ASSESSMENT OF THE APPLICABILITY OF

NONLINEAR
DRUCKER-PRAGER MODEL WITH CAP TO ADOBE

Luis Quiroz Torres

SUMMARY
In order to adequately describe the behavior of materials, it is necessary to use mathematical constitutive models
to represent their response under external loads. The constitutive models are composed of relationship between
the stress tensor and the strain tensor. They represent an idealized description of the actual behavior. The two
basic models frequently used are the elastic and ideally plastic models. Real materials almost never match the
conditions defined by models mentioned before but these were mainly used because of its simplicity, which is
essential for professional practice. In recent years, with the development of the theory of plasticity as well as the
use of increasingly powerful computers and friendly computer programs, new elasto-plastic models have
appeared. These models describe the non-linear characteristics of various materials such as concrete, soil, rock,
etc. In this paper, the Drucker-Prager model with cap is used and its applicability to adobe is evaluated using the
finite element program ADINA. A comparison of results get from numerical simulations and tests carried out at
CISMID was done. It was found that considering Drucker-Prager failure envelope, it is possible to get an
acceptable approximation of the stress-strain behavior in terms of initial stiffness and ultimate strength with
differences in some cases of 10%, but the model presents a less gradual transition between the elastic and plastic
behavior. The model does not reproduce correctly the strength degradation observed in tests.

Keywords: Adobe, constitutive model, plasticity, nonlinearity, Drucker-Prager

1. INTRODUCTION

Many old structures such as Chan-Chan (Figure 1.1) and buildings constructed in the colonial period
(Figure 1.2) were built on adobe and / or mud with material available near the structures and local
workers as people in charge of construction. Over the years, due to their design or function, these
structures are considered historic and they have important cultural significance to society.

Figure 1.1. Chan-Chan, The world's largest clay Figure 1.2. Colonial houses (adobe and quincha.)
city. Trujillo, Perú Lima, Perú

It is known that these structures have high vulnerability (Figure 1.3). In several earthquakes occurred
around the world, these structures presented bad behavior and in some cases they collapsed quickly
due to earthquakes of moderate magnitude (Figure 1.4) causing economic and cultural losses and also
loss of human life.

Figure 1.3. The citadel of Bam after the earthquake of Figure 1.4. Partial collapse of adobe structures. Ica,
2003 (6.2 degrees on the Richter scale) 1997.

To reduce the vulnerability of adobe structures, it is important to understand the behavior of the
material they are made. Adobe is a complex material that due to its differential behavior requires
elaborated constitutive models (Blondet et al (2002)). These models are defined by idealized
parameters that need to be adjusted or calibrated through test. This calibration is very important if we
want to have an acceptable estimation of the results obtained by numerical simulations. The particular
objective of the calibration is to optimize the efficiency of finite element models for the prediction of
displacements and stress as close as possible to the real, though any model inherently has assumptions
and approximations that make a constraint between the real and simulated behavior (Bathe (1996),
Zienkiewiz et al (2005)). When the efficiency of the model is achieved, it can be used for other types
of analysis with different loading histories.

2. DRUCKER-PRAGER MODEL WITH CAP

Figure 2.1. Drucker-Prager model with cap

Figure 2.1 shows the Drucker-Prager model with cap (Kojic et al (2005)). It consists of a fixed line
defined by equation (2.1) and a cap defined by equation (2.2)

fDP  I1  J2D  k  0 (2.1)

Where  and k are constants of material, I1 and J2D are the first invariant of stresses and the second
invariant of deviatoric stresses. X is the location of the cap and it depends on the volumetric plastic
strain.
DiMaggio and Sandler in 1971 proposed a definition for the hardening law of the cap as follow

1  eVP  0
X  ln  1   X (2.3)
D  W

Where W and D are constants of material, ePV is volumetric plastic strain and 0X represents the initial
location of the cap.

3. COMPUTER PROGRAMS BASED ON FINIT ELEMENT METHOD

The finite element method (FEM) is widely applied in engineering practice through a number of
modern programs, giving great potential for solving problems of structural analysis. This family of
programs can be divided into two groups: general purpose programs and specialized programs.

The general purpose programs (such as ANSYS, NASTRAN, ABACUS, ADINA, PAK, DIANA) can
be used to obtain numerical simulations similar to real behavior. The results are then subject to
engineering judgment with empirical data of the materials, service specifications and other criteria.
Specialized programs in the field of civil engineering (such as SAP2000, STAAD, TOWER, ETABS)
are also used to model structures and the results can be transferred to appropriate algorithms for sizing
and design of structural elements according to selected codes. Thus, specialized software improve the
quantitative performance of the designer during the analysis while the general purpose software
changes the point of view of the designer to quality terms (Zoran Bonic Todor Vacev et al (2010a, b)).
In order to model the specimens by considering the problem of nonlinearity of the material and the
different states of stress and strain, a general purpose program is the most logical choice. The program
selected for this purpose was ADINA.

4. OUTLINE OF TESTS

From previous studies conducted at CISMID, it was selected some test results. The tests selected can
be classified as follows: axial compression tests on prisms and diagonal compression test on low walls.
Triaxial compression tests on cylindrical specimens were carried out in the geotechnical laboratory of
CISMID for this investigation.

4.1. Compression tests on adobe prisms

It was selected the results of five compression tests on adobe prisms of the project developed by JICA
and CEETyDES (2009). The purpose of these tests was to understand the behavior of adobe to axial
loads. In Figure 4.1 it is shown one of the specimens tested. The average dimensions of the specimens
were 37 cm side at the base and 47 cm height. Figure 4.2 shows characteristic curves obtained from
test. It was found that the average compression strength due to the axial load is about 8.83 kgf/cm 2.

4.2. Diagonal compression tests on adobe low walls

Of the same project mentioned above, it was selected the results of five diagonal compression tests on
adobe low walls. The purpose of these tests was to understand the behavior of adobe to axial load and
shear. The dimensions of the specimens were 100 cm side and 20 cm thick. The results show greater
variability compared to axial compression tests. It was found that the diagonal compression strength is
in average 0,266 kgf/cm2. Figure 4.3 shows one specimen and Figure 4.4 shows typical curves get
from diagonal compression tests.
 Figure 4.1. Adobe specimen tested to compression Figure 4.2. Typical curve get from compression test
on adobe prisms

Figure 4.3. Adobe specimen tested to diagonal Figure 4.4. Typical curve get from diagonal
compression compression test on adobe low walls

4.3. Triaxial compression tests on adobe specimens

The material used to manufacture the cylindrical specimens were adobe bricks remaining from tests
presented in sections 4.1 and 4.2. The dimensions of the specimens were in average 4.85 cm diameter
and 9.5 cm high, due to the requirements of the triaxial testing machine. It was tested two sets of
specimens (6 specimens) with different level of confinement: 1 kgf/cm 2, 2 kgf/cm2 and 4 kgf/cm2.
Figure 4.5 shows the test of one specimen and Figure 4.6 shows typical curves get from triaxial
compression test.

Figure 4.5. Triaxial test of adobe Figure 4.6. Typical curves get from triaxial compression test
specimen
5. ESTIMATION OF PARAMETERS OF DRUCKER-PRAGER MODEL WITH CAP

5.1. Estimation of elastic parameters (E, )

The curves obtained from compression tests on prisms were used to estimate the first parameter.
Figure 5.1 shows the curves in terms of vertical stress and vertical strain.

10

8
Vertical stress (s - gkf/cm2 )

4
Prism 1
3
Prism 2
2 Prism 3
1 Prism 4
Prism 5
0
0 0.005 0.01 0.015 0.02 0.025 0.03
Vertical strain (e)

Figure 5.1. Curves obtained from compression tests on adobe prisms

The modulus of elasticity (E) was estimated based on the slope of the curves presented above using
two ranges. The first range was set between values of 0% to 25% of maximum compression stress and
the second range was set between values of 25% to 50% of maximum compression stress. In case of
the first range, it was found a value of 724 kgf/cm2 for E, while for second range, it was found a value
of 820 kgf/cm2. Based on the results presented before, a value of 800 kgf/cm 2 was set for parameter E.
Similar procedure was followed with experimental results from triaxial compression test. It was found
that E increased in 15% with a confinement stress of 4 kgf/cm 2.

There is no available information from test to quantify a Poisson's ratio. A value of 0.25 was set
because of the low variation of reference values reviewed (Yamin et al. (2003), Wang et al. (2001),
Roonsson and Boothby (1998), Mroginski et al. (2006), Lopez et al. (2000).)

5.2. Estimation of parameters that define yielding surface (, )

As can be observed in Equation (2.1), the yielding surface of Drucker-Prager is expressed in terms of
the first invariant of stresses (I1), and the second invariant of deviatoric stresses (J 2D). Based on the
results of compression tests on prisms and triaxial compression tests on cylindrical specimens of
adobe, the maximum values of I1 and J2D are estimated and presented in Table 5.1. sc is the stress of
confinement of specimens.

Figure 5.2 shows a graphical representation of Table 5.1. It is observed that the points fit to a line. To
estimate the equation of the line, it was used a linear regression (Motulsky et. al. (2003)). The fitted
line is presented in Equation (5.1) and the values of the parameters that define the yielding surface are
 = 0.3342 y  = 2.3447 kgf/cm2.

√(J2D) = 0.3342(I1) + 2.3447 (5.1)

In Figure 5.2, it is also shown the value of the coefficient of determination, which is close to 1. This
would indicate that failures in the specimens are given in the Drucker-Prager line.
Table 5.1. Maximum values of I1 and √(J2D) get from compression tests on prisms and triaxial compression tests
on cylindrical specimens of adobe
Test sc I1 √(J2D)
(kgf/cm2) (kgf/cm2) (kgf/cm2)
Prism 1 0 7.93 4.58
Prism 2 0 9.20 5.31
Prism 3 0 8.94 5.16
Prism 4 0 9.35 5.40
Prism 5 0 8.70 5.02
Set1-M1 1 17.98 8.65
Set1-M2 1 20.69 10.21
Set2-M1 2 25.41 11.21
Set2-M2 2 25.1 11.03
Set3-M1 4 36.4 14.09
Set3-M2 4 36.07 13.90

18

16

14
R² = 0.9858
12

10
√(J2D) ((kgf/cm2 )

Compression tests
2
Triaxial compression tests
0
0 5 10 15 20 25 30 35 40 45
I 1 (kgf/cm2 )

Figure 5.2. Yielding surface of Drucker-Prager obtained from tests

5.3. Estimation of parameters that define the cap (W, D, 0X)

The values of parameters W and D can be estimated from hydrostatic test results (Desai (1984)). From
the results presented in previous section, it is clear that the failure occurs in the yielding surface and
the cap is not reached. Based on that, it can be said that the initial location of the cap ( 0X) must be
larger than 36.4 kgf/cm2. For the numerical simulations, it was set a value of 40 kgf/cm 2 for parameter
0
X.

From Kojic, Milos; Bathe, Klaus-Jurgen (2005), it was found that the value of W varies between 0.066
and 0.18 while the value of parameter D varies between 0.00953 (kgf/cm 2)-1 and 0.711 (kgf/cm2)-1. In
the present study, it was considered the value of 0.18 for parameter W and the value of 0.711
(kgf/cm2)-1 for parameter D. As it was mentioned before, the cap is not reached in the case of the tests
presented in section 4.

5.4. Estimation of tension cutoff (T)

In order to model the material failure due to the effects of tension, it was set a limit (T). For the present
study, the value of T was set to zero because adobe does not support tension stresses.
6. NUMERICAL SIMULATIONS

6.1. Numerical simulation of compression tests

Three-dimensional solid isoparametric elements of eight nodes were used to simulate the compression
tests. As can be observed in Figure 6.1, a quarter of the specimen was modeled. In the same figure, it is
shown the boundary conditions and loads (vertical displacements () at top face) used in the model.

6.2. Numerical simulation of triaxial compression tests

In this case, two dimensional isoparametric elements of eight nodes based on displacement were used.
The assumption of axisymmetric element was considered. Figure 13 shows a scheme of the model
with its boundary conditions and loads applied (vertical displacement () at top and stress of
confinement (s)).

Figure 6.2. Schema of model for triaxial

6.3 Numerical simulation of diagonal compression tests

In this case, as in the case of the numerical simulation of compression test, it was used three-
dimensional solid isoparametric elements of eight nodes. Figure 6.3 shows the numerical model for
diagonal compression test. In the same Figure is shown the boundary conditions and load (prescribed
displacement) applied.

Figure 6.3. Numerical model for diagonal compression tests

7.1. From compression tests on prisms

Figure 7.1 shows the comparison between tests and numerical simulations in terms of “vertical stress –
vertical strain”. It is observed that there is an acceptable agreement between test results and numerical
simulation for the maximum strength. It is also observed a less gradual transition from elastic to plastic
behavior in the numerical simulation.

The average maximum vertical stress obtained from test is about 8.77 kgf/cm 2 and the maximum
vertical stress obtained from numerical simulation is 9.64 kgf/cm 2. The model estimates the maximum
strength (saxial) with a relative error of 10%. Similar values are also observed in Díaz and Ríos (2005.)
The model does not reproduce the strenght degradation observed in the tests.

7.2. From triaxial compression tests

Figure 7.2 shows the comparison between the tests and numerical simulation for triaxial compression
tests considering a confinement stress of 4 kgf/cm 2. The dashed curve is the average from the curves
obtained from tests. As in the case of compression tests, it is observed that there is a good
approximation of maximum strength but the model does not describe appropriately the behavior of
material after maximum strength.
12 35

30
10

25
8
Vertical stress (kgf/cm2 )
Vertical stress (kgf/cm2 )

20
6
15

4 Prism 1
Prism 2 10
Prism 3 Set3-M1
2 Prism 4 Set3-M2
5
Prism 5 Average from test
Drucker-Prager model Drucker-Prager model
0 0
0.000 0.005 0.010 0.015 0.020 0.025 0 0.02 0.04 0.06 0.08 0.1 0.12
Vertical strain (-) Vertical strain (-)

Figure 7.1. Comparison between tests and Figure 7.2. Comparison between tests and numerical
numerical simulation for compression tests on simulation for triaxial compression tests (Confinement
prisms. stress of 4 kgf/cm2)

Table 7.1 shows the maximum strength obtained from test and numerical simulation for all triaxial
compression tests. As can be observed, the relative error is about 10%.

Table 7.1. Maximum strength obtained from tests and numerical simulation for triaxial compression tests
Test Maximum saxial Averagesaxial saxial 1 Relative error
2 2 2
(kgf/cm ) (kgf/cm ) (kgf/cm ) (%)
Set1-M1 15.98
16.53 14.77 -10.64
Set1-M2 18.69
Set2-M1 21.41
19.86 19.89 0.15
Set2-M2 21.1
Set3-M1 28.4
27.045 30.14 11.43
Set3-M2 28.07
1
Maximum strength obtained from numerical simulation.
7.3 From diagonal compression tests
1.8 18
M01
Drucker-Prager model with cap
M02 16
1.5 Numerical simulation
M03
M04 14
M05
1.2
Numerical simulation 12
Vertical load (tonf)

√(J2D) (kgf/cm2 )
0.9 10

8
0.6
6
0.3
4

0 2
0 0.5 1 1.5 2 2.5 3 3.5
0
-0.3 0 5 10 15 20 25 30 35 40 45
Vertical displacement (mm) I 1 (kgf/cm2 )

Figure 7.3. Comparison between tests and numerical Figure 7.4. Numerical simulation of history of stresses
simulation for diagonal compression tests for diagonal compression tests.

Figure 7.3 shows the results obtained from tests and numerical simulation in terms of vertical load and
vertical displacement. It is observed that the experimental results show large variability but they tend
to present a linear behavior. From the numerical simulation, it is observed that the low walls keep in
the elastic range. To verify this last observation, it was traced the history of stresses in the plane √(J 2D)
– I1. From Figure 7.4, it is observed that the points remained in the elastic zone until the failure.

CONCLUSIONS

In general, all test results have variability; especially those get from diagonal compression tests. In
axial compression test we could observe that the behavior is approximately linear until the
maximum stress. For small deformations, it is proposed a modulus of elasticity of 800 kgf/cm2.
In triaxial compression tests performed in specimens made of the same homogeneous material of the
bricks, it wasn’t observed considerable variation in the modulus of elasticity. There is no available
information from test to quantify a Poisson's ratio. For small deformations, it is assumed that values
around 0.20 to 0.25 might be appropriate according to the reported by other authors. The maximum
strengths obtained in uniaxial compression tests are in the range of 8 to 10 kgf/cm2.

To simulate the behavior of adobe, it was used the Drucker-Prager model with cap. A procedure was
presented to estimate the parameters of the model based on results of axial compression tests and
triaxial compression tests with different confining pressures. It was found that the maximum stresses
fit good to a line defined by the Drucker-Prager parameters  = 0.3342 and  = 2.3447 kgf/cm2.
However, assuming a straight failure envelope, the resistance observed in diagonal compression test is
overestimated. In no one of the tests was observed the necessity of consider a cap, even in triaxial
compression tests with confining stress of 4 kgf/cm 2. Note that in adobe structures are expected lower
confining stresses.

In numerical simulations performed with ADINA program, it was got an acceptable approximation of
the stress-strain behavior in terms of initial stiffness and ultimate strength, with differences in
some cases in the order of 10%. Numerical simulations show a less gradual transition between elastic
and plastic behavior. This gradual transition observed in the tests may be the result of non-uniform
distribution of stresses in the specimens.
Given the limited experimental data available and the large variability observed in the results, it is
necessary to develop a more extensive testing program. It is particularly recommended to perform
more triaxial compression test with low confining stresses.

AKCNOWLEDGEMENT
This paper is part of the master’s thesis presented to the Graduate School of Civil Engineering of the National
University of Engineering, Perú and it is based on test carried out at Japan-Perú Earthquake Engineering and
Disaster Mitigation Center (CISMID). The author would like to thank Dr. C. Zavala for providing the test results
of adobe specimens, Dr. H. Scaletti for his tireless support during the investigation and also to the technical staff
of the Structural Laboratory and Geotechnical Laboratory of CISMID.

REFERENCES

ADINA R& D, Inc. (2003). ADINA user interface command reference manual, Volume I: ADINA model
definition, Report ARD 03-2, Watertown.
ADINA R& D, Inc. (2003). ADINA user interface command reference manual, Volume II: ADINA-T model
definition, Report ARD 03-8, Watertown.
ADINA R& D, Inc. (2003). Theory and modeling guide, Volume I: ADINA, Report ARD 03-7, Watertown.
Bathe, K.-J. (1996). Finite element procedures, Prentice Hall, New Jersey.
Blondet, M., Torrelva, D., Villa Garcia, G. (2002). Adobe in Peru: tradition, research and future. Modern Earth
Building 2002 – International Conference and Fair, Berlin.
Desai. C. S.; Siriwardane, H. J. (1984). Constitutive laws for engineering materials. Pretince-Hall, Englewood
Cliffs, NJ.
Díaz, K.; Ríos, J. (2005). Alternatives of rehabilitation in adobe and rammed earth. Los Andes University,
Faculty of engineering, Bogotá
Motulsky , H. J.; Christopoulos, A. (2003). Fitting models to biological data using linear and nonlinear
regression. A practical guide to curve fitting. Oxford University Press.
JICA y CEETyDES (2009). Ensayos de resistencia en componentes de adobe. Centro Peruano – Japonés de
Investigaciones Sísmicas y Mitigación de Desastres. Lima.
Kojic, M.; Bathe, K.-J. (2005). Inelastic analysis of solids and structures. Springer-Verlag. Berlin.
López, M.; Guaita, M.; Ayuga, F.; Cañas, I. (2000). Asientos en zapatas corridas. Comparación entre teorías
clásicas y el método de los elementos finitos. Univ.de Santiago de Compostela, Dpto. de Ingeniería
Agroforestal. España.
Mroginski, J..; Di Rado, H.; Beneyto, P.; Awruch, A. (2006). Modelado elastoplástico asociado de suelos no
saturados. Universidad Nacional del Nordeste, Comunicaciones Científicas y Tecnológicas. Chaco –
Corrientes.
Ronsson, B.; Boothby, T. (1998). Inelastic behavior of sand-lime mortar joint masonry arches. Engineering
Structures, Vol.20, Nos I-2, pp. 14-24.
X. Wang; K.J. Bathe; J. Walczak (2001). A stress integration algorithm for J3-dependent elasto-plasticity
models. Computational Fluid and Solid Mechanics 2001, pp. 542-545.
Yamin, L.; Rodríguez, A.; Fonseca, L.; Reyes, J. (2003). Comportamiento sísmico y alternativas de
Rehabilitación de edificaciones en adobe y tapia pisada con base en modelos a escala reducida ensayados en
mesa vibratoria. Universidad de los Andes, Facultad de Ingeniería, Bogotá.
Zienkiewicz O.C., Taylor R.L. y Zhu J.Z. (2005). The finite element method. Elsevier Butterworth Heinemann,
Oxford-Burlington.
Zoran Bonić, Verka Prolović, Biljana Mladenović (2010a). Mathematical modeling of materially nonlinear
problems in structural analyses (Part I – Theoretical fundamentals). Facta Universitatis Series: Architecture
and Civil Engineering. Vol. 8: N° 1, pp. 67 – 78.
Zoran Bonić, Todor Vacev, Verka Prolović, Marina Mijalković, Petar Dančević (2010b). Mathematical modeling
of materially nonlinear problems in structural analyses (Part II – Application in contemporary software).
Facta Universitatis Series: Architecture and Civil Engineering. Vol. 8: N° 2, pp. 201 – 210.