You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/275220179

Calibration of a Concrete Damage Material Model in LS-Dyna for a Wide Range


of Concrete Strengths

Conference Paper · November 2009


DOI: 10.13140/RG.2.1.3503.4723

CITATIONS READS

5 3,627

3 authors, including:

Eytan Kochavi G. Ben-Dor


Ben-Gurion University of the Negev Ben-Gurion University of the Negev
64 PUBLICATIONS   299 CITATIONS    409 PUBLICATIONS   6,665 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

shock waves on an object View project

structural mechanics View project

All content following this page was uploaded by Eytan Kochavi on 29 April 2015.

The user has requested enhancement of the downloaded file.


Calibration of a Concrete Damage Material Model in LS-Dyna for a
Wide Range of Concrete Strengths

Natalia Markovich a, Eytan Kochavi a,b, Gabi Ben-Dor a


a
Protective Technologies Research & Development Center, Faculty of Engineering Sciences,
Ben- Gurion University of the Negev, Beer-Sheva, Israel
b
Nuclear Research Center Negev, Beer-Sheva, Israel

ABSTRACT

Constitutive modeling is the mathematical description of material behavior under various loadings. This is one
of the most intensely researched fields within solid mechanics because of the importance of accurate constitutive models
for practical engineering problems. A wide range of available constitutive models for concrete were investigated. One
successful model implemented in the hydrocode LS-DYNA, is the Concrete Damage model, which is able to predict the
behavior of confined plain concrete. This model is calibrated in this work for a wide range of concrete strengths, with
respect to triaxial compression test data found in literature. The adjustment of parameters was performed as function of
the only parameter known to most users: the unconfined compressive strength of a specific concrete. This enables an
occasional user to input only this parameter, while the rest will automatically be calculated. This calibrated model named
herein BGU, provides a flexible and powerful method to predict the performance of confined plain concrete. It is shown
that the present calibration performed well when compared to the automatic parameters generation option currently
available in the code.

1. INTRODUCTION

The rapid development of efficient mathematical algorithms and the increasing availability of powerful
computer resources in the last decades have facilitated the development of realistic constitutive material models. The
Finite Element Method (FEM) together with highly sophisticated constitutive models has become an indispensable tool
in structural engineering for the prediction of the ultimate load and the corresponding failure mechanism. This
encourages the development of more capable and realistic constitutive models, based on a rigorous approach, for the
analysis and design of concrete structures.
Constitutive modeling of concrete materials has been a theme of research for some decades. Nevertheless, the
complex behavior of concrete, due to its composite nature, cannot always be faithfully reflected in any models dedicated
to the constitutive modeling of the material. Although the behavior of concrete can seldom be replicated with perfect
accuracy, one must remain satisfied with those constitutive models that can exhibit behavior of concrete as realistic as
possible until more advanced models replace them.
This work is addresses the Concrete Damage Model which is implemented in LS-Dyna1 code which is one of
the Hydro codes used to compute dynamic problems. The main difficulty of using realistic but complex constitutive
models is that they require a long set of input parameters. In order to obtain the proper values for these parameters, the
properties of concrete should be extracted from uniaxial, biaxial, and triaxial tests (in tension and compression) that
should be performed on samples of concrete. Although there are many types of concrete of various strengths, the
majority of common concretes can be characterized by a single parameter called the uniaxial unconfined compressive
strength, often denoted in literature as fc'. For concretes, the uniaxial compressive strength cannot describe the elastic
response, the inelastic (plastic) response including the shear failure envelope, compressibility (compaction), and tensile
failure. Obviously a single parameter cannot characterize all aspects of all types of concrete. However, frequently
engineers have to perform analyses involving a specific concrete for which little or no information is available in
addition to fc'. Further complicating this lack of material characterization knowledge, it is often the case that engineers
performing the analyses have had no formal training in concrete material response and characterization. Therefore, it is
practical to have a concrete material model that requires minimal input, but provides a robust representation of the many
response characteristics of this complex material, including damage and failure.

2. CONCRETE DAMAGE MODEL in LS-DYNA

The Concrete Damage Model handels the volumetric and deviatoric responses separately, as often done in
explicit codes. It is a three-invariant model, uses three shear failure surfaces, and includes damage and strain-rate effects.
This model is a linear isotropic hypoelastic-plastic model with strain rate scaled elastic domain, a non-associated flow
rule and non-linear anisotropic strain hardening and softening representing stable and unstable cracking. The deviatoric
and isotropic parts of the stress tensor are uncoupled and the isotropic behavior is governed by a compaction curve or
equation of state (EOS). The EOS prescribes a set of pressures, unloading bulk moduli, and volumetric strains. Once the
pressure has been determined from the EOS, a movable surface, or failure surface, limits the second invariant of the
deviatoric stress tensor.
Concrete Damage Model was developed by Javier Malvar2, Jim Wesevich, and John Crawford of K&C*, and
Don Simon of Logicon RDA in support of the Defense Threat Reduction Agency’s programs, USA.
It was developed and modified (Malvar2,3,4,5,6) mainly to analyze concrete structures subjected to blast loading.
This concrete model originally based on the Pseudo-tensor model implemented in LS-DYNA (originally DYNA3D) as
material model type 16 and now available in LS-DYNA as Material model type 72. The modifications that were made to
the original model ensured that the material response followed experimental observations for standard uniaxial, biaxial,
and triaxial tests for both tension and compression type loading. As for now, the last version of K&C concrete material
model presented in LS-DYNA is model type 72 release III. A disadvantage of previous implementation of that material
model for regular user was the overwhelming amount of input that is required. The most significant user improvement
provided by Release III is a model parameter generation capability, based solely on the unconfined compression strength
of concrete. This model is expected to be used primarily with the option to generate the model parameters automatically.
However, there remains an option to modify parameters manually. Since the purpose in this work is to investigate the
model, the manual modification option is used to adjust the parameters with accordance to experimental three-axial
concrete results. Further we use K&C notation to show automatically parameter generated model in simulation results.
Concrete Damage model is a plasticity-based formulation with three independent failure surfaces, which change
shape depending on the pressure. Plasticity model for concrete are represented in principal stress space. The general
characteristics of the failure surfaces are described by its cross-sectional shape on the deviatoric plane and its trace on
meridian planes. The constitutive model defines in terms of stress invariants. The following Figure 1 present the stress-
strain response from an unconfined uniaxial compression test with the surfaces, which define three important regions of
concrete behavior. Here failure surfaces represented in ∆σ−P plane, Figure 1 (b), where ∆σ limits deviatoric stresses and
defined as follows:

∆σ = 3J 2
(1)
where J2 is the second invariant of the deviatoric stress tensor. While P is pressure (pressure is positive in compression).
Figure 1 (b) represent stress-strain concrete response for compression.

*
Karagozian & Case - consulting engineering firm
Figure 1 (a) Failure surfaces in concrete damage material model; (b) Concrete constitutive behavior, modified from
Malvar 4.

The material response is considered linear up to point 1 (Pt.1), the first yield, the path shown in grey in Figure 1
(b). After yielding, a hardening plasticity response occurs until point 2 (Pt.2), the maximum strength, is reached (red).
After reaching the maximum strength, softening occurs until a residual strength (Pt.3), which depends on the amount of
confinement, is obtained, (green). The higher the amount of confinement pressure applied on a specimen, the higher
maximal and residual strengths become.
After the stress reaches the initial yield surface and before reaching the maximum failure surface, the stress is
calculated as a linear interpolation between the two see Figure 1 (a):

∆σ = η ( ∆σ m − ∆σ y ) + ∆σ y (2)
After reaching the maximum failure surface, the current failure is similarly interpolated between the maximum and the
residual stresses:

∆σ = η ( ∆σ m − ∆σ r ) + ∆σ r (3)

where η, parameter indicating the relative location of the current failure surface, varies between 0 and 1, and depends on
the accumulated effective plastic strain parameter λ. The value η equals 0 when λ = 0; this is the initial value of η before
any plasticity has occurred Then, when η = 1 at some value λ = λm the meaning is that the maximum failure surface has
been reached. Finally, when reduces to η = 0 at some larger value of λ − the residual failure surface is reached. This
would permit ∆σ sequentially to take on the values ∆σ Y , ∆σ m and ∆σ r . These ranges of the parameters η & λ are
summarized in Table 1, where λend is some large value of λ.

Table 1 Damage evaluation


η λ Current position
0≤η<1 0 ≤ λ < λm Between yield failure surface to maximum failure surface
η=1 λ = λm On maximum failure surface
0<η≤1 λm < λ ≤ λend Between maximum failure surface to residual failure surface

Consequently, when λ ≤ λm, the current failure surface is calculated using Eqn(2), and when λ ≥ λm, the current
failure surface is calculated using Eqn(3), as shown at Figure 2.
maximum failure surface

0
°m °
yield failure surface residual failure surface

Figure 2 Damage function schematic evaluation in Concrete Damage Model

In summary, a total of eight parameters define three fixed failure surfaces, as follows (Malvar 4):

p
∆σ m = a0 +
a1 + a2 p (maximum failure surface) (4)
p
∆σ r =
a1 f + a2 f p
(residual failure surface) (5)
p
∆σ y = a0 y +
a1 y + a2 y p
(yield failure surface) (6)

The basic parameters necessary to define the Concrete Damage model are described in Table 1. Only principal
parameters are mentioned, the rest of the parameters related to output requests or unit converting factors are not
discussed herein.

Table 1 Concrete Damage Model Parameters in LS-DYNA


ρ mass density
υ Poisson’s ratio
ft tensile strength
a0y, a1y, a2y initial yield surface parameter
a0, a1, a2 maximum shear failure surface parameter
a1f, a2f residual failure surface parameter
η parameter indicating the relative location of
the current failure surface
λ modified effective plastic strain (damage
parameter)
3. THE CALIBRATION OF THE CONCRETE DAMAGE MODEL

The calibration of the Concrete Damage Model was based on a consistent study of the model’s parameters, with
highlight on the compressive states, in which most of the concrete's potential to resist loads presents itself. Then, the
relevant parameters, most of which are shown in Table 1, were presented as functions of the maximal unconfined
concrete strength, fc'. The process of obtaining these functions is actually the calibration which is the purpose of this
work. This process is explained herein.

3.1 Failure surface parameters

As mentioned above, the damage concrete model is based on three independent fixed surfaces to define the
plastic behavior of material. Since the parameters which define the shape of the failure surface are user-provided, one
can input them in form of formulas that are functions of concrete strength.
The points that define failure surfaces in the Concrete Damage Model are the yield point, the maximum strength
and the residual strength of concrete constitutive behavior as shown on Figure 1, which should be determined from
experimental three-axial concrete results. In this work the adjustment of the parameters (or calibration) was done not
according to actual tests performed on concrete specimens, but using experimental data found in literature. We decided
to use the extensive experimental data published by Attard7, 8 and Attard & Steward which covered a wide range of
concrete strengths (20 to 120 MPa) and presented relationships by which test results may be generated numerically. The
test samples used by Attard7, 8 were concrete cylinders 100 mm in diameter and 200 mm high, produced from three kinds
of Australian aggregate and five different mixes of cement with silica fume and binder. The combination of these mixes
resulted with eleven types of high strength concrete (HSC) which were tested uniaxialy 90 days after casting and ranged
with strengths of 60-120 MPa (after 28 days the uniaxial compressive strength range was 45-109 MPa). Then, samples
were tested in a triaxial cell which was able to maintain a lateral pressure of up to 20 MPa. Based on test results, Attard7,8
provided an analytical model of the full stress-strain relationship for uniaxialy and triaxially loaded concrete specimens.
This empirical model, in the form algebraic relations, was shown to be applicable for a wide range of concrete. Attard et
al demonstrated that their model could predict a stress-strain relationship for any of the tests performed with accuracy of
about 10%. Moreover, they demonstrated that the specific combination of cement mix and aggregate had no significant
effect on the shape of the stress-strain curve as long as same maximal strength had been obtained.
Thus, failure surface points were found from Attard's concrete's stress-strain behavior curves. Figure 3 and
Figure 4 represent these curves obtained by Matlab routine based on Attard's formulations for 50 MPa concrete. Figure 3
represent yield failure surfaces points position, which were taken as 0.65 of maximum strengths. Figure 4 represent
maximum strengths points. The points at Figure 3, 4 are indexed from (1) to (5) are the yield/maximum strength
obtained for 0 MPa, 1 MPa, 5 MPa, 10 MPa and 15 MPa confinement pressure, respectively.

140
yield surface points
fr=15 MPa
120

100
fr=10 MPa

80 (5)
(4)
stress, MPa

60 fr=5 MPa
(3)
(2)
40 (1)

fr=1 MPa
20

fr=0 MPa
0
0 0,005 0,01 0,015 0,02
axial strain

Figure 3 Attard concrete constitutive behavior results for 50 MPa concrete strength, yield strength points
140
maximum surface points
(5) fr=15 MPa
120

(4)
100
fr=10 MPa
(3)
80
stress, MPa

60 (2) fr=5 MPa

40 (1)

fr=1 MPa
20

fr=0 MPa
0
0 0,005 0,01 0,015 0,02
axial strain

Figure 4 Attard concrete constitutive behavior results for 50 MPa concrete strength, maximum strength points

The residual strength due to confinement should be obtained for large strains where stress becomes almost constant.
Figure 5 shows Attard results for large strains.

Figure 5 Attard concrete constitutive behavior results for 50 MPa concrete strength, residual strength points

Let us consider meridian plane the hydrostatic pressure and stress differences, this is the form in which Malvar4
described his failure surfaces, see Figure 1. Those data should be determined from Attrard's concrete response. Since for
cylindrical triaxial compression lateral principal stresses are equal, hydrostatic pressure could be written as follows:
1 1 1
p = I1 = (σ 1 + σ 2 + σ 3 ) = (σ 1 + 2σ 2 ) (7)
3 3 3
Further, stress differences or second deviatoric invariant for yield, maximum and residual strength points were defined.
The next step is to map set of stress differences and pressures to the meridian plain in 3J 2 / p axes, thus we receive
points that lie on failure surfaces. Figures 6-8 represent yield, maximum, residual surfaces for 50 MPa strength concrete.
Herein points numbered from (1) to (5) represent different level of confinement: 0, 1, 5, 10, 15 MPa, respectively.
The additional points were decided to take in order to determine residual surface, they are 20 MPa and 25 MPa, see
Figure 8.
80

yield surface
70
yield surface points
(5)
60 (4)

50 (3)

40 (2)
(1)
∆σ, MPa

30

20

10

0
0 10 20 30 40 50 60
P, MPa
Figure 6 Yield surface for 50 MPa concrete in meridian plane, based on Attard results
140

maximum surface
120 maximum surface points
(5)
100
(4)

80 (3)
∆σ, MPa

60 (2)
(1)
40

20

0
0 10 20 30 40 50 60
P, MPa

Figure 7 Maximum failure surface for 50 MPa concrete in meridian plane, based on
Attard results
100
(7)
residual surface
90
residual surface points
80
(6)
70
(5)
60
(4)
50
∆σ, MPa

40
(3)
30

20

10 (2)
(1)
0
0 10 20 30 40 50 60
P, MPa

Figure 8 Residual failure surface for 50 MPa concrete in meridian plane, based on
Attard results
The next stage was to find the fitting curve, which suites failure surfaces points in the best way. The fitting
curves shapes should satisfy the eqns (4 – 6) for yield, maximum and residual failure surfaces, respectively. Using a
nonlinear least squares curve fitting procedure, the failure surfaces for 50 MPa concrete were obtained. Accordingly, we
received eight failure surface parameters for 50 MPa strength concrete.
Further, using the same approach as for 50 MPa concrete, eight failure surface parameters from Attard's curves
for wide range of concrete were found. In this work, we consider an interval from 30 to 120 MPa concrete strengths
every 10 MPa. Thus, more 9 calculations should be done, similar to above.
Table 2 represent eight failure surfaces parameters for wide range of concrete found based on Attard's curves.

Table 2 Failure surface constants for 30-120 MPa range concrete strength
concrete failure surfaces parameters
strength,
a0 a1 a2 a0y a1y a2y a1f a2f
MPa
30 10.34 0.4729 0.003486 6.739 0.7290 0.005329 0.5181 0.002420
40 12.67 0.4491 0.002838 8.227 0.6902 0.004383 0.4740 0.002726
50 14.00 0.4195 0.002580 9.080 0.6448 0.003978 0.4635 0.002410
60 14.45 0.3902 0.002442 9.438 0.6017 0.003736 0.4582 0.002199
70 15.52 0.3774 0.002178 10.078 0.5803 0.003354 0.4606 0.001912
80 16.02 0.3637 0.001993 10.443 0.5600 0.003058 0.4659 0.001640
90 16.62 0.3543 0.001817 10.752 0.5441 0.002802 0.4725 0.001403
100 17.24 0.3473 0.001664 11.163 0.5336 0.002567 0.4808 0.001173
110 17.55 0.3401 0.001541 11.395 0.5231 0.002370 0.4899 0.000960
120 18.12 0.3357 0.001423 11.702 0.5155 0.002196 0.4899 0.000960

Since the parameters which define the shape of the failure surface are user-provided, we can input them in form
of formulas that are functions of concrete strength. The next step, therefore, was to find a fitting curve for each failure
constant as a function of concrete strength fc' for the strength range in Table 2. Appropriate functions were chosen in
order to provide the best fit to the data points. The following equations were chosen for constants:
b
a0 , a1 , a2 , a1 y , a2 y - exponential function of form a f c' ( ) ;
a0y - logarithmic c ⋅ ln( f c' − d ) ;
k ,
a1 f = e ; a 2 f =
f c'
where a, b, c, d, e, k are some constants that should be determined using least square fitting.
The obtained constants (in form of functions) were substituted in LS-Dayna's input file in order to receive first
assumptions simulation results. We found the results in unsatisfactory deviation from Attard's. After some modifications
for parameters a0 and a2, adequate accuracy has been obtained. The set of suitable functions for the failure surfaces
parameters is herein presented:
0.4369
a 0 = 2.442 f c' (8)
' −0.2463
a1 = 1.084 ⋅ f c
(9)
' −0.6416
a 2 = 0.03276 ⋅ f c (10)

a0 y = 2.525 ⋅ ln( f c' − 15.18) (11)


' −0.2706
a1 y = 1.846 ⋅ f c (12)
−0.5355
a 2 y = 0.03181 ⋅ f c' (13)
a1 f = 0.4773 (14)
0.11 (15)
a1 =
f c'
These failure surface parameters can be included in any LS-Dyna input file as functions of compressive strength, which
will enable the simple use of the concrete model even by the occasional user. The following sets of plots Figures 9-16
represent failure surfaces parameters evaluation with wide range of concrete strength and comparison with original
automatic generated parameters model. The first assumption for parameters a0 and a2 are also shown.

K&C
35 BGU 0,5
fitting curve to Attard
Attard points
0,45
25
a0 values

a1 values
0,4
15 K&C
0,35 BGU
Attard points
5 0,3
20 30 40 50 60 70 80 90 100 110 120 130 20 30 40 50 60 70 80 90 100 110 120 130
fc', MPa fc', MPa
Figure 9 Parameter a0 evaluation with wide Figure 10 Parameter a1 evaluation with wide
range of concrete strength range of concrete strength

K&C 30
K&C
0,0035 BGU
fitting curve to Attard
25 BGU
Attard points
Attard points
20
0,0025
15
ay0 values
a2 values

10
0,0015
5

0,0005 0
20 30 40 50 60 70 80 90 100 110 120 130 20 30 40 50 60 70 80 90 100 110 120 130
fc', MPa fc', MPa
Figure 11 Parameter a2 evaluation with wide Figure 12 Parameter ay0 evaluation with
range of concrete strength wide range of concrete strength

0,8 K&C
0,008 K&C
BGU BGU
0,7 Attard points Attard points
0,006
a2y values

0,6
a1y values

0,004
0,5

0,002
0,4
20 30 40 50 60 70 80 90 100 110 120 130
20 30 40 50 60 70 80 90 100 110 120 130
fc', MPa fc', MPa
Figure 13 Parameter a1y evaluation with Figure 14 Parameter a2y evaluation with
wide range of concrete strength wide range of concrete strength
0,55 BGU 0,006
K&C
K&C
Attard points
BGU
0,5 0,004 Attard points
a1f values

a2f values
0,45 0,002

0,4 0
20 30 40 50 60 70 80 90 100 110 120 130 20 30 40 50 60 70 80 90 100 110 120 130
fc', MPa fc', MPa
Figure 15 Parameter a1f evaluation with wide Figure 16 Parameter a2f evaluation with wide
range of concrete strength range of concrete strength

3.2 The Damage function

The change in stress as a function of plastic strain arises from physical mechanisms such as internal cracking,
and the extent of this cracking is affected by the hydrostatic pressure when the cracking occurs. This mechanism gives
rise to the “confinement” effect on concrete behavior. To account for this phenomenon a “damage” function was defined
and incorporated into K&C Concrete Model.
For numerical simulation function η(λ) is input as a discrete function of mostly 13 (η,λ) pairs. The current
value of λ is compares to a proposed set of 13 pairs determined by user. The damage function value η should initiate
with 0, increase to 1 at some value λ=λm, and then decrease to 0 at some large value of λ. Thus, according to the LS-
DYNA manual1 at the beginning of the subroutine the value of λm is defined simply as the value of λ corresponding to
the first relative maximum of η in the. In accordance with our Attard's curves observation the value 0.00009 had been
taken for λm.
Figure 17 shows our suggested values for the discrete damage parameters (labeled BGU) in comparison to the
automatically generated values (labeled K&C). That (η,λ) set is used for all our simulations in model calibration process.

1,0
K&C
BGU
0,8

0,6
°

0,4

0,2

0,0
0,000 0,001 0,002 0,003 0,004 0,005
°

Figure 17 BGU damage function compared to K&C


4. NUMERICAL RESULTS

The model consists of the cylindrical specimen used for experiments in Attard's work: 100 mm in diameter and
of 200 mm height. An eighth of a cylinder with three symmetry planes was defined in LS-DYNA. It is important to note
that since the ends of the real cylinders in Attard's tri-axial compression tests were ground to obtain smooth surfaces, we
did not fix the ends to radial direction in our simulations. Moreover, we found that it was insignificant as fixed ends
resulted in very similar axial forces as function of prescribed displacement. The difference between radially fixed and
free ends results were no more than 2%.
The loading was applied with accordance to Attard's tests, quasi-static loading. The confining pressure was
applied initially in order to prevent crashing of sample and then displacement of the axial loading piston began to move
with defined velocity. Velocity is used to represent displacement control, rather than pressure or force condition, in order
to model softening following peak strength. Simulations last 100 sec with maximum axial displacement is 10 % of the
cylinder height and applying confinement pressure 0 MPa, 5 MPa, 10 MPa and 15 Mpa sequentially.
As mentioned above, all parameters were calibrated as function of uniaxial concrete strength (fc') and
implemented in Concrete Damage Model input file. Figure 18 and 19 shows a comparison of our results (labeled: BGU),
with Concrete Damage Model automatically generated parameter results (labeled: K&C) and Attard's test results
(labeled: Attard) for 50 MPa and 100 MPa concrete respectively. Each plot is for a different confinement pressure (0, 5,
10 and 15 MPa).

140

120

fr=15 MPa
100

fr=10 MPa
80
Attard
BGU
stress, MPa

60 K&C

fr=5 MPa
40

20

unconfinement
0
0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 0,04
axial strain

Figure 19 Concrete Damage Material Model, Release III for 50 MPa strength concrete simulation results, calculated with
K&C, BGU input data are compared to Attard constitutive behavior
200

180

160
fr=15 MPa
140

120
Attard
stress, MPa

100 BGU
fr=10 MPa K&C
80

60
fr=5 MPa
40

20
unconfinement
0
0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 0,04
axial strain

Figure 20 Concrete Damage Material Model, Release III for 100 MPa strength concrete simulation results, calculated
with K&C, BGU input data are compared to Attard constitutive behavior

As we can conclude, the deviation from Attard model is no more than 2% relative to the maximum strength,
which is a more than acceptable deviation. One has to keep in mind that Attard's correlations were within 10% error8.
This is also in the range of error9 expected to be found between test results performed on concrete specimens casted from
the same batch and of the same age. Anyone accustomed to concrete would consider a 10% error as negligible.
Comparing BGU model and K&C model with Attard's stress-strain curves, calculating absorbed energy, we can estimate
errors obtained with following formulas:
energy K &C − energy Attard
errorK &C = ⋅ 100%
energy Attard (16)
energy BGU − energy Attard
errorBGU = ⋅ 100% (17)
energy Attard
The results of absorbed energy calculation for 50 MPa and 100 MPa concretes represented in tables 3 and 4,
respectively.

Table 3 Absorbed energy calculation for 50 MPa concrete


concrete confinement, Energy error, %
strength, MPa Attard K&C BGU K&C BGU
MPa 0 0.21 0.39 0.20 89 2
50 5 2.16 2.42 1.90 12 12
10 3.30 3.89 3.15 18 5
15 4.10 4.87 4.10 19 0.07
Table 4 Absorbed energy calculation for 100 MPa concrete
concrete confinement, Energy error, %
strength, MPa Attard K&C BGU K&C BGU
MPa 0 0.37 2.04 0.41 446 10
100 5 2.74 3.33 2.42 21 12
10 4.60 5.37 4.07 17 11
15 6.05 6.76 5.55 12 8

As one can see BGU model gives more accurate results with smaller errors than automatic K&C concrete
model.

5. CONCLUSIONS

The most important aspects of this work are the study and calibration of the "Concrete Damage" model's
parameters in order to enable simple use of this model using LS-DYNA. Usually, engineers have to perform analyses
involving concrete where little or no information is available other than its unconfined strength. For concrete simulations
requiring a high degree of confidence, the analyst may need to specify as much as 49 material model parameters. This set
of parameters must be defined using experimental data and knowledge of fundamental concrete behavior. For this
purpose, the concrete damage model's parameters were calibrated using published experimental data for a wide range of
concrete strengths. In our case Attard's results were used. As a result of the calibration performed in this research, the
model enables now easy use and shows good agreement with the triaxial tests. The main result of this project is a novel
BGU calibrated model that performers more accurately than the calibrated model implemented by Karagosian&Case into
the last version of LS-DYNA.

ACKNOWLEDGMENTS

We thank Barak Ostraich for his assistance in numerical simulations.

REFERENCES
1. LS-DYNA keyword user’s manual, version 971. Livermore Software Technology Corporation; May 2007
2. Malvar, L.J., Crawford, J.E., Wesevich, J.W. and Simons, D. A new concrete material model for DYNA3D,
Karagozian and Case. Report No. TR94-14.3, 1994.
3. Crawford, J.E. and Malvar, L.J. User’s and theoretical manual for k&c concrete model. Karagozian and Case,
Burbank,CA, TR 97-53.1. December 1997.
4. Malvar, L.J., Crawford, J.E., Wesevich, J.W. and Simons, D. A plasticity concrete material model for DYNA3D.
International Journal of Impact Engineering, 19(9/10), 847-873, 1997.
5. Malvar L.J., Crawford J.E., Wesevich J.W. and Simons D. A new concrete material model for DYNA3 - release II:
shear dilation and directional rate enhancements. Report to the Defense Nuclear Agency, TM-96-2.1. Glendale, CA:
Karagozian and Case Structural Engineers, January 1996.
6. Malvar, L.J., Crawford, J.E. and Morrill K.B. K&C concrete material model release III – automated generation of
material model input. Report to the Defense Nuclear Agency, TR99-24-B1, August 18, 2000.
7. Attard, M.M. and Setunge, S. The stress strain relationship of confined and unconfined concrete. ACI Materials
Journal, 93(5), 432-442, 1996.
8. Attard, M.M. and Stewart. M.G. A two parameter stress block for high strength concrete. ACI Structural Journal,
95(3), 305-317, 1998.
9. ACI 214-77(97) Recommended practice for evaluation of strength test results of concrete. American Concrete
Institute, 1977.

View publication stats

You might also like