You are on page 1of 17

Engineering Fracture Mechanics 71 (2004) 2197–2213

www.elsevier.com/locate/engfracmech

Impact behaviour of concrete: a computational approach


F. Barpi *

Department of Structural and Geotechnical Engineering, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
Received 21 February 2003; received in revised form 2 June 2003; accepted 17 November 2003

Abstract
The purpose of this study is to examine the numerical simulation of concrete specimens under high loading rates in
tension. The data found in the literature are described: they show an increase in compressive and tensile strength as a
function of the loading rate. To study this behaviour, we focused our attention on the assessment of the consistency
model through the simulation of many experimental results related to a wide range of strain rates. In particular, the
tests on unnotched and notched specimens performed at Delft University in recent years are examined. The proposed
model is able to describe the increase in strength due to a high loading rate by an appropriate choice of the viscosity
parameter.
 2004 Elsevier Ltd. All rights reserved.

Keywords: Concrete; Hopkinson bar; Impact; Softening; Strain rates; Viscoplasticity

1. Introduction

Concrete is a rate-dependent material, i.e., its properties (ultimate strength, Young modulus, fracture
energy) are highly dependent on the loading rate. For this reason, knowledge of the constitutive rela-
tionship for a wide range of strain rates is required to develop a realistic material law.
High loading rates are frequently encountered in engineering problems. Some structures have to be
designed directly for impacts or impulsive loading: shelters, ammunition bunkers, structures to protect
mountain roads from falling rocks, structures to protect bridges against impact of ships, accidental
explosions.
The aim of this work is to evaluate the capabilities of the consistency model, developed at Delft Uni-
versity in recent years, concerning the impact tensile behaviour of concrete. In the consistency model the
viscoplastic contribution is incorporated in the loading function by using a scalar parameter. The contri-
bution of viscosity makes it possible to overcome the problems (mesh dependence and, in general, non-
objectivity of the numerical response) encountered by using standard softening plasticity models.

*
Tel.: +39-11-564-4886; fax: +39-11-564-4899.
E-mail address: fabrizio.barpi@polito.it (F. Barpi).

0013-7944/$ - see front matter  2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2003.11.007
2198 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

2. Experimental results

2.1. General

This section presents some experimental results for high loading rates applied to specimens in com-
pression and in tension. The general trend is an increase in dynamic strength as the loading rate increases
but, as shown later, this effect is more pronounced in tension than in compression.

2.2. Compression

We mainly refer to a detailed paper by Bischoff and Perry [4]. In their work, a large amount of
experimental data is collected and compared, as shown in Fig. 1. Despite the different experimental tech-
niques used and the different methods employed to analyse the results, an increase in strength as a function
of the compressive strain rate is generally found. Their paper presents an assessment of the measuring
devices and a discusses various definitions of loading rates, important to the compare results obtained in
many different ways.
Bischoff and Perry [4] also examined the influence of concrete quality (that seems to have a significant
effect), type of aggregate, age, curing and moisture conditions together with the influence of the loading rate
on the elastic modulus (found to increase), critical compressive strain, i.e., the strain corresponding to the
maximum stress (generally found to increase), PoissonÕs ratio (found to increase).

2.3. Tension

The influencing factors that affect the strength of concrete in impact tension tests are investigated in
many papers. Fig. 2 shows 1 the envelope of experimental results from Hatano [10], Birikimer and
Lindemann [3], Zielinski et al. [35,36], Kormeling [11], Soroushian et al. [28], Reinhardt et al. [21], Wee-
rheijm [31], Rossi et al. [23] and Cadoni et al. [5]. These data have been obtained by means of different test
set-ups (not always clearly defined, especially for old tests), specimen dimensions and material properties.
This may partly explain the heavy scatter in the results. The aim of the picture is merely to show that in the
literature we find the same clear trend in tension as in compression: an increase in measured strength is
always observed when the imposed strain (or stress) rate increases.
A useful experimental tool to determine the influence of high loading rates is the Split-Hopkinson bar.
This device was first developed for compression tests, see, for instance [29]. A striker bar impacts the
incident bar and a compressive pulse travels through the incident bar and reaches the specimen. At this
interface, part of the incident pulse is reflected (because of the mismatch of the mechanical impedances of
bar and specimen) and part of the pulse is transmitted into the specimen (where a uniform state of strain
and stress can be obtained).
This type of set-up was modified at Delft University, see [19,20,33,35,36] in order to test concrete
specimens in tension under stress rates from 2 to 60 N/(mm2 ms). These authors also investigated the
influence of maximum aggregate size, water-cement ratio, cement type and humidity.
More recently, Reinhardt et al. [21] extensively investigated the influence of the free water content on the
behaviour of microconcrete (maximum aggregate size 2 mm) under strain rates from 0.25 to 1.25 s1 . It is

1
The data of Fig. 2, that were originally expressed by stress rates, are converted to strain rates by using the simple relationship:
r_
e_ ¼ ; ð1Þ
Eav
where Eav represents an average value of the elastic modulus taken equal to 30,000 N/mm2 .
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2199

Fig. 1. Strength increase vs. strain rate (experimental results in compression). Reprinted from Bischoff and Perry [4].

4.5 Birikimer and Lindemann 1971


Kormeling (5_80_3) 1980
Hatano 1 1960
Hatano 2 1960
4.0 Hatano 3 1960
Reinhardt et al. (w/c= 0.46, wet) 1990
Reinhardt et al. (w/c= 0.46, dry) 1990
Soroushian et al. 1986
Rossi et al. (w/c= 0.50, wet) 1994
3.5
Strength increase ft,d / ft,s (_)

Rossi et al. (w/c= 0.50, dry) 1994


Rossi et al. (w/c= 0.30, wet) 1994
Rossi et al. (w/c= 0.70, wet) 1994
Zielinski 1981
Weerheijm 1992 (notch./unnotch. spec.)
3.0 Cadoni et al. (5mm) 2001
Cadoni et al. (10mm) 2001
Cadoni et al. (dry) 2001
Cadoni et al. (curing) 2001
2.5 Cadoni et al. (wet) 2001

2.0

1.5

1.0
–6 –4 –2 0 2
10 10 10 10 10
–1
Strain rate (s )
Fig. 2. Strength increase vs. strain rate (experimental results in tension).

shown that the content of free water strongly influences the response: dry specimens show a small increase
f f
of strength ft;dt;s ffi 1:45 while wet specimens show a larger increase ft;dt;s ffi 4:1. The influence of free water was
also investigated for miniconcrete (maximum aggregate size 10 mm) by the same authors, see [23], by means
2200 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

of experiments conducted at stress rates from 5 · 105 to 77.2 N/(mm2 ms). The increase in strength is
approximately equal to 1.40 (dry specimens) and 2.10 (wet specimens). Fig. 2 presents these results (dry and
wet specimens) for different values of the water–cement ratio (labelled with w=c). Rossi and Boulay [22] also
investigated the influence of the free water content on fracture toughness.
The influence of rate effects and inertia effects on the biaxial stress is investigated in [31,34]. By using the
Split-Hopkinson bar in combination with a lateral compression device, they tested notched prismatic
specimens with two different types of concrete.
Moreover, a description of some recent experiments, based on a modified version of the Split-Hopkinson
technique, called Hopkinson Bar Bundle, can be found in Cadoni et al. [5].

2.4. Some remarks on experiments and numerical models

A comparison between Figs. 1 and 2 shows that the increase in strength is greater in tension than in
compression. For instance, for a strain rate equal to 101 s1 , the ratio between dynamic and static strength
is 2 for compression and 3.5 for tension. Considering that microcracks are more pronounced in com-
pression than in tension, it is reasonable to conclude that the effects of the loading rate decreases as the
microcracking phase increases. On the other hand, it can also be argued that the role of free water is
different in tension and in compression.
We give here more information and references on different experimental test and a survey on recent
numerical results. Triaxial tests are examined in [1,9] while tests on reinforced-concrete structures are
reported in [12]. Many authors examined similar problems from a numerical point of view: we should
mention, among others, Bicanic and Zienkiewicz [2] (plasticity), Oh [15] (microcrack planes), Dube
et al. [6] (rate dependent damage), Sercombe [24], Sercombe et al. [25] (viscous hardening), Oritz and
Pandolfi [16] (cohesive model) and Gatuingt and Pijaudier-Cabot [8] (viscoplasticity and rate dependent
damage).

3. The consistency viscoplastic model

3.1. General

It is well known that the viscoplastic model is able to predict the increase in strength as a function of the
strain rate. In this model, viscosity is seen as a regularisation parameter or a micromechanical parameter
determined from observed shear band width [13]. Here, viscosity is used to describe high loading rate
problems and, as explained later, is determined from experimental tests. It seemed reasonable to use a so-
called consistency viscoplastic model, see [30], for problems dominated by high loading rates. By doing this,
it is also possible to overcome the problems encountered by using softening-plasticity models (mesh
dependence and, in general, non-objectivity of the numerical response). This is due to the viscoplastic
regularisation provided by the model, see, for instance [13,14,18,27].

3.2. Consistency model

This model is similar to the viscoplastic model proposed by Perzyna [17] and Duvaut and Lions [7] but
the main difference is that the viscoplastic contribution is incorporated in the loading function by using a
scalar parameter. The strain rate e_ ij is split in the elastic and viscoplastic part in the classical form:

e_ ij ¼ e_ eij þ e_ vp
ij ; ð2Þ
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2201

where the viscoplastic flow is defined as

e_ vp _ og ;
ij ¼ k ð3Þ
orij

(k_ being the plastic multiplier, f the loading function and g the plastic potential function). For associative
plasticity, considered here, f ¼ g.
The consistency equation (now represented by a differential equation) takes the form

of
f_ ðrij ; j; jÞ
_ ¼ r_ ij  hk_  m€
k¼0 ð4Þ
orij

with
of j_
h¼ ð5Þ
oj k_
and
of j

m¼ ; ð6Þ
oj €
k
where h represents the hardening–softening modulus, j the equivalent plastic strain and m the viscosity
parameter.
The classical Kuhn–Tucker conditions still remain valid (no overstresses, i.e., stresses outside the loading
function f , are allowed):

k_ P 0; f 6 0; _ ¼ 0:
kf ð7Þ

The main difference between the consistency model and the Perzyna and Duvaut–Lions viscoplastic model
is related to the unloading phase, as explained in [30].

3.3. Numerical implementation

In this section the numerical implementation of the consistency model is presented (for more details, see
[30]). We shift from a tensorial notation to a matrix notation: tensors rij and eij are now replaced by the
corresponding vectors r and e.
The incremental stress–strain relationship and the plastic flow are given by

Dr ¼ De ðDe  Devp Þ; ð8Þ

where De is the elastic stiffness matrix, and


og _
e_ vp ¼ k_ ¼ kq: ð9Þ
or
The discretised consistency condition becomes
f ðrðiþ1Þ ; kðiþ1Þ ; k_ ðiþ1Þ Þ  f ðrðiÞ ; kðiÞ ; k_ ðiÞ Þ þ df ¼ f ðiÞ þ ndr þ hdk þ mdk_ ¼ 0: ð10Þ
Differentiation of Eq. (8) gives
dr ¼ De ðde  devp Þ; ð11Þ
2202 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

while, by estimating the plastic flow and the plastic multiplier through the generalised Euler method, 2 we
obtain
 
de ¼ ð1  #Þ d_et þ#d_etþDt Dt ¼ #d_evp
vp vp vp
tþDt Dt; ð13Þ
|{z}
¼0

 
_ _
dk ¼ ð1  #Þ dkt þ#dktþDt Dt ¼ #dk_ tþDt Dt: ð14Þ
|{z}
¼0

In Eqs. (13) and (14) the quantities d_evp _


t and dkt vanish because we perform local iterations (index i) during a
global iteration (at time t).
By differentiating Eq. (9), we get
 
_ _ oq oq oq _
vp
d_etþDt ¼ dðkqÞ ¼ k dr þ dk þ dk þ qdk_ ð15Þ
or ok ok_
and, substituting Eqs. (13) and (15) into Eq. (11)
2 0 1 3
6 B oq oq oq _ C 7
dr ¼ De de  De 6 _B C _7
4k@ or dr þ ok dk þ ok_ dkA þ qdk5#Dt: ð16Þ
|{z} |{z}
¼0 ¼0

It is assumed that the direction of plastic flow is independent of the plastic multiplier, that is, in Eq. (16),
oq
ok
¼ 0 and oq
ok_
¼ 0.
By arranging Eq. (16) and using Eq. (14), we obtain
dr ¼ Pde  P
ndk; ð17Þ
where P represent a pseudo-elastic stiffness matrix:
 1
1
P ¼ ðDe Þ þ #kDt_ oq ð18Þ
or
and 
n the pseudo-plastic flow direction:

 _ oq þ k_ oq :
n ¼ q þ #kDt ð19Þ
ok ok_
Eliminating dr by substituting Eq. (17) into Eq. (10) we get
¼0
z}|{
ðiÞ T
f þ n P de
dk ¼ T T m : ð20Þ
n þ h þ #Dt
n P
The strain increment de vanishes from Eq. (20) because we are performing local iterations during a global
iteration with fixed strain increment.

2
The implicit time stepping scheme is
DðÞ ¼ ½ð1  #ÞðÞ_ t þ #ðÞ
_ tþDt Dt; ð12Þ
where ðÞ is a generic quantity and # 2 ½0; 1. In this paper we assumed # ¼ 1.
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2203

By substituting Eq. (20) into Eq. (17) we obtain the incremental stress–strain relationship:
0 1
¼0
z}|{
B P nnT P C
nf þP
B C vp
dr ¼ BP  T T m Cde ¼ D de ð21Þ
@ n Pn þ h þ #Dt A

in which the term P nf vanishes because, on a global level, it is assumed that f ¼ 0.


The tangential stiffness matrix Dvp is now given by

nnT P
P
Dvp ¼ P  m : ð22Þ
nT P
nT þ h þ #Dt

The term hvp ¼ h þ #Dt


m
in Eqs. (20)–(22) is greater than zero if

m
Dt < DTlim ¼  ; ð23Þ
#h

that is (assuming # ¼ 1) the condition presented by Simo and Hughes [26], while, with m ¼ 0, the algorithm
results in the classical return mapping algorithm presented by the same authors. The implicit stress-update
algorithm is presented in Table 1.
Recently, a modified version of the consistency model has been used by Winnicki et al. [32].

Table 1
Stress update algorithm at time t (u: displacement vector, B: stress–displacement matrix, rtr : trial stress)
De ¼ BDu; rtr ¼ rt þ De De:
If f ðrtr ; kt ; k_ t Þ P 0 then plastic state. Set
ð0Þ ð0Þ ð0Þ ð0Þ ð0Þ
Dk tþDt ¼ 0; Dk_ ¼ 0; r
tþDt ¼ rtr ;
tþDt ftþDt ¼ f tr ; k_ tþDt ¼ 0:

(H) Start local iteration:


" #1
ðiÞ
e 1 oq
P ðiÞ
¼ ðD Þ þ #Dtk_ ðiÞ ;
or
 ðiÞ
_ oq þ k_ oq ;
nðiÞ ¼ q þ #kDt

ok ok_
f ðiÞ
Dkðiþ1Þ ¼ DkðiÞ þ m ;
nT Pn þ h þ #Dt
ðiþ1Þ
ðiþ1Þ Dk 1#_
k_ tþDt ¼ þ kt ;
#Dt #
h i
Devp ¼ ð1  #Þ_evp _ ðiþ1Þ ðiÞ
t þ #ktþDt qtþDt Dt;

ðiþ1Þ
rtþDt ¼ rt þ De ½De  Devp ;
 
ðiþ1Þ ðiþ1Þ
f ðiþ1Þ ¼ f rtþDt ; kt þ Dkðiþ1Þ ; k_ tþDt ;

if jf ðiþ1Þ j > d i i þ 1 go to ðHÞ


else elastic state: rtþDt ¼ rtr .
2204 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

4. Numerical results

4.1. General

All the numerical calculations have been carried out with the consistency model by using four-noded
quadrilateral elements (plane strain condition) with four Gauss integration points. A Newmark time
integration scheme (time integration parameter b ¼ 0:3025 and c ¼ 0:60) and the Rankine loading criterion
have been used together with the assumption of linear softening (so that h ¼ ft;s =ju ). The discretisation is
related to the specimen only: the bars of the Split-Hopkinson set-up are left out of consideration.
The strain rate is applied at the right end of the model imposing a displacement uL ðtÞ that is a linear
function of time. Under the assumption of small displacements and strains, the strain rate (constant) is
given by e_ L ¼ L1 ouotL ðtÞ, where L is the specimen length. It implies that the acceleration at the right end is equal
to zero.
This hypothesis is simplified: in some papers [5,29] strain vs. time diagrams are presented and all of them
show that the measured strain rate is not linear (directly from the beginning of the load application). We
used such a simplified assumption since detailed information on the applied loading rates in the tests
mentioned in Fig. 2 was not available.

4.2. Assessment of the model

In this section the capabilities of the numerical model discussed above are explored. Instead of focusing
our attention on a particular test, we examine the behaviour of the model over a large range of strain rates
using ‘‘average’’ material properties to obtain data comparable with the results plotted in Fig. 2. First, a
constant viscosity parameter is used and, later, a strain rate dependent viscosity is taken into account in
order to improve the results.
The material constants are presented in Table 2. The geometric and material parameters are as specified
in [36] (length L ¼ 100 mm). The specimen is modelled by using 80 elements. The numerical results obtained
with m ¼ 1:60 N s/mm2 are presented in Fig. 3. It is clear that the consistency model is not able to
approximate the highly scattered envelope of the experimental results over the whole range of strain rates.
The numerical results are reasonably good for high strain rates (101 s1 ), but are unsatisfactory for lower
values: it is necessary to assume that the viscosity m is a function of strain rate e_ L . Following this approach,
by changing m until a good agreement with the average increase in strength 3 is reached, we obtain the
results listed in Fig. 4 for three different values of ju .
A good approximation (shown to be nearly independent of ju ) is expressed by the equation
m ¼ að_eL Þb ð24Þ
with a  8:43 and b  0:75 (see Fig. 4), or, in log–log form:
log m ¼ c1 þ c2 log e_ L ð25Þ
with c1 ¼ log a  0:92 and c2 ¼ b  0:75.
This relationship can be assumed as a constitutive relationship for problems dominated by tension and
can be included in a Finite Element code. It is clear that the assumption m ¼ constant, along a wide range
of strain (or stress) rates, is not acceptable.

3 ft;d
With the term ‘‘average increase in strength’’ we indicate the mean value of ft;s corresponding to various values of e_ obtained from
5 1 4 1 3 1 2
Fig. 2, i.e., 1.10 for e_ ¼ 10 s , 1.20 for 10 s , 1.40 for 10 s , 1.60 for 10 s1 , 1.80 for 101 s1 , 2.20 for 100 s1 and 4.00 for
101 s1 .
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2205

Table 2
Average material properties
E (N/mm2 ) m (–) ft;s (N/mm2 ) . (N s2 /mm4 ) h (N/mm2 )
25,000 0.10 3.50 9
2.4 · 10 )350

4.5
Birikimer and Lindemann 1971
Kormeling (5_80_3) 1980
Hatano 1 1960
Hatano 2 1960
4.0 Hatano 3 1960
Reinhardt et al. (w/c= 0.46, wet) 1990
Reinhardt et al. (w/c= 0.46, dry) 1990
Soroushian et al. 1986
Rossi et al. (w/c= 0.50, wet) 1994
3.5
Strength increase ft,d / ft ,s (_)

Rossi et al. (w/c= 0.50, dry) 1994


Rossi et al. (w/c= 0.30, wet) 1994
Rossi et al. (w/c= 0.70, wet) 1994
Zielinski 1981
Weerheijm 1992 (notch./unnotch. spec.)
3.0 Cadoni et al. (5mm) 2001
Cadoni et al. (10mm) 2001
Cadoni et al. (dry) 2001
Cadoni et al. (curing) 2001
Cadoni et al. (wet) 2001
2.5 Oh curve 1987
Consistency model, m=0.80
Consistency model, m=1.60
–2
Consistency model, k u=1 10
–3
2.0 Consistency model, k u=1 10
–4
Consistency model, k u=5 10

1.5

1.0
–6 –4 –2 0 2
10 10 10 10 10
–1
Strain rate (s )
Fig. 3. Strength increase vs. strain rate (experimental and numerical results).

Similar results were proposed by Bicanic and Zienkiewicz [2] for the experimental results of Hatano in
tension. They proposed a fluidity parameter g (used in the Perzyna model and comparable to the reciprocal
value of m) that is an increasing function of elastic strain rate.

4.2.1. Mesh sensitivity


In this section the mesh sensitivity of the consistency model is analysed. The same procedure presented in
Section 4.2 is applied to a fine mesh (160 elements); the results are presented in Fig. 5. The picture shows a
very small difference between the results obtained using the fine and the coarse mesh; a similar indepen-
dence of mesh size is also obtained with a 100 elements mesh (20 rows and 5 columns). The results obtained
in terms of the constitutive relationship m  e_ L are hence found to be mesh-independent.

4.3. Notched specimen

Specimen dimensions are B ¼ 50 mm, L ¼ 100 mm and T ¼ 80 mm, while notch depths are 5 and 7 mm
[31], see Fig. 6. We examined specimens A5, A7, B0, B5, B7 (‘‘A’’ and ‘‘B’’ refer to the concrete mix, ‘‘5’’
and ‘‘7’’ to the notch depth; ‘‘0’’ is the unnotched specimen). The material constants are presented in
2206 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

4
10 –2
ku=1 10 , 80 el.
–3
ku=1 10 , 80 el.
–4
ku=5 10 , 80 el.
(–0.75)
3 y=8.43 x
10
Viscosity m (N/mm2 s)

2
10

1
10

0
10 –4 –3 –2 –1 0 1
10 10 10 10 10 10
–1
Strain rate (s )
Fig. 4. Viscosity parameter m vs. strain rate (log–log scale).

4
10 –2
ku=1 10 , 80 el.
–2
ku=1 10 , 160 el.
–3
ku=1 10 , 80 el.
–3
3 ku=1 10 , 160 el.
10 –4
ku=5 10 , 80 el.
Viscosity m (N/mm2 s)

–4
ku=5 10 , 160 el.

2
10

1
10

0
10 –4 –3 –2 –1 0 1
10 10 10 10 10 10
–1
Strain rate (s )
Fig. 5. Viscosity parameter m vs. strain rate (log–log scale): comparison between fine and coarse mesh.

Table 3, while the stress rates and the experimental results are presented in Table 4. It is shown that the
influence of the notches on dynamic strength is significantly more important for high quality concrete than
low quality concrete.
In the numerical simulations the right end of the specimen has been loaded with an imposed dis-
placement; for this reason, the stress rates r_ L of Table 4 are converted to the strain rates e_ L of Table 5
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2207

Fig. 6. Specimens A5 and B5 (reprinted from Weerheijm [31]).

Table 3
Material properties (Mix A: high quality concrete, Mix B: low quality concrete)
Mix E (N/mm2 ) m (–) ft;s (N/mm2 ) . (N s2 /mm4 )
A 33,900 0.10 2.98 2.4 · 109
B 30,900 0.10 1.63 2.4 · 109

Table 4
Loading rates and experimental results
Case r_ L (N/(mm2 ms)) ft;d (N/mm2 ) ft;d =ft;s (–)
A5 17.7 4.44 1.49
A7 12.0 3.93 1.32
B0 10.0 2.69 1.65
B5 13.7 2.45 1.50
B7 13.1 2.74 1.68

by using Eq. (1). The level of lateral compression is taken equal to zero. The values of m from Eq. (24)
have been used, while, for both concrete mixes, the same value of ju (equal to 3.5 · 104 ) has been
chosen.
The notched specimens are modelled with 384 elements (see Fig. 7), while the unnotched specimen is
discretised with 50 elements (5 rows and 10 columns).
A comparison between Tables 4 and 5 shows that the numerical results, in terms of the ratio between
dynamic and static strength, are in quite good agreement with the experimental ones. Following the
description of the experimental results given by Weerheijm [31], it is noted that the stresses (calculated on
the basis of the axial deformation of the aluminium bar) are plotted against the total deformation of the
specimen (see Figs. 8, 10, 12, 14 and 16). To obtain a meaningful comparison, the numerical curves are
plotted, on the same figures, in terms of stresses at the right boundary vs. displacements at the same
boundary. The experimental stress–displacement curves show a steeper softening branch than the numerical
2208 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

Table 5
Numerical parameters and results
Case e_ L (s1 ) ju (–) ft;d (N/mm2 ) ft;d =ft;s (–)
4
A5 0.590 3.5 · 10 4.25 1.43
A7 0.400 3.5 · 104 3.60 1.21
B0 0.333 3.5 · 104 2.57 1.58
B5 0.457 3.5 · 104 2.77 1.70
B7 0.437 3.5 · 104 2.66 1.63

Fig. 7. Finite element mesh (384 elements), notch depth 5 mm.

ones. We remind that the simulations presented here are related to the specimen only; we believe that the
simulation of the behaviour of the specimen with the bars of the Split-Hopkinson apparatus may give
results closer to the experimental values. Finally, Figs. 11, 13, 15 and 17 show the equivalent plastic strain j
at the end of the numerical computation (t ¼ 1  102 s for the Mix A and t ¼ 8  103 s for the Mix B).
The analyses shown in the different figures were carried out with different material properties and
loading rates. For instance, the comparison of the maximum equivalent plastic strain (Figs. 9 and 11) must
take into account the different loading rate (r_ ¼ 17:7 and 12 N/(mm2 ms)) and the different notch depth. In
other words, these results are not directly comparable.
In order to examine the influence of notch depth (from a numerical point of view), we ran the case A7
with the same strain rate as case A5. In this way, the numerical results are comparable. We found a value of

4.00
Stress (N/mm )
2

Numerical results
Experiments
2.00

0
0 0.02 0.04 0.06 0.08 0.10
Displacement (mm)

Fig. 8. Stress vs. displacement for case A5.


F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2209

0.008656
0.008122
0.007588
0.007054

EqPlSt
0.00652
0.005986
0.005452
0.004918 0
0.0075 10
0.004384
0.005 20
0.00385 30
0.003316 40
0.0025
0.002782 50
0.002248 60 m)
0.001714 -20 70 x (m
0.00118 -10 80
y (m 0 90
m) 10 100
20 110

Fig. 9. Equivalent plastic strain profile for case A5.

4.00
Stress (N/mm2)

Numerical results

2.00 Experiments

0
0 0.02 0.04 0.06 0.08 0.10
Displacement (mm)

Fig. 10. Stress vs. displacement for case A7.

0.0124692
0.0116384
0.0108076
0.00997675
EqPlSt

0.00914593
0.00831512
0.00748431
0.00665349 0
0.01 10
0.00582268
0.00499187 20
0.00416106 0.005 30
40
0.00333024 50
0
0.00249943 60 m)
0.00166862 -20
-10
70 x (m
0.000837803 80
y (m 0 90
m) 10 100
20 110

Fig. 11. Equivalent plastic strain profile for case A7.


2210 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

4.00

Stress (N/mm 2)
Numerical results
Experiments
2.00

0
0 0.02 0.04 0.06 0.08 0.10
Displacement (mm)

Fig. 12. Stress vs. displacement for case B0.

0.000752375
0.00074775
0.000743125
0.0007385
0.000733875
EqPlSt

0.00072925 0
0.000724625 10
20
0.00072 30
0.00075
0.000715375 40
0.00071075 0.0007 50
0.000706125 60 m)
x (m
0.0007015 -20 70
0.000696875 -10 80
0.00069225
y (m 0 90
m) 10 100
0.000687625 20 110

Fig. 13. Equivalent plastic strain profile for case B0.

4.00
Stress (N/mm2)

Numerical results
Experiments
2.00

0
0 0.02 0.04 0.06 0.08 0.10
Displacement (mm)

Fig. 14. Stress vs. displacement for case B5.

ft;d  3:30 N/mm2 (ft;d =ft;s ¼ 1:11): with the same loading rate, the more damaged specimen (A7) shows a
lower carrying capacity.
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2211

0.00457501
0.00427001

EqPlSt
0.00396502
0.00366003
0.00335503
0.00305004
0.005
0.00274504
0.004 0
0.00244005 10
0.00213506 0.003
20
0.00183006 0.002 30
0.00152507 0.001 40
0.00122008 0 50
0.000915082 60 m)
0.000610088
-20
-10
70 x (m
0.000305095 80
y (m 0 90
m) 10 100
20 110

Fig. 15. Equivalent plastic strain profile for case B5.

4.00
Stress (N/mm2)

Numerical results
Experiments
2.00

0
0 0.02 0.04 0.06 0.08 0.10
Displacement (mm)

Fig. 16. Stress vs. displacement for case B7.

0.00425625
0.00397251
0.00368876
EqPlSt

0.00340502
0.00312127
0.00283753
0.00255378 0.004
0.00227004 0
0.003 10
0.00198629 20
0.00170255 0.002
30
0.0014188 0.001 40
0.00113506 0 50
0.000851312 60 m)
0.000567567
-20
-10
70 x (m
0.000283822 80
y (m 0 90
m) 10 100
20 110

Fig. 17. Equivalent plastic strain profile for case B7.


2212 F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213

5. Conclusions

• The model presented here is able to describe high loading rate problems. Tension tests were carried out in
order to determine the parameters of the model.
• By choosing the value of viscosity m as a function of the strain rate, it is possible to obtain a good agree-
ment in terms of the ratio between dynamic and static strength over a wide range of strain rates. At the
same time, this approach makes it possible to obtain the constitutive relationship m–_e necessary to per-
form realistic simulations of high loading rate tests.
• The consistency model does not suffer from mesh dependence when a softening stress–strain constitutive
law is used. This is due to the viscoplastic regularisation offered by adding viscosity to the constitutive
relationship.
• At this stage, the model is not able to explain the different sensitivity found in tension and compression
tests (as shown in Figs. 1 and 2) neither the increase in the elastic modulus observed experimentally. This
is one of the aspects to be improved. The capabilities of the model in terms of size-effects on the consti-
tutive relationship (Eq. (25)) will also be investigated.

Acknowledgements

The author wishes to thank professors Alberto Carpinteri, Rene de Borst and Bert Sluys who made this
cooperation possible. Financial support from the European Community (contract ERBFMRXCT 960062,
‘‘Spatio-temporal instabilities in deformation and fracture’’) is gratefully acknowledged.

References

[1] Bazant ZP, Oh BW. Strain-rate effect in rapid triaxial loading of concrete. J Engng Mech (ASCE) 1982;108(EM5):764–82.
[2] Bicanic N, Zienkiewicz OC. Constitutive model for concrete under dynamic loading. Earthquake Engng Struct Dyn 1981;11:689–
710.
[3] Birikimer DL, Lindemann R. Dynamic tensile strength of concrete materials. Am Concr Inst J 1971:47–9.
[4] Bischoff P, Perry SH. Impact behaviour of plain concrete loaded in uniaxial compression. J Engng Mech (ASCE) 1995;121(6):685–
93.
[5] Cadoni E, Albertini C, Labibes K, Solomos C. Behavior of plain concrete subjected to tensile loading at high strain rate. In: de
Borst R, Mazars J, Pijaudier-Cabot G, van Mier JGM, editors. Fracture mechanics of concrete structures. Amsterdam: Balkema
Publishers; 2001. p. 341–7.
[6] Dube J, Pijaudier-Cabot G, La Borderie C, Glynn J. Rate dependent damage model for concrete––wave propagation and
localisation. In: Mang H, Bicanic N, de Borst R, editors. Computational modelling of concrete structures (EURO-C 1994).
Swansea: Pineridge Press Limited; 1994. p. 313–22.
[7] Duvaut G, Lions JL. Les inequations en mecanique et en physique. Paris: Dunod; 1972.
[8] Gatuingt F, Pijaudier-Cabot G. Computational modelling of concrete structures subjected to explosion and perforation. In: VI
International Conference on Computational Plasticity (COMPLAS), 11–14 September 2000, Barcelona. Barcelona: CIMNE;
2000.
[9] Gran JK, Florence AL, Colton JD. Dynamic triaxial tests of high-strength concrete. J Engng Mech (ASCE) 1989;115(5):891–
904.
[10] Hatano T. Dynamical behaviour of concrete under impulsive tensile loads. Technical Report C-6002. Tokyo: Central Research
Institute of Electric Power Industry; 1960.
[11] Kormeling HA. Strain Rate and Temperature Behaviour of Steel Fibre Concrete in Tension. PhD thesis, Delft University of
Technology; 1986.
[12] Krauthammer T, Shanaa HM, Assadi-Lamouki A. Response of reinforced concrete structural elements to severe impulsive load.
Comput Struct 1994;53(1):119–30.
[13] Loret B, Prevost JH. Dynamic strain localization in elasto-(visco-)plastic solids, Part 1. General formulation and one-dimensional
examples. Comput Meth Appl Mech Engng 1990;83:247–73.
F. Barpi / Engineering Fracture Mechanics 71 (2004) 2197–2213 2213

[14] Needleman A. Material rate dependence and mesh sensitivity in localization problems. Comput Meth Appl Mech Engng
1989;67:69–85.
[15] Oh BW. Behaviour of concrete under dynamic tensile loads. Am Concr Inst J 1987:8–13.
[16] Oritz M, Pandolfi A. A class of cohesive elements for the simulation of three-dimensional crack propagation. Int J Solids Struct
1998;44:1267–82.
[17] Perzyna P. Fundamental problems in viscoplasticity. In: Recent advances in applied mechanics, vol. 9. New York: Academic
Press; 1966. p. 243–377.
[18] Prevost JH, Loret B. Dynamic strain localization in elasto-(visco-)plastic solids, Part 2. Plane strain examples. Comput Meth Appl
Mech Engng 1990;83:275–94.
[19] Reinhardt HW. Concrete under impact loading. Tensile strength and bond. Heron, p. 1–48.
[20] Reinhardt HW, Kormeling HA, Zielinski AJ. The Split Hopkinson bar, a versatile tool for the impact testing of concrete. Mater
Struct 1986;19/109:55–63.
[21] Reinhardt HW, Rossi P, van Mier JGM. Joint investigation of concrete at high rates of loading. Mater Struct 1990;23:213–6.
[22] Rossi P, Boulay C. Influence of free water in concrete on the cracking process. Mater Struct 1990;42(152):142–6.
[23] Rossi P, van Mier JGM, Toutlemonde F, Le Maou F, Boulay C. Effect of loading rate on the strength of concrete subjected to
uniaxial tension. Mater Struct 1994;27:260–4.
[24] Sercombe J. Modelisation du Comportament du Beton en Dinamique Rapide. PhD thesis, Ecole Nationale de Fonts et Chaussees;
1997.
[25] Sercombe J, Ulm FJ, Toutlemonde F. Viscous hardening plasticity for concrete in high-rate dynamics. J Engng Mech (ASCE)
1998;124(9):1050–7.
[26] Simo JC, Hughes TJR. In: Computational inelasticity [Interdisciplinary Applied Mathematics], vol. 7. New York: Springer-
Verlag; 1998.
[27] Sluys LJ. Wave Propagation, Localisation and Dispersion in Softening Solids. PhD thesis, Delft University of Technology; 1992.
[28] Soroushian P, Choi KB, Fu G. Tensile strength of concrete at different strain rates. In: Mindess S, Shah SP, editors. Cement-based
composites: strain rate effects on fracture, vol. 64. USA: MRS Publication; 1986. p. 87–92.
[29] Tang T, Malvern LE, Jenkins DA. Rate effects in uniaxial dynamic compression of concrete. J Engng Mech (ASCE)
1992;118(1):108–24.
[30] Wang W. Stationary and Propagative Instabilities in Metals––a Computational Point of View. PhD thesis, Delft University of
Technology, 1997.
[31] Weerheijm J. Concrete Under Impact Tensile Loading and Lateral Compression. PhD thesis, Delft University of Technology;
1992.
[32] Winnicki A, Pearce CJ, Bicanic N. Viscoplastic Hoffman consistency model for concrete. Comput Struct 2001;79:7–19.
[33] Zielinski AJ. Fracture of Concrete and Mortar Under Uniaxial Impact Tensile Loading. PhD thesis, Delft University of
Technology; 1982.
[34] Zielinski AJ. Compression under biaxial loading: static compression–impact tension. Technical Report 5-85-1. The Netherlands:
Stevin Laboratories, Delft University of Technology; 1985.
[35] Zielinski AJ, Reinhardt HW, Kormeling HA. Experiments on concrete under repeated uniaxial impact tensile loading. Mater
Struct 1981;14(81):163–9.
[36] Zielinski AJ, Reinhardt HW, Kormeling HA. Experiments on concrete under uniaxial impact tensile loading. Mater Struct
1981;14(80):103–12.

You might also like