Professional Documents
Culture Documents
To cite this article: A H Alwathaf, W A Thanoon, M S Jaafar & J Noorzaei (2012) Mathematical
Modelling of Stress-Strain Curves of Masonry Materials, Australian Journal of Structural
Engineering, 13:3, 219-230
Article views: 7
AH Alwathaf†
Civil Engineering Department, Faculty of Engineering, Sana’a University, Sana’a, Yemen
WA Thanoon
College of Engineering and Architecture, University of Nizwa, Nizwa, Oman
2. In Desayi & Krishnan’s (1964) equation, the ratio of different curvature on the ascending and also in
initial tangent modulus (Eo) to the secant modulus the descending part. This is because the material
(E s = σ o/ ε o) should be equal to 2, otherwise parameter, A, is the power of the third term of
the expression yields higher or lower than the the denominator. This is a significant feature in
maximum stress at the peak of the test data. this equation because the other relations have
3. It can be seen that the Desayi & Krishnan (1964) constant curvature. Equation (6) is the derivative
equation is a special case from the Saenz (1964) of equation (5) at the origin point when εo = 0,
equation in which Eo = 2Es. However, the ratio of however, this equation is not reliable to estimate
initial tangent modulus to the secant modulus the material parameter, A. The accuracy of this
(E o/ E s) should be equal or greater than 2 in
method depends on the accurate estimation of the
the Saenz (1964) equation to avoid a concave slope of the stress-strain curve at the origin and
curve at the lower stress level. Saenz’s formula hence many trails are needed to find the suitable
is frequently used in the concrete modelling initial slope. Carreira & Chu (1985) proposed
under biaxial stress conditions when the material different empirical expressions to estimate the
non-linearity is included (Chen, 1982; Cerioni material parameter. However, their proposed
& Doinda, 1994). However, the formula is not formulas are not applicable to the test specimens in
suitable for the descending part because the this study due to some restrictions and conditions
inflection point cannot be provided by the that mostly concern specimen shape.
formula. Therefore, some authors have used 6. Thanoon (1997) observed that the best fit results
linear formula in the descending part instead of are obtained by his equation defining two levels
Saenz’s formula in their stress-strain modelling of stress, one at 0.45σo stress level and the other
(Cerioni & Doinda, 1994). at 0.85σo. The present coefficient values in table
4. An attempt was done by Wang et al (1978) to fit 1 for Thanoon’s equation are assigned for plain
concrete obtained from his test result.
the concrete data, however, this equation can only
fit test data having certain values of compressive Some concluding remarks on the reviewed studies
strength and corresponding strain within a can be drawn as follows:
specific range. 1. The uniaxial stress-strain relations with constant
5. Equation (5) used by Carreira & Chu (1985) is coefficients are not suitable for biaxial modelling
capable of simulating the stress-strain relation because these coefficients restrict using those
for different concrete materials that may show expressions in non-linear analysis where different
family curves are produced due to the different standard specifications. For masonry blocks,
biaxial stress states during the incremental loading. however, this choice is limited because concrete
2. Estimating the initial tangent modulus at the block units are moulded by compaction and
origin from the test data accurately is quite vibration simultaneously, rather than usual casting.
difficult, particularly for some stress-strain curves At the beginning, the uniaxial compression test was
where the curvature is varied. conducted on coupons (40×40×100 mm) cut from
3. Equation (5) seems to be able to fit properly the block units face-shell to obtain stress and strain as
test data provided that a suitable method is used uniform as possible, and also to minimise the end
to estimate the material parameter, A. This formula platen restrain. However, the coupons results were
can be also incorporated easily in the modelling quite different and even contradictory. This is due
of masonry with different material properties. to the non-uniform cross-sections of the face-shell,
which commonly tapered. Unfortunately, there is no
In this study, compression tests of concrete masonry sufficient thick intersection between the face-shell and
blocks units and concrete grout cylinders have web that could be cut and used in the coupon test.
been carried out to obtain the experimental stress-
strain diagrams. Numerical manipulation has been A comprehensive investigation conducted by Guo
conducted on the experimental results to find the (1991) showed that the strength ratios of block
best fit formulas for the experimental diagrams. coupons to the corresponding block units varied
Non-linear regression analysis technique has been from 0.96 to 0.93, and the accuracy can be expected
used to obtain the proper material parameters. The within that limit. Moreover, uniaxial compression
relationship has been extended to masonry biaxial test using block units were conducted by some
stress modelling. Masonry block and grouted prisms researchers to investigate the stress-strain behaviour
have been analysed by finite element programs to and to establish the strength characteristic of the
verify the stress-strain model. block material (Page, 1978; Ali & Page, 1988; Guo &
Drysdal, 1989; Guo, 1991). Based on the above, the
uniaxial compression test was conducted on the block
2 COMPRESSION TEST OF units instead of the coupons.
BLOCK AND GROUT
Specimens of this test consist of different individual
2.1 Test specimens concrete block units (15 stretchers, 12 corners, and
10 halves) and concrete grout cylinders (5 cylinders).
The standard concrete cylinder or cube is most Figures 1(a) and 1(b) shows the block and grout
commonly used for concrete material tests in specimens. The test block called interlocking hollow
(a)
(b)
Figure 1: Compression test specimens – (a) corner, stretcher and half block units; and
(b) grout cylinders.
block was obtained by interlocking keys to ensure highly affected by the block density, in which higher
efficient construction formation with well-aligned density revealed higher compressive strength.
masonry structures without mortar. The blocks were The variation in the block density is due to the
prepared for test by cutting the extended projections unavoidable variation in the machine compaction
from the top. To assure planarity and smoothness of and vibration during casting process. Elastic
the top and bottom surfaces of the specimens, silicon modulus (Ec) in this table is the slope of a straight line
carbide stone was used for grinding of the surfaces from the origin to 0.45 of the compressive strength of
and also mortar plaster was cast as a capping on the the material (0.45σo) on the stress-strain curve.
surfaces of the specimens. The tests were started 28
The resulted stress-strain curves for all block unit
days after curing and storing of block units and grout
specimens are grouped into three typical curves
specimens inside the lab. Preparing and testing the
according to their strength (higher, medium and
block units and grout cylinders were according to the
lower) as shown in figure 2. All block unit types
requirements of ASTM C140-99b (ASTM, 1999) and
exhibited stress-strain behaviour as shown in
ASTM C 469-94 (ASTM, 1994), respectively.
figure 2, in which each type has relatively three
modes of deformation. The stress-strain curves in
2.2 Compression test setup and test procedure the figure are non-linear, and consist of ascending
and descending branches. As shown in the figure,
All specimens were tested under displacement the specimen units revealed different behaviours
control mode to allow constant strain rate, which according to the strength of the block units. These
yields a complete loading history for the specimens curves also indicate that the stronger material has a
in the ascending and descending portions. The brittle behaviour compared to weaker material. This
load was applied continuously with a constant is consistent with the observed failure during the test
stroke of 0.005 mm/s until failure, as suggested by in which quite sudden failure was observed in blocks
Neville (1995). The specimens were instrumented having higher strength. It can be also observed that
to measure the axial deformation using two linear the modulus of elasticity increases with the increase
variable displacement transducers of 0.001 mm of compressive strength. For structural analysis of
precision located vertically on two opposite sides on masonry wall, it is necessary to define the strain (εo)
the specimens using special glue. Load and vertical at the level of maximum stress (σo) and the elastic
deformations were measured directly by the machine
acquisition system. The gauge length that used for all
block units was 100 and 120 mm for grout cylinders.
3 SELECTION OF STRESS-STRAIN
RELATION
where A is the material parameter that was were obtained. Both of the estimated initial tangent
determined according to equation (6) as follows: moduli yield an inaccurate stress-strain curve as
shown in the fitted curve in figure 8, because they
1
A (6) provide improper material parameters. Therefore, if
1 (Es Eo ) the initial tangent modulus method is used for the
present test curve, many trials are needed to find the
The material parameter expression (equation (6)) is appropriate range of the initial slope, Eo. Therefore, a
not suitable for curve fitting of the tested masonry different method should be proposed to evaluate the
materials conducted in this study because it depends material parameter based on the entire experimental
on the initial slope at the origin, Eo. To explain this data, not only on the data at the origin point.
shortcoming, figures 7 and 8 show how the initial
modulus could yield inappropriate values for the
material parameters via equation (6). The data used 4 PROPOSED TECHNIQUE FOR MATERIAL
in the figures is for the typical stress-strain test data of PARAMETER DETERMINATION
the grout specimen that is selected to clarify the idea.
The maximum compressive stress of the specimen is While some stress-strain test curves are quite difficult
22.29 MPa and the corresponding strain is 0.0021. The to estimate their initial slopes and since the stress-
initial tangent modulus is expected, graphically, to be strain data near zero loads is very sensitive to the
within two values (30,000 and 23,000 MPa) because setting of the deformation measurement, the entire
the curvature is not clear, and varies in the vicinity of stress-strain diagram should be much less sensitive
the origin point as shown in figure 7. It can be seen to the setting. Based on this consideration, it seems
from the figure that the third value of Eo (17,336 MPa) reasonable to determine the material parameter, A,
is out of the expected range because it seems that this based on the whole stress-strain test diagram.
line cannot be considered as a tangent for the curve at In this study, non-linear regression analysis is used to
the origin point. Substituting these two initial tangent obtain the material parameter (A) of the stress-strain
moduli into equation (6) to obtain the corresponding relation in equation (5) based on the entire diagram
material parameters and using equation (5) to draw of the test data. Non-linear regression is a method of
the possible stress-strain curves, two fitted curves finding a non-linear model of the relationship between
the dependent variable and a set of independent
variables (Bates & Watts, 1998). The dependent
variable in equation (5) is the stress ratio (σ/σo) after
normalising the stress using the maximum stress at
the peak, σo. Furthermore, the independent variable
is the strain ratio (ε/εo). Non-linear regression can
estimate models with arbitrary relationships between
independent and dependent variables. This is
accomplished using iterative estimation algorithms.
The non-linear regression analysis technique was
applied for the previous test curve to find a more
accurate value for the material parameter and
A = 2.56 was obtained. The resulted value was used to
find the best fit curve by applying equation (5). Figure
Figure 7: Initial tangent modulus estimation at 9 shows a comparison between the fitted curves using
the origin for a specimen. the non-linear regression analysis results (A = 2.56)
and the previous results obtained by equation (6)
(A = 1.54 and 1.85 for different initial tangent moduli).
As shown in figure 9, employing the proposed
technique yields an excellent fitted curve with
coefficient of determination of R2 = 0.994. It is worth
to mention here that the initial tangent modulus
that gives the accurate curve fitting is the third line
in figure 7, which is 17,336 MPa. This value is the
initial slope of the best fit curve and can be calculated
from the derivative of equation (5) (equation (10)
in section 5) or from equation (6) using A = 2.56.
Therefore, the initial tangent modulus at the origin
is not a reliable method to evaluate the material
parameter and hence the proposed technique based
Figure 8: Stress-strain curve based on the on performing non-linear regression analysis of the
estimated initial tangent moduli. whole experimental data exhibited accurate results.
Figure 9: Stress-strain data fitting comparison. Figure 10: Stress-strain data fitting of block unit.
30
/mm2
Compressive stress NMPa
curve fitting has been obtained. The test data and 20
Figure 12 shows two equivalent uniaxial stress-strain element code is developed to implement the
curves of the principal directions of an element loaded proposed constitutive model. Moreover, numerical
by biaxial compression stresses σ1 and σ2 (|σ1| < procedure and solution algorithm are proposed
|σ2|). The equivalent uniaxial strain for direction i is to solve the non-linear system. More details about
εiu (i = 1, 2), σip and εip are the equivalent uniaxial peak the model and solution algorithm can be obtained
and the corresponding strain, and Eit is the tangent elsewhere (Alwathaf, 2006).
modulus of elasticity at the studied principal stress σi.
The equivalent uniaxial strain εiu essentially removes
Poisson’s effect, whereas the strengthening due to the 6 FINITE ELEMENT ANALYSIS
micro-cracking confinement in biaxial compression
stress and softening in compression-tension stress Masonry concrete block unit and grouted masonry
fields are incorporated in σip and εip, respectively. Thus prism were simulated under vertical loads and non-
a single relation (equation (8)) can represent the infinite linear finite element analysis (FEA) was conducted.
variety of monotonic biaxial loading curves. It can be The stretcher block unit was used in the simulation
seen in figure 12, higher confinement is provided by of the block and grouted prism. The simulated prism
σ2 in direction 1. As a result, this higher confinement was constructed from three courses without mortar
increases the concrete strength in the orthogonal and hence mortarless (dry) joints were formed. The
direction 1 in which σ1p > σ2p. dry joint properties were established experimentally
and incorporated in the simulation. However, its
The secant modulus at any point on the curve can be characteristics were not added to the study because
found by dividing equation (8) by εiu and expressed this study focuses on the masonry material stress-
as follows: strain characteristic. More details about the dry joint
A(H iu H ip )V ip characteristics can be found elsewhere (Jaafar et al,
Eis (9) 2006; Thanoon et al, 2008a; 2008b).
[ A 1 (H iu H ip )A ]H iu
Previous studies on hollow block masonry systems
Also, the tangent moduli E1t and E2t for a given under compression showed that the differences in
principal stress directions are found as the slopes ultimate strength using 2D (plane stress and plane
of the σ1 versus ε1u and σ2 versus ε2u curves for the strain) and 3D FEA were not more than 5% (Suwalski
current ε1u and ε2u as follows: & Drysdale, 1986; Guo, 1991). The accuracy can
be increased if the FE mesh is used in the critical
AEs ª¬ A 1 (H iu / H ip )A A(H iu / H ip )A º¼ section where the failure is initiated. Also using
Eit 2 (10) 3D discretisation for non-linear analysis involves
ª¬ A 1 (H iu / H ip )A º¼
higher computational cost. Therefore, plane stress
2D continuum is adopted in this study.
where Es is the secant modulus at the peak (maximum
stress) σip/εip. The properties used in the analysis of block and
grout material are shown in table 2. In table 2,
A detailed model for masonry analysis is proposed, the initial tangent modulus at the origin (Eo) was
and an incremental-iterative 2D non-linear finite obtained by substituting the compressive strength
(σo) and corresponding strain (εo) into equation (10).
-ıi To examine the stress-strain model, the FEA results
were compared with those found experimentally for
2
ı1p 1
the unit block and grouted prism (Alwathaf, 2006).
ı2p
ı2 ı2 6.1 Block unit
ı1
ı1
As the entire compressive loads of a block unit are
carried by the block face-shells due to the removal
of top projections, the FEA was conducted on the
face-shell only as shown in the 2D mesh of the block
- İiu
İ1u İ2u İ2p İ1p face-shell in figure 13. The boundary condition at the
top and the bottom was also modelled as shown in
Figure 12: Equivalent uniaxial stress-strain of an figure 13, which simulates the end platen restrain
element under biaxial compression. against lateral deformation during the test.
350
/mm 2
20 Demec
MPa
Points
Compressive load kN
300
Compressive Stress N
15 250
200
10
FE model 150 FE model
Test result - - - - Test Results
PRG1
5 100
PRG2
50
PRG3
0 0
0 0.001 0.002 0.003 0.004 0 0.4 0.8 1.2 1.6
Strain mm/mm Axial deformation mm
Figure 14: Stress-strain curves of block unit. Figure 16: Mode of deformation of grouted prism.
y y
x
z x
(a) (b)
test), which is 8.0 % higher than the average strength Carreira, D. J. & Chu, K. C. 1985, “Stress-Strain
of the tested prisms. Relationship for Plain Concrete in Compression”,
ACI Journal, Vol. 82, pp. 797-804.
7 CONCLUSION Cerioni, R. & Doinda, G. 1994, “A Finite Element
Model for the Nonlinear Analysis of Reinforced and
Stress-strain behaviour for masonry materials Prestressed Masonry Wall”, Computer and Structures,
under axial compression has been investigated Vol. 53, pp. 1291-1306.
experimentally and analytically in this study. The
investigation includes the material hardening and Chen, W. F. 1982, Plasticity in reinforced concrete,
softening parts to obtain the complete structural McGraw-Hill Book Company, USA.
response.
It has been found that the stress-strain data of the Darwin, D. & Pecknold, D. 1977, “Nonlinear biaxial
masonry materials (block and grout) has been stress-strain law for concrete”, Journal of Engineering
accurately fitted by the employed stress-strain Mechanics Division, ASCE, Vol. 103, No. 4, pp. 229-241.
equation after using the whole diagram data to obtain
the material parameter. The non-linear regression Desayi, P. & Krishnan, S. 1964, “Equation for the
analysis for the entire experimental stress-strain Stress-Strain Curve of Concrete”, ACI Journal, Vol.
curve yielded accurate material parameters. A perfect 61, No. 3, pp. 345-350.
formula for any concrete material has been obtained
by this technique. Foster, S. & Gilbert, R. 1996, “Rotating crack finite
element model for reinforced concrete structures”,
The best fit stress-strain relation proposed for Computers & Structures, Vol. 58, pp. 43-50.
masonry block and grout material was implemented
effectively in the FE modelling in order to estimate Guo, P. 1991, “Investigation and Modelling of the
accurately the non-linearity of different masonry Mechanical Properties of Masonry”, PhD Thesis,
materials under loading in the uniaxial and biaxial McMaster University, Hamilton, Ontario, Canada.
stress states. Comparable deformation and strength
was obtained by the FE mode compared to the test Guo, P. & Drysdal, R. G. 1989, “Stress-Strain
result of the block unit and prism. Relationship for Hollow Concrete Block in
Compression”, Proceeding of 5th Canadian Masonry
REFERENCES Symposium, Dept. of Civil Engineering, University
of British Colombia, Vancouver, BC, Canada, pp. 599.
Ali, S. S. & Page, A. W. 1988, “Finite Element Model
Hognestad, E. 1951, “A Study Of Combined Bending
for Masonry Subjected to Concentrated Loads”,
And Axial Load In Reinforced Concrete Member”,
Journal of Structural Engineering, ASCE, Vol. 114, No.
University of Illinois Engineering Experimental
8, pp. 1761-1784.
Station, Bulletin Series No. 399.
Alwathaf, A. H. 2006, “Development of Finite Jaafar, M. S., Alwathaf, A. H., Thanoon, W. A.,
Element Code for Non-linear Analysis of Interlocking Noorzaei, J. & AbdulKadir, M. R. 2006, “Behaviour of
Mortarless Masonry System”, PhD thesis, University Interlocking Mortarless Block Masonry”, Proceedings
Putra Malaysia, Malaysia. of ICE, Construction Materials, Vol. 159, No. 3, pp.
111-117.
American Society for Testing and Materials (ASTM),
1994, ASTM C 469-94 Static Modulus of Elasticity Neville, A. M. 1995, Properties of Concrete, Longman
and Poisson’s Ratio of Concrete in Compression, Group, London, UK.
Philadelphia, PA.
Page, A. W. 1978, “A Finite Element Model for
American Society for Testing and Materials (ASTM), Masonry”, Journal of Structural Division, ASCE, Vol.
1999, ASTM C140-99b Sampling and Testing Concrete 104, No. 8, pp. 1267-1285.
Masonry Units and Related Units, Philadelphia, PA.
Park, P. & Paulay, T. 1975, Reinforced Concrete
Structures, John Wiley & Sons Inc., USA.
Ayoub, A. & Filippou, F. 1998, “Nonlinear finite
element analysis of RC shear panels and walls”, Saenz, L. P. 1964, “Discussion of ‘Equation for the
Journal of Structural Engineering, ASCE, Vol. 124, No. stress-strain curve of concrete’, by Desayi and
3, pp. 298-308. Krishnan”, ACI journal, Vol. 61, No. 9, pp. 1229-1235.
Bates, D. M. & Watts, D. G. 1998, Nonlinear Regression Suwalski, P. & Drysdale, R. 1986, “Influence of
and Its Applications, Wiley, New York. Slenderness on the capacity of concrete Block Walls”,
AHMED ALWATHAF
WALEED THANOON
MUHAMMAD JAAFAR
JAMALODIN NOORZAEI