECTIONS
NUMBER OF SECTIONS
SECTION-NO. TYPE-NO.
1 1
MATERIAL I1O,(S)
DIMENSION(S)
O.760000E+OO 0. 122000E+O1
IIXXXIXXXIIXXXIXIXISIIXXXXZIIXXIXXXXXIIXXIIIXIIIXXXIIIIIIIXXXIIXIIIIXX
ELEMENT GROUPS
GROUP SECTION-NO.
1 1
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
309
ELEMENT COHNECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
310
9 1 9 10 15
10 1 10 11 15
11 1 11 12 16
12 1 12 7 16
13 1 7 13 1
14 1 8 13 2
15 1 9 13 3
16 1 10 13 4
17 1 11 13 5
18 1 12 13 6
+++++++++++,+++++++++++++++++++++++++++.++++++++++++,+++++++++++++++++
RESTRA I HTS
NUMBER OF RESTRA I hITS
36
RESTRAINT NODE DIRECTION
1 1
36 6 6
XXXX XXXZX XX X XXX XXX XX XXX XXXXXXXXXXXXXX XXXIXXXXXXXXXXXX XXXXXX XXXXX XX XXXI
LOAD I HG
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
HEUMARKS PARAMETERS
BETA GAMMA
0.250000E+OO O.500000E+OO
+++++++++++++++ +++++++++++++++++++++++++++++++++++++++++++++++++++++++
311
HUMBER-OF-LOAD-CURUES START I HG-TI liE
1 O.000000E+OO
EQUILIBRIUM STAGES
NUMBER OF STAGES
ITERATIUE STARTEGY
ITERATIONS INITIAL-REFORMS. STEP-REDUC. ITER.-DIU.
10 5 5 5
+++++++++++++++++++++++++++++++++++++++,++++++++++++++++++++++++++++++
CONUERGENCE CRITERIA
TOLERANCE FORCE/TRANS. MOMENT/ROT. I1AX.-CONU.
O.100000E-05 0. 100000E+09 O.100000E+09 O.100000E+04
+++++++++++++++.++++++++++++++++++++++++++++++++++++++++++++++++++++++
OUTPUT
FREQUENCY STRESSES
1 0
X XX X X XX XX XXXXXX XXXXXXXXXXXXX XXXXXXXXXXXX XXX XXXXXXXX XX XX XXXX XXX XXXXXXXX
IXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXIXXXXXXX
20_Elements/3D_Elements (2/3)
2
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
ANALYSIS TYPE
EIGEN-UALUE STATICS DYNAMICS
0 0 1
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
312
MATER I ALS
SECT I OHS
++++++++++++++++++++++++++++++4+++++++++++++++++++++++++++++++++++++++
NUMBER OF SECTIONS
2
MATERIAL NO.(S)
DIMENSION(S)
0.350000E+OO O.250000E-01 O.350000E+OO O.250000E-01
O.350000E+OO 0. 150000E-01
SECTION-HO. TYPE-HO.
2 3
MATERIAL NO.(S)
DIMENSION(S)
O.200000E+OO O.250000E-01 O.200000E+OO O.250000E-01
O . 500000E+OO 0.1 50000E-01
XXX XXXXX XZXXXXX XXXX X XX X XXXZXXXIZXXXZX XXZZX XXXXXXXXZXXXXXX X X XXXX XXX XX XX
ELEMENT GROUPS
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
313
SUBDIUISION PATTERNS
IIUI1BER OF PATTERNS
GROUP MASS
1 O.300000E.06
+++++++++++++++++++++++++++++++++++++++++,+++++++++,.+++,+++,++++,++++
ELEMENT CONNECTIUITY
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
314
RESTRA I HIS
NUMBER OF RESTRAINTS
4
RESTRAINT NODE DIRECTION
1 1 2
2 1 3
3 f 2
4 1 3
ZZX%XXZZIXXXXXZXXXXZXXXXXXZXXZXXX%XXXXXXZIXIIZIXIIIXX%X%X%111%XX%ZXXXZ
LOADING
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NEIJMARKS PARAMETERS
BETA GAMMA
O.250000E+OO O.500000E+OO
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NUMBER-OF-LOAD-CURUES STARTING-TIME
1 -. 100000E-01
LOAD-CURUE TIME-POINTS
1 251
TItlE-POINT TIME LOAD-FACTOR
1 O.000000E+OO 0. 1 20000E-02 (Earthquake record)
2 O.200000E-01 0.1 10000E-02
EQUILIBRIUM STAGES
NUMBER OF STAGES
2
STAGE END-OF-STAGE STEPS
1 O.000000E+00 1
2 0.500000E+01 250
315
ITERRTIUE STARTEGY
ITERATIOHS INITIAL-REFORMS. STEP-REDUC. ITER.-DIU.
10 5 5 6
CONUERGENCE CRITERIA
TOLERAHCE FORCE/TRRKS. I1OI1ENT/ROT. I1AX.-COIIU.
0.100000E-05 O.100000E+O? 0.100000E.07 O.100000E+O4
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
OUTPUT
FREQUENCY STRESSES
2 0
XX XXX XXX XX X XXX X XXIX XXXXXXXXXXXXXXXXXX XXX XXX XXXXXXXXXX XXXXXX XXIX XXXXXXX
316