Finite Elements in Analysis and Design 97 (2015) 1–19

Contents lists available at ScienceDirect

Finite Elements in Analysis and Design

journal homepage: www.elsevier.com/locate/finel

Calibration of the continuous surface cap model for concrete

Hua Jiang a,b,n, Jidong Zhao a
a
Dept of Civil and Environmental Engineering, the Hong Kong University of Science and Technology, Clearwater Bay, Kowloon, Hong Kong SAR, China
b
Key Laboratory for Bridge and Tunnel of Shaanxi Province, Chang’an University, Xi’an, China

art ic l e i nf o a b s t r a c t

The continuous surface cap (MAT 145) model in LS-DYNA is known by its elegant and robust theoretical
Keywords: basis and can well capture many important mechanical behaviors of concrete. However, it appears to be
Concrete less popular than many other constitutive models in engineering application due to many material
Constitutive model parameters involved in the model formulation which are difficult to calibrate. This study presents an
Numerical simulation effective calibration method to determine the material parameters for this model as functions of uniaxial
LS-DYNA compression strength and the maximum aggregate size of concrete according to formulas from CEB-FIP
MAT 145 code and concrete test data from other published literatures. The obtained parameters can be
MAT 159
conveniently used for occasional users with little or no information on concrete in hand. We further
compare the predictions of stress–strain relationship in tension and compression under different
confining pressures as well as hydrostatic compression by the model, and validate the model based
on impact test of RC beams. Besides, the model is further compared against a similar model-MAT 159 in
terms of model performance. The results demonstrate that the model based on the calibrated
parameters is capable of offering reasonable and robust predictions.
& 2014 Elsevier B.V. All rights reserved.

1. Introduction how the plastic deformation is calculated. Category one normally

adopts an associated flow to calculate the plastic strain increment,
Arguably one of the most widely used man-made materials, and may capture the plastic volume expansion (dilatancy) [6]
concrete underpins the performance and safety for key structures caused by shear loading at low confining pressure. It considers
relevant to civil engineering, onshore and offshore engineering, coupled volumetric and shear behavior (i.e. shear enhanced
nuclear facility protection and many others. The dynamic compaction and pressure dependence of shear strain) of concrete.
responses of concrete when subjected to impact or blast loads Typical examples of this category include the geologic cap (MAT
are of particular importance in many of these engineering fields, 25), Schwer Murray Cap, also called continuous surface cap (MAT
and relevant research has hence attached much interest. Computer 145), CSCM Concrete (MAT 159), Mohr Coulomb (MAT 173), and
modeling has now been widely adopted as a cheap and effective Druker Prager (MAT 193) [3]. Category two generally employs the
way in assisting the design of (reinforced) concrete structures Prandtl–Reuss flow theory (where the Von Mises criterion is used
under those extreme loads. Among many key ingredients that as the plastic potential) to calculate the plastic strain increment.
affect reliable and accurate predictions by a numerical tool, an The plastic volume strain is obtained from the equations of state
appropriately developed and calibrated constitutive model to (EOS), and the plastic volume strain increment is independent of
describe the dynamic behavior of concrete subjected to dynamic the incremental flow rule (the Prandtl–Reuss flow theory). Since
loads plays a core role. shear and volumetric behaviors are decoupled, the phenomenon
There have been a good number of constitutive models devel- of shear dilation cannot be captured. Typical models belong to this
oped in the literature for concrete, with forms ranging from category are the soil and form (MAT 5/14) [3], pseudo tensor (MAT
relatively simple to more sophisticated (see a recent review in 16) [3], concrete damage (MAT 72) [3,7], Winfrith concrete (MAT
[1,2]). They have also been implemented in commercial software 84/85) [3,8], Johnson Holmquist concrete (MAT 111) [3,9], RHT
such as LS-DYNA [3], AUTODYN [4] and ABAQUS/explicit [5]. These (MAT 272) [3,10] and so on. Those models have been widely used
models can be generally divided into three categories according to to model concrete under high impact loads. The third category
commonly assumes a non-associated flow in calculating the plastic
strain increment. With a different plastic potential surface than
n
Corresponding author at: Dept of Civil and Environmental Engineering, the
the yield surface, the shear dilatancy can be well controlled.
Hong Kong University of Science and Technology, Clearwater Bay, Kowloon, Hong
Kong SAR, China.
A typical example of this category is the plastic-damage model
E-mail addresses: [email protected] (H. Jiang), [email protected] (J. Zhao). [5,11].

http://dx.doi.org/10.1016/j.finel.2014.12.002
0168-874X/& 2014 Elsevier B.V. All rights reserved.
2 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

Notation D1 ; D2 parameters determining the shape of pressure–

volume in Eq. (6)
MAT 145 continuous surface (Schwer Murray) cap σ; σ stress tensor and effective stress tensor in Eq. (7)
7 ! "
MAT 159 CSCM concrete d; d ; G τ 7 scalar damage variable in Eqs. (7) and (8)
MAT 72 concrete damage r 07 ; τ 7 damage threshold and undamaged energy norm in
MAT 72R3 concrete damage REL3 (K&C concrete) Eq. (8)
0 0
MAT 25 geologic cap f c ; f bc uniaxial and biaxial compression strength of concrete
0 0
MAT 111 Johnson Holmquist concrete f t ; f bt uniaxial and biaxial tension strength of concrete
MAT 16 Pseudo tensor τ0 ; σ 0 shear strength and normal strength in Eq. (10), used
MAT 272 RHT ! " ! by Mills " and Zimmerman [20]
MAT 84 Winfrith concrete g σ ij ; g I 1 ; J 2 Gibbs free energy density (per unit volume) in
EOS equations of state ! " Eqs.
! " (22) and (24)
TXC triaixal compression g 1 J 2 ; g 2 I 1 deviatoric and volumetric part of Gibbs free
TOR torsion energy density in Eq. (22)
TXE triaxial extension E; γ Yong’s modulus and Poisson’s ratio
CDM continuum damage mechanics K; G bulk modulus and shear modulus
$1
DERR damage energy release rate used by Simo and Ju [31] Dijkl fourth-order linear-elastic compliance matrix tensor
RC reinforced concrete of the intact material in Eq. (21)
I 1!; J 2 ; J 3" three invariants of stress tensor I ijkl ; I dijkl fourth-order identity tensor and deviatoric tensor in
R I 1 ; J 3 Rubin scaling function in Eq. (1), used by Rubin [27] Eq. (21)
κ; κ0 cap hardening parameter in Eqs. (3) and (5) GF ; GF0 the mode I fracture energy and base value of fracture
F f ðI 1 Þ; α; β; γ; θ strength in triaxial compression in Eqs. energy per unit area in Eqs. (36)–(31)
(1) and (2a) σ(w), ω the stress and displacement in Eqs. (26)–(32)
Q i ; αi ; βi ; γ i ; θi ði ¼ 1; 2Þ strength in torsion and triaxial exten- Gc the compression fracture energy in Eqs. (32)–(36)
sion in Eqs. (2b) and (2c) dmax the maximum aggregate size in Table 2
l a characteristic length of the finite element in Eqs.
n
F c ðI 1 ; κ Þ the cap surface in Eqs. (1) and (3)
XðκÞ; LðκÞ; Xðκ 0 Þ the cap surface parameters in Fig. 1. (29), (31), (35), (36)
S ratio of the major to minor axes of the cap surface in A 7 ; B 7 damage parameters determining the strain softening
Eqs. (4), (18a), (18b) and (19e) curve in Eqs. (8) and (31), (36)
εpv ; W plastic volumetric strain and the maximum value in σ ij ; σ~ ij the viscid and inviscid stress tensor in Eqs. (37)–(39)
Eq. (6) η a fluidity coefficient parameter in Eqs. (37)–(39)

When concrete is subject to low velocity impact, there are torsion (TOR), triaxial extension (TXE) and hydrostatic test to be
typical features needing to be captured by a model, such as shear conducted for their calibration, which greatly limits the practical
enhanced compaction, dilatency before and after peak strength, applicability of the model. On the other hand, a “sister” model-
pre-peak hardening, post-peak softening, modulus reduction/stiff- MAT 159 [3,14,15], with internal material parameters generation
0
ness degradation under cyclic loading, irreversible deformation, based upon the unconfined compression strength f c of concrete,
and localized damage accumulation [12,13]. With a sound theore- aggregate size and the units has been included in LS-DYNA since
tical basis, the MAT 145 available in LS-DYNA can well capture version 971. This model uses the same methodology as the MAT
those behaviors together [3]. However, it has not been as popular 145 to predict the behavior of concrete before peak strength, and is
as simple ones such as the MAT 72, MAT 84/85 and the MAT 111 in different from the latter in terms of strain (post peak) softening
engineering application, due primarily to the complexity of the portion for example the evolution of the both brittle damage and
model with many material parameters involved. For example, a ductile damage norm. A comparison between the MAT 159 and
total of 17 material parameters is required to be provided by the MAT 145 will be provided in this research.
user in this model to define the shear and cap surface, which Indeed, similar issue exists for the MAT 72 which has a total
demands exceedingly complicated experiments ranging from uni- of 49 user defined parameters. Karagozian & Case [16,17], Marko-
axial compression, uniaxial tension, triaxial compression (TXC), vich et al. [18] managed to offer an approach of automatically

J F ( I ,κ )
1

Elliptic C ap

I1

0 L (k ) X (k ) 0 L(k) X(k) I
Fig. 1. Compressive meridional profile of the yield surface in the MAT 145: (a) smooth cap failure function, (b) non-dimensional function used for cap portion.
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 3

generating these parameters based on the unconfined compres- Q 2 F f ðI 1 Þ respectively.

sive strength of concrete, which considerably facilitates the easy ! "
Q 1 ¼ α1 $ γ 1 exp $ β1 I 1 þ θ1 I 1 ð2bÞ
use of the MAT 72 model.
The present study aims to develop a calibration procedure to
! "
facilitate the MAT 145 to be effective used for occasional users of Q 2 ¼ α2 $ γ 2 exp $ β2 I 1 þ θ2 I 1 ð2cÞ
LS-DYNA, in attempting to achieve the same goals as mentioned
above of various studies for the MAT 72. A systematic approach where the eight material parameters α1 ; β1 ; γ 1 ; θ1 and
will be proposed to determine the material parameters of the MAT α2 ; β2 ; γ 2 ; θ2 are evaluated by fitting pffiffiffithe Q 1 F f ðI 1 Þ and Q 2 F f ðI 1 Þ
0
145 based on f c of a concrete, and the maximum aggregate size from TOR and TXE test, and 1= 3 rQ 1 r 1; 0:5 r Q 2 r 1 is
according to the formulas provided in the CEB–FIP code [19] as required to render the deviatoric plane shape of yield surface
well as other empirical formulas in the literature [20,21] which changes from triangular in brittle regime to circle in ductile
will be entailed in the sequel. regime as the confining pressure changes from tensile to high
compressive.
The cap surface [29,30] is a two-part function that is either
2. Theoretical background of the continuous surface cap unity or ellipse (see Fig. 1b):
model (
1 $ ½I 1 $ LðκÞ'2 =½X ðκ Þ $ LðκÞ'2 ; I 1 4 κ
F c ðI 1 ; κ Þ ¼ ð3Þ
To facilitate the calibration of the MAT 145, a brief overview of the 1; otherwise
model is presented in this section. The MAT 145, proposed originally
by Schwer and Murry [22–26], is also called the Schwer Murry Cap or where κ is an internal state variable denoting for hardening of the
continuous surface cap model in LS-DYNA material library. It is indeed cap, by which the intersection of the cap with the I 1 axis X ðκ Þ and
a three-invariant extension of the MAT 25 [3]. This model includes a the transitional point Lðκ Þ (see Fig. 1b) is determined; X ðκÞ $LðκÞ
smooth failure surface and employs damage mechanics to model the is the length of major axe of cap, which is proportional to F f ðκÞ
strain softening and modulus degradation in both tensile and com- with the material constant S (the ratio of its major to minor axes)
pression regimes as well as viscoplasticity for strain rate effects. It
X ðκÞ ¼ LðκÞ þ SF f ðκ Þ ð4Þ
treats plastic flow and damage accumulations as separate processes
based on effective stress concept and the hypothesis of strain
equivalence in continuum damage mechanics (CDM). It is assumed The transitional point LðκÞ is defined by
that plastic flow, controlled by the shear stresses, may result in (
permanent deformation without causing degradation of elastic mod- κ; if κ 4 κ 0
Lðκ Þ ¼ ð5Þ
uli. Meanwhile, damage is assumed to result in progressive degrada- κ0 ; otherwise
tion of the moduli and strength observed on the macroscale due to the
propagation and coalescence of microcracks, microvoids and similar where κ 0 is the value of I 1 at the initial intersection of the cap and
defects in the microstructural level of the material. An elliptical cap shear failure surface.
surface is added to model the plastic volume change related to pore The evolution of the cap’s motion (the cap is only permit to
collapse in concrete material. Besides concrete, it is capable of expand for concrete) is defined by the isotropic hardening rule as
modeling geomaterials including soils and rock [3] and hence is follows, while without cap motion the pressure–volumetric strain
indeed robust and versatile. The paper highlights its use for curve is perfect plastic
concrete only. n o
2
The model is featured by a combined yield surface of a shear εpv ¼ W 1 $e $ D1 ½X ðκÞ $ X 0 ' $ D2 ½X ðκÞ $ X 0 ' ð6Þ
failure surface F f ðI 1 Þ and a cap surface F c ðI 1 ; κÞ, with a continuous
and smooth connection between the two as shown in Fig. 1a. The R
where εpv ¼ trεpij ¼ 3λ∂f =∂I 1 dt is the plastic volumetric strain due
influence of the third deviatoric stress invariant on the shear
to its porosity reduction under compaction, W is the maximum
failure of a material, as suggested by Rubin [27], has also been
plastic volumetric strain, X 0 ¼ X ðκ 0 Þ is the initial abscissa intercept
taken into account.
of the cap surface, D1 and D2 are material parameters determining
The yield surface shown in Fig. 1a can be mathematically
the shape of pressure volume curve.
described by a combination of two functions as follows:
Strain softening and modulus reduction of concrete are mod-
! " ! "
f I 1 ; J 2 ; J 3 ; κ ¼ J 2 $R2 I 1 ; J 3 F 2f ðI 1 ÞF c ðI 1 ; κ Þ ð1Þ eled via an isotropic damage formulation in the model. Strain
softening corresponds to a post-peak decrease in strength and
where I 1 denotes the first invariant of stress tensor defined as modulus reduction denotes a reduction of elastic modulus in cyclic
I 1 ¼ σ ii , J 2 is the second invariant of deviator stress tensor defined loading case. The damage criterion is based on the damage energy
as J 2 ¼ Sij Sij =2 and the deviator stress tensor Sij can be linked to the release rate-based approach, proposed by Simo and Ju [31], which
stress tensor σ ij by Sij ¼ σ ij $ σ ii δij =3, and J 3 ¼ Sij Sjk Ski =3 is the third was considered more reasonable than the equivalent strain-based
invariant of deviator stress tensor; κ is cap hardening parameter, and the stress-based ones [32]. The damaged stress tensor σ
! "
R I 1 ; J 3 is the Rubin scaling function, the detailed documentation (nominal stress tensor, defined as force divided by the total area)
can be found in Ref. [27]. is linked to effective stress tensor σ (undamaged stress tensor or
In the model, shear failure along the compression meridian is true stress tensor, defined as force divided by the total area)
defined in terms of I 1 as an exponential function according to the effective stress concept in CDM by
F f ðI 1 Þ ¼ α $ γexpð $ βI 1 Þ þ θI 1 ð2aÞ σ ¼ ð1 $ dÞσ ð7Þ
where the material parameterspα;ffiffiffiffiβ; γ and θ are evaluated by where the scalar damage variable dð0 r d r 1Þ grows from zero
fitting experimental peak stress J 2 versus I 1 from TXC test. (virgin undamaged material) to unity (completely damaged mate-
The shear curve for the stress
! states
" other than TXC is scaled by rial with effective area reduced to zero). 1 $ d is a reduction factor
the Rubin scaling function R I 1 ; J 3 (a
! general" Lode dependence in associated with the amount of damage at a material point. To
the deviatoric plane) [28], e.g., via R I 1 ; J 3 F f ðI 1 Þ. Then the strength account for different damage responses and to capture the
in TOR and TXE can be expressed by the product Q 1 F f ðI 1 Þ and unilateral effect for concrete material, two distinct expressions
4 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

Table 1
Material parameters of the MAT 145 for different concrete grades (Units: mm, ton, s, N).

Concrete grade C10 C20 C30 C40 C50 C60

0
f c (MPa) 10 20 30 40 50 60
0
f t (MPa) 1.4 2.2 2.9 3.5 4.1 4.6
0
f bt (MPa) 1.4 2.2 2.9 3.5 4.1 4.6
α (MPa) 2.2887 4.8040 7.7088 11.1489 14.9516 19.6383
θ 0.3490 0.3454 0.3400 0.3333 0.3272 0.3191
γ (MPa) 0.1881 1.2821 2.9705 5.3623 8.1299 11.9416
β (MPa $ 1) 0.3513 0.1027 0.0540 0.0338 0.0240 0.0177
α2 0.76 0.76 0.76 0.76 0.76 0.76
θ2 (MPa $ 1) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
γ2 0.42742 0.26443 0.26033 0.25616 0.25616 0.25616
β2 (MPa $ 1) 0.0166 0.0168 0.0115 0.0089 0.0071 0.0059
α1 0.83 0.82 0.82 0.82 0.82 0.82
θ1 (MPa $ 1) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
γ1 0.2560 0.2471 0.2440 0.2407 0.2407 0.2407
β1 (MPa $ 1) 0.037455 0.019743 0.013497 0.010396 0.008317 0.006931
X0 (MPa) 32.14 51.14 70.14 89.14 108.14 127.14
D1 (MPa $ 1) 6.11E $ 04 6.11E $ 04 6.11E $ 04 6.11E $ 04 6.11E $ 04 6.11E $ 04
D2 (MPa $ 2) 2.225E $ 06 2.225E $ 06 2.225E $ 06 2.225E $ 06 2.225E $ 06 2.225E $ 06
W 0.065 0.065 0.065 0.065 0.065 0.065
S 3.8 2.7 2.3 2.1 2.0 1.9

þ $
d and d are used for brittle damage and ductile damage, where J
brittle damage corresponds to tensile mean stress ðP o0Þ, and the θ
Compressive
ductile corresponds to compressive mean stress ðP 40Þ. meridian
The two damage variables are defined as [3,24]
! " r7! "
0 rr 07 rτ 7
7 7 7
¼ G τ 7 ¼ 1 $ 07 1 $ A 7 $ A 7 expB ðr0 $ τ Þ
7
d
γ
τ
α

ð8Þ α−γ
I
where symbol “ 7” denoting “ þ” or “ $ ”, as appropriate,
0
Gð(Þð0 rGð(Þ r1Þ is a monotonically increasing scalar functions
of variable τ 7 , a undamaged energy norm, defined as Fig. 2. Compressive meridian line.
pffiffiffiffiffiffiffiffiffiffiffiffiffi
τ 7 ¼ σ ij : εij in the effective stress space, and r 07 is the damage
threshold, and the initial damage threshold is coincident with the respectively. Parameter β is the coefficient of exponential function
shear plasticity surface; damage d ¼ 0 if τ o
7
and
7
r 07 , as shown in Fig. 2.
7 ! " 7 In order to satisfy the smooth and convex requirements for the
d ¼ G τ 7 4 0 if τ þ 4 r 0þ ; two pairs of parameters A ; B 7
failure surface, the following constraints are imposed on allowable
determine the shape of softening curve. It is worth nothing that
values for theses shear parameters:
parameters A 7 are non-dimensional, and the units for parameters
qffiffiffiffiffiffiffiffiffiffi
B 7 are 1= F=L2 because d in Eq.(8) are non-dimensional.
7 (1) The convexity requirement of the failure surface in both
In summary, there are 12 model parameters αi ; βi ; γ i and meridian plane and deviatoric plane. In compressive meridian
θi ði ¼ ; 1; 2Þ used for shear failure surface, 5 parameters W, D1, D2, plane it is required that
X0 and S for cap surface, and 4 parameters A 7 ; B 7 to describe F″f ðI 1 Þ ¼ $β2 γexpð $ βI 1 Þ r 0
damage. A detailed description how to calibrate and determine these
parameters for different grades of concrete together with a summar-
ization of all those calibration equations will be provided in the In consideration of expð $ βI 1 Þ Z 0; β2 4 0, the following con-
following sections. The calibration work in this research is based on straint is obtained
the units (mm, ton, s, N), the application can be extended to other γ Z0
units when the proposed material parameters are converted to other
units according to the dimensional analysis results provided in
(2) Positive slope at low pressures regimes
Table 1, or FE models in other units are converted into ones in
current unit through Precessor software i.e. ETA femb. F 0f ðI 1 Þ ¼ βγexpð $βI 1 Þ þ θ 4 0

3. A calibration of the continuous surface cap model for When I 1 ¼ 0, it is readily seen that
concrete materials βγ þ θ 4 0

3.1. Constraints on allowable values of shear parameters

(3) Positive slope at a high pressures or the asymptotic slope of
compression line has a positive value. When I 1 -1, we have
In view of the shear failure functions in Eq. (2a), these
expressions can be regarded as a linear combination of a liner θ40
function F f 1 ðI 1 Þ ¼ α þ θI 1 and an exponential function F f 2 ðI 1 Þ ¼ (4) The apex of the meridian has a negative value
$expð $ βI 1 Þ with a combination coefficient γ. Parameters α and
θ are the intercept value in F f 1 axis and the slope of linear function, F f 1 ¼ α $ γexpð $ βI 1 Þ þ θI 1 I1 ¼ 0 4 0
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 5

Compressive
meridian
Triaxial
Bxiaxial compression extension
Triaxial
compression
Tensile
meridian line

Uniaxial
compression

Biaxial Uniaxial tension

compression
Triaxial
tension
Triaxial
tension

Fig. 3. Meridian lines of failure surface: (a) compression, (b) tension.

Thus The four parameters α; β; λ; θ in Eq. (2a) can then be obtained

α4γ Z0 by solving four equations corresponding to the four stress states
0
since f c is known. The results for different concrete grades are
summarized in Table 1.
(5) Positive requirement of shear stress in meridian plane except To determine the parameters for concrete grades other than the
on apex six listed in the table, the best fitting method is adopted to obtain
0
F f 1 ¼ α $ γexpð $ βI 1 Þ þθI 1 4 0 their values as functions of f c according to the existing results in
Table 1. Reasonably good fitting is observed as shown in Fig. 4, and
the fitting formulas are given by
$ 0 %
Hence fc
α ¼ 13:9846exp $ 13:8981 ð11aÞ
0 r expð $ βI 1 Þ r 1 and β40 68:8756

In summarizing above constraints, the requirements for com- θ ¼ 0:3533 $3:4105 ) 10 $ 4 f c $ 3:7150 ) 10 $ 6 f c 2
0 0
ð11bÞ
pressive meridian line parameters are non-negative values
$ 0 %
α 4 0; β 4 0; γ Z 0; θ40 fc
γ ¼ 3:6855exp $ 4:7345 ð11cÞ
40:0239
Following the same method, the constraints on allowable
values for the tensile and shear meridian parameters in Eqs. (2b) β ¼ 18:2146f 0 c
$ 1:7171
ð11dÞ
and (2c) can be also obtained.

3.2. Shear surface parameter 3.2.2. Parameters for the tensile meridian in TXE
Typical stress sates in TXE are uniaxial tension (Point B), biaxial
3.2.1. Parameters for the compressive meridian in TXC compression (Point D), triaxial tension (Point A) and triaxial
As shown in Fig. 3a, the strength data in four stress states, extension (Point E) as seen in Fig. 3b. According to experimental
0
including uniaxial compression (Point C), biaxial tension (Point B), results in Ref. [6], the biaxial compressive strength f bc is related to
0
triaxial tension (equal tension in three directions, see Point A) and f c by
triaxial compression (Point D), are used to determine four para- 0 0
f bc ¼ 1:15f c ð12Þ
meters α; β; γ and θ. According to CEB-FIP code [19], the
The triaxial extension strength is also taken from test results
following relationship holds among the biaxial tensile strength
0 0 reported by Mills and Zimmerman [20]
f bt , the uniaxial tensile strength f t and the uniaxial compressive
strength f c
0 τ0 0:147 þ0:550σ 0
0 ¼ ð13Þ
0 0 ! 0 "2=3 fc f c'
f bt ¼ f t ¼ 1:4 f c =10 MPa ð9Þ
Using the same method for the TXC state, the parameters in
The triaxial compressive strength is obtained from test results TXE for different concrete grades are shown in Table 1 and
reported by Mills and Zimmerman [20], wherein 107 tests includ- demonstrated good fits for parameter β2 as function of f c in
0

ing uniaxial, biaxial and true triaxial with concrete grade varying Fig. 5a. The four parameters α2 ; β2 ; γ 2 ; θ2 are respectively
from C21 to C40 were made. The proposed compressive strength expressed as
for concrete under multiaxial loading condition is
α2 ¼ 0:76 ð14aÞ
τ0 0:199 þ 0:843σ 0
0 ¼ ð10Þ
fc
0
fc θ2 ¼ 0 ð14bÞ

where shear strength defined by τ0 ¼ γ 2 ¼ 0:26 ð14cÞ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! "2 ! "2 pffiffiffiffiffiffiffiffiffiffiffiffi
2
σ x $ σ y þ ðσ x $ σ z Þ þ σ y $ σ z =3 ¼ 2J 2 =3, normal strength $ 0:94843
β2 ¼ 0:285f 0 c ð14dÞ
defined by σ 0 ¼ I 1 =3
6 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

20
0.350

0.345
15
0.340
α value

θ value
10 0.335

0.330

5 0.325

0.320

0 0.315
10 20 30 40 50 60 10 20 30 40 50 60

14 0.40

12 0.35

10 β value 0.30

0.25
8
γ value

0.20
6
0.15
4
0.10
2 0.05

0 0.00
10 20 30 40 50 60 10 20 30 40 50 60

Fig. 4. Evaluation of triaxial compression parameters for a wide range of concretes: (a) parameter α, (b) parameter θ, (c) parameter γ, and (d) parameter β.

a b
0.040
0.018
0.035
0.016
0.030
0.014
0.025
0.012
0.020
0.010
0.015
0.008
0.010
0.006
0.005
20 30 40 50 60 10 20 30 40 50 60

Fig. 5. Evaluation of triaxial extension and shear parameters for a wide range of concretes: (a) parameter β2, and (b) parameter β1.

3.2.3. Parameters for shear meridian in TOR θ1 ¼ 0 ð15bÞ

Relatively few experimental results have been reported for
concrete under pure TOR states. The four parameters α1 ; β1 ; γ 1 ; θ1 γ 1 ¼ 0:2407 ð15cÞ
are obtained
pffiffiffi here according to the Rankine criterion with
Q 1 ¼ 1= 3, as well as Willans–Warnke Lode dependence with β1 ¼ 0:33565f 0 c
$ 0:95383
ð15dÞ
Q 1 corresponding to biaxial compression and triaxial extension
states. The parameters in TOR for different concrete grades are
listed in Table 1. The fitting results for parameter β1 are plotted in 3.3. Cap surface parameters
Fig. 5b, with the following fitting formulas for α2 ; β2 ; γ 2 ; θ2
There are five parameters W, D1, D2, X0 and S used on defining
α1 ¼ 0:82 ð15aÞ the cap surface in the model, among which the first four can be
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 7

Fig. 6. The equation of state for porous material.

4 Shear failure
Experiment data surface
J
(Green and Swanson)
3 B
Fitting curve
F Final Cap surface
E
2 Cap expansion
Intial Cap

T 0 X A C I
1
Fig. 8. Determination the shape parameter S of cap.

0
0 200 400 600 800 1000 1200 The last parameter S for the shape of the cap can be determined
ξ (MPa) from triaxial compression test. The process is summarized as follows:
Fig. 7. Evaluation of hydrostatic compression parameters. ( A hydrostatic load is applied to the specimen leading the
stress state to position A first (see Fig. 8), which results in a plastic
obtained from pressure–density (p–ρ) curves measured in hydro- volumetric strain εpA v according to Eq. (6), and X0 is the elastic limit
static compression test. The p–ρ curve can be further adapted to subject to the hydrostatic load. The hardening of cap surface with
the pressure–volumetric strain p $εv curves through the following the increase of the hydrostatic load can be observed.
relationship εv ¼ ρ=ρ0 $ 1, where ρ0 denotes the initial density. As ( Holding the confining pressure σ 2 ¼ σ 3 ¼ const, applying the
shown in Fig. 6, the p–ρ and p $ εv relationships
! are
" separated
! " into axial compression load Δσ 1 until the specimen fails in shear
three
! regions
" with three turning points A ρA ; P HTL , B ρB ; P HEL and indicated as position B and with the plastic volumetric strain
C ρC ; pC . The first region is marked by AB denoting the elastic Δεpv ¼ Δεv $ Δσ 1 =3 K. The total plastic volumetric strain can be
deformation part with an elastic limit P HEL ¼ X 0 =3. The second one expressed by εpB pA
v ¼ εv þ Δεv .
p

is marked by BC for the compaction part where the pore is being Then I 1c , the intersection of the cap with the I 1 at position C,
crushed out. The last one defines the relationship for fully dense can be estimated from the bellow quadratic equitation according
material in solidification case. to Eq. (16) if parameters D1 ; D2 ; W are known
The elastic limit X0 is set according to the pure hydrostatic !
compression test for concrete, X 0 ¼ 123:9 MPa is reported for con- 2 1 $ εpB
v
D1 ðI 1c $ X 0 Þ þ D2 ðI 1c $ X 0 Þ ¼ $ In ð18aÞ
0 0
crete with f c ¼ 48:4 MPa in [21], X 0 ¼ 70 MPa for f c ¼ 35 MPa and W
0
X 0 ¼ 280 MPa for f c ¼ 140 MPa according to Ref. [4]. A linear fitting
0
curve is followed to fit the relationship between X 0 and f c .
W, D1 and D2 are set using the best fitting method according to Further, parameter S can be expressed by
the p $ εv curves reported by Green and Swanson [21] for concrete
0 ðI 1c $I 1B Þ
with f c ¼ 48:5 MPa (see Fig. 7). The fitting formula is S¼ pffiffiffiffiffiffi ð18bÞ
$ 0% J 2B
ε
$ ln v ¼ D1 ξ þD2 ξ2 ð16Þ In the absence of triaxial compression test data, S can be also
W
estimated from uniaxial compression test. It is known that the
where ξ ¼ XðκÞ $ X 0 and volumetric strain occupied by void is plastic volumetric strain expansion occurs in the shear failure
defined byε0v ¼ W $ εpv ¼ We $ ξðD1 þ D2 ξÞ according to Eq. (6). ε0v surface according to an associated flow rule, while the plastic
decreases from W to 0 as the hydrostatic pressure increases volumetric strain compaction occurs in the cap surface. Position E
(Fig. 6c). (initial cap intersection point) can then be regarded as the critical
Derivative of ε0v with respect to ξ leads to point separating plastic volumetric expansion from plastic volu-
metric compaction. Because the plastic volumetric strain expan-
dε0v
¼ $ WD1 ð17Þ sion is found in uniaxial compression of concrete [6], so the
dξξ ¼ 0
intersection point of cap curve with the shear failure curve lies
Parameters D1 can be considered proportional to the initial before point E ! is reached,
& p and
ffiffiffi' the minimum value of S can be
0"
slope of the ε0v $ ξ curve.
0
estimated by X 0 $f c = f c = 3 .
8 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

300 4.0
Experiment data
250 Fitting curve
3.5

200
3.0

S
150
2.5

100
2.0
50
20 40 60 80 100 120 140 10 20 30 40 50 60

Fig. 9. Evaluation of the cap parameters for a wide range of concretes: (a) parameter X0, (b) parameter S.

! " ! "
Fig. 10. Damage threshold surface in principal undamaged stress space with σ ¼ I 1 I =3 þ S. (a) g σ ij and (b) g 1 J 2 .

Table 1 summarizes the model parameters related to the cap exceeds the current damage threshold It is called the damage
for different grades of concretes, the following fitting formulae criterion here [31,32]. The evolution of the damage variables is
presents the cap parameters for different grade of concrete specified in accordance with the normality rule.
0 The choice of free energy function is critical to the damage
X 0 ¼ 17:087 þ 1:892f c ð19aÞ
modeling of a material. In isothermal conditions, an equilibrium
state can be described by a scalar thermodynamic potential-the
D1 ¼ 6:11 ) 10 $ 4 ð19bÞ ! "
Gibbs free energy density (per unit volume) g σ ij by the following
expression
D2 ¼ 2:225 ) 10 $ 6 ð19cÞ
! " 1 1 $1 1þγ γ
g σ ij ¼ σ ij : εij ¼ σ ij : Dijkl : σ kl ¼ σ : σ $ tr 2 σ ð20Þ
W ¼ 0:065 ð19dÞ 2 2 2E 2E
! 0 " where Eandγ are the Yong’s modulus and Poisson’s ratio, respec-
S ¼ 4:45994exp $ f c =11:51679 þ 1:95358 ð19eÞ $1
tively. Dijkl is the fourth-order linear-elastic compliance matrix
tensor of the intact (undamaged) material, defined as
The linear fitting results for parameter X0 is shown in Fig. 9a, ( )
$1 1 1 þ γ! "
and the fitting result for parameter S are shown in Fig. 9b, which Dijkl ¼ δik δjl þ δil δjk $ γδij δkl
E 2
appears to be rather consistent with the existing results listed in $ %
1 I*I 1 Idijkl Iijkl
Table 1. ¼ I4 $ þ I * I¼ þ ð21Þ
2G 3 K 2G K
3.4. Damage parameters where δij is “Kronecker delta”, the Bulk modulus K and Shear
modulus G can be expressed in terms of E; γ as K ¼ E=3ð1 $ 2γ Þ,
Damage is defined within the framework of CDM. In CDM, the G ¼ E=2ð1 þ γ Þ, Iijkl and Idijkl are the fourth-order identity tensor and
damage and damage energy release rate (DERR) are commonly deviatoric tensor, respectively. I ij is the second-order identity
considered as the thermodynamic conjugates [33–35]. A damage tensor. Eq. (20) represents an ellipsoidal damage shape in principal
surface is defined to determine whether damage loading occurs or undamaged stress space centered at the origin, as demonstrated in
not, and damage initiates and accumulates when the energy norm Fig. 10a for γ ¼ 0:2 (a typical value for concrete). It also represents
(a scalar measurement of strain energy), a function of DERR, a circular shape for γ ¼ 0.
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 9

unchanged, and the curve becomes a horizontal line correspond-

ing to an undamaged case when B þ ¼ 0.
It is known to all that the strain softening response will not be
E objective upon mesh refinement, and will be mesh-dependent
[36]. Objective can be achieved by modifying the constitutive law
and making it depend on mesh size by introducing a parameter
E called “crack band width” [37,38] or “characteristic length”[39].
Damage parameter A 7 ; B 7 , considering the characteristic length
E
[40,41], can be determined as follow:

O O
3.4.1. Brittle damage
The definition of the tension fracture energy GF (per unit area)
Fig. 11. Stress–strain curve for uniaxial tension. is written according to the fictitious crack model [40], which is
also used in the Mat 159 model to regulate mesh size dependence
1.0 1.0 [14]
Z 1 Z 1
! þ" 0
0.8 0.8 GF ¼ σ ðoÞdo ¼ 1 $ d f tdo ð26Þ
o0 o0

0.6 0.6 where σ ðoÞ and w are respectively the stress and displacement, w0
1−d +

0
is the displacement at peak tension strength f t.
0.4 0.4 Substituting brittle damage definition in Eq. (8) into Eq. (26),
gives
0.2 0.2 Z 1( þ )
r0 ! þ"
þ A þ exp $ B ðτ $ r0 Þ do
0 þ þ þ
GF ¼ f t þ 1$A ð27Þ
0.0 o0 τ
0.0
0.0 0.5 1.0 1.5
Energy norm τ
The integration on the first term
Fig. 12. Effects of brittle damage parameter A þ on strain softening curve. Z 1 þ
0 r0 ! þ" 0 ! "
f t þ 1 $A do ¼ f t 1 $ A þ o0 1 ð28Þ
o0 τ
By further decomposing the stress tensor into hydrostatic and
deviatoric parts as σ ¼ I 1 I =3 þ S, the Gibbs free energy density
! " ! " In order to obtain a reasonable value, A þ ¼ 1 is required in
g σ ij in Eq. (20) can
! "be recast into a deviatoric part g 1 J 2 and a Eq. (28), which means that elastic-brittle behavior with no
volumetric part g 2 I 1 as þ
residual strength in tension will appear because limτ þ -1 d -1
! " ! " 1þγ 1 $ 2γ 2 J I ! "
2
! " is obtained from Eq. (8).
g σ ij ¼ g I 1 ; J 2 ¼ J þ I ¼ 2 þ 1 ¼ g1 J2 þ g2 I1 When A þ ¼ 1, we can obtain
E 2 6E 1 2G 18K
l r 0þ
ð22Þ n n 0
l ft
! " GF ¼ þ ¼ þ pffiffiffi ð29Þ
where g 1 J 2 also presents an ellipsoidal shape in the principal B B E
stress space shown in Fig. 10b for γ ¼ 0:2. where l is a characteristic length of the finite element, and
n

The brittle damage energy norm τ þ is defined as typically set equal to the cube root of the element volume in
qffiffiffiffiffiffiffiffiffiffiffiffi
three-dimensional (3D) [7,40].
τ þ ¼ Eε2max ð23Þ
In the absence of experimental data for a particular concrete, GF
where εmax is the maximum principal strain. Substituting can be estimated from CEB–FIP code [19]
0
εmax ¼ f t =E in uniaxial tension into p Eq. (! ")0:7
ffiffiffi (23), the initial damage f ck þ 8
threshold r 0þ can be estimated as f t = E.
0
GF ¼ GF0 ð30Þ
10
The ductile damage energy norm τ $ is defined as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where f ck is the characteristic compressive strength of concrete,
2
pffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! "ffi J2 I1 and GF0 is base value of fracture energy (see Table 2), depends on
τ ¼ σ ij : εij ¼ 2g σ ij ¼
$
þ ð24Þ
G 9K the maximum aggregate size, dmax.
Then, ductile damage threshold r 0$ can be estimated from
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
"ffi pffiffiffi Table 2
0! 0 0
uniaxial compression from Eq. (24) as r 0$ ¼ f c f c =E ¼ f c = E. Base values of fracture energy GF0 [14,19].
If substituting Eq. (8) into Eq. (7) and applying the result to a
dmax (mm) GF0 (N mm/mm2)
uniaxial tension case, another expression of the stress can be
written as 8 0.025
h ! " pffiffi þ i 16 0.030
σ ¼ E ε0þ 1 $ A þ þ A þ ε þ expB Eðε0 $ ε Þ ¼ Eε
þ þ
ð25Þ 32 0.038

where ε is the effective strain (as shown in Fig. 11),

! þε0 is þthe
þ
" strain
corresponds to peak tension strength and ε ε Z ε0 is the
þ Note: a linear fit is used for other size of aggregate, the fit
strain in strain softening section. Two parameters A þ and B þ help equitation is GF0 ¼0.021þ 5.357 ) 10 $ 4dmax.
to regulate the curve shape after peak strength. From Eq. (29), B þ can be obtained by
Fig. 12 shows the effect of the brittle damage parameter B þ on n 0
the strain softening response. It is found that the curve becomes l ft
Bþ ¼ pffiffiffi ð31Þ
increasingly softening with the increase of B þ while A þ ¼ 1 is kept GF E
10 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

4.0 MAT 159

4.0 Experimental data
3.5 3.5 25 mm
MAT 145 50 mm
3.0 3.0
Experimental data 83 mm
Stress (MPa)

Stress (MPa)
2.5 25 mm 2.5
50 mm
2.0 83 mm 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Strain (1.0E-3) Strain (1.0E-3)
Fig. 13. Comparisons of simulation results with unixial tension: (a) MAT 145, and (b) MAT 159.

3.4.2. Ductile damage

The compression fracture energy Gc (per unit area) is written 5 MAT 145
by MAT 159

Peak tensile strength (MPa)

Z 1 Z 1
$ 0 4
Gc ¼ σ ðwÞdw ¼ ð1 $ d Þf cdo ð32Þ
o0 o0
3
Substituting ductile damage definition in Eq. (8) into Eq. (29),
gives 2
Z 1h $ i
0 r0 $ B $ ðτ $ $ r 0$ Þ
$ ð1 $ A Þ þ A exp
$ $
Gc ¼ f c do ð33Þ Application range
o0 τ 1
of the MAT 159

The integration on the first term 0

Z 1 $ 10 20 30 40 50 60 70
r0 $ 0 pffiffiffiffiffi! pffiffiffiffiffi"
0
f c $ ð1 $ A Þdo ¼ 2ð1 $ A Þf c x0 1 $ x0
$
ð34Þ Concrete grade (MPa)
o0 τ
Fig. 14. Comparisons of peak tensile strength predicted by the MAT 145 and
A $ ¼ 1 is required in Eq. (34) to obtain a reasonable value. MAT 159.
When A þ ¼ 1, we obtain the following expression of Gc
with a fluidity coefficient parameter, denoted by η, was employed.
qffiffiffiffiffiffiffiffiffi
22l
n
0 n pffiffiffiffiffiffi It is indeed a three-dimensional generalization of the Duvaut–
Gc ¼ $ 2 þ $ f cl o0 ð35Þ
B B Lions viscoplastic strain rate formulation [23].
Solving quadratic equation, we finally derive the expression of σ ij $ σ~ ij ¼ ηC ijkl ε_ vp
kl
ð37Þ
B $
where σ ij and σ~ ij are the viscid and inviscid stress tensor, C ijkl is the
pffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
0
f c = E þ 2Gc =l þ f c 2=E
n 0
fourth-order linear elastic tensor and ε_ vp kl
is the viscoplastic strain
B $
¼ ð36Þ rate tensor.
Gc =l
n

Using an implicit backward Euler algorithm for the viscoplastic

where GC Unfortunately, only limited research works on GC can be strain rate, the update for the stress tensor can be written as
found [42,43], and there is no GC provided in CEB–FIP code [19].
η η ! " η & trial '
In the absence of GC , B $ can be determined by comparing the σ n þ 1 $ σ~ n þ 1 ¼ CΔεvp ¼ C Δε $ Δεe ¼ σn þ 1 $ σn þ 1 ð38Þ
Δt Δt Δt
predicted uniaxial compression curve with the test data. Besides,
considering that several constitutive models (i.e. the MAT 14, MAT where the elastic trial stress σ trial n þ 1 ¼ σ n þCΔε, and stress tensor
84 in LS-DYNA) with elastoplatic response for uniaxial compres- σ n þ 1 ¼ σ n þ CΔεe . Readily it is seen that the viscid stress tensor can
sion [44–46] perform well for predicting reinforced concrete (RC) be expressed as
structures subject to impact loading, B $ ¼ 0 can be simply adopted
σ trial
nþ1 þσ ~ n þ 1 Δt=η
for the MAT 145, which recovers elasto-plastic behavior for σn þ 1 ¼ ð39Þ
1 þ Δt=η
concrete under uniaxial compression. For ! 0 convenience, the sug-
0"
gested range
! 0 0"of B $
can be 0 rB $
rB þ
= f c =f t , where the upper σ n þ 1 ¼ σ~ n þ 1 is obtained from the above equation by seeting η ¼ 0,
limit B þ = f c =f t is an empirical parameter. which corresponds to the rate-independent situation. σ n þ 1 ¼ σ trial
nþ1
is obtained by setting η ¼ 1, which corresponds to the elastic
3.5. Strain rate effect parameters situation.
Regarding the strain rate effect for concrete, it has been argued
Existing experimental data shows that the peak strength that the 3D FE model is capable of calculating the strength
attained during the direct pull and unconfined compression test increase due to the confining stresses generated by the inertia of
is sensitive to the strain rate, which is also specified by the CEB– the structure, and there is a risk of overestimating the strength if
FIP code [19]. To account for the effect, a viscoplastic formulation FE calculates the failure of concrete by using both strain rate
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 11

Table 3
Material input card of the MAT 145 for the element size of 83 mm.

Table 4
Material input card of the MAT 159 for the element size of 83 mm.

dependent strength increase factors and confining stresses due to simulations, a single cubic element model with three different size
the inertia [47–50]. In consideration of this, η ¼ 0 is suggested in (25 mm, 50 mm and 83 mm) was used to show the element size
this study. dependent of the stress–strain curve, where C44 concrete with a
maximum aggregate size of 10 mm was used to generate material
parameters for the MAT 145 and MAT 159. Tables 3 and 4 show the
4. Numerical validations input material parameters of the MAT 145 and MAT 159 for the
element size of 83 mm.
To validate the applicability of the proposed material para- Evidently, the model predictions by the MAT 145 based on
meters for the MAT 145 to concrete in tension and compression mesh size of 83 mm (the same as the test length) match well with
under different confining pressures, experimental strain–stress the pre-peak test data and show generally acceptable agreements
data from existing literatures for the quasi-static case is used. for the post-peak response, whilst the predictions using the MAT
Furthermore, falling weight impact test on RC beams is used to 159 show appreciable discrepancies for both pre-peak and post-
check the dynamic performance of the MAT 145 with the proposed peak responses no matter what mesh size is used.
parameters. The parameters of the MAT 145 are generated based Fig. 14 further shows a comparison of the peak tensile strength
on the calibration equations summarized in Appendix. In order to for different grades of concrete predicted by the MAT 145 and MAT
compare with a similar model-MAT 159, validations of the MAT 159. The predictions by the MAT 145 are apparently higher than
159 with default parameters (internal generation parameter based those by the MAT 159. Not presented here, a further prediction of
on uniaxial compression strength, the maximum aggregate size the peak tensile strength by the MAT 72 show nearly identical
and units) is also carried out. values with those by the MAT 145 for different grade of concrete.
In the numerical simulation, the compression or tension are Besides, the predicted peak tensile strength by the MAT 159 shows
applied to the top surface of the element via nBoundary Prescribed very small value when f c r 20 MPa, and gives decreasing value
0

Motion keywords with a constant velocity of compression or tension, 0

when f c Z 50 MPa because the default material parameters gen-
while the bottom surface is constrained on the motion direction. For erated by the MAT 159 is for concrete about 20 MPa r
triaxial compression, the confining pressure is applied to the side and f c r58 MPa, with emphasis on the midrange between 28 and
0

ends of the element via nLoad Segment keywords [3]. 48 MPa [14]. Hence the MAT 159 may be not suitable for high
0
strength concrete with f c Z 48 MPa. In comparison, the predicted
4.1. Uniaxial tensile test peak tensile strength by the MAT 145 based on the proposed
material parameter performs well for all grades of concrete.
Fig. 13 shows a comparison of the predicted stress–strain relation
curves by the MAT 145 and MAT 159 against data of uniaxial tension
tests on concrete reported by Geopalaeratnam and Shah [51]. The test 4.2. Uniaxial and triaxial compression tests
material parameters included 28 days compression strength
0 0
f c ¼ 43:88 MPa ¼ 6364 psi, tension strength f t ¼ 3:62 MPa ¼ Fig. 15 presents a comparison of the model predictions with
525 psi, fracture energy GF ¼ 0:00564 N=mm ¼ 0:322 lb=in: and uniaxial and triaxial compression test data reported by Green
stress–strain curve measured in 83 mm length. In the following and Swanson [21], where the compression tests for C48.4
12 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

100 100
MAT 159
80 80

Stress Difference (MPa)

Stress Difference (MPa)
60 60

40 40

20 20

0 0
0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6
Strain (1.0E-2) Strain (1.0E-2)
Fig. 15. Comparisons of simulation results with triaxial compression.

Table 5
Material input card of the MAT 145.

70 used to generate the material parameters for the MAT 145 and
MAT 159. Ductile damage parameters A $ ¼1.0, B $ ¼0.0 was used
Peak compressive strength (MPa)

60 MAT 145 in the MAT 145 (Table 5), and the suggested range of B $ is
MAT 159 0 r B $ r 0:09. The predicted stress is the average value in the
50 mid-section of the specimen shown in Fig. 15a, and the strain is
calculated from the total displacement in top section of the
specimen. It is evident that the predicted results by the MAT 145
40
compare well with test data in the case of high confining pressure,
whilst the predictions by the MAT 159 shows large difference from
30
the test data in all three cases.
Fig. 16 shows a comparison of peak compressive strength for
20 different concrete (C10 to C70) produced by the MAT 145 and MAT
159. The results given by the MAT 145 agree well with the
10 theoretical value for all those concrete, whilst the results predicted
10 20 30 40 50 60 70 by the MAT 159 are slightly smaller than the theoretical value for
Concrete grade (MPa) most concrete types though C20 concrete lies in its application
Fig. 16. Comparisons of peak compressive strength produced by the MAT 145 range.
and MAT 159.
4.3. Hydrostatic compression tests
0
(f c ¼ 48:4 MPa ¼ 7:02 ksi) concrete were conducted on 68.6 mm
(2.7 in.) diameter by 152.4 mm(6.0 in.) long cylinders at three Fig. 17 presents the comparisons of experimental and analytical
confining pressure of 0, 6.8959 MPa (1 ksi) and 13.79 MPa (2 ksi), results of hydrostatic load–unload tests. In the hydrostatic com-
respectively. The maximum aggregate size was 76.2 mm (3/8 in.), pression test, conducted by Green and Swanson [21], the size of
and the elastic modulus E was about 41368.6 MPa (6000 ksi). The specimen and concrete mix are the same as those introduced in
simulations adopted a multi-element structural model with the Section 4.2, and the measurable permanent compaction began
same size of the test specimens, which include 2139 hexahedral about 41.4 MPa (6 ksi) pressure. The simulated diagrams of the
elements with a mesh size of 8 mm (Fig. 15a). C48.4 concrete was MAT 145 compares well with the experimental data especially in
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 13

500
Experimental data 700 Experimental Data
MAT 145
Hydrostatic pressure p (MPa)

MAT 159

Hydrostatic pressure p (MPa)

400 600

500
300
400

200 300

200
100
100

0 0
0 2 4 6 8 10 0 2 4 6 8 10
Volume Strain εv % Volume Strain ε v %
Fig. 17. Comparison of experimental and simulated hydrostatic stress–strain response: (a) MAT 145 and (b) MAT 159.

Table 6
Material input card of the MAT 145.

loading test, whilst the predictions by the MAT 159 differs a all impact cases, while those predicted by the MAT 159 tend to
lot from the experimental data in both loading and unloading overestimate the peak force and have short impact duration.
tests. Focusing on mid-span displacement, the results predicted by both
models agree well with the test curve before peak displacement,
and the residual displacements from both models are slightly
4.4. Reinforced concrete beams subject to impact loads larger than the test value.
Fig. 19 presents the comparisons of the predicted damage by
Fujikake [52,53] conducted tests on RC beams subject to the two models and experimental results, wherein the contours of the
failing weight impact which will be used in this study to validate damage range from 0 (blue) to 1 (red). In the numerical simula-
the performance of the Mat 145 for concrete under impact loads. tion, both vertical cracks and diagonal shear cracks caused by
In the test, a rigid hammer with a mass of 400 kg was dropped overall failure of RC beams are predicted reasonably well by both
freely onto the top surface of RC beams at mid-span from four models, and the damage predicted by the MAT 145 appears to be
different heights (0.15 m, 0.30 m, 0.60 m, and 1.20 m). The uniaxial slightly larger than that by the MAT 159.
compressive strength of concrete was 42 MPa in the test, and the The above comparisons indicate that the MAT 145 performs
estimated tensile strength of concrete according to Japan Society better in predicting the impact force, while the MAT 159 may be
of Civil Engineers code was 2.8 MPa [52]. better in predicting mid-span displacement and damage distribu-
In the FE model, the concrete was meshed with 57852 eight-noded tion of the RC beam with moderate damage.
solid elements with size of 12.5 mm. C42 concrete with the maximum
aggregate size of 10 mm was used to generate material parameters for
the MAT 145 and MAT 159. The material input card of the MAT 145 is 5. Conclusions
shown in Table 6, where ductile damage parameter B $ ¼ 0:1 with the
estimated range 0 r B $ r0:164 is used. The MAT 145 in LS-DYNA is an advanced constitutive model for
Fig. 18 shows comparisons of predicted and testing impact geomaterials with sounded theoretical background, but contains
force and mid-span displacement for the four impact cases. For too many parameters which limit its wide applicability. To facil-
impact force history, the numerical results predicted by the MAT itate this model for convenient use by more users, a detailed
145 agree rather well with the test data in terms of peak force in procedure of calibration of this model for concrete was presented
14 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

300 30
Case 1 (0.15m height impact) Case 1 (0.15m height impact)

Midspan displacement ( mm)

25
Experiment data Experiment data
MAT 145
Impact force (KN)

200 MAT 145 20

MAT 159 MAT 159
15

100 10

0 0
0.000 0.005 0.010 0.015 0.020 0.000 0.005 0.010 0.015 0.020
Time (s) Time (s)

300 30
Case 2 (0.30m height impact)
Case 2 (0.30m height impact)

Midspan displacement ( mm)

Experiment data 25 Experiment data
MAT 145 MAT 145
Impact force (KN)

200 MAT 159 20 MAT 159

15

100 10

0 0
0.000 0.005 0.010 0.015 0.020 0.025 0.000 0.005 0.010 0.015 0.020 0.025
Time (s) Time (s)

400 50
Case 3 (0.60m height impact)
Case 3 (0.60m height impact)
Midspan displacement( mm)

40 Experiment data
Experiment data
300 MAT 145
Impact force (KN)

MAT 145
MAT 159
MAT 159 30
200
20

100
10

0 0
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.000 0.005 0.010 0.015 0.020 0.025 0.030
Time (s) Time (s)

400
Case 4 (1.20m height impact) 50
Case 4 (1.20m height impact)
Experiment data
Midspan displacement( mm)

300 40
Impact force (KN)

MAT 145
MAT 159
30
200
20
Experiment data
MAT 145
100 MAT 159
10

0 0
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.000 0.005 0.010 0.015 0.020 0.025 0.030
Time (s) Time (s)
Fig. 18. Comparisons of analytical and test results for four impact cases.
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 15

Fig. 19. Damage distribution of four RC beams after the impact test: (a) experiment, (b) MAT 145, (c) MAT 159. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)

in this study. The material parameters other than those related to subjected to the failing weight impact using the MAT 145 and MAT
the damage were obtained as functions of the uniaxial compres- 159 were carried out. The results show that the MAT 145 with
sion strength of concrete based on the CEB–FIP code and empirical proposed parameters performs better than the MAT 159 for
formulas reported in the literature. The damage parameters are predicting the impact force, and the difference between two
determined with consideration of a characteristic length of the predicted mid-span displacements in each case is negligible.
finite element. The proposed formulae and relationships greatly
expedite the engineering application of the model.
The proposed method was further verified by a single element
and a multi-element simulation by the MAT 145 which offered
reasonably good predictions for the stress–strain relationship as Acknowledgements
compared to experimental results. Compared with the MAT 159,
the MAT 145 with proposed material parameters shows nearly The first author is grateful for the support from the National
0
identical peak tensile strength f t as the MAT 72, and can be used Science Foundation of China (Grant 51308054). The opinions,
for
! 0 different "grade of concrete including high strength concrete findings, and conclusions do not reflect the views of the funding
f c Z 48 MPa . Furthermore, numerical simulations of RC beams institutions or other individuals.
16 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

Appendix. Input parameter card for the MAT 145

Note: Parameters marked in rectangle are important parameters calibrated in this paper. Parameters marked in ellipse are ignored
parameters with default value. The remaining parameters are the output opinions for users.
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 17

Associated calibration equations of the MAT 145

RO (mass density) [19]

RO ¼ 2:4 ) 10 $ 9 (Card 1-2)
SHEAR (shear modulus) [19]
& 0 '1=3 (Card 1-3)
fc
G ¼ 2ð1Eþ γ Þ ¼ 9137:5
1þγ 10 MPa
BULK (bulk modulus) [19]
& 0 '1=3 (Card 1-4)
E 6092 fc
K ¼ 3ð1 $ 2γ Þ ¼ 1 $ 2γ 10 MPa
GRUN (Gruneisen ratio) [3]
GRUN¼ 0 (Card 1-5)
SHOCK (Shock velocity) [3]
GRUN¼ 0 (Card 1-6)
PORE (flag for pore collapse for constant bulk modulus)
PORE ¼1 (Card 1-7)
ALPHA (shear failure parameter)
! 0 "
α ¼ 13:9316exp f c =68:7383 $ 13:8380 (Card 2-1)
THETA (shear failure parameter)
θ ¼ 0:3533 $ 3:3294 ) 10 $ 4 f c $3:8182 ) 10 $ 6 f c 2
0 0 (Card 2-2)
GAMMA (shear failure parameter)
! 0 "
γ ¼ 3:6657exp f c =39:9363 $ 4:7092 (Card 2-3)
BETA (shear failure parameter)
β ¼ 18:17791f 0 c
$ 1:7163 (Card 2-4)
EFIT (dilitation damage mechanics parameter) [3]
EFIT¼ 1 (no dilatation damage) (Card 2-5)
FFIT (dilitation damage mechanics parameter) [3]
FFIT¼ 0 (no dilatation damage) (Card 2-6)
ALPHAN (kinematic strain hardening parameter)
ALPHAN ¼0 (no hardening) (Card 2-7)
CALPHAN (kinematic strain hardening parameter)
CALPHAN ¼0 (no hardening) (Card 2-8)
R0 (initial cap surface ellipticity)
! 0 "
S ¼ 4:45994exp $ f c =11:51679 þ 1:95358 (Card 3-1)
X0 (initial cap surface J1 axis intercept)
X 0 ¼ 17:087 þ 1:892f c
0
(Card 3-2)
IROCK [3]
IROCK ¼0 (soils-cap can contract) (Card 3-3)
IROCK ¼1(rock/concrete-cap cannot contract) (Card 3-3)
SECP (shear enhanced compaction)
SECP ¼0 (no consideration of shear enhanced compaction) (Card 3-4)
AFIT (ductile damage mechanics parameter)
A$ ¼ 1 (Card 3-5)
BFIT (ductile damage mechanics parameter)
pffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f 0c = E þ 2Gc =ln þ f 0c 2=E (Card 3-6)
B$ ¼ Gc =l
n
! 0 0"
Or 0 rB r B = f c =f t
$ þ (Card 3-6)
RDAM0 (ductile damage mechanics parameter)
f
0
f
0
r 0$ ¼ pcffiffiE ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c (Card 3-7)
0 1=3
18275ðf c =10Þ
W (plastic volume strain parameter)
W ¼ 0:065 (Card 4-1)
D1 (plastic volume strain parameter)
D1 ¼ 6:11 ) 10 $ 4 (Card 4-2)
D2 (plastic volume strain parameter)
D2 ¼ 2:225 ) 10 $ 6 (Card 4-3)
CFIT (brittle damage mechanics parameter)
Aþ ¼ 1 (Card 4-6)
DFIT (brittle damage mechanics parameter)
l f
B þ ¼ G pt ffiffiE
n 0
(Card 4-7)
F

TFAIL (tensile failure stress)

0 ! 0 "2=3 (Card 4-8)
f ¼ f t ¼ 1:4 f c =10 MPa
DBETA(rounded vertices parameter)
18 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19

DBETA ¼0 (Card 5-2)

DDELTA (rounded vertices parameter)
DDELTA ¼0 (Card 5-3)
VPTAU (viscoplasticity relaxation time parameter, η)
η ¼0 (Card 5-4)
ALPHA1 (torsion scaling parameter)
α1 ¼ 0:82 (Card 6-1)
THETA1 (torsion scaling parameter)
θ1 ¼ 0 (Card 6-2)
GAMMA1 (torsion scaling parameter)
γ 1 ¼ 0:2407 (Card 6-3)
BETA1 (torsion scaling parameter)
β1 ¼ 0:33565f 0 c
$ 0:95383 (Card 6-4)
ALPHA2 (tri-axial extension scaling parameter)
α2 ¼ 0:76 (Card 6-5)
THETA2 (tri-axial extension scaling parameter)
θ2 ¼ 0 (Card 6-6)
γ 2 ¼ 0:26 (Card 6-7)
BETA2 (tri-axial extension scaling parameter)
β2 ¼ 0:285f 0 c
$ 0:94843 (Card 6-8)

References [23] L.E. Schwer, Viscoplastic augmentation of the smooth cap model, Nucl. Eng.
Des. 150 (1994) 215–223.
[24] Y.D. Murray, B.A. Lewis, Numerical Simulation of Damage in Concrete.
[1] K. Yonten, T.M. Majid, D. Marzougui, A. Eskandarian., An assessment of Technical Report Submitted to the Defense Nuclear Agency by APTEK,
constitutive models of concrete in the crashworthiness simulation of road- Contact No.DNA001-94-C-0075, DNA-TR-95-190, 1995.
side safety structures, Int. J. Crashworthiness 10 (1) (2005) 5–19. [25] L.E. Schwer, Demonstration of the Continuous Surface Cap Model With
[2] Z. Tu, Y. Lu, Evaluation of typical concrete material models used in Damage: Concrete Unconfined Compression Test Calibration. LS-DYNA Geo-
hysdrocodes for high dynamic response simulations, Int. J. Impact Eng. 36 material Modeling Short Course Notes, 2001.
(2009) 132–146. [26] L.E. Schwer, Y.D. Murray, Continuous surface cap model for geomaterial
[3] LS-DYNA Keyword User’s Manual Volume II Material Models. LS-DYNA R7.1, modeling a new LS-DYNA material type, in: Proceedings of Seventh inter-
Livermore Software Technology Corporation (LSCT), 2014. national LS-DYNA User Conference, 2002.
[4] ANSYS Autodyn User’s Manual, Release 15.0. ANSYS, Inc., 2013. [27] M.B. Rubin, Simple, convenient, isotropic failure surface, J. Eng. Mech. 117 (2)
[5] SIMULIA.Abaqus 6.13 Analysis User’s Manual. Providence, Rhode Island: (1991) 348–569.
Dassault Systèmes, 2013. [28] J.P. Bardet, Lode dependences for isotropic pressure sensitive elastoplastic
[6] H. Kupfer, H.K. Hilsdorf, H. Rusch, Behavior of Concrete Under Biaxial materials, J. Appl. Mech., ASME 57 (3) (1990) 498–506.
Stresses, ACI J. 66 (8) (1969) 545–666. [29] I.S. Sandler, F.L. Dimaggio, G.Y. Baladi, Generalized Cap model for geological
[7] L.J. Malvar, et al., A plasticity concrete material model for dyna3d, Int. J. materials, J. Geotech. Eng. Div., ASCE (1976) 638–699.
Impact Eng. 19 (9-10) (1997) 847–873. [30] C.D. Foster, R.A. Regueiro, A.F. Fossum, R.I. Borja, Implicit numerical integra-
[8] B.J. Broadhouse, The Winfrith Concrete Model in LS-DYNA (Report: SPD/D tion of a three-invariant, isotropic/kinematic hardening cap plasticity model
(95)363), Structural Performance Department, AEA Technology, Winfrith for geomaterials, Comput. Methods Appl. Mech. Eng. 194 (2005) 5109–5138.
Technology Centre, UK,, 1995. [31] J.C. Simo, J.W. Ju, Strain- and stress-based continuum damage models—I:
[9] T.J. Holmquist, G.R. Johnson, W.H. Cook, A computational constitutive model Formulation, Int. J. Solids Struct. 23 (7) (1987) 821–840.
for concrete subjected to large strains, high strain rates and high pressures, [32] J.W. Ju, On energy-based coupled elastoplastic damage theories: constitutive
in: Proceedings of 14th International Symposium on Ballistics, Quebec, 1993, modeling and computational aspects, Int. J. Solids Struct. 25 (7) (1989)
pp. 591–600. 803–833.
[10] W. Riedel, K. Thoma, S. Hiermaier, Penetration of reinforced concrete by [33] J. Lemaitre, A continuous damage mechanics model for ductile fracture, J.
BETA-B-500 dnumerical analysis using a new macroscopic concrete model Eng. Mater. Technol. ASME 107 (1985) 83–89.
for hydroc-odes, in: Proceedings of 9th International Symposiumon Interac- [34] W.Y. Zhou, J.D. Zhao, Y.G. Liu, Q. Yang, Simulation of localization failure by
tion of the Effect of Munitions with Structures. Berlin—Strausberg, Germany, strain-gradient-enhanced damage mechanics, Int. J. Numer. Anal. Methods
1999, pp. 315–22. Geomech. 26 (8) (2002) 793–813.
[11] J. Lubliner, J. Oliver, S. Oller, E. Onate, A plastic-damage model for concrete, [35] J. Mazars, A description of micro and macroscale damage of concrete
Int. J. Solids Struct. 25 (3) (1989) 299–326. structures, Int. J. Fract. 25 (5/6) (1986) 729–737.
[12] W.F. Chen, Plasticity in Reinforced Concrete, McGraw-Hill Book Company, [36] Z.P. Bažant, Crack band theory for fracture of concrete, Mater. Struct. 16
1982. (1983) 155–177.
[13] W.F. Chen, D.J. Han, Plasticity for Structural Engineers, Springer-Verlag, New [37] Z.P. Bažant (Edited), Fracture mechanics of concrete structures, in: Proceed-
York, USA, 1988. ings of the First International Conference on Fracture Mechanics of Concrete
[14] Y.D. Murray, User’s Manual for LS-DYNA Concrete Material Model 159, Report Structures, Breckenridge,Colorado, USA, 1992.
No. FHWA-HRT-05-062, Federal Highway Administration, 2007. [38] J. Oliver, A consistent characteristic length for smeared cracking models, Int.
[15] Y.D. Murray, A. Abu-Odeh, R. Bligh, Evaluation of Concrete Material Model J. Numer. Methods Eng. 28 (1989) 461–474.
159, Report No. FHWA-HRT-05-063, Federal Highway Administration, 2007. [39] A. Hillerborg, M. Modéer, P.E. Petersson, Analysis of crack formation and
[16] L.J. Malvar, J.E. Crawford, K.B. Morill, 2000. K&C Concrete Material Model crack growth in concrete by means of fracture mechanics and finite
Release III: Automated Generation of Material Model Input, Karagozian & elements, Cem. Concr. Res. 6 (1976) 773–782.
Case, 2000TR-99-24-B1. [40] J. Oliver, M. Cervera, S. Oller, J. Lubliner, Isotropic damage models and
[17] L. Schwer, L.J. Malvar, Simplified Concrete Modeling with Mat_concrete_da- smeared crack analysis of concrete, in: Second International Conference on
mage_rel3, JRI LS-Dyna User Week, August 2005. Computer Aided Analysis and Design of Concrete Structures, 1990, pp. 945–
[18] N. Markovich, E. Kochavi, G. Ben-Dor, An improved calibration of the 948.
concrete damage model, Finite Elem. Anal. Des. 47 (2011) 1280–1290. [41] M. Cervera, M. Chiumenti, R. Codina, Mesh objective modeling of cracks
[19] CEB-FIP Model Code 1990, Design Code. Lausanne: Thomas Telford, 1993. using continuous linear strain and displacement interpolations, Int. J. Numer.
[20] L.L. Mills, R.M. Zimmerman, Compressive strength of plain concrete under Methods Eng. 87 (2011) 962–987.
multiaxial loading Conditions, ACI J. Proc. (1970) 802–807. [42] R.A. Vonk, Softening of Concrete Loaded in Compression (Ph.D. Dissertation),
[21] S.J. Green, S.R. Swanson, Static constitutive relations for concrete. Air Force Delft University of Technology, The Netherlands, 1992.
Weapons Laboratory (Technique Report No. AFWL-TR-72-2), Kirtland Air [43] D.C. Jansen, S.P. Shah, Effect of length on compressive strain softening of
Force Bace, 1973. concrete, ASCE J. Eng. Mech. 123 (1) (1997) 25–35.
[22] L.E. Schwer, Y.D. Murray, A three-invariant smooth cap model with mixed [44] A.Q. Bhatti, N. Kishi, Impact response of RC rock-shed girder with sand
hardening, Int. J. Numer. Anal. Methods Geomech. 18 (10) (1994) 657–688. cushion under falling load, Nucl. Eng. Des. 240 (2010) 2626–2632.
 H. Jiang, J. Zhao / Finite Elements in Analysis and Design 97 (2015) 1–19 19

[45] N. Kishi, S.G. Khasragh, H. Kon-No, Numerical simulation of reinforced [49] D.M. Cotsovos, M.N. Pavlovic, Numerical investigation of concrete subjected to
concrete beams under consecutive impact loading, ACI Struct. J. 108 (2011) high rates of uniaxial tensile loading, Int. J. Impact Eng. 35 (5) (2008) 319–335.
444–452. [50] S. Saatci, F.J. Vecchio, Nonlinear finite element modeling of reinforced concrete
[46] D.-K. Thai, S.-E. Kim, Failure analysis of reinforced concrete walls under impact structures under impact loads, ACI Struct. J. 106 (5) (2009) 717–725.
loading using the finite element approach, Eng. Fail. Anal. 45 (2014) 252–277. [51] V.S. Geopalaeratnam, S.P. Shah, Softening response of plain concrete in direct
[47] Q.M. Li, H. Meng, About the dynamic strength enhancement of concrete-like tension, ACI J. 85 (3) (1985) 310–323.
materials in a split Hopkinson pressure bar test, Int. J. Solids Struct. 40 (2) [52] K. Fujikake, B. Li, Impact response of reinforced concrete beam and its
(2003) 343–360. analytical evaluation, J. Struct. Eng. ASCE 135 (8) (2009) 938–950.
[48] D.M. Cotsovos, M.N. Pavlovic, Numerical investigation of concrete subjected to [53] H. Jiang, X.W. Wang, S.H. He, Numerical simulation of impact tests on
compressive impact loading—Part 1: A fundamental explanation for the reinforced concrete beams, Mater. Des. 39 (2012) 111–120.
apparent strength gain at high loading rates, Comput. Struct. 86 (1-2)
(2008) 145–163.