International Journal of Impact Engineering 36 (2009) 283–293

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

Numerical assessment for impact strength measurements in concrete materials

Werner Riedel a, *, Nobuaki Kawai b, Ken-ichi Kondo b
a
Fraunhofer Institut für Kurzzeitdynamik, Ernst-Mach-Institut, Eckerstraße 4, D-79104 Freiburg, Germany
b
Materials & Structures Laboratory, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8503, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Measurement of dynamic strength of concrete at impact relevant strain rates and pressures is the pur-
Received 27 June 2007 pose of the described study. Therefore, an experimental design of direct planar impact experiments with
Received in revised form 19 December 2007 longitudinal and transverse strain gauges is analyzed in predictive hydrocode simulations using an
Accepted 19 December 2007
elastic–plastic damage model for concrete. The calculations and first experimental results on mortar
Available online 10 May 2008
show decreasing phase velocities of stress waves both in longitudinal and lateral gauges. The model
clearly associates it with the onset of damage, possibly interpreted as a failure wave. Numerical analysis
Keywords:
is furthermore used to compare a monolithic target block to a thoroughly assembled concrete sample in
Concrete strength
Strain rate
order to include flat gauges in the material. The planned experimental procedure to derive wave speeds,
Failure wave particle velocities and strain rates from stress measurements is anticipated and validated on the basis of
Damage simulated gauge signals. The most important finding is the prediction and first experimental confir-
mation that concrete ultimate strength and damaged yield stress can be derived at strain rates in the
order of 104/s from the proposed type of experiments. This technique promises new insight into the
strength and failure processes of concrete in the challenging loading region around the characteristic
minimum of the shock particle velocity relationship.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction measurements and scattering are very challenging around this

1.1. Dynamic strength analysis and modeling approaches
Analysis and modeling of concrete impact behavior has been an
active research area over the last decades. A number of authors
have measured the longitudinal stress wave behavior ranging from
low velocities up to strong shock loading as high as 20 GPa pressure
and strain rates of 106/s. Fig. 1 gives an overview of available lit-
erature data by Gregson [14], Grady [11,12,13], Kipp and Chhabildas
[24], Ockert [28], Ishiguchi [21], Tsembelis [36,37,38] Gebbeken [9],
Riedel [31,32]. The dashed lines point out the remarkable drop in
the compression wave velocity Us, or better phase velocity cp [41],
from particle velocities up between 50 and 300 m/s. The reduction
from the acoustic 1D strain longitudinal sound speeds above
4000 m/s to less than 2000 m/s for conventional strength concrete
[24,32] is generally assigned to the compaction of the concrete
pores and the onset of damage. Higher loading intensities (up above
>400 m/s) override the energy consumption of these processes. As
the impedances of aggregate and mortar differ most in this regime,
* Corresponding author. Tel.: þ49 7628 9050 692; fax: þ49 7628 9050 677.
E-mail address: riedel@emi.fraunhofer.de (W. Riedel).
minimum.
Moreover, concrete exhibits a complex distortional behavior.
Fig. 1, right, shows equivalent stresses at failure loads as function of
hydrostatic pressures p for four different concrete qualities be-
tween 35 and 140 MPa compressive strength. Eq. (1) defines the
equivalent stress seq as function of the principal stresses s1, s2, s3
for general load cases. In the special case of cylindrical symmetric
loading with s2 ¼ s3, it is equal to the stress difference Ds.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1h i cyl: symm:
seq ¼ ðs  s2 Þ2 þðs2  s3 Þ2 þðs3  s1 Þ2 ¼ js1  s2 j
2 1
¼ Ds ¼ 2s (1)
1
p ¼  ðs1 þ s2 þ s3 Þ (2)
3
Before failure initiation inelastic stresses start for conventional
concrete at about 85% of the strength in uniaxial tension and above
35% in uniaxial compression. Irreversible pore compaction is ob-
served under hydrostatic loading. After failure shear stresses are
only supported under confined conditions but are still substantial,
for example, during penetration into thick concrete targets [17]. A
number of models (e.g. Chen [5], Riedel [31], Itoh et al. [20], Geb-
beken and Ruppert [10], Malvar et al. [25]) describe elasticity,
hardening up to failure and post-failure shear resistance by three

0734-743X/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijimpeng.2007.12.012
284 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293

Fig. 1. Left: shock properties of concrete showing a decisive minimum in stress propagation speed at up ¼ 250 m/s. Data ‘Riedel 03 Conventional’ [31] used in Sections 3 and 4.
Right: static distortional strength as a function of pressure.

limit surfaces, as shown in Fig. 2. A point of discussion remains the
closure of the elastic surface towards high hydrostatic pressures by
‘cap’ formulations.
Strain rate dependence on these limit surfaces is so far mostly
measured along the lines of uniaxial compressive and tensile
stresses seq ¼ 3p (dashed lines in Fig. 4, right). Dynamic experi-
ments loading concrete specimens in dynamic uniaxial compres-
sion can reach strain rates of several 100/s with strength increase
factors up to 2.5 (Bischoff [1]). Split Hopkinson-Bar measurements
on scaled concrete or mortar can reach 1700/s and increase factors
up to 4 (Grote et al. [15]). The strain rate dependence in uniaxial
tension, for example using a Hopkinson-Bar in spallation configu-
ration, is even stronger with dynamic strength increase factors of
4–12 at maximum strain rates from 50 to 120/s (Klepaczko [23],
Schuler et al. [33]).
Shock waves generate strain rates from 104 to 106/s. Dynamic
shear strength of brittle materials under shock loading has been
measured and simulated mainly for glass and ceramics by a number
of authors, e.g. Kanel et al. [22], Brar et al. [4], Clifton [6], Espinosa
et al. [7,8], Millet et al. [26], Bourne et al. [3]. By mounting flat stress
gauges in normal and lateral direction into specimens subjected to
plate impact, shear stresses s of the materials are measured. Planar
stress and shock waves are compressing in uniaxial strain with
seq ¼ 3p(1  2n)/(1 þ n) in the elastic regime (broken line in Fig. 2,
right). Higher loading with permanent deformation causes
deviation from this line along the failure surface towards higher
pressures. Time delayed failure has been shown to result in a two-
wave structure referred to as ‘failure wave’, which is especially
apparent in the lateral stress signal. Rise in longitudinal strain has
been observed with the passing of the failure front, while lateral
strains remain negligible [26]. A distinct influence on the propa-
gation of the failure wave by additional target assembly surfaces
was reported by Bourne et al. [3].
Micro-concrete and mortar materials have been studied with
stress gauges in the material by Tsembelis et al. [37,39] and Grote
et al. [16] to derive deviatoric strength and failure. Grote reports the
limit of elastic longitudinal and transverse stresses to be at 167 m/s
for PMMA plate impact which corresponds to 130 MPa compressive
stress. He shows that stronger loading does not provide shear
stresses above the theoretical elastic limit. Tsembelis measures the
stress differences Ds from 0.4 to 1.45 GPa for higher longitudinal
stresses from 0.8 up to 6 GPa. They are almost identical for micro-
concrete and cement paste.
In this work the impact shear strength analysis method is ap-
plied to large mortar and concrete samples in order to measure
strength and failure at very high strain rates. The 200 mm caliber
allows parallel use of axial and lateral stress gauges at the same
distance from the impact plane. Numerical simulation is used be-
fore and after testing to predict and analyze stress and damage
evolution in monolithic and assembled concrete targets.

Fig. 2. Three surface concept for the concrete strength with hardening, failure and residual friction resistance.
 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293 285

1.2. Configuration referential (moving with the material) and solved together with the
equation of state (Eqs. (6) and (11)) linking pressure p, density r and
The experimental configuration under development by Kawai internal energy e. Stresses and strains are decomposed into hy-
et al. [41] is shown in Figs. 3 and 4. A U-shaped aluminium pro- drostatic pressure p and deviatoric stresses Sij. In Eq. (4a) com-
jectile (diameter 200 mm, bottom plate thickness 25 mm, collar pressive pressures and tensile stresses sij, Sij are defined positive.
thickness 15 mm, overall length 100 mm) is impacted with its flat The Jaumann tensor of deviatoric stress rates (Eq. (4b)) is used to
bottom onto a concrete block of overall dimension L ¼ 100 mm, integrate the strength portion of material behavior for small strain
B ¼ 100 mm, H ¼ 185 mm. In the present paper impact velocities of increments d3ij/dt (Eq. (5a)) per time step. Stress contributions from
170 m/s are considered, the impact facility is designed to allow up rotation rates drij/dt (Eq. (5b)) are eliminated to obtain an objective
to 280 m/s. Polyvinylidene fluoride (PVDF) stress gauges 1–3 with material description, independent of the choice of coordinate sys-
measurement areas of 6  6 mm are placed in longitudinal di- tem and rigid body motions. Explicit time integration fulfilling the
rection at 10, 20 and 30 mm beneath the impact surface along the Courant–Friedrich–Levy criterion yields stability of the numerical
axis of impact. Gauge 4 measures transverse stresses in the same scheme. This combination of stress and strain referentials and time
plane as number 1 but is located 70 mm off the symmetry axis. The integration is often called ‘update Lagrange–Jaumann scheme’ and
concrete target is assembled from a number of precisely ground allows consistent large strain material description when integrated
blocks and tiles to mount the gauges between them. Tiles Ia, Ib, Ic over several time steps. Coordinates xi, velocities vi and accelera-
measure each 10  100  100 mm and are backed by block II tions are nodal variables in the computational grid, while material
(70  100  100 mm). Blocks III (100  100  20 mm) and IV properties such as mass, pressure, deviatoric stresses, internal en-
(100  100  65 mm) are attached below the resulting cube. The ergy are evaluated in the cell centre.
blocks are precisely rectified and connected with a fine film of
epoxy resin thinner than 200 mm. An additional block V of concrete Dr vv
þr i ¼ 0 (3a)
is added onto the assembly. Dt vxi
The following questions arise in context with the experimental
configuration:
Dvi vp vSji
1. Does the assembly of the concrete target influence the mea- r ¼  þ þ rfi (3b)
Dt vxi vxj
surement of longitudinal and transverse stresses in comparison
to a monolithic sample?
2. When and to which extent does the assembly alter the damage
evolution and propagation? De vv vv
r ¼ rq_  p i þ Sji i (3c)
3. Can it be assumed that the lateral stress measurements in Dt vxi vxj
gauge position 4 are the same as in position 1, where the lon-
gitudinal stress is taken?
4. Which interpretation will be possible from the stress gauge  
1
data? sij ¼ Sij  pdij S_ ij ¼ Sij r_ jk þ Skj r_ ik þG 3_ ij  dij 3_ kk (4a,b)
3
rotation correction
To clarify these points the experimental configuration is nu-
merically modeled in the following.
! !
1 vvi vvj 1 vvi vvj
3_ ij ¼ þ r_ ij ¼  (5a,b)
2. Numerical simulation approach 2 vxj vxi 2 vxj vxi

2.1. Lagrangian hydrocode

Basis of the simulations is the finite difference scheme of the
commercial hydrocode AUTODYN-V6.0 [27], which also in-
corporates finite element and smooth particle hydrodynamics
discretisations. The differential Eqs. (3a)–(3c) are derived from
conservation of mass, momentum and energy in Lagrangian
2.2. Constitutive models for aluminium and epoxy
Both materials are described using an equation of state in Mie–
Grüneisen form (Eq. (6a)) with a Hugoniot reference curve pH(r)
and the Grüneisen coefficient G ¼ V(vp/ve)V to extrapolate to

Fig. 3. Experimental configuration of single stage accelerator and concrete sample at TokyoTech.
286 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293

Fig. 4. Sample geometry with concrete block structure containing longitudinal and transverse gauges. Cylindrical symmetric model of impactor and concrete sample. Mesh res-
olution of 0.2 mm describing the adhesive layer for assembled sample.

rffiffiffiffiffiffiffiffiffiffiffiffiffi
general material states (V denotes the volume). The Aluminium
p vf 2 p p spij seeq  Y
Hugoniot curve is expressed in terms of shock velocity Us and d3ij ¼ dl 3peq ¼ 3 3 D3peq ¼ ¼
vsij 3 ij ij 3G 3G
particle velocity up using the linear approximation with the mate-
rial parameters bulk sound speed cB and slope S (Eq. (6b)). A (9a,b,c)
polynomial formulation (Eq. (6c)) is chosen for the epoxy. The " 0
! 0
! #
Yp p GT n
equation of state parameters are summarized in Table 1, left Y ¼ Y0 1 þ þ ðT  300Þð1 þ b3Þ (10)
columns. Y0 h1=3 G0
  
1 r
pðr; eÞ ¼ Gre þ pH 1  G 1 (6a) Elastic stresses are limited by the flow rule (Eq. (8)) based on the
2 r0
von Mises equivalent stress seq (Eq. (1)). The flow rule (Eq. (9a)) is
associated with the flow surface, so that plastic strain increments
hð1 þ hÞ are calculated using ‘radial return’ (Eq. (9c)). Steinberg and Gui-
pH ðhÞ ¼ r0 c2B pH ðhÞ nan’s model describes the aluminium flow surface depending on
½1  hðS  1Þ2
r isotropic strain hardening, pressure and temperature (Eq. (10)). All
¼ A1 h þ A2 h2 þ A3 h3 with h ¼ 1 (6b,c) material parameters are summarized in Table 1, right. A constant
r0
yield stress is assumed for the epoxy adhesive.
Elastic stress rates dSij/dt are derived via Hooke’s law in
Jaumann’s formulation (Eq. (4b)) from strain and rotation rates
(Eqs. (5a,b)). Steinberg et al.’s [34,35] model describes the alu- 2.3. Constitutive model for concrete
minium with the shear modulus G depending on pressure and
temperature. A constant shear modulus is assumed for epoxy. 2.3.1. Equation of state
" ! # The concrete equation of state for the pore free matrix material
 0
G0p p GT is based on the Mie–Grüneisen form (Eq. (6a)) with a polynomial
G ¼ G0 1 þ ßþ ðT  300Þ (7)
G0 h1=3 G0 Hugoniot curve (Eq. (6c)). Herrmann’s model [18] describes po-
rosity a ¼ rmatrix/rporous as an additional state variable (Eq. (11a)).
      The compaction path (Eq. (11b)) is defined using the initial porosity
f sij ¼ F sij  Y ¼ 0 F sij ¼ seq (8) ainit, pore crush pressure pel, lockup pressure pcomp and exponent N.

Table 1
Equation of state and strength parameters to describe the Al2024-T4 projectile and the epoxy adhesive

Al2024-T4 [34,35]
Equation of state Shock Shear modulus 2.86000Eþ07 (kPa)
Reference density 2.78500Eþ00 (g/cm3) Yield stress 2.60000Eþ05 (kPa)
Gruneisen coefficient 2.00000Eþ00 (–) Maximum yield stress 7.60000Eþ05 (kPa)
Parameter CB 5.32800Eþ03 (m/s) Hardening constant 3.10000Eþ02 (–)
Parameter S 1.33800Eþ00 (–) Hardening exponent 1.85000E01 (–)
Reference temperature 3.00000Eþ02 (K) Derivative dG/dP 1.86470Eþ00 (–)
Specific heat 8.63000Eþ02 (J/kgK) Derivative dG/dT 1.76200Eþ04 (kPa)
Melting temperature 1.22000Eþ03 (K) Derivative dY/dP 1.69500E02 (–)

Epoxy adhesive
Equation of state Polynomial Shear modulus 4.08000Eþ06 (kPa)
Reference density 1.18600Eþ00 (g/cm3) Yield stress 4.00000Eþ04 (kPa)
Bulk modulus A1 8.83900Eþ06 (kPa) Eq. plastic failure strain 3.00000Eþ00 (–)
Parameter A2 1.75500Eþ07 (kPa) Reference temperature 2.93000Eþ02 (K)
Parameter A3 1.51600Eþ07 (kPa)
Parameter B0 1.13000Eþ00 (–)
Parameter B1 1.13000Eþ00 (–)
 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293 287

The employed equation of state data is derived from plate impact positive) with parameters A and N. All measures of hydrostatic
tests in [31,32] and specified in Table 2. pressure and the deviatoric strength are normalized over the
uniaxial compressive strength fc, when denoted with *. Towards
porous
p ¼ f ðrmatrix ; eÞ ! p ¼ f ðra; eÞ with low pressures the exponential form is continued by linear
" #N interpolations from the uniaxial compressive strength (seq*,p*) ¼
pcomp  p (fc/fc,1/3) to shear strength (fs/fc, 0) and uniaxial tensile strength
a ¼ 1 þ ðainit  1Þ ð11a;bÞ
pcomp  pel (ft/fc, ft/3fc) and extrapolated to the intersection HTL* ¼ HTL/fc with
the hydrostatic axis. The auxiliary variable HTL0 * only occurs in
Eq. (13a) to provide continuity of YTXC* at the uniaxial compres-
2.3.2. Failure surface depending on pressure, triaxiality sive strength (seq*,p*) ¼ (fc/fc,1/3). Hydrostatic pressures in ten-
and strain rate sion are limited to HTL* with correction of the internal energy for
The strength model proposed and developed in [30,31] is in- the release of the tensile pressure portion below HTL*. For such
spired by Chen’s [5] concept of three limit surfaces in stress space. states the deviatoric strength is zero, until pressures above HTL*
Fig. 2, left, shows the three surfaces describing the elastic limit Yel, occur. All above and following input parameters are summarized
failure Yfail and residual shear strength Yfric of the damaged con- in Table 2.
crete under confined conditions. Rotation of the compressive meridian around the hydrostatic
      axis spans the complete failure surface in stress space (Fig. 2, left).
f p; seq ; q; 3_ ¼ seq  Yfail p; q; 3_ ¼ seq  YTXC ðpÞR3 ðqÞFRate 3_ ¼ 0 To describe reduced strength on shear and tensile meridians it is
(12a) multiplied with a factor Q2  R3(q)  1 (Fig. 5, right). The lode angle
q describes stress triaxiality and depends on the third invariant J3
    of the stress tensor (Eq. (14b)). The dimensionless function R3
Yfail p; q; 3_ ¼ YTXC ðpÞR3 ðqÞFRate 3_ (12b) (Eq. (14a), [40]) scales referring to the compressive meridian YTXC
with R3  1. The lowest value Q2 > 0.5 is reached on the tensile

n 1 meridian with principal stresses combined as sI > sII ¼ sIII. The
*
YTXC ¼ A p*  HTL0* for p*  (13a) ratio Q2 of tensile to compressive meridian decreases with
3
increasing pressure (from 0.684 to 0.705 for 1/3 < p/fc < 7/3). This
effect is called ‘brittle to ductile transition’ and is described by
*
YTXC ¼ 0 for p* < HTL* (13b)
Eq. (14c).

 h i1
2
2 1  Q22 cos q þ ð2Q2  1Þ 4ð1  Q22 Þcos2 q þ 5Q22  4Q2
R3 ðq; Q2 Þ ¼
 (14a)
4 1  Q22 cos2 q þ ð1  2Q2 Þ2

The failure surface Yfail, described by Eq. (12b) and shown in Figs. pffiffiffi
2 and 5, can be measured on hydraulic machines with triaxial 3 3 J3 27detðSÞ
cos 3q ¼ ¼ (14b)
control; some results are shown in Fig. 1, right. It is therefore taken 2 J 3=2 2s3eq
2
as reference to construct all limit surfaces. The compressive me-
ridian YTXC(p) (Eq. (13)) describes the pressure dependence for
principal stress conditions sI < sII ¼ sIII (tensile stresses defined 0:5 < Q2 ¼ Q2;0 þ Bp*  1 (14c)

Table 2
Employed material data for 37.7 MPa concrete, AUTODYN [27] input to ‘RHT concrete’

Equation of state P alpha Tens./comp. meridian ratio (Q) 6.80500E01 (–)

Reference density 2.75000Eþ00 Brittle to ductile transition 1.05000E02 (–)
Porous density 2.16000Eþ00 (g/cm3) G (elas.)/(elas.–plas.) 2.00000Eþ00 (–)
Porous sound speed 2.92000Eþ03 (m/s) Elastic strength/ft 7.00000E01 (–)
Initial compaction pressure 2.51000Eþ04 (kPa) Elastic strength/fc 5.30000E01 (–)
Solid compaction pressure 6.00000Eþ06 (kPa) Fractured strength constant B 1.60000Eþ00 (–)
Compaction exponent 3.00000Eþ00 (–) Fractured strength exponent m 6.10000E01 (–)
Solid EOS Polynomial Comp. strain rate exp. a 3.20000E02 (–)
Bulk modulus A1 3.52700Eþ07 (kPa) Tensile strain rate exp. d 3.60000E02 (–)
Parameter A2 3.95800Eþ07 (kPa) Failure RHT Concrete
Parameter A3 9.04000Eþ06 (kPa) Damage constant, D1 4.00000E02 (–)
Parameter B0 ¼ G 1.22000Eþ00 (–) Damage constant, D2 1.00000Eþ00 (–)
Parameter B1 ¼ G 1.22000Eþ00 (–) Minimum strain to failure 1.00000E02 (–)
Reference temperature 3.00000Eþ02 (K) Residual shear modulus fraction 1.30000E01 (–)
Specific heat 6.54000Eþ02 (J/kgK) Tensile failure Hydro (Pmin)
Compaction curve Standard Erosion Geometric Strain
Strength RHT Concrete Erosion strain 2.00000Eþ00 (–)
Shear modulus 1.67000Eþ07 (kPa) Tens./comp. meridian ratio (Q2,0) 6.80500E01 (–)
Compressive strength (fc) 3.77000Eþ04 (kPa) Brittle to ductile transition B 1.05000E02 (–)
Tensile strength (ft/fc) 1.00000E01 (–)
Shear strength (fs/fc) 1.80000E01 (–)
Intact failure surface constant A 1.60000Eþ00 (–)
Intact failure surface exponent n 6.10000E01 (–)
288 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293

Fig. 5. Definition of the failure surface by the compressive meridian and the deviatoric section. Data, e.g. by [17].

The term FRate(d3/dt) in Eqs. (12a,b) accounts for the rate en- Tension
hancement of deviatoric strength. CEB-Bulletin 187 [2] proposes
 d
empirical fit functions (Eqs. (15a,b) and (16a,b)) for uniaxial tensile ftd 3_
and compressive loading. In the model the strain rate influence for FRate ¼ ¼ 3_  30 s1 (16a)
ft 3_ 0
hydrostatic pressures above fc/3 is calculated according to Eq. (15a),
for pressures below ft/3 Eq. (16a) is applied. Strain rate enhance-
ment factors FRate between these limits are linearly interpolated. By p ffiffi
ftd
¼ h 3_ 3_ > 30 s1
3
this formulation the complete failure surface (Eq. (12)) is radially (16b)
dilated around the origin (p ¼ 0,seff ¼ 0) for higher strain rates, as ft
indicated by the dashed line and square measurement points in 1
Fig. 5, left. The input parameters for a and d are specified in Table 2. 3_ 0 ¼ 3  106 s1 ; d ¼
10 þ 12 fc
The units in Eqs. (15) and (16) are [MPa] for ft and fc and [1/s] for
strain rates.
Compression
log h ¼ 7d  0:492
 a
f 3_
FRate ¼ cd ¼ 3_  30 s1 ð15aÞ
fc 3_ 0
2.3.3. Elastic limit and hardening
p ffiffi The initial elastic surface Yelastic of the virgin material is derived
fcd 3
¼ g 3_ 3_ > 30 s1 ð15bÞ from the failure surface Yfail using the ratio of elastic tensile and
fc
compressive stress over the respective ultimate strength (ft,el/ft and
1 fc,el/fc). For pressures below ft,el/3ft the elastic scaling function Felastic
3_ 0 ¼ 30  106 s1 ; a ¼ takes the value ft,el/ft. Above fc,el/3fc it is equal to fc,el/fc. Between
5 þ 34 fc
these bonds it is linearly interpolated with respect to the pressure.
log g ¼ 6a  0:492 The elastic surface is closed consistently with the porous equation

Fig. 6. Comparing simulated stress signals in block assembled (light colours) with monolithic target (strong colours). Comparison of lateral measurement in gauge positions 4 and 1.
 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293 289

Fig. 7. Damage plots at 15 ms (scaling 0–0.1) and 80 ms (scaling 0–1.0). Comparison of assembled target (middle) and monolithic concrete block (right) shows good agreement.

of state (Eq. (11)) towards higher pressures involving pore com-

 !
paction using a parabolic cap function Fcap (Eq. (17b)). The upper Yfail  Yelastic Gelastic
pl;hard
3eq ¼ (18b)
cap pressure po at which deviatoric stresses are reduced to zero 3G Gelastic  Gplastic
starts for the virgin material at pel. During continued compaction it
grows consistently with the current pore compaction pressure of Between the initial elastic surface and the ultimate failure sur-
the equation of state. The lower pressure pu for cap influence is fc/3. face hardening is described using Eqs. (18a,b). The loading surface
Fig. 2, right, illustrates the resulting elastic surface. Yhard is scaled between Yelastic and Yfail controlled by the equivalent
plastic strain. The plastic stiffness is specified by the input ratio
Yelastic ¼ Yfail Felastic Fcap Gelastic/(Gelastic  Gplastic).

8  2.3.4. Damage evolution and residual shear strength

>
> 1 for p  pu ¼ fc 3
< rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

 When hardening states reach the ultimate strength of the con-
2
Fcap ¼ 1 ppu
for pu < p < po (17a,b) crete on the failure surface Yfail damage is accumulated during
>
> po pu
: further inelastic loading controlled by plastic strain. The model by
0 for p  po ¼ pel
Holmquist and Johnson [19] is taken as basis for the evolution law.
Eqs. (19) and (20) specify how the plastic increments are normal-
ized over the effective strain to failure 3pl,fail
eq . The effective strain to
3pl

Yhard ¼ Yelastic þ
eq
Yfail  Yelastic (18a) failure is depending on the hydrostatic pressure p with the shape
pl;hard parameters D1, D2 and a lower limit efmin. Contrary to the original
3eq

Fig. 8. Influence of effective gauge size for measurement of lateral stress in position 4.
290 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293

Fig. 9. Damage initiation and associated wave structure in axial and lateral stress profiles (only rise shown). Right: derived dynamic strength data of from (simulated) longitudinal
and transverse stress measurements.

paper only deviatoric plastic increments contribute to the damage 200 mm by one cell size (Fig. 4, right, 80,000 cells) The whole target
evolution. Volumetric plastic increments only affect the equation of configuration is modeled as one domain.
state and the elastic cap function, a detail not described by Holm-
quist. The effect of damage is modeled as a loss in deviatoric 3.1. Comparison of assembled against monolithic concrete target
strength by interpolating as Eq. (21b) between the failure surface
Yfail and residual friction resistance surface Yfric (Eq. (21a)). This As a first predictive step, the described numerical modeling
surface can so far not be measured; this is one aim of the described approach is used to provide insight to the questions raised in
work. Until this is achieved the same pressure dependence as for Section 1.2 about the influence of the target block structure on
the intact limit surface is assumed (B ¼ A, m ¼ n). experimental measurements. Therefore, the ‘assembled’ simulation
configuration with explicit description of the epoxy adhesive layers
Z 3eq
D3pl
pl;fail
1 p incremental
X eq by one cell layer of 200 mm is compared to one single monolithic
D ¼ d3eff ¼ (19) ‘block’ of concrete without adhesive. Fig. 6, left, shows the com-
3 pl;hard
eq 3pl;fail
eq ðpÞ pl;fail
3eq ðpÞ
parison of longitudinal stress signals in gauges 1–3 for both con-
figurations. The overall agreement is very good. Minor differences


D2 occur for times later than 15 ms after impact. The agreement in the
pl;fail
3eq p* ¼ D1 p*  HTL*  efmin (20) uniaxial compression and release phase is excellent.
Fig. 6, right, depicts the lateral stress calculations in gauge 4.

 Again we note minor differences between both simulation config-
Yfric ¼ Bpm Ydamaged ¼ Yfail þ D Yfric  Yfail (21a,b) urations. For the one-dimensional compression and release phase
up to 15 m differences are negligible with a very small deviation of
the second stress step at about 80 MPa.
More important is the difference of measuring lateral stresses in
3. Analysis of impact tests positions 4 and 1. For the experimental analysis of concrete shear
strength, they shall be assumed congruent, as the distance from the
The analysis of stresses will focus on the first phase of com- impact phase is identical. However, simulations show that this
pressive loading under uniaxial strain. Therefore, the configuration assumption holds only for the rising of stresses and still for the peak
can be simplified as cylindrical symmetric, which allows very high values. Release is different, which is caused by the proximity of the
mesh resolutions in order to resolve the bonding layers with projectile edge and the resulting arrival of the lateral shear release
wave. Later on the impactor cylindrical collar is keeping up the
compressive stress at gauge 4 more than the flat U-bottom of
25 mm thickness at the level of gauge 1. Although still appropriate
for the analysis of pressure and peak values, gauge 4 is moved to-
wards the projectile centre line in an improved configuration [41].
Bourne et al. [3] measured differences in failure propagation in
glasses when joining two high precision plates to replace one
monolithical. The here described model predicts that concrete does
not show this behavior. Fig. 7, left column, compares damage pro-
files from the impact face into the concrete near the symmetry axis.
The early phase of uniaxial compression (upper line) shows no or
negligible influence in the damage propagation. It is worthwhile to
notice that calculated damage levels of 8.5–11% are very small at
this early stage. More deviations occur when later times are con-
sidered with various release waves and multiaxial stress and shear
states acting on the concrete material. Important deviations are
found when analyzing both damage patterns profiles at 80 ms
Fig. 10. Evaluation of wave speeds in elastic regime, at ultimate strength and damaged. (Fig. 7, lower line). Still, the overall damage pattern is similar: A
Slope for evaluation of strain rate of shock wave. large shear damage zone occurs close to the projectile edge.
 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293
Fig. 11. Peak particle velocities and strain rates evaluated from stress signals compared to continuous signal from simulation.
Localized shear and bending cracks extend into the centre of the
target. A distinct spallation plane is formed at 60–70 mm in the
concrete target towards the rear surface. At stages later than 120 ms
the target is totally damaged and disrupted.
Summing up, the assembled concrete target with gauges be-
tween precisely joined blocks using thin films of epoxy resin is very
representative for the behavior of a monolithic target during one-
dimensional compression and release. Careful consideration is
necessary when analyzing later stress states, especially combining
longitudinal measurements on the symmetry axis and lateral
measurements off-axis (e.g. gauges 1 and 4).
3.2. Influence of gauge size
Resolving wave propagation details in concrete can be disturbed
from the following sources:
– The heterogeneity of the sample causes internal reflections and
strain localizations during stress wave propagation. This
problem has been numerically analyzed by Park et al. [29],
Riedel et al. [32] and many other authors.
– Large gauge sizes may be necessary to derive sufficiently ho-
mogenized state variables for unscaled concrete. This problem
is most significant when the gauge extends in the direction of
the wave propagation as transverse gauge 4 in the present case.
The size influence of the gauge size is studied by averaging the
signal along 1–30 cells spread over 0.2–6 mm in Fig. 8. Local analysis
Fig. 12. First experimental measurements of stress signals using mortar samples. Longitudinal stress signals indicated a failure wave in experiment ‘A’, left. Successful lateral
measurements in experiment ‘B’, right, allow strength analysis in Table 4.
291
of lateral stress measurements in only one cell of 0.2 mm length
shows a two-wave structure separating at 100 MPa with a time delay
of about 0.5 ms for the following wave. This detail is lost when in-
creasing the gauge size Lgauge. The minimal resolvable time window
can be approximated as the time tres ¼ Lgauge/cp,wave the wave front
needs to travel across the gauge. For a gauge length of 6 mm and the
compressive wave traveling with cP ¼ 2230 m/s (see Section 3.3)
wave features down to tres ¼ 2.7 ms can be resolved. This is not suf-
ficient to resolve the small two-wave structure in Figs. 8 and 9.
3.3. Evaluation of dynamic strength, failure wave speed
and strain rate
The procedure of evaluating experimental gauge signals is pre-
pared by numerical predictions. As all state variables are computed
in every cell, assumptions on measurements signals can be verified.
First the interpretation of the wave structures during arrival of the
compression wave (Figs. 6 and 9) is addressed. The elastic and
plastic waves separate at 48 MPa stresses (I), but the second wave
structure (II) at 280 MPa longitudinal and 80 MPa lateral stress is of
particular interest. In Fig. 9 it is clearly associated with the onset of
damage at the ultimate strength of the concrete.
The stress state in terms of hydrostatic pressure and equivalent
(shear) stress can be evaluated from the longitudinal and trans-
verse stress signals in Fig. 9, left, using Eqs. (1) and (2). The analysis
can be continuously applied across the different limits of elastic
loading (I), peak strength (II) and compression of damaged material
(III). Plotting as equivalent stress versus hydrostatic pressure (Fig. 9,
292 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293
right) shows elastic loading and the beginning of hardening inside
the static failure surface. The peak stress at failure (II) lies outside
the static failure surface due to strain rate hardening. After onset of
damage the material is further compressed along the dynamic yield
surface of the damage material. The ultimate stress state (III) is
defined by impact velocities and impedances.
The phase speeds of the different wave portions can be evalu-
ated as in experimental practice: the arrival time and the loading
rate are deduced at the point of highest gradient of the wave por-
tion. Fig. 10 illustrates the derivation of the longitudinal elastic
wave speed (cp,elastic ¼ 4228 m/s), the wave speed of the intact
material at ultimate strength (cp,intact ¼ 2567 m/s), the damaged
(cp,fail ¼ 2229 m/s) and the release wave speed (crel ¼ 2000 m/s),
calculated as cp¼(x2  x1)/(t2  t1).
The longitudinal stress in the damaged material can be read as
637 MPa from the maximum plateau of gauge 1 in Fig. 10. Using the
momentum Eq. (22a) the particle velocity can be derived to 140 m/s.
Input into mass conservation (Eq. (22b)) results in a strain of 6.27%
at ultimate compression. Comparing the peak particle velocity
deduced from the stress signal to 140 m/s and the locally simulated
particle velocities in the cells (Fig. 11, left, 131 m/s), we note
reasonable agreement with a 6.8% over prediction for the value
deduced as in an experiment.
s u 3
up ¼ 3 ¼ p 3_ y
r0 cp cp trise
In experimental practice the strain rate will be derived by using
the rise time of the slope again at the steepest point (see inclined
dashed lines in Fig. 10). In gauge 1 this rise time corresponds to 1.75
ms which leads to a strain rate of 3.58  104/s when using Eq. (22c).
The rate derived in gauge 2 is very similar with 2.85  104/s.
Comparison to local strain rate histories in the simulation cells
(Fig. 11, right) reveals very good agreement. Interestingly, the
highest strain rates are simulated at the elastic–plastic transition
and near the end of damage evolution.
3.4. First experimental results
Fig. 12, left, shows the experimental measurement ‘A’ of the
configuration in Figs. 3 and 4. Mortar with a maximum grain size of
4 mm and uniaxial compressive strength of 37.7 MPa is used as
a starting point with the perspective to testing full scale concrete in
the 200 mm caliber test facility. Impact velocities were 193 m/s in
both experiments ‘A’ and ‘B’.
The longitudinal stress signals of gauges 1–3 show all decreasing
phase velocities of stress wave portions above 210 MPa. This ob-
servation is very consistent with the simulations and is attributed
to different wave speeds across the failure front. The compression
wave properties before and after failure evaluated by Eq. (22) as
previously described are given in Table 3. Both measured and
simulated values lie in the large scatter range of available experi-
ments in Fig. 1. The differences underline the importance of im-
proved measurements and analysis at the minimum of wave speeds
at particle velocities between 30 and 300 m/s behind the failure
wave.
Table 3
Compression wave properties, derived from experiments A and B using Eq. (22) on
gauge signals 1 and 2
Experiment/time Wave velocity Particle Comp.
cP (m/s) velocity up (m/s) strain (–)
A, at ultimate strength 3623 26.1 0.00719
B, at ultimate strength 2810 37.1 0.0132
A, damaged 1730 153 0.0888
B, damage 1610 140 0.0865
Table 4
Experimental concrete pressure and strength during impact
Position/time Longitudinal Lateral stress Pressure Equivalent
stress (MPa) (MPa (MPa) stress (MPa)
1, at ultimate strength 225 96,6 139 128
2, at ultimate strength (122) 49 (73) (73)
1, failed 487 133 251 354
2, failed (482) (65) (197) (397)
Fig. 12, right, shows experimental measurements ‘B’ with two
lateral gauges numbered 4 and 5 at the offsets of the longitudinal
gauges 1 and 2, but closer to the impact centre. The longitudinal
gauges show again the two-wave structure which we assign to the
onset of damage. The lateral gauges do not display distinctly this
feature because of the effective gauge length of 6 mm along the
impact direction (see Section 3.2 and Fig. 8). Nevertheless, the
stress tensor can be evaluated in terms of pressure and equivalent
stress according to Eqs. (1) and (2). The derived values at ultimate
strength (II) under a strain rate of 14,350/s and at maximum
compression (III) beyond the failure threshold with 23,100/s are
given in Table 4. Larger measurement series of such strength
measurements at highest strain rates promise unique input data for
dynamic concrete models.
(22a,b,c)
4. Conclusions
The present paper describes predictive numerical simulations
and first impact compression and shear stress measurements in
concrete materials. The experimental configuration is modeled in
a Lagrangian hydrocode together with equation of state and dy-
namic strength descriptions for concrete developed earlier and
summarized in Section 2. The simulations predict that thoroughly
assembled targets with integrated gauges are representative for
monolithic concrete samples, especially during the loading phase of
a planar compression or shock wave. The numerical analysis fur-
thermore supports derivation of stress wave and particle velocities,
strains and strain rates from time-resolved stress measurements.
Decreasing phase velocities for compression wave portions
above 200–300 MPa are observed both in simulations and first
experiments. Based on the model, the slower wave portions can be
clearly attributed to the onset of damage and thus the material
strength at high strain rates of 25,000/s. It is experimentally con-
firmed by first experiments on mortar, further justification and
extension to unscaled concrete has to follow. For this purpose
guidelines for the gauge size influence are given by numerically
analyzing highlighting to the rule of thumb tres ¼ Lgauge/cp,wave.
In the proposed configuration the triaxial strength is measured
at strain rates of 5000–15,000/s with hydrostatic pressures from 70
to 140 MPa and deviatoric stresses from 70 to 130 MPa. Higher
stresses may be reached by impact velocities exceeding 200 m/s in
near future.
The technique can furthermore provide valuable data on the
yield surface of damaged concrete at strain rates of about 25,000/s.
This is currently a major lack for dynamic concrete models in
hydrocodes. Signal predictions and first measurements analyze the
damaged strength surface up to 250 MPa of hydrostatic and
400 MPa equivalent stress at a strain rates of 3  104/s. Further
measurements with more data, detailed evaluation and transition
to full scale concrete are very promising perspectives.
Strain
rate (1/s)
5490 Acknowledgements
14,350
28,500 The described work results from cooperation of Tokyo Institute
23,100
of Technology, Materials and Structures Laboratory and the Ernst-
 W. Riedel et al. / International Journal of Impact Engineering 36 (2009) 283–293
Mach-Institute of Fraunhofer Society, Germany. The author is
grateful for the temporary assignment as visiting professor at
TokyoTech, which made the numerical study possible. Our special
thanks are due to Dr. Katayama, Itochu Technical Solutions Cor-
poration, Japan, for his continued support of the scientific exchange
and work. We furthermore thank Mr. Hasegawa for his valuable
work in constructing the test configuration. The support in sample
preparation by K. Tanaka and K. Inoue, Tokyo Institute of Technol-
ogy, is gratefully acknowledged.
References
[1] Bischoff PH. Materials & Structures 1999;24:425–50. RILEM.
[2] Bischoff PH, Schlüter F-H, editors. Concrete structures under impact and im-
pulsive loading, synthesis report. Bulletin d’information, No. 187. Dubrovnik:
Comité Euro-Internationale du Beton; September 1988.
[3] Bourne N, Millet J, Rosenberg Z, Murray N. J Mech Phys Solids 1998;46(10):
1887–908.
[4] Brar NS, Bless SJ, Rosenberg Z. Appl Phys Lett 1991;59:3396–8.
[5] Chen WF. Plasticity in reinforced concrete. New York: McGraw Hill, ISBN 0-07-
010687-8; 1982.
[6] Clifton RJ. Analysis of failure waves in glasses. Appl Mech Rev 1993;46:540.
[7] Espinosa HD, Xu Y, Brar NS. J Am Ceram Soc 1997;80:2061.
[8] Espinosa HD, Xu Y, Brar NS. J Am Ceram Soc 1997;80:2074.
[9] Gebbeken N, Greulich S, Pietzsch A. Int Jn Impact Engineering 2006;32:2017–31.
[10] Gebbeken N, Ruppert M. Arch Appl Mech 2000;70:463–78.
[11] Grady DE, Furnish MD. Shock compression of condensed matter; 1989. p. 621–3.
[12] Grady DE. Impact compression properties of concrete. In: Sixth international
symposium on interaction of nonnuclear munitions with structures, Panama
City, Florida, May 3–7; 1993.
[13] Grady DE. Dynamic decompression properties of concrete from Hugoniot
States – 3 to 25 GPa. Experimental Impact Physics Department; Feb. 1996.
Technical Memorandum TMDG0396.
[14] Gregson VR. A shock wave study of Fondu-Fyre WA-1 and concrete, General
Motors Materials and Structures Laboratori, Report MSL-70-30; 1971.
[15] Grote DL, Park SW, Zhou M. Int Jn Impact Eng 2001;25:869–86.
[16] Grote DL, Park SW, Zhou M. Jn Appl Physics 2001;89:2115–23.
[17] Hanchak SJ, Forrestal MJ. Perforation of concrete slabs with 48 MPa (7 ksi) and 140
MPa (20 ksi) unconfined compressive strengths. Int J Imp Eng 1992;12(1):1–7.
[18] Herrmann W. Jn Appl Physics 1969;40(6):2490–9.
[19] Holmquist TJ, Johnson GR. A computational constitutive model for concrete
subjected to large strains, high strain rates, and high pressures. In: Fourteenth
International Symposium on Ballistics, Québec; 1993.
293
[20] Itho M, Katayama M, Mitake S, Niwa N, Beppu M, Ishikawa N. Numerical study
on impulsive local damage of reinforce concrete structures. In: Proceedings of
the structures under shock and impact, Cambridge; 2000.
[21] Ishiguchi M, Yoshida M, Nakayama Y, Matsumura T, Takahashi I, Miyake A,
et al. A study of the Hugoniot of mortar. Journal of Japan Explosives Society,
Kayaku Gakkaishi 2000;61(6):249–53.
[22] Kanel GI, Rasorenov SV, Fortov VE. Shock compression of condensed matters;
1991. p. 451–4.
[23] Klepaczko JR, Bara A. Int J Impact Eng 2001;25:387–409.
[24] Kipp ME, Lalit Chhabildas C. CP429, Shock compression of condensed matter;
1997 p. 557–60.
[25] Malvar LJ, Crawford JE, Wesevich JW, Simons D. Int Jn Impact Eng 1997;19:
847–73.
[26] Millet JCF, Bourne NK, Rosenberg Z. CP505, Shock compression of condensed
matter; 1999 p. 607–610.
[27] N.N.. AUTODYN, theory manual. Horsham, UK: Century Dynamics Ltd.; 2003.
[28] Ockert J. Ein Stoffgesetz für die Schockwellenausbreitung von Beton, disser-
tation, TH Karlsruhe; 1997.
[29] Park SW, Xia Q, Zhou M. Int J Imp Eng 2001;25:887–910.
[30] Riedel W, Thoma K, Hiermaier S, Schmolinske E. Penetration of re-
inforced concrete by BETA-B-500. In: Proc. 9. ISIEMS, Berlin, Strausberg,
Mai; 1999.
[31] Riedel W. Beton unter dynamischen Lasten – Meso- und Markomechanische
Modelle. In: Ernst-Mach-Institut, editor. Freiburg: Fraunhofer IRB, ISBN 3-
8167-6340-5; 2004. Available from: http://www.irbdirekt.de/irbbuch/; 2004.
[32] Riedel W, Wicklein M, Thoma K. Shock properties of conventional and high
strength concrete, experimental and mesomechanical analysis. Int Jn Impact
Engineering 2008;35:155–71.
[33] Schuler H, Mayrhofer Chr, Thoma K. Int. Jn. Impact Eng. 2006;32:1635–50.
[34] Steinberg DJ, Cochran SG, Guinan MW. J Appl Phys 1980;51:3.
[35] Steinberg DJ. Equation of state and strength properties of selected materials.
LLNL; Feb 1991.
[36] Tsembelis K, Millett JCF, Proud WG, Field JE. CP505, Shock compression of
condensed matter; 1999 p. 1267–70.
[37] Tsembelis K, Proud WG, Field JE. CP620, Shock compression of condensed
matter; 2001 p. 1414–17.
[38] Tsembelis K, Proud WG, Willmott GR, Cross DLA. CP706, Shock compression of
condensed matter; 2003 p. 1488–91.
[39] Tsembelis K, Proud WG. Shock compression of condensed matter. AIP; 2005. p.
1496–99.
[40] Willam KJ, Warnke EP. Constitutive model for the triaxial behaviour of con-
crete. In: International Association for Bridge Structural Engineering, seminar
on concrete structure subjected to triaxial stresses, IABSE Proc. 19, Italy; 1975.
[41] Kawai N, Inoue K, Misawa S, Riedel W, Tanaka K, Hayashi S, et al. The dynamic
behavior of mortar under impact loading. AIP-CP955. In: Elert M, Furnish M,
Chau R, Holmes N, Nguyen J, editors. Shock-compression of condensed matter,
SCCM 2007. p. 549–52.