Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

TECHNICAL NOTES

Material Identification Procedure for Elastoplastic


Drucker–Prager Model
Juan José López Cela1
Downloaded from ascelibrary.org by University of Exeter on 07/16/15. Copyright ASCE. For personal use only; all rights reserved.

Abstract: A procedure to identify the material parameters of an elastoplastic model is presented. It aims to predict uniaxial stress–strain
curves in tension or compression. The Drucker–Prager model is chosen because of its simplicity. The procedure takes into account the
hardening/softening regime by varying two physical parameters related to the model: the cohesion and the friction angle.
DOI: 10.1061/共ASCE兲0733-9399共2002兲128:5共586兲
CE Database keywords: Material properties; Elastoplasticity; Models; Stress–strain curves.

Introduction the yield surface are directly related to variations, during the plas-
tic flow, of two physical parameters: cohesion and friction angle,
The identification of the material parameters defining the consti- which can be easily related to the Drucker–Prager material pa-
tutive laws is an important problem in solid mechanics. Some rameters. These variations have been analyzed independently.
authors propose inverse methods based on optimization tech- Because the numerical examples presented in this work have
niques to adjust material parameters so that the calculated re- been performed with only one element, no attention is paid to the
sponse 共obtained from a numerical method such as the finite- localization phenomena that are usually associated with softening
element method兲 matched the measured one 共Schnur and Zabaras behavior. In other, more general calculations, some regularization
1992; Ghouati and Gelin 2001兲, although direct evaluation of the procedures should be considered 共De Borst et al. 1993兲.
parameters from the experimental data is also used 共Ahadi and
Krenk 2000兲. Constitutive Model
In this work, a procedure to obtain the hardening/softening The constitutive model used in this study is the elastoplastic
characteristics of an elasto–plastic model from a given uniaxial Drucker–Prager model with hardening or softening. A Drucker–
stress–strain curve is presented. The mathematical identification Prager model is represented by the two-parameter yield surface
procedure has been developed in order to derive the material data 共Loret and Prevost 1986兲:
defining the yield surface and its evolution from uniaxial experi- F 共 ␴,␣,c 兲 ⫽␣␴⫹␶⫺c⫽0 (1)
mental stress–strain curves. The procedure is based on particular-
izing the yield surface for uniaxial conditions and then to obtain where ␴⫽hydrostatic stress; ␶⫽shear stress; and ␣,c⫽material
the derivatives of the material parameters with respect to the parameters.
equivalent plastic strain. The model is implemented on a finite-element code that uses a
The procedure is applied to a Drucker–Prager model. It is a central difference, explicit time marching algorithm which, com-
relatively simple model because it only needs two parameters to bined with suitable lumping of the mass matrix, leads to a fully
define the yield surface and one more parameter to define the explicit implementation.
plastic potential function and, consequently, the flow rule, but it The numerical implementation of the model is based on the
can be used to model some important structural materials. For radial return algorithm 共Wilkins 1969兲 generalized for the case of
example, it can reproduce some features considered typical in strain hardening or softening 共Krieg and Key 1976兲, as summa-
concrete: softening and different behavior in tension and com- rized for a von Mises model by Hughes 共1984兲, but by taking into
pression. Also, it is a model widely used by engineers and imple- due account the particularities of a Drucker–Prager model. The
mented in many commercial finite-element codes. plastic flow and the plastic potential function are defined as fol-
The hardening or softening regime can be obtained by differ- lows:
ent evolutions of the yield surface. In this study, the evolutions of ⳵Q
␧˙ p ⫽␭˙ (2)
⳵␴
1
ETS Ingenieros Industriales, Univ. de Castilla–La Mancha,
13071 Ciudad Real, Spain. E-mail: jlopez@ind-cr.uclm.es Q 共 ␴,␤,c 兲 ⫽␤␴⫹␶⫺c⫽0 (3)
Note. Associate Editor: Arup K. Maji. Discussion open until October
1, 2002. Separate discussions must be submitted for individual papers. To
where ␧˙ p ⫽rate plastic strain tensor; ␭˙ ⫽consistency parameter;
extend the closing date by one month, a written request must be filed with and ␤⫽material parameter related to the dilatancy, representing
the ASCE Managing Editor. The manuscript for this technical note was an inelastic volume increase.
submitted for review and possible publication on October 12, 2000; ap-
proved on October 2, 2001. This technical note is part of the Journal of
Identification of Material Parameters
Engineering Mechanics, Vol. 128, No. 5, May 1, 2002. ©ASCE, ISSN The objective of this section is to determine the material param-
0733-9399/2002/5-586 –591/$8.00⫹$.50 per page. eters of the Drucker–Prager constitutive law in order to predict,

586 / JOURNAL OF ENGINEERING MECHANICS / MAY 2002

J. Eng. Mech. 2002.128:586-591.


Downloaded from ascelibrary.org by University of Exeter on 07/16/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Simplified uniaxial compressive and tensile stress–strain curves

as closely as possible, the behavior of the material. The present ⳵c ⳵c ⳵k


identification procedure aims at predicting experimental uniaxial ⫽ (7)
⳵␧
¯ p ⳵k ⳵␧
¯p
curves, for example, for concrete in uniaxial compression 共Wis-
chers 1978兲, or in uniaxial tension 共Hughes and Chapman 1966兲. The value of ⳵c/⳵k depends on the type of matching between the
As a preliminary step, it is common practice to approximate the Drucker–Prager and the Mohr–Coulomb surfaces and can be ob-
continuous experimental curves with piecewise linear ones, as tained from Eq. 共6兲.
shown in Fig. 1. To get ⳵k/⳵␧¯ p , we derive the generic value of the normal
The Drucker–Prager criterion, characterized by a conical yield stress component along the material curve in, e.g., direction 1, ␴ 1
surface and defined by the material parameters ␣ and c, can be with respect to the cohesion k while keeping the friction angle ␾
considered as a smooth approximation to the Mohr–Coulomb cri- constant:
terion, which uses an irregular hexagonal pyramid yield surface ⳵k ␴˙ 1
and is defined by cohesion k and friction angle ␾. The procedure p⫽ (8)
⳵␧
¯ ⳵␴ 1 p
is the following: ថ␧
The starting point is the Mohr–Coulomb yield surface, where ⳵k
the relationships between the material parameters 共k and ␾兲 and
the yield stresses in compression and tension 共f c and f t 兲 are given Cohesion k Constant and Friction Angle ␾ Variable
by Mohr’s circles 共Chen 1982兲:
In this case, both Drucker–Prager material parameters are vari-
f c⫺ f t able. We are looking for
sin ␾⫽ (4)
f c⫹ f t
⳵c ⳵c ⳵␾
f c 共 1⫺sin ␾ 兲 1 ⫽ (9)
k⫽ ⫽ 冑f c f t (5) ⳵␧
¯ p ⳵␾ ⳵␧¯p
2 cos ␾ 2
⳵␣ ⳵␣ ⳵␾
Then, by matching the Drucker–Prager surface to the Mohr– ⫽ (10)
⳵␧
¯ p ⳵␾ ⳵␧
¯p
Coulomb surface, expressions relating the material parameters of
both surfaces are obtained 共Chen 1982兲: The values of ⳵c/⳵␾ and ⳵␣/⳵␾ are obtained from Eq. 共6兲.
To get ⳵␾/⳵␧ ¯ p , we derive the uniaxial tension ␴ 1 with respect
6 sin ␾ 6k cos ␾ to the friction angle ␾ while keeping cohesion k constant:
␣⫽ , c⫽ (6)
) 共 3⫾sin ␾ 兲 ) 共 3⫾sin ␾ 兲 ⳵␾ ␴˙ 1
⫽ (11)
where the ⫾ sign depends on whether the Drucker–Prager cone is ⳵␧
¯ p ⳵␴ 1 p
completely inscribed in 共tensile meridians兲, or completely circum- ថ␧
⳵␾
scribes, the Mohr–Coulomb hexagon 共compressive meridians兲.
In the general case of nonperfect plasticity, some kind of yield To compute the evolutions of the cohesion and friction angle
surface evolution should be considered to obtain the expressions with respect to the equivalent plastic strain, the values of ⳵␴ 1 /⳵k,
of the material parameters in terms of the equivalent plastic strain ⳵␴ 1 /⳵␾, and ¯␧ p are needed and are obtained as follows.
共defined as ␧ថ p ⫽ 32 冑␧˙ p :␧˙ p 兲. We have decided to consider the varia- The expressions of the Mohr–Coulomb yield surface particu-
tion of the Mohr–Coulomb material parameters because they larized for uniaxial tests are the following: for compression tests,
have a clear physical interpretation. In this work, only one of the 2k cos ␾
two possible material parameters k and ␾ has been considered ␴ 1⫽ 共 compressive meridians兲 (12)
1⫺sin ␾
variable, while the other one has been kept constant.
and for tension tests,
Cohesion k Variable and Friction Angle ␾ constant 2k cos ␾
␴ 1⫽ 共 tensile meridians兲 (13)
1⫹sin ␾
In this case c is variable while ␣ is constant. Thus, we have to
obtain the value of ⳵c/⳵␧¯ p: The expressions for compression tests obtained with tensile

JOURNAL OF ENGINEERING MECHANICS / MAY 2002 / 587

J. Eng. Mech. 2002.128:586-591.


meridians and those for tension tests with compressive meridians
are not written because, as it will be shown in the numerical
examples, it is not possible either to predict a compression test
冉 冊⳵c
⳵␧
¯ p
i
⫽ 共 c ⬘ 兲 i ⫽E
␤⫺) 1⫺sin ␾ 6 cos ␾
冑2␤ ⫹3 2 cos ␾ ) 共 3⫺sin ␾ 兲
2

using tensile meridians or the opposite 共tension test with compres-


sive meridians兲. ⌬␴
⫻ 共 compression FP兲
Now, the compression test and compressive meridians will be ⌬␴⫺E⌬␧
named the ‘‘compression fitting procedure’’ 共compression FP兲

冉 冊
and, in the same manner, the tension test and tensile meridians,
⳵c ⫺␤⫺) 1⫹sin ␾ 6 cos ␾
‘‘tension fitting procedure’’ 共tension FP兲. ⫽ 共 c ⬘ 兲 i ⫽E
To obtain ␧ថ p , we write the constitutive rate equation. For ex- ⳵␧
¯p i 冑2␤ ⫹3
2 2 cos ␾ ) 共 3⫹sin ␾ 兲
ample, for an uniaxial compression test along direction 1, we
have ⌬␴
⫻ 共 tension FP兲

冋 册
⌬␴⫺E⌬␧
␭˙
␴˙ 1 ⫽E ␧˙ 1 ⫹ 共 ␤⫺) 兲
Downloaded from ascelibrary.org by University of Exeter on 07/16/15. Copyright ASCE. For personal use only; all rights reserved.

(14)
3 Step 4
On the other hand, the plastic strain rate tensor is defined accord- Compute the value of the increment of equivalent plastic strain
ing to the flow rule as for each increment of the c parameter:

␧˙ p ⫽␭˙
⳵Q

⳵␴ 冉
⫽␭˙
1 ␴⬘ 1
2 ␶
␭˙
冊 )
⫹ ␤I ⫽ diag ␤⫺),␤⫹ ,␤⫹
3 ␤ 2
)
2 冋 册 共 ⌬␧
¯ p 兲i⫽
⌬c i
共 c⬘兲i
(15)
and with the definition of the equivalent plastic strain
Cohesion k Constant and Friction Angle ␾ Variable
3␧ថ p
␭˙ ⫽ (16) Step 1
冑2␤ 2 ⫹3 Compute the Mohr–Coulomb material parameters.
Substituting in Eq. 共14兲, the desired expression is obtained: 1. Compute cohesion k, which is a constant value from Eq. 共5兲;
and
␴˙ 1 ⫺E␧˙ 1 冑2␤ 2 ⫹3 2. For each value of the piecewise linear curve, compute the
␧ថ p ⫽ (17)
E ␤⫺) corresponding values of friction angle ␾ from Eq. 共12兲 共com-
pression FP兲 or 13 共tension FP兲.
In a similar manner, for tension tests we obtain

␴˙ 1 ⫺E␧˙ 1 冑2␤ 2 ⫹3 Step 2


␧ថ p ⫽ (18) Compute, by using Eq. 共6兲, the Drucker–Prager material param-
E 共 ⫺␤⫺) 兲
eters ␣ i and c i , from the Mohr–Coulomb ones. In this case, both
parameters ␣ i and c i are variable.

Proposed Material Identification Procedure


Step 3
Compute the slope of the Drucker–Prager material parameters c
The necessary steps to obtain the Drucker–Prager material pa-
and ␣ in terms of the equivalent plastic strain, that is, compute
rameters are summarized in this section.
¯ p ) i ⫽(c ⬘ ) i and (⳵␣/⳵␧
(⳵c/⳵␧ ¯ p ) i ⫽(␣ ⬘ ) i for each ⌬␴,⌬␧ of the
curve:
Cohesion k Variable and Friction Angle ␾ Constant

Step 1 冋 册
⳵c
⳵␧ p ⫽ 共 c ⬘ 兲 i ⫽E
¯ i
␤⫺) 1⫺ 共 sin ␾ 兲 i 6k 1⫺3 sin ␾
冑2␤ 2 ⫹3 2k ) 共 3⫺sin ␾ 兲
2
Compute the Mohr–Coulomb material parameters.
1. Compute the friction angle ␾ that is a constant value from ⌬␴
Eq. 共4兲; and ⫻ 共 compression FP兲
⌬␴⫺E⌬␧
2. For each value of the piecewise linear curve ␴ i , compute the

冋 册
corresponding values of the cohesion k i from Eq. 共12兲 共com-
⳵c ␤⫹) 1⫹ 共 sin ␾ 兲 i ⫺6k 1⫹3 sin ␾
pression FP兲 or 13 共tension FP兲. ⫽ 共 c ⬘ 兲 i ⫽E
⳵␧
¯p i 冑2␤ 2 ⫹3 2k ) 共 3⫹sin ␾ 兲
2

Step 2
⌬␴
Compute, by using Eq. 共6兲, the Drucker–Prager material param- ⫻ 共 tension FP兲
eters ␣ and c i , from the Mohr–Coulomb ones. ␣ parameter is ⌬␴⫺E⌬␧

冋 册
constant, while c i is variable.
⳵␣ ␤⫺) 1⫺ 共 sin ␾ 兲 i 6) cos ␾
⫽ 共 ␣ ⬘ 兲 i ⫽E
Step 3 ⳵␧
¯p i 冑2␤ 2 ⫹3 2k 共 3⫺sin ␾ 兲 2
Compute the slope of the Drucker–Prager material parameter c in
¯ p) i
terms of the equivalent plastic strain, that is, compute (⳵c/⳵␧ ⌬␴
⫻ 共 compression FP兲
⫽(c ⬘ ) i for each ⌬␴,⌬␧ of the curve: ⌬␴⫺E⌬␧

588 / JOURNAL OF ENGINEERING MECHANICS / MAY 2002

J. Eng. Mech. 2002.128:586-591.


Downloaded from ascelibrary.org by University of Exeter on 07/16/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. General description of numerical uniaxial tests

Fig. 3. Numerical tests

JOURNAL OF ENGINEERING MECHANICS / MAY 2002 / 589

J. Eng. Mech. 2002.128:586-591.


冋 册
⳵␣
⳵␧ p ⫽ 共 ␣ ⬘ 兲 i ⫽E
¯ i
␤⫺) 1⫺ 共 sin ␾ 兲 i
冑2␤ 2 ⫹3 2k
plastic strain have been obtained. These expressions take into
consideration the third parameter of the model ␤ that define the
plastic potential function.
From the numerical examples it is concluded that using the
6) cos ␾ ⌬␴ compression fitting procedure the compressive stress–strain curve
⫻ 共 tension FP兲 is correctly predicted for every value of ␤ but it is not possible to
共 3⫹sin ␾ 兲 2 ⌬␴⫺E⌬␧
predict the tensile one. The same can be said for the tension fitting
Step 4 procedure: only good prediction of the tensile stress–strain
Compute the value of the increment of equivalent plastic strain curves. This phenomenon appears with the two evolutions inves-
for each increment of the ␣ and c parameters tigated in this work that are achieved by making change indepen-
dently the cohesion k and the friction angle ␾ of the material.
⌬␣ i
共 ⌬␧
¯ p 兲i⫽
共 c⬘兲i
Downloaded from ascelibrary.org by University of Exeter on 07/16/15. Copyright ASCE. For personal use only; all rights reserved.

Notation

⌬c i The following symbols are used in this paper:


共 ⌬␧
¯ p 兲i⫽ c ⫽ second material parameter defining yield surface
共 ␣⬘兲i
and plastic potential function;
E ⫽ Young’s modulus;
Note. Because in the expression of 关 ⳵␾/⳵␧ ¯ p 兴 i appears sin ␾ F(␴,␣,c) ⫽ Drucker–Prager yield surface;
that it is variable along the curve some discretization is needed in f c ⫽ yield stress in compression;
the linear branches of the curve. f t ⫽ yield stress in tension;
k ⫽ cohesion;
Q(␴,␤,c) ⫽ plastic potential function;
␣ ⫽ first material parameters defining yield surface;
Numerical Examples ␤ ⫽ first material parameters defining plastic poten-
tial function;
In this section some numerical tests are presented. All of them
␧ 1 ⫽ generic value of normal strain component
aim at predicting stress–strain curves in uniaxial tension or com-
along, e.g., direction 1;
pression. The values of the material parameters 共␣ and c兲 are
¯␧ p ⫽ equivalent plastic strain;
obtained for the two kinds of yield surface evolutions and accord-
␧p ⫽ plastic strain tensor;
ing to the two fitting procedures 共tension FP and compression FP兲
␭ ⫽ consistency parameter;
described in the previous section.
These tests have been performed just using one three- ␴ ⫽ hydrostatic stress;
dimensional finite element with eight nodes and one integration ␴ 1 ⫽ generic value of normal stress component
point. The length of the side is 1 m. The boundary conditions are along, e.g., direction 1;
chosen to represent 81 of the complete specimen, i.e., the model ␴ ⫽ stress tensor;
has three symmetry planes. The load is a nodal-imposed displace- ␶ ⫽ shear stress; and
ment in tension or compression varying slowly in time. In Fig. 2 ␾ ⫽ friction angle.
are represented the general material characteristics, the load, and
the boundary conditions of the problem.
The simplified uniaxial compressive and tensile stress–strain References
curves are those of Fig. 1.
All cases have been computed for different values of the ␤ Ahadi, A., and Krenk, S. 共2000兲. ‘‘Characteristic state plasticity for
parameter, ␤⫽0.0, ␤⫽0.2, and ␤⫽0.3. granular materials. Part II: Model calibration and results.’’ Int. J. Sol-
The results are shown in Fig. 3. With the compression fitting ids Struct., 37, 6361– 6380.
procedure, uniaxial compressive stress–strain curves 关Fig. 3共a兲兴 De Borst, R., Sluys, L. J., Pamin, J., and Mühlhaus, H. B. 共1993兲. ‘‘Fun-
are correctly predicted for every value of ␤, and for the two yield damental issues in finite-element analysis of localization of deforma-
surface evolutions. The same can be said for the uniaxial tensile tion.’’ Eng. Comput., 10, 99–121.
stress–strain curves with the tension fitting procedure 关Fig. 3共b兲兴. Chen, W. F. 共1982兲. Plasticity in reinforced concrete, McGraw-Hill, New
Note that in both cases the six curves are practically coincident. York.
However, neither compression curves with the tension fitting pro- Ghouati, O., and Gelin, J. C. 共2001兲. ‘‘A finite-element-based identifica-
tion method for complex metallic material behaviours.’’ Comput.
cedure 关Fig. 3共c兲兴 nor tensile curves with the compression fitting
Mater. Sci., 21, 57– 68.
procedure 关Fig. 3共d兲兴 are correctly predicted.
Hughes, T. J. R. 共1984兲. ‘‘Numerical implementation of constitutive mod-
els: Rate-independent deviatoric plasticity.’’ Theoretical foundations
for large-scale computations for nonlinear material behavior, S.
Conclusions Nemat-Nasser et al., eds., Martinus Nujhoff, Boston, 26 –57.
Hughes, B. P., and Chapman, G. P. 共1966兲. ‘‘The complete stress–strain
A procedure to identify the material parameters of an elastoplastic curve for concrete in direct tension.’’ RILEM Bull., 30, 95–97.
Drucker–Prager model has been presented. This procedure is Krieg, R. D., and Key, S. W. 共1976兲. ‘‘Implementation of a time-
based on predicting uniaxial stress–strain curves in tension or independent plasticity theory into structural computer programs, Con-
compression. Expressions to evaluate the two material parameters stitutive equations in viscoplasticity: Computational and engineering
␣ and c of the model as well as their evolution with the equivalent aspects.’’ 共AMD-20兲, J. A. Stricklin and K. J. Saczalski, eds., Ameri-

590 / JOURNAL OF ENGINEERING MECHANICS / MAY 2002

J. Eng. Mech. 2002.128:586-591.


can Society of Mechanical Engineers, New York. Methods Eng., 33, 2039–2057.
Loret, B., and Prevost, J. H. 共1986兲. ‘‘Accurate numerical solutions for Wilkins, M. L. 共1969兲. ‘‘Calculation of elastic–plastic flow.’’ Rep. No.
Drucker–Prager elastic–plastic methods.’’ Comput. Methods Appl. UCRL-7322 Rev. 1, Lawrence Livermore National Laboratory, Univ.
Mech. Eng., 54, 259–277. of California.
Schnur, D. S., and Zabaras, N. 共1992兲. ‘‘An inverse method for determin- Wischers, G. 共1978兲. ‘‘Application of effects of compressive loads on
ing elastic material properties and material interface.’’ Int. J. Numer. concrete.’’ Betontech. Ber., Nos. 2 and 3, Dusselford.
Downloaded from ascelibrary.org by University of Exeter on 07/16/15. Copyright ASCE. For personal use only; all rights reserved.

JOURNAL OF ENGINEERING MECHANICS / MAY 2002 / 591

J. Eng. Mech. 2002.128:586-591.

You might also like