Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: The Concrete Damage Model that is implemented in the LS-Dyna code is capable of simulating the
Received 11 February 2011 behavior of plain concrete under complex static and dynamic loading conditions. However, the values
Accepted 19 May 2011 for the numerous parameters, which are required as an input, are left for the user to provide. In this
Available online 2 July 2011
study the Concrete Damage Model was calibrated for a wide range of strong concretes, using triaxial-
Keywords: compression-test data that were obtained from the literature. In contrast, the adjustment of the
Concrete parameters of the present model is provided as a function of the unconfined compressive strength of
Constitutive model the concrete. Although, not enough validation has been done either in higher pressures or in actual
Numerical simulation tests, it is evident that the presently calibrated model shows better agreement with published test
LS-Dyna
results than the model currently available in LS-Dyna.
Damage
& 2011 Elsevier B.V. All rights reserved.
High-strength concrete
High-performance concrete
0168-874X/$ - see front matter & 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.finel.2011.05.008
N. Markovich et al. / Finite Elements in Analysis and Design 47 (2011) 1280–1290 1281
η
stress point, (Pt. 1). After yielding, a strain hardening plasticity
response occurs until the maximum strength, (Pt. 2), is achieved.
After reaching the maximum strength, softening occurs until the
residual strength is reached, (Pt. 3), all of which depend on the
amount of confining stress.
A total of eight parameters, ai, define the three initial failure 0
λm λ
surfaces, in the Concrete Damage Model:
yield failure surface residual failure surface
Maximum failure surface: Fig. 4. Schematic evolution of the damage function.
p
Dsm ¼ a0 þ ð2Þ
a1 þ a2 p
use the basic concrete damage model, the user needs to supply 49
parameters while in most cases the only known property is the
Residual failure surface: unconfined compressive strength of the concrete. Ideally these
p parameters should be determined from available laboratory data
Dsr ¼ ð3Þ
a1f þa2f p obtained from unconfined compression tests and several triaxial
compression tests that must be performed at different confining
Yield failure surface: stresses. However, in the following section, a method for estimat-
ing all of the above parameters as a function of the user specified
p
Dsy ¼ a0y þ ð4Þ unconfined strength of the concrete is presented. For this purpose,
a1y þ a2y p
the empirical relations of Attard and Setunge [13] from test
data are used as representative of the compressive behavior of
Each of the above surfaces is based on the second invariant of concrete over a wide range of strong concretes.
the deviatoric stress:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ds ¼ 32ðS21 þS22 þS23 Þ ð5Þ 4. The damage function
Once the values of the three points are known from Fig. 3a, the
The change in the stress as a function of plastic strain arises
shape of the curve in Fig. 3b should be obtained. After the stress
from physical mechanisms such as internal cracking, and the
reaches the initial yield surface and before it reaches the max-
extent of this cracking is affected by the magnitude of the
imum failure surface, the current stress is calculated by means of
hydrostatic pressure when cracking occurs. This mechanism is
a linear interpolation between the two surfaces, i.e.,
commonly known as the ‘‘confinement’’ effect on concrete beha-
Ds ¼ ZðDsm Dsy Þ þ Dsy ð6Þ vior. To account for this phenomenon a ‘‘damage’’ function was
After reaching the maximum failure surface, the current failed defined by [8] and implemented in the Concrete Damage Model of
stress is similarly interpolated between the maximum and the LS-Dyna. The damage function is defined as
residual stress surfaces, i.e., Z ep
dep
l¼ when p Z0 ð8Þ
Ds ¼ ZðDsm Dsr Þ þ Dsr ð7Þ 0 rf ð1 þðp=rf ft ÞÞb1
where Z is a parameter, which indicates the relative amount of Z ep
damage and the current failure surface region, and is a function of dep
l¼ when p o0 ð9Þ
the effective plastic strain, l. As can be seen in Fig. 4 where the 0 rf ð1 þðp=rf ft ÞÞb2
evaluation of the damage function is shown schematically, Z ¼0 at
where the effective plastic strain increment is given by
l ¼0, and Z increases to 1 for l ¼ lm (the effective strain for which qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the deviatoric second invariant is a maximum). Then Z decreases dep ¼ ð23depij depij Þ ð10Þ
back to 0 at some larger value of l. Whenever l r lm the state lies
between the yield and the maximum failure surfaces; for l Z lm and b1 and b2 are damage scaling parameters for the cases of
the state lies between the maximum failure and the residual uniaxial compression and tension, respectively. The parameters
surfaces. b1 and b2, in Eqs. (8) and (9), determine the rate at which the
Since the full details of the material model are well established damage occurs; rf is the dynamic increase factor that accounts for
and documented in references [8–12], only the above formulae strain rate effects. It is important to note that in more recent
are essential for explaining the calibration method that is pro- versions (2007) of LS-Dyna, e.g., Version 971 [1], the following
posed in the present study. enhanced form of the above equations has been introduced:
The data that define the compressive behavior of a specific dep
concrete, which the user needs to supply, are the eight values for dl ¼ for p Z 0 ð11Þ
½1 þ ðs=100Þðrf 1Þð1 þðp=rf ft ÞÞb1
the constants, ai, of Eqs. (1)–(3), i.e., the parameters of the failure
surfaces. In addition, the function ZðlÞ for Eqs. (6) and (7) should
dep
be supplied in the form of a table (a set of 13 value pairs), an dl ¼ for p o 0 ð12Þ
equation of state and other features of the material. In order to ½1 þ ðs=100Þðrf 1Þð1 þðp=rf ft ÞÞb2
1284 N. Markovich et al. / Finite Elements in Analysis and Design 47 (2011) 1280–1290
(3)
comparison motivated the authors of the present study to better 80
calibrate the Concrete Damage Model. This improved calibration fr = 5MPa
60 (2)
40 (1)
fr = 1 MPa
20
fr = 0 MPa
0
0 0.005 0.01 0.015 0.02
axial strain
Fig. 8. A&S concrete constitutive behavior results for a 50 MPa strong concrete.
Points 1–5 are the maximum strength points for 0, 1, 5, 10 and 15 MPa
confinement pressures, respectively.
laboratory data of unconfined compression and triaxial compres- (5) represent different levels of confinement: 0, 1, 5, 10 and
sion tests over a range of confining pressures. In the present 15 MPa, respectively.
study, the empirical relations of Attard and Setunge [13] are used. The third step of the calibration process was to fit functions to
These relations, from whom the estimated tests results may be the points of each curve. To use Malvar’s model, these fitting
generated numerically, cover a wide range of strong concretes functions had to be in the form of Eqs. (2)–(4) for yield, and
(20–120 MPa). maximum and residual failure surfaces, respectively. A nonlinear
As shown in Fig. 3b, three stress–strain points define each least squares fitting procedure was used to obtain the eight
curve: the yield, the maximum strength and the residual strength. appropriate constants (ai) of Eqs. (2)–(4) that best fit the failure
The calibration process consists of a few steps: surfaces obtained for 50 MPa strong concrete.
The first step of the calibration process is to extract these The fourth step of the calibration process used the same
three points from the A&S concrete stress–strain curves [13]. approach applied for the 50 MPa strong concrete (repeating
Figs. 7 and 8 show curves obtained from A&S relations for a
50 MPa strong concrete. Fig. 7 shows yield points and Fig. 8 shows 80
maximum strengths points. The points indexed from (1) to (5) are yield surface (5)
70 yield surface points
the yield/maximum strengths obtained for 0, 1, 5, 10 and 15 MPa
confinement pressures, respectively. 60 (4)
Unlike other materials that have a linear elastic stress–strain (3)
50
, MPa
response, concrete has a fairly nonlinear elastic behavior. Differ-
ent sources suggest that the elastic response can be assumed for 40 (2)
axial stress of up to 0.33–0.65 of the unconfined compressive (1)
30
strength. However, for high strength concrete (HSC) the ascend-
ing branch is fairly linear. A few values (0.4, 0.5 and 0.6) of 20
unconfined compressive strength were used to define the slope of 10
the curve that is treated as linear, and the resulting differences
0
were insignificant. For the final calibration, yield at 0.65 of the
0 10 20 30 40 50 60
unconfined compressive strength of concrete was assumed. The
residual strength due to confinement should be obtained for large P, MPa
strains where the stress difference becomes almost constant; the Fig. 10. Yield surface for a 50 MPa strong concrete in meridian plane, based on
results are shown in Fig. 9. A&S results.
The shear surfaces in Malvar’s Concrete Damage Model [8–11]
are described in the meridian plane, as stress difference vs. mean 140
stress (see Fig. 3a). Once the stress–strain curves of a specific
maximum surface
concrete are obtained using A&S results, it is possible to present 120 maximum surface points
the three shear surfaces in the meridian plane. This is done by (5)
100
transforming from terms of axial stress and confining pressure, to (4)
, MPa
stress difference and mean stress. Since for triaxial compression 80 (3)
of cylinders, the lateral and the radial principal stresses are equal, (2)
the mean stress could be written as follows: 60
(1)
p ¼ 13 I1 ¼ 13 ðs1 þ s2 þ s3 Þ ¼ 13ðs1 þ 2s2 Þ ð16Þ 40
The second step of the calibration process was to map the set 20
identified points into the meridian plane. Figs. 10–12 show the 0
resulting yield, and maximum and residual surfaces for a 50 MPa 0 10 20 30 40 50 60
strong concrete. In these figures, the points numbered (1)– P, MPa
Fig. 11. Maximum failure surface for a 50 MPa strong concrete in meridian plane,
based on A&S results.
90
80 residual surface
residual surface points (5)
70
60
(4)
, MPa
50
40
(3)
30
20
(2)
10 (1)
0
0 10 20 30 40 50
P, MPa
Fig. 9. A&S concrete constitutive behavior results for 50 MPa strong concrete, Fig. 12. Residual failure surface for a 50 MPa strong concrete in meridian plane,
residual strength points. based on A&S results.
1286 N. Markovich et al. / Finite Elements in Analysis and Design 47 (2011) 1280–1290
Fig. 16. Evaluation of the parameter a0y for a wide range of strong concretes.
Fig. 13. Evaluation of the parameter a0 for a wide range of strong concretes.
Fig. 14. Evaluation of the parameter a1 for a wide range of strong concretes.
Fig. 17. Evaluation of the parameter a1y for a wide range of strong concretes.
Fig. 18. Evaluation of the parameter a2y for a wide range of strong concretes.
Fig. 15. Evaluation of the parameter a2 for a wide range of strong concretes.
Table 1
Input data of the damage function.
k g
Table 2
EOS input data for a 50 MPa strong concrete. the compressive strength. Nevertheless, it is very important to find
an appropriate correlation, which will fit experimental data in order
Volume strain Pressure (MPa)
to obtain good computational results. Generally, if the concrete is of
0 0
normal weight, the general equation for Young’s module could be
0.007034 96 written as follows:
0.02896 240
0.050981 390 Ec ¼ qðfc0 þ cÞw þ k ð27Þ
0.062837 565
0.072613 737 where q, c and w are some constant coefficients and k is a correcting
0.144392 2000 factor. The coefficients could be found empirically from test results.
0.216174 3263 The correcting factor, k, is sometimes neglected. Since the A&S
0.28795 4526
results have been used in this study as a substitute for experimental
0.359729 5789
results, the best fit for the module of elasticity (Young’s module) as a
function of the strength of concrete was found to be
Table 3
Absorbed energy calculation for a 50 MPa strong concrete.
Confinement (MPa) 0 5 10 15
Energy
A&S 0.21 2.16 3.30 4.10
K&C 0.39 2.42 3.89 4.87
BGU 0.20 1.90 3.15 4.10
Error (%)
K&C 89 12 18 19
BGU 1.7 11.8 4.5 0.1
Table 4
Fig. 24. Concrete Damage Material Model, Release III for a 50 MPa strong concrete Absorbed energy calculation for a 100 MPa strong concrete.
simulation results, calculated with K&C, and BGU input data are compared to A&S
constitutive behavior. Concrete Strength (MPa) 100
Confinement (MPa) 0 5 10 15
Energy
A&S 0.37 2.74 4.60 6.05
K&C 2.04 3.33 5.37 6.76
BGU 0.41 2.42 4.07 5.55
Error (%)
K&C 446 21 17 12
BGU 10 12 11 8
axial displacement loading, the confinement pressure was main- An improvement in the calibration of the LS-Dyna Concrete
tained constant. Damage Model was obtained while retaining the option of the
As mentioned above, all the parameters were calibrated as user-friendly automatic parameter generation model feature. The
functions of an unconfined strong concrete (fc0 ) and implemented in improved calibration is in better agreement with three triaxial test
the LS-Dyna’s input file. Figs. 24 and 25 show results computed for 50 results published by Attard and Setunge [13], which are available for
and 100 MPa strong concretes, respectively. They show a comparison a wide range of strong concretes (20–120 MPa). This option is very
of the present results (BGU) and the results obtained using the K&C useful for the occasional users of LS-Dyna who need to perform
generated parameters (K&C) with the A&S ‘‘test’’ results). Each curve analyses involving concrete when little or no information is available
is for a different confinement pressure (0, 5, 10 and 15 MPa). on the specific concrete aside from its unconfined strength. The main
Considering the maximum strength, the deviation of the K&C result of this study is the BGU calibrated model that behaves more
and BGU calibrations s from the A&S ‘‘test’’ results is no more than similarly to the A&S ‘‘test’’ results than the calibrated model imple-
2%, which is acceptable, especially in view of the fact that the A&S mented by Karagosian & Case into the latest version of LS-Dyna. This
relations were within 10% error [13]. This is also in the range of is of course expected since the calibration was done according to the
the error expected when comparing test results on concrete A&S correlations. It should be further noted that the A&S experiments
specimens cast from the same batch and of the same age [21]. were limited to a low confinement level of only 20 MPa. For high
Anyone familiar with concrete testing would consider a 10% error strength concrete, this is a low confinement. Further experiments
as acceptable. To establish an objective and meaningful compar- or published data are needed to improve the calibration. Radial
ison between the BGU and the K&C calibrations to the A&S stress– displacements of the specimens should also be measured in order to
strain curves, the authors suggest calculating the energy absorbed better describe the behavior of the concrete. Unfortunately such data
by the specimens during compression tests, i.e., the area under were not provided by Attard and Setunge [13].
the stress–strain curves. Then, the errors from the A&S results
could be estimated using the following definitions:
Acknowledgments
energyK&C energyA&S
errorK&C ¼ 100% ð29Þ
energyA&S
The authors would like to express their thanks to Barak Ostraich
for assistance in numerical simulations and guidance in the use of
energyBGU energyA&S LS-Dyna. Dr. Leonard (Len) Schwer was kind enough to revise and
errorBGU ¼ 100% ð30Þ
energyA&S add some very enlightening comments, and for this the authors are
1290 N. Markovich et al. / Finite Elements in Analysis and Design 47 (2011) 1280–1290
deeply thankful. Many thanks also to Dr. Javier Malvar for clearing up [10] L.J. Malvar, J.E. Crawford, J.W. Wesevich, D. Simons, A new concrete material
some aspects of his model and providing helpful data. model for DYNA3D Release II: shear dilation and directional rate enhance-
ment, Karagozian & Case, 1996 TR-96-2.2.
[11] L.J. Malvar, J.E. Crawford, K.B. Morill, K&C Concrete material model release III:
automated generation of material model input, Karagozian & Case, 2000
References TR-99-24.3.
[12] L. Schwer, L.J. Malvar, Simplified concrete modeling with Mat_concrete_
damage_rel3, JRI LS-Dyna User Week, August 2005.
[1] LS-Dyna User manual, Version 971, Livermore Software Tech. Corp., Livermore, [13] M.M. Attard, S. Setunge, Stress–strain relationship of confined and uncon-
California, USA, 2007. fined concrete, ACI Mater. J 93 (5) (1996) 432–442.
[2] AUTODYN Theory Manual, Century Dynamics, Inc., 2005. [14] N. Markovich, Investigation and calibration of constitutive models of con-
[3] MSC/Dytran User’s Manual Version4.7, vols. 1 and 2, MSC Software Corp., LA, crete, M.Sc. thesis, Ben-Gurion Univ. Negev, Dept. Mech. Eng., February 2009.
CA, 1999. [15] M. Unosson, Modeling of concrete material behavior with application to
[4] X.C. Zhou, V.A. Kuznetsovb, H. Haoa, J. Waschlb, Numerical prediction of concrete reinforced concrete beams subjected to impact, FOI-R-0167-SE, 2001.
slab response to blast loading, Int. J. Impact Eng. 35 (2008) 1186–1200. [16] J. Magnusson, M. Unosson, A. Carlberg, High performance concrete ‘‘HPC’’-field
[5] Z. Tu, Y. Lu, Evaluation of typical concrete material models used in hysdrocodes experiments and production, FOI-R-0256-SE, 2001.
for high dynamic response simulations,, Int. J. Impact Eng 36 (2009) 132–146. [17] H. Hansson, P. Skoglund, M. Unosson, Structural protection for stationary/
[6] E. Kochavi, Y. Kivity, I. Anteby, O. Sadot, G. Ben-Dor, Numerical model of mobile tactical behavior, FOI-R-0281-SE, 2001.
composite concrete walls, in: Proceedings of the 9th Biennial ASME Con- [18] F.A. Tavarez, Simulation of the behavior of composite grid reinforced
ference on Engineering Systems Design & Analysis, ESDA2008, Haifa, Israel, concrete beams using explicit finite element methods, M.Sc. thesis, Univer-
July 2008. sity Wisconsin-Madison, USA, 2001.
[7] E. Kochavi, Y. Kivity, I. Anteby, O. Sadot, G. Ben-Dor, Numerical modeling of [19] N. Gebbeken, S. Greulich, A. Pietzsch, Hugoniot properties for concrete
composite concrete-Durisol walls, in: Proceedings of the Military Aspects of determined by full-scale detonation experiments and flyer-plate-impact
Blast & Shock, MABS20, Oslo, Norway, 2008. tests, Int. J. Impact Eng 32 (2006) 2017–2031.
[8] L.J. Malvar, J.E. Crawford, J.W. Wesevich, D. Simons, A plasticity concrete [20] M. Unosson, Numerical simulations of penetration and perforation of high
material model for DYNA3D, Int. J. Impact Eng. 19 (9–10) (1997) 847–873. performance concrete with 75 mm steel projectile, FOA-R-00-01634-311-SE,
[9] L.J. Malvar, J.E. Crawford, J.W. Wesevich, D. Simons, A new concrete material 2000.
model for DYNA3D, Karagozian & Case, Glendale, California, USA, 1994 [21] ACI 214-77(97) Recommended practice for evaluation of strength test results
TR-94-14.3. of concrete. American Concrete Institute, 1977.