Journal of Building Engineering 44 (2021) 103239

Contents lists available at ScienceDirect

Journal of Building Engineering

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

A concrete damage plasticity model for predicting the effects of

compressive high-strength concrete under static and dynamic loads
Hoang- Le Minh a, Samir Khatir b, Magd Abdel Wahab c, d, *, Thanh Cuong-Le a, **
a
Faculty of Civil Engineering, Ho Chi Minh City Open University, Ho Chi Minh City, VietNam
b
Department of Electrical Energy, Metals, Mechanical Constructions, and Systems, Faculty of Engineering and Architecture, Ghent University, 9000, Gent, Belgium
c
CIRTech Institute, Ho Chi Minh City University of Technology (HUTECH), Ho Chi Minh City, Viet Nam
d
Soete Laboratory, Faculty of Engineering and Architecture, Ghent University, Technologiepark Zwijnaarde 903, B-9052, Zwijnaarde, Belgium

A R T I C L E I N F O A B S T R A C T

Keywords: In this paper, a concrete damage plasticity model is presented for simulating the behaviour of high-strength
Nonlinear dynamic analysis concrete; C60, C80 and C110 under static and dynamic loading conditions. This model is based on the
Reinforced concrete ground-breaking studies recorded in recently completed studies to improve and develop the original concrete
Finite element method
damage plasticity (CDP) model in ABAQUS software. In this CDP model, the stress-strain curves in compressive
3D detailed modelling
Damage factor
behavior and tensile behavior during the softening phase (after stress reaches peak strength) consider the effect
Concrete damaged plasticity of mesh size in Finite Element model. Through recently published results, these curves show many advantages
Dynamic impact factor compared to the previous studies. At the same time, the tensile damage variable (dt ) and the compressive damage
variable (dc ) are presented using an exponential function to replace the values reported in previous researches.
Especially, in the case of dynamic problems, the Dynamic Increase Factor (DIF) according to the fib MODEL
CODE 2010 (MC2010) is used in this study as a significant parameter to determine the effects of strain rate on
compressive strength of high-strength concrete. To prove the reliability and effectiveness of the proposed model,
numerical simulation of the static compressive tests is implemented to verify its suitability in case of static
loading problems. Then, Split-Hopkinson pressure bar (SHPB) test is simulated to determine the level of accuracy
of DIF in simulating the dynamic loading problems. The results show that the CDP model introduced in this study
has an acceptable agreement with the experimental results for both cases; static and dynamic loading conditions,
with high reliability.

1. Introduction stress-strain relationship, yield surface, the damage variants in

compression (dc ) and tension (dt ), which register the changes of slope in
Nowadays, the dynamic loads become the most significant factors in the comparison between elastic phase and damaged phase. Because of
concrete structures because of their unpredictability and powerful the complicated nonlinear behavior, it is difficult to select a suitable
destructiveness. The behavior of concrete under dynamic loads shows a model of concrete using numerical simulations. In the related works,
critical difference in comparison with static loads due to the effects of concrete damage plasticity (CDP) model is widely used and is recognized
the strain rate [1]. The compressive strength in case of dynamic loading with high-reliability in the analysis of concrete behavior. Many pro­
was reported to be larger than that of static loading due to inelastic posed models related to the nonlinear behavior of concrete were
deformation delaying effects in concrete. Therefore, it is particularly developed and improved from the original CDP [2–7]. An integral dy­
important to find a suitable concrete model for simulating the perfor­ namic strength criterion for concrete was proposed [8]. This model ac­
mances of concrete structures under dynamic loads. At the same time, counts for the short duration stress history in the range idealized as
this model should also achieve high reliability in case of static loads. In elastic based on concept of critical value of generalized stress impulse,
general, the significant behavior of concrete is determined through the proposed by Campbell for steel. Many scholars have conducted in-depth

* Corresponding author. CIRTech Institute, Ho Chi Minh City University of Technology (HUTECH), Ho Chi Minh City, Viet Nam.
** Corresponding author. Faculty of Civil Engineering, Ho Chi Minh City Open University, Ho Chi Minh City, VietNam.
E-mail addresses: [email protected] (S. Khatir), [email protected], [email protected] (M. Abdel Wahab), [email protected]
(T. Cuong-Le).

https://doi.org/10.1016/j.jobe.2021.103239
Received 13 March 2021; Received in revised form 4 July 2021; Accepted 1 September 2021
Available online 4 September 2021
2352-7102/© 2021 Published by Elsevier Ltd.
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

investigations on the changes of strength of concrete in the softening and Table 1

hardening phases to describe and establish the damage variables dc , dt . Average compressive characteristics of concrete C60, C80, C110 for static
Besides, the combination of theories of plasticity and the investigation of loading tests.
the effects of stress components in establishing the plastic surface is also Grade of E (GPa) ν Average compressive strength (MPa)
a primary research topic. Ref. [9] reported a new theoretical concept of a concrete
Φ100 × 200 100 mm Φ75 × 150
formulation for anisotropic tensile damage model based on developing mm Cylinder cubes mm
and modifying a “generalized pseudo-Rankine” model. In this model, a Cylinder
new closed-form solution was presented for loading cases of pure C60 37.1 0.21 68.1 (±3.9)a 70.5 69.2 (±0.4)
distortion, pure shear, and uniaxial tension after loading and unloading (±0.4) (±0.01) (±0.4)a a

phases in perpendicular direction. The models, including the develop­ C80 37.1 0.20 74.0 (±4.3) a
85.0 77.5 (±2.6)a
ment of some nonlinear formulations of stress-strain relationship in the (±0.4) (±0.002) (±5.3)a
C110 40.7 0.19 110.4 115.1 107.9
effective stress space, can be found in Refs. [10–12]. Ref. [13] intro­ (±1.1) (±0.02) (±7.4)a (±3.0)a (±2.2)a
duced a damage plasticity model related to the formulation of plasticity
a
in stress space. The model made use of a new yield criterion accounting Average compressive strength is obtained from three tests with the tolerance
in parentheses.
for the effects of both elastic and plastic stiffness degradation. The ac­
curacy of this model was validated using some tests and the results
obtained agreed with the experimental data quite well. At the same time, dynamic compressive behavior of high-strength concrete with three
Ref. [13] also presented a new approach based on the theory presented different strengths: 60 MPa, 80 MPa and 110 MPa, with labels; C60, C80,
in Ref. [14] for adjusting the concrete damage plasticity model under the C100, respectively. SHPB technique was used to register the effects of
effects of confinement having a uniform and non-uniform conditions. strain rate on compressive strength. Especially, the dynamic increase
Ref. [10] proposed the new formulation of damage elastic-plastic model factor (DIF) recorded the increase of the concrete strength, depending on
to overcome the limitations of pure elastic damage models or pure the strain rate. DIF is an important factor in building a concrete damaged
elastic-plastic ones, which do not satisfy the un-loading phase. Three plasticity model in cases of dynamic loading. However, up to now, there
tests were conducted including the following features; triaxial loading, is no study that has evaluated the best DIF curve in the related works.
tension, and cyclic compression to check the accuracy of this model. The The value of DIF still depends on the specific problems [2,20,27–29]. In
results showed that the proposed model had a good agreement with the this paper, the DIF curve is employed from MC2010 to establish the
data obtained from tests. Ref. [15] reported a new approach to compute stress-strain curves in compression and tension of high-strength concrete
the damage variables (dc ) and (dt ) accounting for mesh elements. The in case of dynamic loading. The theory of these curves is based on the
analysis of concrete behavior is usually investigated under dynamic and original CDP model. Thus, the stress-strain curve in the softening phase
static loadings. The previous studies have shown that the mechanical in the original CDP is replaced by a new curve according to Ref. [26]. To
failure and damage features of concrete in the two cases were different describe in detail the damage of each concrete element in the numerical
[2,16]. In the dynamic loading case, finding suitable model with the model, the original CDP introduces two variables; tensile damage vari­
dynamic behavior and depends on the strain rate is attractive [17–20]. able (dt ) and compressive damage variable (dc ). In previous studies,
Many related studies have been conducted in the literature. Ref. [2] these variables have normally been proposed with values in the range
presented the stress-strain relationship of concrete in compression be­ [0, 1]. In this study, the variables dt and dc are proposed according to
tween a low loading rate of around 10 micro-strains per second and a Ref. [15]. For the first time, the combination between DIF according to
higher loading rate between 5 and 10 strains per second. Ref. [21] used MC2010 and the damaged variables (dt dc ) calculated according to
an elastic-viscoplastic model to analyze the behavior of concrete under Ref. [15] together with CDP model revised in softening phase is used to
dynamic loading as a first step, then the finite element method was establish a new damaged concrete model. This model is used to simulate
employed to solve the failure localization problem. In the final step, the the experiments in Ref. [26] using 3D Finite Element software (ABA­
comparison of the proposed model and experimental data obtained from QUS). The obtained results are compared with the data from measure­
a real experimental reinforced concrete beam was presented. The nu­ ments to illustrate the effectiveness of the proposed model. Through the
merical results showed that an elastic-viscoplastic model performed well comparison, the model presented in this paper provides a highly reliable
with experimentally measured forces and deflections and it seemed that numerical solution and can be used as a potential approach to analyze
the concrete behavior would have to be considered as viscoelas­ the behavior of high-strength concrete in both static and dynamic cases.
tic–viscoplastic in order to improve the model prediction. Ref. [22] used In the next section in this paper, the high-strength concrete experiments
numerical methods to study the influence of impact loading at high of Ref. [26] under static and dynamic loadings are introduced.
strain rate on concrete. The analyses focused on the load-capacity, the
effect of the microstructure, and energy absorbency. Then, a damage 2. High-strength concrete experimens
model of concrete was presented by editing some parameters for satis­
fying the results obtained from experiments. Many cases with a random 2.1. Quasi-static compression tests
distribution of aggregates were conducted to confirm the reliability of
this model. Ref. [23] presented a new approach in numerical simulation In this section, the experiments of Ref. [26] are introduced. The re­
of dynamic tensile behavior of concrete using a rate-dependent cohesive sults obtained from Ref. [26] are presented with full details such as;
model. Through the results obtained from simulation, good agreement sample sizes, average compressive strength for the static and dynamic
with the test data was also observed as validation of the proposed loading tests with different strain rates. Three grades of concretes with
approach. Ref. [24] reported the testing results in a total of 110 cylin­ static uniaxial compressive strengths of ~60 MPa, ~80 MPa and ~110
drical specimens to investigate the mechanical behavior of mortar and MPa are denoted as C60, C80 and C110, respectively. Three categories,
concrete under various strain rate. SHPB technique and a proposed with specimens having sizes; 100 mm cubes, Φ75 × 150 mm (cylinder
compressive constitutive model were employed to investigate the failure with 75 mm diameter and 150 mm height) and Φ100 × 200 mm (cyl­
samples. Ref. [25] presented the effect of initial static and dynamic loads inder with 75 mm diameter and 150 mm height) in each type of C60,
on concrete dynamic compressive failure. Following the conclusions, the C80 and C110, are reported in this section. To determine the average
compressive strength under different dynamic loads gradually decreased static compressive strength of concrete, a very low strain rate registered
with the increase of initial static load. At the same time, the influence of at 1 × 10− 5 s− 1 was applied in uniaxial compression tests; C60, C80 and
the initial static load on dynamic compressive strength of concrete could C110. Strain gauges were located at the mid-sections in samples having
be reduced when the strain rate was increased. Ref. [26] investigated the geometry Ø100 × 200 mm (cylinder with 100 mm diameter and 200 mm

2
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Fig. 4. The compressive stress-strain curve of C110 at various strain rates [26].

Fig. 1. The compressive stress-strain curve of C60, C80 and C110 under static (SHPB) technique at various strain rates ranging from 40 s− 1 to 110 s− 1.
loading [26]. The results in each grade of concrete are expressed as a stress-strain
curve in Fig. 2 through Fig. 4. It can be seen that the three high-
strength concretes demonstrate a similar trend. The compressive
strength of concrete is significantly increased when taking into account
the effect of strain rate in comparison with the cases of static loading.
The figures presented in this section are used as reference’s data to
verify the proposed model. In the next sections, a concrete damage
plasticity model will be presented. To validate the proposed model, the
experiments in this section, including static and dynamic tests, will be
simulated by ABAQUS software.

3. A concrete damage plasticity model for simulating high-

strength concrete under static and dynamic loadings

A concrete damage plasticity model for high-strength concrete C60,

C80, C110 in this study is based on the combination of the modification
from the original CDP model and DIF calculated according to MC2010.
Fig. 2. The compressive stress-strain curve of C60 at various strain rates [26]. There is a primary modification from the original CDP model that can be
described as follows:

• The stress-strain relationship in compression in the softening phase

(after the compressive strength reaches a peak) is calculated ac­
cording to Ref. [26].
• Compressive damage variable (dc ) is calculated according to
Ref. [30].
• Tensile damage variable (dt ) is calculated according to Ref. [30].
• The ratio between biaxial compression and uniaxial compression is
selected as 1.2 to agree with the experimental data for compressive
high-strength concrete C60, C80, C100.

The full details of the proposed model are described in this section,
including determining the parameters of yield function and plastic flow

Fig. 3. The compressive stress-strain curve of C80 at various strain rates [26]. Table 2
The equations for calculating the characteristics of the concrete base on the
height) to measure Modulus of elasticity (Young’s modulus E and Pois­ average compressive strength.
son’s ratio ν). The results of measurements are shown in Table 1. Be­ Equation Units
sides, four strain gauges were set up as two diametrically opposite pairs,
Average compressive strength according to CEB- fcm = fck + 8 MPa
i.e. one pair for measurement of axial strain and the other for strain in FIP Model Code 1990 [31]
the circumferential direction. The stress-strain curves relationship of Average tension strength according to CEB-FIP ftm = max(0.3fck 0.67 , MPa
samples with size Ø100 × 200 mm cylinder in categories C60, C80 and Model Code 1990 [31] 0.9fsp )
C110 specimens are plotted in Fig. 1. Modulus of elasticity according to fib Model Code ( )1/3
fcm GPa
2010 [32] Eci = Ec0 αE
10
Peak compression strain according to EN1992-1-1 εtm = 0.0007(fcm )0.31
2.2. Dynamic compression tests [33]
Fracture energy GF according to fib Model Code GF = 73(fcm )0.18 N/m
The specimens with geometry Ø77 × 40 mm (cylinder with 77 mm 2010 [32]
( )2
diameter and 40 mm height) in each of C60, C80 and C110 are selected Crushing energy Gch according to Ref. [30]
Gch =
fcm
GF
N/m
to test under dynamic loading using Split Hopkinson Pressure Bar ftm

3
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Fig. 5. Concrete yield surface in (a) deviatoric plane and (b) plane stress.

potential, defining the compressive and tensile stress-strain curves, and

Table 3
calculating the dc and dt .
The parameters of yield function and flow potential in the CDP model.
Dilation angle Eccentricity (ε) σb0 /σc0 Kc
3.1. Definition of plastic flow potential and yield function
300–400 0.1 1.2 0.667

Concrete is a complex material characterized by many different pa­

rameters such as; average compressive strength, average tension the p − q plane at high confining pressure. ∈ denotes the eccentricity of
strength, crushing energy, fracture energy, etc. Usually, the values of the potential plastic surface, which is defined as the rate at which the
these parameters are obtained through experiments. In this paper, we function reaches the uniaxial tensile. In ABAQUS the default value ∈ =
present the calculation for the values of material parameters according 0.1 was used. And σ t0 is the uniaxial tensile stress at failure condition.
to the standards of CEB-FIP Model Code 1990 [31], fib Model Code 2010 The yield function presented in Ref. [14] is shown in Eq. (3) and Eq.
[32] and EN1992-1-1 [33]. Especially, crushing energy Gch was refer­ (4).
enced in the proposal of Alfarah [30]. These equations are described in
detail in Table 2. F=
1
(q − 3αp + β〈σmax 〉 − γ − 〈σmax 〉) − σc = 0 (3)
It can be noticed that the important concrete characteristics pre­ 1− α
sented in Table 2 can be calculated directly from the average strength fcm
(fb0 /fc0 ) − 1 σc 3(1 − Kc )
, which can be easily obtained through experiments of the static uniaxial α= ; β= (1 − α) − (1 + α); γ = (4)
compressive test. This will create favorable conditions to avoid errors
2(fb0 /fco ) − 1 σt 2Kc − 1
because unnecessary experiments are avoided. Through these important
Where 〈.〉 is the Macaulay bracket, and p and q follow the definition in
characteristic values, the damage concrete plasticity model (CDP) is also
Eq. (2). fb0 and fc0 are the biaxial and uniaxial compressive yield
easily established.
strengths, respectively, σ max is the maximum effective principal stress
There are several models that have been proposed to describe the
accounting for the effects of the damage variables dc (for compression)
behavior of concrete from the un-crack/un-crush stage to the failure
and dt (for tension). Also σ c and σ t are defined as σc = σc /(1 − dc ) and
stage. In all models, the damage concrete plasticity model (CDP) is
σ t = σt /(1 − dt ), where dc and dt denote damage variables at
widely used. This model is employed in the ABAQUS manual. The yield
compression and tension stages. Kc is the ratio between the magnitudes
function in CDP is modified to account for the increased strength due to
of deviatoric stress on tensile and compressive meridians, and Kc is
the effects of confinement, i.e. the ratio between biaxial and uniaxial
bounded in from 0.5 to 1. Thus, if Kc = 0.5, the Rankine yield surface is
compression and tension in the research of Ref. [14], with the modifi­
established, and if Kc = 1 is von-Mises yield surface. In ABAQUS, Kc = 2/
cations suggested in Ref. [12].
3 was used as the default value.
In the CDP model, the following non–associated potential plastic
The yield surfaces in the different deviatoric plane and in the plane
flow rule is used following Eq. (1)
stress are depicted in Fig. 5.
∂G The yield surface of the concrete was reported in Ref. [26], having
ε˙p = λ̇ (1)
∂σ the coefficient σb0 /σ c0 = 1.16. This coefficient causes the compressive
stress to increase due to the effective confinement and it can be obtained
Where σ and ε˙p denote the stress and plastic strain rate tensors, from the maximum of the experimental range, including a uniaxial
respectively, λ̇ is a plastic multiplier, and G defines Drucker–Prager compression test and biaxial compression test. However, the coefficient
hyperbolic function in the model given by Eq. (2) σ b0 /σ c0 = 1.16 is only suitable to simulate the problems affected by
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )2 ̅ normal strength. For compressive high-strength concrete, this coeffi­
G= ∈ σ t0 tan ϕ + (q)2 − ptan ψ cient is modified based on the experimental study of equal
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (2) biaxial-to-uniaxial compressive strength ratio in Ref. [34]. Thus, we
1 3
p = − tr(σ), q = ‖dev(σ )‖ propose σ b0 /σ c0 = 1.2 to establish the yield surfaces of concrete in static
3 2
and dynamic loading for compressive high-strength concrete. Other
Where p and q are the hydrostatic stress and the von-Mises equivalent parameters of the yield surface are given in Table 3.
stress, respectively. The parameter ψ is the dilation angle registered in

4
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Table 4
Material parameters for uniaxial compression under static loading.
Concrete C60 C80 C110

εf 0.016 0.018 0.02

ζ 0.25 0.16 0.11

Fig. 6. Concrete behaviour in compression.

3.2. Compressive behaviour

The compressive behaviour of concrete is characterized by the stress-

strain relationship. Thus, if the stress is between 0 and 0.4fcm, the stress- Fig. 7. Stress-strain relationships of the proposed model under static loading
strain relationship is linear, and if the stress is between 0.4fcm to fcm, the with C60, C80 and C110.
stress-strain relationship is nonlinear. After reaching the peak value fcm ,
the compressive damage variable dc is used for the strain at equivalent merical simulation and experiments. According to Ref. [26], the values
compression plastic εpl el in
c , and elastic εc . The crushing strain εc denotes the of εf and ζ are used in the case of high-strength concrete C60, C80, C110
damaged extent of concrete in compression loading. In this paper, the in both static and dynamic cases are given in Table 4.
compression behaviour is introduced, including three phases as shown The stress-strain relationships using the model presented of high-
in Fig. 6. strength concrete C60, C80 and C110 using Eqs. (5)–(8) and other pa­
Phase 1: rameters in Table 2 are shown in Fig. 7.
In this phase, the stress and strain relationship of concrete has a
linear relationship. At the end of the straight line, the value of stress is 3.2.1. Compressive damage variable
recorded according to MC2010 with value σ c = 0.4fcm . The stress-strain This parameter is used to define the compression hardening data for
relationship in this phase is expressed as in Eq. (5) the concrete damaged plasticity material model. And it can be used to
specify the compressive stiffness degradation. In the study of Ref. [26],
σ 1c = E0 εc (5) compressive damage variable (dc ) is assumed that damage initiates only
This linear relationship can be determined based on the secant after the maximum stress is attained. Since it is challenging to measure
modulus of concrete materials, which is determined by Eq. (6) damage quantitatively in a compression test. Thus, dc was proposed in
( ) Ref. [26] as Eq. (9)
fcm
E0 = 0.8 + 0.2 Eci f
88 (6) dc = 1 − (9)
fcm
1/3
Eci = 10000fcm
However, Eq. (9) suggested by Ref. [26] did not consider the effec­
Phase 2: tiveness of mesh size to the simulation results. To improve the damage
Phase 2 is defined when the concrete reaches the crushing strain. The model of concrete, this paper introduces the calculation of dc according
stress-strain relationship of concrete is nonlinear and is given by Eq. (7) to Ref. [30], where dc is more accurate because the effects of mesh size
( )2 are mentioned and given by Eq. (10)
Eci fεcmc − εεctmc
σ 2c = ( ) fcm (7)
1 + Eci fεcmc − 2 εεctmc

Where Eci is Modulus of elasticity of concrete expressed as in Eq. (6)

Phase 3:
In this phase, concrete stiffness starts to decrease due to the
appearance of dc . The stress-strain relationship in this phase is calculated
according to Eq. (8) obtained from Ref. [26].
[ ( )]
( ) ε − εcm
σ 3c = fcm + σf − fcm exp ζ 1 − f (8)
ε − εcm

Where σ f , εf are the stress and strain at the complete crushing. In this
paper, σf = 0.4fcm is assumed, and the parameter ζ and εf are not ob­
tained from experiments; their values are proposed to match the nu­
Fig. 8. Compressive damage variable (dc ) vs. crushing strain.

5
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Fig. 10. Stress-strain curve of the proposed model under static loading for C60,
C80 and C110.

Fig. 9. Uniaxial model of concrete behavior in tension.

1 [ ( ) ( ) ( ]
dc = 1 − 2 1 + ac exp − bc εch
c − ac exp − 2bc εch
c (10)
2 + ac

Where ac , bc can be determined from Eq. (11) and Eq. (12).

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ac = 2(fcm / fc0 ) − 1 + 2 (fcm /fc0 )2 − (fcm /fc0 ) (11)

fc0 leq ( ac )
bc = 1+ (12)
Gch 2 Fig. 11. Tensile damage variable (dt ) vs. cracking strain.

Where fc0 denotes uniaxial compressive yield strength, leq denotes the effects of the mesh size, it is assumed that the crack occurs in each
mesh size in the numerical model. Fig. 8 shows the relationship of element after tensile stress reaches a peak, the strain at tensile strength
compressive damage variable-crushing strain of C60, C80 and C110 with εtm can be calculated from crack opening, given in Eq. (15) where leq
leq = 50 mm. denotes length of element (mesh size). Based on this assumption, the
stress-strain relationship is illustrated in Fig. 9
w
3.3. Tension behavior εt = εtm + (15)
leq
At the first phase, when the concrete has no cracks yet, the stress- In comparison with the tensile stress-tensile strain curve in Ref. [26],
strain relationship is considered linear and determined through the the tensile stress-strain presented in this paper is more suitable when
modulus of elasticity. The behavior of the concrete in this phase does not using in F.E. analysis because of considering the effect of element’s
depend on the mesh size. After cracks take place at the phase ‘strain- length, i.e. mesh size, which was not mentioned in Ref. [26]. Fig. 10
softening, the stress-strain relationship becomes nonlinear and depends shows the tensile stress-strain curve calculated according to Eq. (13)
on the FE mesh size. Cracks appear and grow until the concrete through Eq. (15).
completely fails. The influence of the mesh size (leq ) in the FE. model can
cause convergence problems when the model is implemented. There­ 3.3.1. Tensile damage variable
fore, Ref. [35] presented the curve of stress-crack opening given by Eq. The tension damage variable (dt ) introduced in this study is defined
(13) to replace the stress-strain curve. according to Ref. [30] as shown in Eq. (16). The relationship of tensile
Regarding tensile behavior, the ratio between tensile stress σ t (w) damage variable and cracking strain of C60, C80 and C110 with leq = 50
where w is crack width, and maximum tensile strength ftm is given by Eq. mm is shown in Fig. 11.
(13).
1 [ ( ) ( ) ( ]
[ ( )3 ] dt = 1 − 2 1 + at exp − bt εch − at exp − 2bt εch (16)
σ t (w) w w( ) 2 + ac t t
(13)
w
= 1 + c1 e − c 2 wc − 1 + c31 e− c2
ftm wc wc
Where at , bt can be determined from Eq. (17) and Eq. (18)
Where c1 = 3, c2 = 6.93 and wc is the critical crack opening. Therefore, √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
wc can be considered as the fracture crack opening according to Eq. (14) at = 2(ftm / ft0 ) − 1 + 2 (ftm /ft0 )2 − (ftm /ft0 ) (17)

GF
wc = 5.14 (14) ft0 leq ( at )
ftm bt = 1+ (18)
GF 2
In this paper, the real crack spacing is ignored. To account for the

6
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Table 5
Numerical simulation methods.
Terms User Size of Interaction method
element meshing

Concrete test C3D8R 25 mm Tangential behavior: fms =

samples 0.47
Normal behavior: “hard
contact"
Rigid steel plates C3D8R 50 mm Tangential behavior: fms =
0.47
Normal behavior: “hard
contact"

of loading expressed in Eq. (19). And the strain rate – DIF curve is shown
Fig. 12. Increase of concrete strength under high strain rates for compression. in Fig. 12.
⎧ ( ε̇ )0.014
⎪
⎪ for ε̇ ≤ 30s− 1
⎨ ε˙0
DIF = ( )0.014 (19)
⎪
⎪ ε̇
⎩ ˙ −1
0.012 for ε > 30s
ε˙0

Where ε˙0 = 30 × 10− 6 s− 1 .

The influence of different strain rates on compressive stress-strain
curves of typical C60 is shown in Fig. 13. Thus, the compressive peak
strength in the dynamic condition is calculated by multiplying the
compressive peak strength in the static condition with the DIF given in
Eq. (19). After reaching the new compressive peak strength, the curve of
stress-strain is obtained using Eq. (5) to Eq. (8). As a result, the higher
the strain rates, the more the compressive strength is. And its corre­
sponding peak strength, showing that there is an obvious rate-dependent
behavior in compressive stress-strain curves of concrete material,
Fig. 13. Compressive stress-strain curves of typical concrete C60 under especially under high strain rates.
different strain rates.
4. Numerical simulation

4.1. Numerical simulation under static loading

3.4. The dynamic increase factor (DIF)
To evaluate the reliability of the model presented in this paper, nu­
The dynamic increase factor (DIF), defined by the ratio of the dy­
merical simulations are used. The experiments of Ref. [26] including the
namic strength to the quasi-static strength in uniaxial compression, has
compressive tests of high-strength concrete known as C60, C80, C110
been widely accepted as an important parameter to measure the strain-
having a shape Φ100 × 200 mm (cylinder with 75 mm diameter and
rate effect on the strength of concrete-like materials. In the case of the
150 mm height) were implemented as shown in Fig. 14(a). The average
dynamic loading, the parameters of the original CDP model and yield
compressive strength and initial elastic Modulus for C60, C80 and C110
surface of concrete are still the same as the static loading given in
are given in Table 1 in section 2. The other material parameters are
Table 1. Meanwhile, the average stress in compression and tension are
determined using the equations in Table 2. The Finite Element software
modified by multiplying them with the DIF. Therefore, finding a
ABAQUS is employed for the numerical simulations. C3D8R element is
formulation of DIF is important to simulate the real-concrete like ma­
used to model the concrete test samples and the two rigid steel plates
terial under dynamic loading. This paper introduces the DIF according to
arranged on top and bottom of the specimen. C3D8R is solid eight-node
MC2010 [32] for the compressive strength of concrete under high rates

Fig. 14. (a) Experimental compression tests, (b) Model boundaries, (c) F.E. mesh.

7
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Where Δl is the displacement between the two rigid steel plates, F is the
static compressive force obtained from the numerical simulation
method, L0 and D0 is the initial size of the test samples.
The rigid steel plates and concrete interface is simulated in ABAQUS
using contact property options. Thus, the tangential behavior is
described with friction formulation, and normal behavior is described
with the “Hard” contact method. For tangential behavior, the friction
coefficient is a significant factor, and this coefficient will depend on the
surface conditions. In this study, in order to determine the value of the
friction coefficient, C60 will be selected to simulate the different friction
coefficients μ = 0.1, μ = 0.47, μ = 0.7. And the results in comparison
between simulation and experiment is illustrated in Fig. 15.
It can be seen that the friction coefficient crucially affects the
simulation results, especially during the softening phase (after stress
Fig. 15. Comparison between simulation and Y.B. Guo’s experiment for C60 reaches peak strength). According to Fig. 15, the friction coefficient μ =
for different friction coefficients. 0.47 shows the best performance compared to the experiment. So, in this
study, we select μ = 0.47 to simulate the rest of the examples, including
brick element with reduced integration and hourglass control. The static and dynamic experiments. The compressive stress-strain curve
boundary conditions are applied by constraining the displacements in obtained from numerical simulation and experiments is presented in
the X and Y directions as shown in Fig. 14(b)). The concrete test samples Fig. 16(a), (b) and (c) for C60, C80 and C110, respectively. The crushing
using a mesh of 25 mm elements size is shown in Fig. 14(c)). pattern in the numerical model is obtained by defining the compressive
The friction coefficient between the concrete test samples and the damage variable (dc ) as shown in Fig. 17. The results show that the
rigid plates is assumed to be fms = 0.47 in this study. The parameters figures obtained from the numerical simulation are in good agreement
used in the numerical simulations are summarized in Table 5. with the experimental results. In general, the proposed model intro­
The stress-strain relationship in the simulation method is calculated duced in this paper shows high reliability for simulating the behaviour of
based on Eq. (5) high-strength concrete C60, C80 and C110 under static loading.

Δl F
ε= , σ= / (20)
L0 πD20 4

Fig. 16. Comparison between simulation and Y.B. Guo’s experiment for (a) C60, (b) C80 and (c) C110.

8
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

u2 = f3 (x − cb t) = uT (23)

Where f1 , f2 , f3 are three arbitrary functions for the displacement

caused by the incident wave (uI ), reflected wave (uR ) and transmitted
wave (uT ). The strain can be defined by:
∂u
ε= (24)
∂x
Differentiating Eq. (22) with respect to x and t, we have

(25)
′ ′
ε = f1 + f2 = εI + εR
( ′)
(26)
′
u˙1 = c − f1 + f2 = c( − εI + εR )

Differentiating Eq. (23) we obtain

u˙2 = − cb εT (27)
The average strain rate ε̇ is calculated from the strain of the test
Fig. 17. Crushing pattern using simulation. specimen during the wave transmission.
u˙1 − u˙2
ε̇ = (28)
Hs

Where Hs is the test specimen height. Combination Eq. (26) and Eq. (27)
in Eq. (28), we have a new formulation as in Eq. (29).
u˙1 − u˙2 cb
ε̇ = = ( − εI + εR + εT ) (29)
Hs Hs
The force at two ends of the input and iutput bars are obtained from a
Fig. 18. Split Hopkinson Pressure Bar test set up. two-wave and one-wave method in Eq. (30) and Eq. (31), respectively.
F1 = Ab Eb (εI + εR ) (30)
4.2. Numerical simulation under dynamic loading
F2 = Ab E b ε T (31)
4.2.1. Principles of Split Hopkinson Pressure Bar tests
Split-Hopkinson pressure (SHPB) is one of the most techniques used Where Ab and Eb are the cross-section and Young’s Modulus of the
in dynamic testing as shown in Fig. 18. The principle of this method is Hopkinson bars. Eventually, with a short height of the sample, the force
based on one-dimensional stress wave propagation straight bar having a in Eq. (30) and Eq. (31) are equal. Then, We obtain Eq. (32).
thin thickness. A general review of this testing technique, including full εI + εR = εT (32)
details of the method, can be found in Ref. [36]. The general working
steps are as follows: From Eq. (29) and Eq. (32), the strain rate can be rewritten as in Eq.
(33)
• The input bar receives the pressure covered obtained from initial 2cb εR
velocity. This creates a stress wave at the impacted end position of ε̇ = (33)
Hs
the input bar. Next, this stress wave transmits through the input bar
to create incident strain (εI ). This wave impacts the concrete test Assuming isotropic material and a constant cross-section of the
sample having a position between 2 bars. At this position (free end of samples, the values of stress and strain at the concrete samples are
input bar) a part of the stress wave passes through the test specimen calculated according to Eq. (34) and Eq. (35)
into the output bar to create transmitted strain (εT ). And another part Ab Eb εT
reflects into the input bar to determine the reflected strain (εR ). The σs = (34)
A
strains εI , εT , εR are recorded by placing strain gauges on the
perimeter of the 2 bars. ∫t
2cb
• The three strains εI , εT , εR are used to calculate the strain rate εs = εR dt (35)
Hs
curves and the dynamic stress-strain curves. 0

Thus, by arranging the sensors in the input bar and output bar at the
The differential equation of 1D wave propagation is shown in Eq.
position needed to be measured, we can record the strain εI , εR , εT at
(21)
any time in each bar. Based on Eq. (21) to Eq. (35), the stress-strain
∂2 u 1 ∂2 u relationship will be determined for different strain rates.
= (21)
∂x2 c2b ∂t2
4.2.2. Numerical simulation using SHPB tests
Where x is the displacement along the length of a straight bar, t is the To illustrate the suitability of the CDP presented in the numerical
time of impact-loading and cb is the elastic wave speed of input and model under dynamic loading, whose strain rate is a significant factor
output bars. For input bar, the solution of Eq. (21) can be written as Eq. affecting the compressive strength of concrete, ABAQUS software and
(22) the proposed model in this paper are employed to simulate the dynamic
experiments in Ref. [26]. The concrete test samples of C60, C80 and
u1 = f1 (x − cb t) + f2 (x + cb t) = uI + uR (22)
C110 have the same size of 70 mm diameter (D = 70 mm), and 40 mm
The solution for the output bar can be written as in Eq. (23) height (H = 40 mm). The characteristics of C60, C80 and C110 are given

9
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Table 6 numerical simulation method is presented in Table 7, and the simulation

The characteristics of C60, C80 and C110 in dynamic experiments according to model in ABAQUS is shown in Fig. 19.
Ref. [26]. The Dynamic Explicit Algorithm with total time (t = 0.0023 s) is
Concrete fcm (N/ fctm (N/ Elastic Modulus Density (kg/ used to analyze the SHPB test in ABAQUS. The free end of the input bar
sample mm2) mm2) (GPa) m3) and output bar is constrained in the vertical direction. An impulse force
C110 110.4 7.0 40.70 2450
is applied at the end of the Input bar as shown in Fig. 20. And the value
C80 74.5 6.2 37.00 2450 of impulse force is converted from the initial velocity as shown in
C60 68.1 5.5 37.10 2450 Fig. 21.
The reflected and transmitted strain of C110 at the strain rate 40 s− 1
recorded at the mid-length of each bar are selected as the first example
Table 7 to compare the reflected and transmitted signals obtained from experi­
Parameters used in the numerical model of dynamic experiment. ments. The process of the stress wave transmission in the input bar and
output bar are shown in Fig. 22. And the process creates the incident
terms User Size of Interaction method
element meshing strain (εI ) in the input bar, the transmitted strain (εT ) in the output bar
and the reflected strain (εR ) in the input bar based on the propagation of
Concrete test C3D8R 5 mm Tangential behavior: fms =
sample 0.47 stress waves between the input bar and output bar are shown in Fig. 23.
Normal behavior: “hard
contact"
The Input Bar C3D8R 50 mm Tangential behavior: fms =
0.47
Normal behavior: “hard
contact"
The Output Bar C3D8R 50 mm Tangential behavior: fms =
0.47
Normal behavior: “hard
contact"

in Table 6. The diameter of the impulse pressure bar (input Bar) and the
resulting bar (output Bar) are the same with 80 mm diameter (D = 80
mm), and the length in each bar is 5 m. Two bars uses steel material with
E = 203 GPa, υ = 0.3, Density ρ = 7840 kg/m3.
The C3D8R element is used for the concrete test samples, the input
bar and the output bar. The interaction between the surface of concrete
test samples and the surface of the straight bar makes use of the “hard
Fig. 21. Input pressures for various strain rate.
contact” method with the coefficient of friction fms = 0.47. The

Fig. 19. Numerical simulation SHPB test in ABAQUS.

Fig. 20. The process transmission pressure to input bar.

10
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

σ c0 = 1.2 are suitable for simulations in two cases of dynamic and static
problems.

5. Conclusion

In this paper, the original CDP, which is available in ABAQUS

manual, is modified based on the advantages reported in recently
completed studies. Thus, the compressive stress-strain curve in the
softening phase (after the stress reaches the peak) in CDP is replaced by a
new curve obtained from Ref. [26]. At the same time, two damaged
variables (dc ) and (dt ) are introduced and combined with the
stress-strain curve. For the first time, these combinations create a new
version of CDP applied in simulating the behavior of high-strength
concrete C60, C80 and C100 under static loading. In cases of dynamic
loading, the increasing compressive strength is implemented through
the DIF. To simplify the application, DIF is admitted according to
Fig. 22. Comparison of predicted reflected and transmitted strain between
simulation and experiment for C110 at strain rate 40 s− 1.

The values of displacement and strain of the elements having the

same location with the strain gauge arranged in the experiment of input
Bar can be found in the history output in ABAQUS. Based on these re­
sults, the stress-strain relationship of the concrete test sample under
dynamic loading conditions is calculated from Eq. (21) to Eq. (35). The
stress-strain curve in dynamic loading obtained from simulation are
compared with experimental results of samples C100, C80, C60 as
shown in Figs. 24–26. At high strain rate 110 s− 1, the simulation results
of C60 and C80 and C110 show a good agreement with the results ob­
tained from the experiment. Meanwhile, at lower strain rates 40 s− 1, 60
s− 1, this comparison is measured in accordance with acceptable errors. It
can be illustrated that the application of DIF according to MC2010 and Fig. 24. Comparison of stress-strain responses determined from the simulation
the adjustment of yield surface in the CDP model using coefficient σb0 / with experimental results of C110 at the different strain rate.

Fig. 23. (a) The process of the incident strain (εI ) in the input bar, and (b) the transmitted strain (εT ) in the output bar and the reflected strain (εR ) in the input bar.

11
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

Health Monitoring of Vietnamese infrastructures’ funded by the Flemish

Government.

References

[1] Q. Li, H. Meng, About the dynamic strength enhancement of concrete-like

materials in a split Hopkinson pressure bar test, Int. J. Solid Struct. 40 (2) (2003)
343–360.
[2] P.H. Bischoff, S. Perry, Compressive behaviour of concrete at high strain rates,
Mater. Struct. 24 (6) (1991) 425–450.
[3] M.S. Williams, Modeling of local impact effects on plain and reinforced concrete,
Struct. J. 91 (2) (1994) 178–187.
[4] H. Fu, M. Erki, M. Seckin, Review of effects of loading rate on concrete in
compression, J. Struct. Eng. 117 (12) (1991) 3645–3659.
Fig. 25. Comparison of stress-strain responses determined from the simulation
[5] H. Fu, M. Erki, M. Seckin, Review of effects of loading rate on reinforced concrete,
with experimental results of C60 at strain rate 110/s. J. Struct. Eng. 117 (12) (1991) 3660–3679.
[6] T. Jiang, et al., Three-dimensional nonlinear finite element modeling for bond
performance of ribbed steel bars in concrete under lateral tensions, Int. J. Civ. Eng.
(2020) 1–23.
[7] V. Broujerdian, H. Karimpour, S. Alavikia, Predicting the shear behavior of
reinforced concrete beams using nonlinear fracture mechanics, Int. J. Civ. Eng. 17
(5) (2019) 597–605.
[8] A. Stolarski, Dynamic strength criterion for concrete, J. Eng. Mech. 130 (12) (2004)
1428–1435.
[9] I. Carol, E. Rizzi, K. Willam, On the formulation of anisotropic elastic degradation.:
II. Generalized pseudo-Rankine model for tensile damage, Int. J. Solid Struct. 38
(4) (2001) 519–546.
[10] L. Jason, et al., An elastic plastic damage formulation for concrete: application to
elementary tests and comparison with an isotropic damage model, Comput.
Methods Appl. Mech. Eng. 195 (52) (2006) 7077–7092.
[11] J. Ju, On energy-based coupled elastoplastic damage theories: constitutive
modeling and computational aspects, Int. J. Solid Struct. 25 (7) (1989) 803–833.
[12] J. Lee, G.L. Fenves, Plastic-damage model for cyclic loading of concrete structures,
J. Eng. Mech. 124 (8) (1998) 892–900.
[13] J. Zhang, J. Li, Investigation into Lubliner yield criterion of concrete for 3D
simulation, Eng. Struct. 44 (2012) 122–127.
[14] J. Lubliner, et al., A plastic-damage model for concrete, Int. J. Solid Struct. 25 (3)
(1989) 299–326.
Fig. 26. Comparison of stress-strain responses determined from the simulation
[15] F. Lopez-Almansa, B. Alfarah, S. Oller, Numerical simulation of R.C. frame testing
with experimental results of C80 at strain rate 110/s. with damaged plasticity model. Comparison with simplified models, in: Second
European Conference on Earthquake Engineering and Seismology, Istanbul,
Turkey, 2014.
MC2000. However, the ratio between biaxial compression and uniaxial
[16] R.H. Evans, Effect of rate of loading on the mechanical properties of some
compression will be adjusted from 1.16 to 1.2 according to the results materials, J. Inst. Civil Eng. 18 (7) (1942) 296–306.
obtained from the experiment in Ref. [34]. This adjustment together [17] T.V. Do, T.M. Pham, H. Hao, Dynamic responses and failure modes of bridge
columns under vehicle collision, Eng. Struct. 156 (2018) 243–259.
with the DIF makes an increase in concrete strength due to the increase
[18] G. Cusatis, Strain-rate effects on concrete behavior, Int. J. Impact Eng. 38 (4)
in strain rate. To evaluate the reliability of the current version of CDP, (2011) 162–170.
ABAQUS software is employed to simulate the experiments of Ref. [26] [19] M. Li, et al., Specimen shape and size effects on the concrete compressive strength
in both static and dynamic loading. Through the comparison between under static and dynamic tests, Construct. Build. Mater. 161 (2018) 84–93.
[20] Z. Ye, Y. Hao, H. Hao, Numerical study of the compressive behavior of concrete
numerical simulation and experimental data, it can be concluded that material at high strain rate with active confinement, Adv. Struct. Eng. 22 (10)
the model presented in this paper achieves high reliability and can be (2019) 2359–2372.
referenced to simulate the behavior of high-strength concrete in related [21] J. Georgin, J. Reynouard, Modeling of structures subjected to impact: concrete
behaviour under high strain rate, Cement Concr. Compos. 25 (1) (2003) 131–143.
works. [22] S. Park, Q. Xia, M. Zhou, Dynamic behavior of concrete at high strain rates and
pressures: II. Numerical simulation, Int. J. Impact Eng. 25 (9) (2001) 887–910.
Authors’ statement [23] W. Zhou, et al., Mesoscopic simulation of the dynamic tensile behaviour of
concrete based on a rate-dependent cohesive model, Int. J. Impact Eng. 95 (2016)
165–175.
Hoang-Le Minh: Investigation, Methodology, Validation and Writing [24] X. Chen, S. Wu, J. Zhou, Experimental and modeling study of dynamic mechanical
– original draft. properties of cement paste, mortar and concrete, Construct. Build. Mater. 47
(2013) 419–430.
Samir Khatir: Investigation. [25] L. Jin, W. Yu, X. Du, Effect of initial static load and dynamic load on concrete
Magd Abdel Wahab: Supervision, Funding Acquisition, Writing – dynamic compressive failure, J. Mater. Civ. Eng. 32 (12) (2020), 04020351.
review & editing. [26] Y. Guo, et al., Response of high-strength concrete to dynamic compressive loading,
Int. J. Impact Eng. 108 (2017) 114–135.
Cuong Le Thanh: Conceptualization, Supervision, Validation and
[27] Y. Hao, H. Hao, Numerical evaluation of the influence of aggregates on concrete
Writing – review & editing. compressive strength at high strain rate, Int. J. Prot. Struct. 2 (2) (2011) 177–206.
[28] Y. Hao, H. Hao, Influence of the concrete DIF model on the numerical predictions
of R.C. wall responses to blast loadings, Eng. Struct. 73 (2014) 24–38.
Declaration of competing interest [29] M. Ferraioli, Dynamic increase factor for nonlinear static analysis of rc frame
buildings against progressive collapse, Int. J. Civ. Eng. 17 (3) (2019) 281–303.
The authors declare that they have no known competing financial [30] B. Alfarah, F. López-Almansa, S. Oller, New methodology for calculating damage
variables evolution in Plastic Damage Model for R.C. structures, Eng. Struct. 132
interests or personal relationships that could have appeared to influence
(2017) 70–86.
the work reported in this paper. [31] C.E.-I.d. Béton, CEB-FIP Model Code 1990: Design Code, Thomas Telford
Publishing, 1993.
[32] I. Concrete, Fib Model Code for Concrete Structures 2010, Ernst & Sohn publishing
Acknowledgements
house2013, 2013.
[33] R. Narayanan, A. Beeby, Designers’ Guide to EN 1992-1-1 and EN 1992-1-2.
The authors acknowledge the financial support of VLIR-UOS TEAM Eurocode 2: Design of Concrete Structures: General Rules and Rules for Buildings
Project, VN2018TEA479A103, ‘Damage assessment tools for Structural and Structural Fire Design, vol. 17, Thomas Telford, 2005.

12
H. Le Minh et al. Journal of Building Engineering 44 (2021) 103239

[34] W. Dong, et al., Experimental study of equal biaxial-to-uniaxial compressive [35] H. Cornelissen, D. Hordijk, H. Reinhardt, Experimental determination of crack
strength ratio of concrete at early ages, Construct. Build. Mater. 126 (2016) softening characteristics of normalweight and lightweight, Heron 31 (2) (1986)
263–273. 45–46.
[36] W.W. Chen, B. Song, Split Hopkinson (Kolsky) Bar: Design, Testing and
Applications, Springer Science & Business Media, 2010.

13