R. A.

Clegg et al / International Journal of Impact Engineering

Hypervelocity Impact Damage Prediction in Composites

Part I - Material Model and Characterisation
*

R. A. Clegg**, D. M. White*, W. Riedel+, W. Harwick+

Century Dynamics Ltd, Dynamics House, Hurst Road, Horsham, West Sussex, RH12 2DT, England
+
FhG - Ernst-Mach-Institut, Eckerstrae 4, D-79104, Freiburg, Germany

Abstract
This paper reports on the development of an extended orthotropic continuum material model and associated
material characterisation techniques for the simulation and validation of impacts onto fibre reinforced composite
materials. This Part I of the paper focuses on the details of the numerical model and the material characterisation
experiments. Part II Experimental Investigation and Simulations [1] provides details of hypervelocity impact
damage experiments and simulations performed to assess the capabilities of the developed model. Here a detailed
description of the material model [2], as implemented in AUTODYN [3], is provided. The model is an extension
of that presented in [4], which allowed the correct thermodynamic response of orthotropic materials to be
simulated under shock wave loading. Improved capabilities to allow prediction of the extent of damage and
residual strength of fibre composite materials after impact are described along with set of quasi-static and
dynamic experiments used to characterise the directional non-linear strength and extent of damage. Application of
the model and derived material data is demonstrated through the simulation of a hypervelocity impact event of an
aluminium sphere impacting the Columbus module shielding system of the International Space Station (ISS).
Keywords: Hypervelocity, Impact, Material Model, Material Characterisation, Kevlar, CFRP

1. Introduction
The study of impacts on composite materials is a requirement for both manned and unmanned
spacecraft structures. On manned spacecraft, such as the ISS, fibre reinforced composite materials are
used in the primary shielding system used to mitigate the effects of orbital space debris impacts. On
unmanned spacecraft carbon fibre reinforced composite materials are commonly employed for major
structural components. Although such components are not designed to mitigate the effects of
hypervelocity impact, the consequences in terms of secondary debris after such impacts is of concern in
relation to the survivability of on-board equipment. This paper describes an advanced material model,
and set of material characterisation experiments, for orthotropic materials that has been developed to
*

Corresponding author. Tel.: +44 1403 270066; fax:+44 1403 270099

E-mail address: richard.clegg@centurydynamics.co.uk

R. A. Clegg et al / International Journal of Impact Engineering

improve the predictive capabilities of numerical simulations of hypervelocity impacts on a range of

composite spacecraft structures/materials [2].
The foundation for the material modelling and characterisation work presented here is described in
detail in [4]. This work led to the identification of the following phenomena, which we believe are of
primary interest for composite and textile materials subject to high/hypervelocity impact:
material anisotropy
shock response
anisotropic strength degradation (damage).
The material model [2] developed to address these aspects of composite material simulation under
impact loading is now described. The material model development was carried out hand-in-hand with
material characterisation experiments which are also described alongside the model features.
2. Basic Orthotropic Stiffness and Shock Response
In anisotropic materials, the traditional independent approach for the solution of the equation of
state and constitutive relations in a hydrocode is complicated because these two sub-models are
strongly coupled; volumetric strain leads to deviatoric stress, and similarly deviatoric strain leads to
spherical stress. A methodology [5] was implemented and further developed in AUTODYN as reported
in [4]. Consider a linear elastic orthotropic material for which the total incremental stress, ij, can be
related to the total incremental strain, ij, through the orthotropic stiffness matrix, Cij. The coefficients
of Cij being functions of the orthotropic elastic material constants, Eii, ij and Gij.

ij = C ij ijd + v
3

(1)

To facilitate the coupling of the deviatoric and volumetric material response the total strain has
been decomposed into volumetric, v, and deviatoric, ijd, components. Since the pressure is the average
of the three direct stresses, from (1) we can obtain:

P =

1
(C11 + C22 + C33 + 2C12 + 2C23 + 2C31 ) v
9
1
1
1
d
d
d
(C11 + C21 + C31 )11
(C12 + C22 + C32 ) 22
(C13 + C23 + C33 ) 33
3
3
3

(2)

For an isotropic Hookean material the first term on the right-hand side of (2) is equivalent to a
linear equation of state, whilst the remaining deviatoric strain terms would be zero. Thus for an
orthotropic material we can replace the first term with a non-linear Mie-Gruneisen equation of state and
the remaining terms act as a correction due to deviatoric strains.
A model based on the above was implemented in AUTODYN [3] and has subsequently been used
for simulating high to hypervelocity impacts on a wide range of applications and materials [6], [7], [8],
[9], [10]. This experience identified potential weaknesses of the model especially with respect to
anisotropic strength degradation. Characterising and modelling these aspects is the main focus of the

R. A. Clegg et al / International Journal of Impact Engineering

current work.
The phases of anisotropic strength degradation in a composite material can be categorised into
Nonlinear hardening (maximum stress increases with strain)
Nonlinear softening (maximum stress decreases with strain)
The observed material characteristics, and tests to derive them, are now outlined along with the
macro-mechanical material modelling features developed to represent the behaviour.
2.1 Non-linear Hardening
Many composite materials exhibit non-linear stress strain behaviour when subject to tensile uniaxial
stress loading. The most common example of such behaviour is found when loading a unidirectional
fibre reinforced laminate at 45 degrees to the fibre direction (Fig. 1). The observed non-linear response
results in a reduction in the tangent material stiffness with strain.
200

H ydraulic piston
Stepwise Loading
Monotonic Loading
150

Stress [MPa]

100

h
Specimen

l0

h
w
l
l0
t

50

Load cell

= 60 mm
= 10 mm
= 230 mm
= 165 mm
= 20 mm

0
0

0.02

0.04

0.06

0.08

0.1

0.12

Strain [-]

Fig. 1. Uniaxial tension test, cyclic stress strain curve for Kevlar-epoxy (0/90) loaded at 45 to the fibre direction [2].

The envelope to the stress strain curves (Fig. 1) clearly shows non-linear hardening behaviour.
Cyclic unloading and re-loading also reveals that the non-linear hardening results in irreversible
deformation. This terminology is consistent with that used for dislocation plasticity. However, in this
case the underlying micro/meso mechanical behaviour may include a number of different mechanisms
such as matrix plasticity, fibre re-orientation and matrix/fibre damage. The baseline material model [4]
employs a linear elastic orthotropic stiffness matrix under small tensile strains hence is not capable of
capturing this type of behaviour.
To enable phenomenological representation of this hardening behaviour, a generalised limit surface
[11] has been adopted. This surface is quadratic in stress space and is generally applicable to composite
materials. For example, no a priori assumptions are made in respect of the influence of hydrostatic
stress on inelastic deformation as in Hills orthotropic yield criteria [12].
2
2
2
f = a11 11
+ a22 22
+ a33 33
+ 2a12 11 22 + 2a23 22 33 + 2a13 11 33

(3)

+2a44 + 2a55 +
2
23

2
31

k 0

2
2a66 12

where the stresses ij refer to the principal material directions, and k is a state variable defining the
instantaneous value of the limit surface. The nine coefficients, aij define the extent of anisotropy in the

R. A. Clegg et al / International Journal of Impact Engineering

material response and are constants, implying isotropic hardening. It can be shown [11] that these
constants are related to the Plastic Poisson Ratio (PPR) in the material via

a11 = a22

P
21
12P

, a33 = a22

P
23
P
32

, a11 = a33

P
P
a12 = a22 21
, a23 = a22 23

31P
13P

(4)

P
, a13 = a33 31

The values of the coefficients can thus be obtained from a series of well-instrumented tension and
shear experiments. Fig. 2 shows the results of a typical in-plane and through thickness instrumented
tension test on an aramid-fibre-epoxy specimen [2]. Note the non-linear characteristics of both the inplane and out-of-plane Poisson ratio. For the purpose of the numerical model, this data must be
linearised into an elastic and inelastic Poisson ratio.
Applied strain gauges on sample side
and front surface

= 0.08

In-plane PPR

= 0.70

m
m

Out-of-plane-plane PPR

Fig. 2. Uniaxial tension test, measurement of elastic and plastic Poissons ratio [2].

In terms of numerical implementation, an iterative backward-Euler procedure [13] is used to return

trial elastic stresses, which lie outside the limit surface, normally back to the limit surface. The
associated incremental inelastic strains are defined using the Prandtl-Reuss equations, often called an
associated flow rule, and state that the plastic strain increments are proportional to the stress gradient of
the yield function. The proportionality constant, d, is known as the plastic strain-rate multiplier.
Written out explicitly the inelastic strain increments are given by
d 11p 2a11 11 +
p
d 22 2a12 11 +
d 33p 2a13 11 +
p=
d 23
d 31p
p
d12

2a12 22
2a22 22

+
+

2a23 22 +
4a44 23
4a55 31
4a66 12

2a13 33
2a23 33

2a33 33
d

From which an equivalent inelastic strain measure can be derived [2]

(5)

R. A. Clegg et al / International Journal of Impact Engineering

(d )

P 2

8
f d2
3

(6)

This in turn feeds back into the hardening function, k, of the limit surface.
The developed non-linear hardening option, in combination with the baseline model, allows for
much improved simulation of the non-linear in-plane hardening behaviour of fibre composite materials
as shown in Fig. 3.

Fig. 3. Aramid-fibre-epoxy uniaxial stress tension tests - experiment and simulation. (a) Loading at 0 to fibre direction;
(b) Loading at 45 to fibre direction.

2.2 Non-linear Softening (Damage)

Considering failure in fibre composite materials, delamination caused by through thickness strain,
shear strain causing matrix cracking, and fibre failure are the dominant modes of damage for impact
loading. These modes of failure lead to a reduction in load carrying capacity in one or more material
directions we shall term this phase of deformation as softening (reduction in capability to hold stress
with strain). A key characteristic here is the anisotropic nature of the effects of failure on the residual
strength of the material. For example delamination leaves the material with zero tensile strength across
the fibres, but significant tensile strength remains in the plane of the fibres.
To understand and quantify this type of phenomena a range of material characterisation tests must
be performed to establish; the point at which softening starts to occur, the rate at which the material
softens to zero strength in a particular material direction. These features determine the energy absorbed
during the damage process.
The modelling approach taken here introduces additional failure initiation criteria and associated
limit surfaces in material stress space, to represent specific modes of failure. Modified forms of the
well-known Hashin failure criteria [14] have been adopted for both the failure initiation criteria and
additional limit surfaces.
2

ii-plane::

eiif2

ij
ii
ik
1
+
+
=
ii , fail (1 Dii ) ij , fail (1 Dij ) ik , fail (1 Dik )

(7)

R. A. Clegg et al / International Journal of Impact Engineering

The above failure initiation criteria are checked at each material integration point each cycle. If one
of the failure criteria is exceeded (eiif>1), the orthotropic damage model is activated for that integration
point. The damage model has several aspects:
2.2.1

An Updated Elastic Material Stiffness Matrix

C11
C12 (1 Max( D11 , D22 ) ) C13 (1 Max( D11 , D33 ) ) 0

C (1 Max( D , D ) )
C 22
C 23 (1 Max( D22 , D33 ) ) 0
11
22
12
C13 (1 Max( D11 , D33 ) ) C 23 (1 Max( D22 , D33 ) )
0
C33
Cij =
0
0
0
C
44

0
0
0
0

0
0
0
0

2.2.2

0
0
0
0
C55
0

0
0

0
0

C66

(8)

Limit Surfaces in Material Stress Space and Damage Law

Additional limit surfaces are defined in material stress space through the failure initiation criteria in
each material plane. If a stress point lies outside a limit surface, it is associatively returned to the limit
surface using an iterative backward-Euler procedure [13]. Returning a stress point to a softening limit
surface results in inelastic crack strain.
The resulting inelastic strain ijcr is then used as input to an orthotropic damage evolution law
Lij Fij2 ijcr

Dij =
2Gijf Fij

(9)

where Fij represents the initial failure stress in the three material directions and shear, Gijf is the
fracture energy for each mode of failure and Lij is a local characteristic dimension of the numerical
integration point in each direction. This damage equation is formulated such that the work required to
extend a crack by a unit length is relatively insensitive to the local element size in each direction and is
an extension of the model developed for isotropic materials [15].
2.2.3

Damage Characterisation

The material data requirement for the orthotropic softening model was purposely limited to the
direct and shear failure initiation stress (Fij) in each of the material planes, and the associated fracture
energy release rates (Gfij) for mode I and II crack growth. A variety of material characterisation
experiments were developed and used to populate the parameters of the failure surfaces.
Static Damage Characterisation Tests
For the cross woven aramid-fibre-epoxy laminate under consideration (Kevlar/129-epoxy),
transverse isotropy was assumed. In this case the parameters for the failure surfaces in the two material
planes perpendicular to the fibres could be derived from the in-plane uniaxial tension tests, as detailed
above: The tensile failure initiation stress could be obtained directly from the experiment; the in-plane
shear failure stress was derived through numerical simulation and calibration to match the experimental

R. A. Clegg et al / International Journal of Impact Engineering

results for in-plane tension at 45 to the fibres (Fig. 3).

The transverse shear failure stress (mode II) was assumed to be equal to the inter-laminar shear
failure stress and derived from a Double Notched Shear test as descried in Fig. 4(a) (performed
according to standard proposal ASTM D 3846-79 and optimised according to [16] to compensate for
normal stresses). The mode II delamination energy was derived using an End Notch Flexure test as
described in Fig. 4(b) (performed in accordance with standard proposal EN 6034). A predefined crack
propagates as a result of the specimen bending and associated shear forces at the crack tip in the Mode
II loading. The total fracture energy GIIc is calculated from the initial crack length, the critical load to
start the crack and the associated piston displacement at onset of crack propagation.
The through thickness (mode I) delamination energy was obtained using a Double Cantilever Beam
test arrangement as described in Fig. 4(c) (performed in accordance with the standard proposal
EN6033). An initial delamination or pre-crack is introduced at the tip of the specimen with a diamond
saw. The pre-cracked specimen is loaded continuously by the loading device until a total propagated
crack length of 100 mm is achieved. The mode I fracture energy can be derived from the resulting load
displacement curves.
A major challenge for the characterisation and modelling of fibre reinforced composites subject to
impact is the through thickness (transverse shear) mode of deformation. Little attention is paid in the
literature to this mode of deformation, most composite research being focussed on in-plane and bending
response (structural behaviour). In this work a Short Beam Bending test [2] was used to try to
characterise the through thickness behaviour (Fig. 4(d)).
(a)

(b)

hydraulic piston

h
specimen

h
w
l
l0
t

load cell

= 100 mm
= 25 mm
= 35 mm
= 40 mm
= 5.7 mm

Piston
Specimen

Initial crack

l0 l

l
w
a
b
t

= 10 mm
= 8 mm
= 40 mm
=
= 5.7 mm

(c)

a
t
Load cell

Force

(d)

specimen dimensions:

5.7 mm

l=3.35 mm

fixed support

l=3.35 mm

- length: 140 mm
- width: b=20.24 mm
- thickness: t=5.7 mm

50 mm
wt

specimen

a
w

L
w
h
a

= 200 mm
= 25 mm
= 5.7 mm
= 50 mm

loading fork

Fig. 4. (a) Double notched shear test; (b) end notched flexure test; (c) double cantilever beam test; (d) short beam bending test.

Dynamic Damage Characterisation Tests

Planar Plate Impact tests were used here to estimate the spall strength (through thickness tensile
strength) of the material. The tests consisted of a 3mm thick aluminium alloy flyer impacting 5.65mm
thick aramid-fibre-epoxy laminate at velocities between 100m/s and 255m/s. A laser interferometer was

R. A. Clegg et al / International Journal of Impact Engineering

used to measure the particle velocity on the rear surface of the target.
To derive the through thickness tensile strength at high strain rates a superposition of release waves
is generated in the target plate. Therefore the ratios of plate thicknesses and shock velocities are adapted
to superimpose release waves from the target free surface and from the projectile rear side in the sample
(target) material. Fig. 6 shows the velocity signals recorded at the target free surface of the aramidfibre-epoxy composite. Note that in comparing experiment with numerical simulations of the
configurations using accurate shock data for the material [4], mismatches are observed with respect to
rise time and the shock plateau level. So far, shock dispersion in the composite sample is assumed to be
the reason for this unexpected experimental behaviour. Although spallation measurements with the
aramid-fibre-epoxy composite are not as successful as for metals, estimates for input to the material
model could be obtained.
2.3 Final Material Model, Data and Required Material Characterisation
The full set of capabilities of the developed model for fibre composite materials can be summarised
Orthotropic elastic stiffness
Non-linear and energy dependent volumetric response for high compressions
Non-linear orthotropic strength with isotropic hardening
Stress based failure initiation on each material plane
Orthotropic damage based on fracture energy
The required input parameters and the matrix of material characterisation tests used to derive each
parameter is presented in Table 1. Data derived for the aramid-fibre-epoxy laminate considered here is
also provided.
Table 1. Model input data and derived material constants for Kevlar-epoxy
EQUATION OF STATE : Orthotropic
Parameter
= 1.65000E+00 g/cm3
E11 = 1.94800E+06 kPa
E22 = 1.79898E+07 kPa
E33 = 1.79898E+07 kPa
12 = 7.56000E-02
23 = 7.56000E-02
31 = 6.98000E-01
G12 = 2.23500E+05 kPa
G23 = 1.85701E+06 kPa
G31 = 2.23500E +05 kPa
Tref = 3.00000E+02 K
Cv = 1.42000E+03 J/kgK)

Source
Direct Measurement
Uniaxial tension tests*
0o tension test
0o tension test
0o tension test
0o tension test
Confined compression test [4]
45o tension test
Short beam shear test
Short beam shear test
Literature

Sub Equation of State:: Polynomial

A1 = 5.89499E+06 kPa
A2 = 5.00000E+07 kPa
T1 = 5.89500E+06 kPa

Planar plate impact tests

Planar plate impact tests
Assumed equal to A1

STRENGTH MODEL : Orthotropic Yield (non-linear hardening)

Parameter
a11 = 1.50000E+00
a22 = 1.00000E+00
a33 = 1.00000E+00
a12 = -6.80000E-01
a13 = -6.80000E-01
a23 = -2.60000E-01
a44= a55=a66 = 4.00000E+00
Plastic Strain = 0.0, 9.0e-6, 6.2e-4,
1.9e-3, 2.5e-3, 5.0e-3, 8.8e-3, 1.2e-2,
2.6e-2
Effective stress = 1.55e5, 1.55e5,
1.67e5, 1.78e5, 1.87e5, 1.93e5,
2.10e5, 2.35e5, 2.52e5, 3.16e5 kPa

Source
0o tension test
o
0 tension test
0o tension test
0o tension test
0o tension test
0o tension test
45o tension test
0o tension test
0o tension test

FAILURE MODEL : Orthotropic Damage (non-linear softening)

Parameter
fail 11 = 4.50000E+04 kPa
fail 33 = 2.45000E+05 kPa
fail 12 = 1.40000E+04 kPa
fail 23 = 2.00000E+04 kPa
fail 31 = 1.40000E+04 kPa
Gf11 = 5.44710E+02 J/m2
Gf 22 = 3.00000E+01 J/m2
Gf12=Gf23=Gf31=1.46130E+03 J/m2

Source
Calibration to match PPI Spall and DCB tests
fail 22 = 2.45000E+05 kPa
0o tension test
Interlamina shear test
Calibration to match 45o tension test
Interlamina shear test
Double cantilever beam test
Calibrated to 0o tension test
End notched flexure test

R. A. Clegg et al / International Journal of Impact Engineering

3. Validation

Initial validation and assessment of the developed material model and derived input was performed
using dynamic planar plate impact experiments [2]. Here we shall discuss two examples of this type of
test. Further examples are described in Part II of the paper [1] along with hypervelocity impact
validation.
3.1 Inverse Plate Impact
During development of the baseline model [4] inverse plate impact experiments were reported from
which the uniaxial compression behaviour at strain rates up to 104 /s was studied for aramid-fibre-epoxy
composites. In these sets of experiments a 50mm diameter projectile composed of aramid-fibre-epoxy
with a C45-steel backing plate impacts a stationary witness plate also made of C45-steel. A VISAR
laser interferometer system recorded the velocity trace of the witness plate during the impact.
Simulations of these experiments are repeated here using the enhanced material model and data to
ensure that the addition of the non-linear strength and damage aspects of the model did not prejudice
the models ability to reproduce the transient shock wave propagation phenomena. The simulations were
performed using the Lagrange solver of AUTODYN [3] in a uniaxial strain configuration with element
size 0.1mm. Fig. 5 demonstrates the ability of the model to reproduce experimental VISAR signals at
three velocities.
Projectile

KevlarEpoxy
C45 Steel
Stationary witness
plate

C45 Steel

2.0mm 5.7mm 2.0mm

Fig. 5. Aramid-fibre-epoxy inverse flyer plate experiments and simulation comparisons (572m/s, 788m/s, 1055m/s).

3.2 Plate Impact Damage

Both direct and symmetric planar plate impact tests were performed at different velocities to
provide a basis for validating the ability of the numerical model to predict damage levels for given
impact conditions [2]. 2D axisymmetric simulations of these tests were conducted using the Lagrange
solver of AUTODYN [3] and the material model and data as described in [2]. Elements of side 0.1mm
were used throughout. The results are presented and compared with the experiment, and results using
the baseline material model and data [4] (Fig. 6).

R. A. Clegg et al / International Journal of Impact Engineering

(Expt.)

117m/s

(Simulation)

(Expt.)

150m/s

(Simulation)

Fig. 6. Planar plate impact test. (a) target rear surface velocity; (b) recovered samples and simulated delamination.

The key item to correlate between simulation and experiment is the magnitude of the pull-back
spall signal observed in the tests. For the higher velocity, the experiment shows a pull-back spall signal
of 62 m/s. The simulation with the original baseline model and data does not show any velocity
reduction after the peak velocity. The new extended model and data gives a reduction of 58m/s by
release waves, which is similar to the observed signal.
The samples recovered from the experimental tests all showed extensive delamination, illustrating
that the spall strength was exceeded in all tests. Contours of simulated through thickness damage are
consistent with experimental observations (Fig. 6).
The results of simulations of a symmetric planar plate impact test with the enhanced model and data
show a more marked improvement in the prediction levels of damage levels over the original base
model. This configuration is further discussed in detail in Part II [1].
4. HVI Simulation Columbus module shielding

Hypervelocity impact simulations on the Columbus shielding configuration using the baseline
material model have previously been conducted and reported in [4]. These 2D axisymmetric
simulations have been repeated here using the enhanced material model and data. Comparisons with the
previous simulation work and experiments [2] are discussed. The case of a 15mm diameter Al2007
sphere impacting the shielding configuration at 6500m/s (Test case A8611, [4]), is described here.
The projectile, first bumper (aluminium) and the second bumper (Nextel and Kevlar-epoxy) are
represented using the SPH solver with smoothing length of 0.25mm. The rear wall is represented using
the Lagrange solver with element edge size of 0.25mm out to a radius of 60mm. Gradual transition to
larger elements is then made to a final radius of 200mm for efficiency purposes.
Fig. 7 highlights the through thickness damage (delamination) predicted. It can be seen that during
the impact extensive delamination/damage occurs in the immediate vicinity of the impact (Light grey
regions Fig. 7). However the lateral growth in delamination is significantly less for the latest model and
data. This correlates better with experimental observations compared with previous simulations [4]. As
observed in experiment, the new simulation predicts cratering of the backwall with no. This represents a

R. A. Clegg et al / International Journal of Impact Engineering

refined prediction of rear-wall damage over equivalent simulations performed using the baseline
material model and data [4], as quantified in Table 2.

No
perforation
of backwall

(a) Final damage, old model and data[4]

(b) Final damage, new model and data[4]
Fig. 7. 2D Axisymmetric simulation of Alenia/EMI test A8611 [2].

Table 2. Columbus shield HVI simulations, quantitative damage comparison

Test / Ident
Experiment
Original
Simulation [4]
New Simulation [2]

Kevlar/Epoxy

Back wall

Bumper
Hole

Nextel
Hole

Front

Back

Hole

Crater

Pl. Def.

26.55
29

95
96

68
94

52
46

N/A
6

1.2
N/A

227.5
174

30

88

77

75

N/A

1.5

180

5. Conclusions

An improved continuum material model suitable for modelling impact on composite materials
under low to hypervelocity impact conditions has been developed and implemented in AUTODYN [3].
The model is now capable of representing: orthotropic elastic stiffness, non-linear shock effects,
orthotropic non-linear hardening, orthotropic non-linear damage. The features of the model can be
turned on/off as required for the material being simulated and the available input data. The model is
designed to be applicable to a wide range of orthotropic materials under low to hypervelocity impact
conditions.
An extensive experimental programme covering a wide range of static and dynamic tests was
conducted hand-in-hand with the numerical modelling work. This included both material
characterisation experiments for; directional strength and failure, delamination energy, equation of state
measurement. Additionally new impact experiment configurations were developed to allow quantitative
assessment of the extent of damage development in composite materials under different loading

R. A. Clegg et al / International Journal of Impact Engineering

conditions.
Throughout the model development and experimental programme, an extensive simulation
programme was used to support the model development and assess and validate the improved
composite material model. This included; simulation of material characterisation tests to verify that the
model could reproduce the underlying material response observed in experiments, simulation of impact
damage experiments to assess the evolving material model and supplied input data, simulation of
hypervelocity impact events on simplified and representative shielding configurations to assess the
performance of the new model and data on the target application.
Further application and validation of the final model and input data is described in Part II of this
paper [1].
Acknowledgements

This work was funded by ESA/ESTEC under contract No.12400/97/NL/PA(SC), CCN No. 2. The
authors gratefully acknowledge the direction and advice provided by Michel Lambert of ESTEC.
References
[1] Riedel W, Nahme H, White DM, Clegg RA, Hypervelocity Impact Damage Prediction in Composites, Part
II Experimental Investigation and Simulations, paper submitted to Int. J. Impact. Engng. 2005.
[2] Riedel W, Harwick W, White DM, Clegg RA, Advanced Material Damage Models for Numerical
Simulation Codes, Fhg-EMI report no. I 75/03, ESA CR(P) 4397, 2003.
[3] Century Dynamics, AUTODYN Theory Manual, 2005.
[4] Hiermaier SJ, Riedel W, Hayhurst CJ, Clegg RA, Wentzel CM, Advanced Material Models for
Hypervelocity Impact Simulations, EMI report no. E43/99, ESA CR(P) 4305, 1999.
[5] Anderson CE, Cox PA, Johnson GR, Maudlin PJ, A Constitutive Formulation for Anisotropic Materials
Suitable for Wave Propagation Computer program-II, Comp. Mech. 1994; 15: 201-223.
[6] Hayhurst CJ, Livingstone IH, Clegg RA, Destefanis R, Faraud M, Ballistic Limit Evaluation of Advanced
Shielding using Numerical Simulations, Int. J. Impact Engng. 2001; 26: 309-320.
[7] Clegg RA, Hayhurst CJ, Nahme H, Validation of an Advanced Material Model for Simulating the Impact
and Shock response of Composite Materials, APS SCCM 2001.
[8] White DM, Taylor EA, Clegg RA, Numerical Simulation and Experimental Characterisation of Direct
Hypervelocity Impact on a Spacecraft Hybrid Carbon Fibre/Kevlar Composite Structure, Int. J. Impact
Engng., 2003; 29 : 779-790.
[9] Silva MAG, Cismasiu C, Chiorean CG, Numerical Simulation of Ballistic Impact on Composite Laminates,
Int. J. Impact Engng. 2004.
[10] Riedel W, Thoma K et al, Vulnerability of composite aircraft components to fragmenting warheads
experimental analysis, material modeling, numerical studies, 20th Int. Symp. Ballistics, 2002.
[11] Chen JK, Allahdadi FA, Sun CT, J. Comp. Mat., 19997; 31: 788-811.
[12] Hill R., Proc. Royal Society, London A, 1948; 193: 281-297.
[13] Crisfield MA, Non-linear Finite Element Analysis of Solids and Structures, John Wiley, 1997.
[14] Hou JP, Petrinic N, Ruiz C, Prediction of impact damage in composite plates, Comp. In Sc. and Tech., 2000;
60: 273-281.
[15] Clegg RA, Hayhurst CJ, Numerical modelling of the compressive and tensile response of brittle materials
under high pressure dynamic loading, Shock Compression of Condensed Matter, AIP press, 1999.
[16] Thielicke B, Mechanical properties of C/C composites, Key Engineering Materials, 164-165: 145-160.

W. Riedel et al. / International Journal of Impact Engineering

Hypervelocity Impact Damage Prediction in Composites

Part II - Experimental Investigations and Simulations
Werner Riedel*+, Hartwig Nahme+, Darren M. White**, Richard A. Clegg**
+
FhG - Ernst-Mach-Institut, Eckerstrae 4, D-79104, Freiburg, Germany
Century Dynamics Ltd., Dynamics House, Hurst Road, Horsham, West Sussex, RH12 2DT, England

**

Abstract
The extension of damage in composites during hypervelocity impact (HVI) of space debris is controlled by
failure thresholds and subsequent energy consumption during damage growth. Characterisation and modelling of
the material under partially and fully damaged states is essential for the prediction of HVI effects on fibrecomposite structures. Improved experimental and numerical analysis techniques have been developed [5] and are
summarised in paper part I [1]. The present part II deals with the establishment of two precise damage
experiments under hypervelocity impact conditions as a validation basis for numerical simulations: The first type
consists of space debris impact configurations optimised for damage evaluation, the second experiments
reproduce HVI strain rates and compressions in plate impact. Coupling of experimental damage analysis
techniques (visual, ultrasonic, residual strength) to quantify different aspects of failure has been achieved.
Numerical simulations using the commercial hydrocode AUTODYN [4] in mesh-based and SPH formulations are
presented using the material model and data described in part I [1]
Keywords: Delamination, Composite, Damage Modelling, Residual Strength, Plate Impact, ISS

1. Introduction
Model approaches for dynamic behaviour and damage in composite structures caused by
hypervelocity transverse shock loading need to replicate several key aspects of mechanical behaviour:
Non-linear equation of state properties must describe the shock impedances for accurate prediction of
compression and release states with phase changes and spallation [1]. Orthotropic strength is a key
aspect of composite in-plane behaviour to be coupled to equation of state properties. Characterisation
and modelling methods coupling these aspects of directional non-linear mechanical properties have
been developed recently [5] and are described in paper part I Material Model and Characterisation [1].
In part I, application of the numerical technique is demonstrated on the examples of the space
debris protection shield of the International Space Station (ISS) Columbus module involving aramid*Corresponding author. Tel.: +49 761 2714 335; fax: +49 761 2714 316.
E-mail address: riedel@emi.fhg.de

W. Riedel et al. / International Journal of Impact Engineering

fibre reinforced plastics (AFRP) as an intermediate bumper. The aim of the present study was to enlarge
the validation basis by adding controlled damage experiments and loading conditions relevant for
hypervelocity impact applications. Combinations of non-destructive and destructive testing of the
damaged samples should enable in-depth analysis of the aspects of damage and couple different
evaluation techniques. Two types of HVI damage experiments have been developed and used for
validation of the new model approach:
Table 1. Overview and damage experiment types
Damage generated by
HVI debris cloud

Plate impact

Discussion (+ advantages, - limitations)

+ very close to applications
+ damage gradient adapted for validation
- locally statistical impact conditions
+ precise and recorded impact conditions
+ relevant strain rates and pressures
- mainly plane compression and release waves

All tested composite material consisted of 18 layers of 0/90 woven Kevlar 129/812 fabric
(aramid) with 38% mass content of epoxy resin Ciba 914 (Hexcel prepreg: 914/38%/812). The
slightly cylindrical panels, curved to an external radius of 2163.2 mm correspond to the intermediate
composite bumper configuration of the European Columbus module of ISS.
2. Debris Cloud Damage
2.1 Impact Conditions
Whipple shield configurations consisting of an aluminium bumper layer and a composite plate
impacted by an aluminium projectile have been optimised in view of validation of computational
damage models. Experimental knowledge in hypervelocity impact testing and predictive simulations
were used (see Fig. 1) for pre-test optimisation of the design with respect to:
Ballistic limit: Only a small central perforation should occur.
Shatter region: Smooth transition of hit densities from the central impact region to the
outer damage area should be reached by avoiding a large central fragment.
Cloud extension: By maximising this parameter a large damage transition area should be
produced but fully contained in intact surroundings of the sample plate.
Table 2. Impact conditions for debris cloud damage experiments
Experiment No.
Projectile
- material
- density
- diameter
- mass (weighed)
- impact velocity
- impact angle

4354
Al 99.98%
2.70 g/cm3
7.00 mm
474.44 mg
4.75 km/s
0

4355
AlMg3 (Al5754
H34/ hard)
2.70 g/cm3
8.20 mm
770.88 mg
4.68 km/s
0

4356
AlMg3 (Al5754
H34/ hard)
2.70 g/cm3
7.0 mm
474.86 mg
4.81 km/s
0

W. Riedel et al. / International Journal of Impact Engineering

The resulting three configurations consisted of an aluminium sphere (diameter 7.0 to 8.2 mm)
fragmented during perforation of a 2 mm aluminium plate at 4.7 km/s (Table 2). A clear distance of
150 mm to the 5.7 mm composite plate permitted the fragment cloud to expand to a lateral extension of
about 360 mm diameter. High speed shadowgraphs (exposure time 100ns) are shown in Fig. 1, right. It
can be seen that only a small central area is fully perforated, the rest of the plate is impacted and
damaged by a wide range of fragment densities.

Fig. 1. Left: Predictive simulation using base line model [3].

Right: High speed shadowgraphs (28 s, 55s; exposure 100ns) of fragment cloud impact onto composite plate (exp. 4355).

2.2 Visual Damage Inspection

Visual inspection of the samples proved good pre-test prediction: all three fragment clouds loaded
the composite plates slightly above the ballistic limit in the very centre of the impact cloud. The
impacted faces of the plates are shown in Fig. 6. Fig. 2 gives additional views of plate 4355.
Table 3. Visible composite damage after debris cloud impact; damage measure defined in Fig. 2 (right)
No.

4354
4355
4356

Front
Primary damage
d1,f
d2,f
[mm]
[mm]
110
110
110
135
110
105

Rear
Secondary damage
D1,f
D2,f
[mm]
[mm]
360
345
130
390
365
360

d1,r
[mm]
75
390
125

d2,r
[mm]
125
110
75

D1,PS,r
[mm]
100
230
132

D2,r
[mm]
130
10
130

Perforation in Composite / Effect on Witness Plate

4354
4355
4356

some composite layers completely destroyed, no clear hole / few scratches on witness panel
clear perforation hole visible / two bulges; distance from centre and several small bulges in the middle;
cratered area about 100 x 90 mm.
some composite layers completely destroyed, no clear hole / few scratches on witness panel
Notation: di,PS diameter of damaged composite layers; Index: f front, r rear, 1 vertical, 2 horizontal

In experiments 4354 and 4356 with 7 mm spheres no clear hole was noticed but scratches on
witness plates behind the composite indicated composite fibres ejected from the rear surface. Loading
with an 8.2 mm sphere produced a small but clear hole in the composite.

W. Riedel et al. / International Journal of Impact Engineering

In all cases, damage can be categorised into two zones:

Primary damage in the central area with high hit densities and clearly visible surface
delaminations on the front face.
The adjacent area of secondary damage shows isolated, local impacts without obvious
coalescence of damage or delamination on the surface. The observed damage patterns are
summarised in Table 3.
D1,PS
front:d1,PS,f
rear:d1,PS,r

front:d2,PS,f D2,PS
rear:d2,PS,r

Fig. 2. Visible damage on impact and rear face. Definition of damage measures.

2.3 Non-destructive Damage Characterisation

Through thickness failure was analysed by ultrasonic measurements before sectioning of the target
for further evaluations described below. The composite panels and the ultrasonic detector were placed
in water for transmission of longitudinal waves from the source into the test object and back. The pulse
travels through the water and is partially reflected at every impedance boundary such as the front side,
discontinuities in the material or the rear surface of the sample.
0%

0 mm

d3

d1

d2
d1=1,568mm; d2=5,91mm

100 mm

100%
d1=1,86mm; d2=5,91mm

d3=19%

100 mm

d3=16%

Fig. 3. Left pair: Depth of delamination measured from impact and rear side: cone shaped delamination shape.
Right pair: Intensity of reflection from impact and rear side.

The scanning area captured 320 x 320 mm with a scanning step of 0.2 mm. An analysis gate
defined the range d1 to d2 through thickness between the front side and back wall echo in order to
avoid misinterpretation of the plate free surfaces. The values refer to front side, the position of which is
following the curvature of the composite panels. The location of the intermediate echo, caused by

W. Riedel et al. / International Journal of Impact Engineering

delamination, correlates with the depth of the defect. A lower sensitivity threshold value d3 of 16-19%
intensity was used to suppress background noise (see Fig. 3).
Fig. 6 shows scan results of the three panels from the impact side. Fig. 3 gives additionally the rear
view, as only the depth of the first delamination detected in the analysis gate can be displayed in any
scan. The gray scaling of the plate refers to the delamination depth measured between the specified
limits d1 and d2. The following observations could be stated:
Large areas of delamination are shown also in the outer zones where only isolated impacts
occurred. The delaminated areas clearly exceed the visible ranges of heavy damage di,f.
Beginning from a central region of delamination through the complete thickness, the
damage in the secondary region is cone shaped toward the rear surface.
The damage patterns are mostly circular, slightly flattened normal to the laminate axis.
The large sphere (8.2 mm, No. 4355) creates a bigger central damage zone especially on
the rear surface compared to the smaller projectile of equal strength (7.0 mm, no. 4354).
But the total area around the outer zones of damage is of the same size.
Comparing effects created by the small spheres (No. 4354, 4356), the harder alloy (No.
4356) provides a smaller extent of damage with a sharper transition zone. The delamination
cone is steeper through thickness.
2.4 Destructive Testing: Visible Inspections and Residual Strength
Sectioning in strips of 20 mm width, visual inspection and residual strength testing provided
deepened insight to the amount of damage. Hydrocutting was used to minimise additional damage.
Visual inspection of all sections, as exemplarily shown in Fig. 4 showed very fine delaminations in the
zone of secondary damage, proving high sensitivity of ultrasonic scanning.
0

delamination

12

15

18

21

total failure

24

27

30

delamination

Fig. 4. Visual inspection of in-depth damage after sectioning: example central section 4355_8 (out of 16).
Positions of transverse shear testing 0, 3, 6, , 30 of each section.

Advanced damage measures in constitutive models not only describe the areal extension but also
the quantitative effects in terms of strength or stiffness loss. In order to establish the link between
observed damage and its strength effects, local shearing was applied to every section as shown in
Fig. 5. Testing every 30 mm (see scale above Fig. 4) provided a strength analysis grid with 20 mm x
30 mm resolution in the central area of 300 x 300 mm of each composite panel. A surface plot of
damage in terms of shear strength degradation is shown in Fig. 5, right.

W. Riedel et al. / International Journal of Impact Engineering

0
3

2000
Force [N]

30

27

1500
1000

9
24

0.8-1
0.6-0.8
0.4-0.6
0.2-0.4
0-0.2

0.8
0.6

500

12

0.4
0.2

21
15
18

4
5
distance [mm]

30
27
24
21
18
15
12

Damage [-]
1

cm
]

2500

4355_8_0
4355_8_3
4355_8_6
4355_8_9
4355_8_12
4355_8_15
4355_8_18
4355_8_21
4355_8_24
4355_8_27
4355_8_30

strength

th [

stiffness

9
6
3

len
g

3000

0
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
width [cm] (sections)

Fig. 5. Left: shear test curves of section 4355_8. Right: Damage distribution (shear strength ratio) in plate 4355.

30
24

3000-3500
2500-3000
2000-2500
1500-2000
1000-1500
500-1000
0-500

18
12
6
0

10

14

18

22

26

30
24
18
12

26

22

18

14

10

2500-3000
2000-2500
1500-2000
1000-1500
500-1000
0-500

30
24

2000-2500
1500-2000
1000-1500
500-1000
0-500

18
12

30

26

22

18

6
0

Fig. 6. Coupling evaluation techniques: left: visible damage, centre: ultrasonic scan of depth of defects, right: shear strength.

W. Riedel et al. / International Journal of Impact Engineering

Fig. 6 correlates visual surface damage, depth of delaminations and strength reduction for all three
plates. Quantitative comparison of all three techniques provides:
Secondary damage by isolated cloud fragments causes substantial delamination around the
primary damage zone.
Although hardy visible in sections, delamination in the secondary damage zone reduces the
shear strength by 10% to 30%.
Ultrasonic scanning can detect secondary damage with high resolution.
Primary damage results in strength and stiffness losses from 60% to 100%.

2.5 Simulation of HVI Damage Experiments

The axisymmetric simulation model of test 4355 was established using Smooth Particle
Hydrodynamics discretisation (AUTODYN [4], Fig. 7). A smoothing length of 0.2 mm throughout
allowed to resolve the composite panel with 29 particles through thickness. The Al2017A bumper
shield of thickness 2.0 mm was modelled to a radial extent of 75 mm (3750 particles), the nominally
5.7 mm thick Kevlar-epoxy target to 200 mm (29725 particles).

0%
0

0
0
1

10%

0.5
1

0.5
1
15%

20%

b) new model [1],[5]

c) sample
d) details damage, plastic strain
a) impacting fragment cloud
Fig. 7. (a) Simulated debris cloud impact , (b) delamination extension and central hole after HVI impact damage; (c)
comparison to experiment; (d) details of damage (D=0-1) and plastic hoop strain (0-20%).

More realistic simulations of the outer area with loading by isolated fragment impacts were
achieved by additional 3D models. However, meshing with 0.6 mm diameters or only three particles
through composite thickness already resulted in a total of 590 000 particles. Details on composite
material model and data for both approaches are given in part I [1]. Analysis of the simulation allows
the following conclusions:
The central perforation hole is very accurately predicted by the modelling approach. Earlier

W. Riedel et al. / International Journal of Impact Engineering

model approaches [6] with shock and orthotropic strength descriptions but simpler failure
models provided too large perforation holes (Fig. 1, left).
Extents of surface damage on front and rear face are well captured.
In-depth delamination is correctly modelled in the primary damage zone. Coalescence in
the secondary zone is partly replicated as damage and plastic strain in the fine
axisymmetric model with tendency to underpredict the extent.
The implementation of the model in 3D SPH discretisation provides in the coarse model
qualitatively good results. Higher mesh refinement with the associated computational effort
would be required to validate in detail the extent of delamination.
0

0.4

0.5
1

0.8

Fig. 8. Coarse 3D model with three nodes through thickness; left to right: mode I delamination damage (0-1) on front face, at
2.5mm depths and on rear face; representation in plate sections.

3. Plate Impact Damage

Locally varying loading conditions are an inconvenience of the above described debris cloud
damage experiments. Statistics become more important for the smaller damage amounts in the
secondary zone, making validation of limited degradation more difficult.
Therefore, small amounts of damage at relevant stresses and strain rates for hypervelocity impact
applications have been generated in additional experiments of uniaxial strain loading. Two types of
plate impact damage experiments have been performed on the basis of established flyer plate
methods. The first type of axial compression and subsequent axial tension (spallation) is described and
validated in paper I [1]. The following description will focus on the second type of multiple dynamic
compression and later radial extension.
A symmetric configuration of composite and aluminium plates was used to prevent heavy
delamination caused by superposition of release waves. In this way pre-damaged but solid, monolithic
plates could be produced to perform residual strength tests. Properties of plates and impact velocities
for all tests are summarised in Table 4.
Table 4. Impact conditions for plate impact damage experiments (plate diameters 50 mm)

Test No.
Projectile thickness
(Al+comp.)
Target thickness
(comp.+Al)
Impact velocity

2606
10.00 mm +
5.63 mm
5.63 mm +
10.00 mm
124.4 m/s

2607
9.96 mm +
5.63 mm
5.56 mm +
9.95 mm
206.6 m/s

2608
10.01 mm +
5.64 mm
5.57 mm +
10.01 mm
276.1 m/s

2609
10.00 mm +
5.56 mm
5.58 mm +
10.00 mm
255.6 m/s

2610
9.96 mm +
5.52 mm
5.52 mm +
10.00 mm
241.7 m/s

W. Riedel et al. / International Journal of Impact Engineering

Due to the symmetric impact conditions half of the projectile momentum is transferred to the target.
Considering the different impedances of composite (AFRP) and aluminium (Al) the up and the X-t
diagram can be derived according to Fig. 9. Pressure in the composite samples increases stepwise up to
the state (3,III) before release waves (5) and (V) from the backing and target free surface arrive inside
the composite plates. Subsequently the compressive stress is iteratively decreased by complex wave
superposition, but no tensile loading in through thickness direction occurs during the whole process.
However, tensile loading and complex stress-strain states occur when lateral release waves arrive from
the circumference. At these late stages, no one-dimensional analysis of stress and strain states is
possible, but the loading history can be numerically simulated in two dimensions, e.g. using
axisymmetric modelling approaches.

6,E

VI,
6

F
D

5
4

V
3

VI

III

IV

II

II,

2,

D,4

IV,
1,I

V2606 124m/s
V2608 276m/s

300

AUTODYN - ADAMMO

276 m/s

AUTODYN - ORIG. AMMHIS DATA

250
200
150
100
50

124 m/s

A
0

350

Free Surface Velocity [m/s]

Projectile (Al) compression

Projectile (Al) release
Projectile (AFRP) compression
Projectile (AFRP) release
Target (Al) compression
Target (Al) release
Target (AFRP) compression
Target (AFRP) release
3,III

stress

time

v0

0
0,

v0
F

0,A

5000

15000

20000

Time [ns]

particle velocity up

location x

10000

Fig. 9. Lagrange and X-t-diagrams of symmetric uniaxial strain damage experiment. Right: Measured and simulated free
surface velocities. Similar stress wave levels occur for baseline model AMMHIS [3] and new ADAMMO approach [1].

The free surface velocity measured with the VISAR during early stages of the plate impact damage
experiments is shown in Fig. 9, right. They were not used to derive directly constitutive data, but to
validate simulated free surface velocities of the early arriving stress waves. For the lowest impact
velocity of 124.4 m/s (No. 2606), almost no damage is apparent on the recovered samples. With
increasing impact velocity more stiffness loss can be assumed from visible inspection. For intermediate
to fast impacts up to 276 m/s delamination of the discs is observed.
3000
2500

2000
1500

124.4 m/sec stiffness

206.6 m/sec stiffness
241.7 m/sec stiffness
255.6 m/sec stiffness
276.1 m/sec stiffness
intact stiffness
average stiffness
124.4 m/sec Fmax
206.6 m/sec Fmax
241.7 m/sec Fmax
255.6 m/sec Fmax
276.1 m/sec Fmax
intact Fmax
average Fmax

1000

500

1500
1000
500

0
0

50

100

150

200

250

300

velocity [m/sec]

Fig. 10. Left: Plate impact damage samples recovered and sectioned after loading with 124.4 m/s and 276 m/s.
Right: Strength degradation in terms of stiffness and maximum shear loading.

Fmax [N]

V2608
visible
delaminations

2500
2000
stiffness [N/mm] .

V2606
no visible
delaminations

W. Riedel et al. / International Journal of Impact Engineering

In the same manner as the debris cloud damage panels, the preloaded plate impact samples were
sectioned to 20 mm strips and shear tested. Fig. 10 shows the measured shear stiffness and strength
decrease. Already 124 m/s impact loading caused strength and stiffness decreases by 25 to 30%. Above
220 m/s, the plates are mostly damaged with a residual strength of about 20%. Again, delaminations
hardly observable in sections have major effects on the residual strength.
The experiments were simulated using the composite model approach [1]. 0.125mm square
Lagrangian cells were used to model the thickness of the composite plates in axial symmetry. Fig. 15
shows the comparison of simulated free surface velocities of the aluminium target plate. Differences
were observed on details of shock amplitudes and deceleration by release waves. But the overall match
of acceleration during the first 20 s proved good replication of the momentum balance between
projectile and target plates. Interestingly, the earlier model approach [3] describing non-linear equation
of state with linear-orthotropic strength and instantaneous failure simulates very similar free surface
velocities. Obviously, the uniaxial strain compression properties are not strongly influenced by
additional dissipation from non-linear inelastic deformations as modelled in the new approach.
Exp.

Model [1] ,[5]: Through

thickness damage D11=0-1

Model [1],[5]: Plastic strain

0-10%
0%

Model [3]: Material status

elastic

1.8%
2.7%

2606
failure 11
124.4 m/s

5%

failure 11

0.25

2608
276.1 m/s

7%

multiple
failure modes

Fig. 11. Simulated plate impact damage tests: negigible damage, permanent strains below 3% for 124 m/s impact; strains up to
10% for 276 m/s impact, no delaminations simulated. Overpredicted failure simulation with earlier model status [3].

W. Riedel et al. / International Journal of Impact Engineering

Detailed consideration of simulated deformation and damage contours (Fig. 11) shows that the new
damage model replicates the intact composite samples. Delamination damage is not directly predicted
(as ORT DAM 11). But the slowest impact velocities cause plastic deformations up to 3% and up to
10% local deformation are reached in the samples impacted at 276 m/s. Compared to the previous
model approach with linear-orthotropic strength up to failure, much improved prediction of material
states is observed. With the earlier model, both loading types caused complete through thickness
delamination, which was not observed experimentally.
4. Summary and Outlook
Delamination damage in the aramid-weave-epoxy composite panels impacted by space debris
clouds extends far beyond the area of primary damage into areas, where only isolated impacts are
visible on the surface. In the secondary zone of the debris cloud damage test ultrasonic testing detected
very fine areal delaminations, which were hardly visible in sections of samples.
However, subsequent destructive shear testing across the samples showed the important mechanical
effects of small delaminations on shear stiffness and strength. The following damage quantities could
be identified:
The zone of primary damage with high densities of impacting debris resulting in obvious
surface delamination correlated with strength and stiffness losses from 100% to 70%.
In the zone of secondary damage with low hit densities and isolated impacts, delaminations
extended almost to the external limit of debris impacts. They caused strength and stiffness
losses from 20% to 60%.
Loading by hypervelocity debris clouds results in strong local variations, especially in the
secondary zone of the debris cloud damage experiments. These variations make validation of numerical
methods in the area of limited degradation more difficult. Therefore, plate impact damage experiments
were designed to create smaller damage amounts under well defined loading conditions at relevant
strain rates. A number of samples ranging from weak damage (<30%) to strong damage (80%) could be
produced and recovered. Sectioning, visible inspection and shear testing again underlined the effect of
hardly visible in-depth delaminations on strength degradation. Free surface velocities were recorded
using VISAR techniques to validate numerical simulations.
Numerical Simulations with the SPH and mesh based hydrocode AUTODYN [4] of both debris
cloud damage and plate impact damage experiments were performed with the new composite model
introduced in [1] and [5]. For both loading types improved damage and directional plasticity patterns
could be described. In contrary to excessive delamination in earlier approaches [3], the new model
provides contained delamination with tendency to slight underprediction in areas of weaker loading.
But summing up, deformations and damage of the aramid-weave-epoxy composite plates (ISS
configuration) could generally be well simulated.
The modelling approach seems generally applicable to high strength composite structures, as they
show similar deformation phenomena, but more brittleness. Predictive simulations of aluminium
honeycomb structures with face sheets of carbon fibre reinforced plastics (CFRP) gave promising
deformation and damage patterns (see Fig. 12 and [5], [7]). These types of structures are currently
applied in numerous satellites and spacecraft.

W. Riedel et al. / International Journal of Impact Engineering

3
1
2
4
5

Fig. 12. Preliminary application of new composite model approach [5] to a CFRP-aluminium honeycomb structure
(ENVISAT). Analysis of hit point influence on debris cloud dispersion [7].

Acknowledgements
The authors gratefully acknowledge the important technical contributions of Frank Schfer, Holger
Voss and Jochen Peter. We would like to express our thanks to Michel Lambert from ESA/ESTEC for
funding and directing the described research.
References
[1] Clegg RA, White DM, Riedel W, Harwick W. Hypervelocity Impact Damage Prediction in Composites,
Part I Material Model and Characterisation, submitted to Jn. Impact Engn. 2005
[2] Clegg RA, White DM, Hayhurst CJ, Riedel W, Harwick W, Hiermaier S, Advanced Material Models and
Material Characterisation Techniques for Composite Materials subjected to Impact and Shock Wave
Loading, Journal de Physique IV 2003, 110: 311-316
[3] Hiermaier SJ, Riedel W, Hayhurst CJ, Clegg RA, Wentzel CM, Advanced Material Models for
Hypervelocity Impact Simulations, EMI report no. E43/99, ESA CR(P) 4305, 1999.
[4] N.N., AUTODYN, Theory Manual, Century Dynamics Ltd. Horsham, UK, 2003
[5] Riedel W, Harwick W, White DM, Clegg RA. ADAMMO Advanced Material Damage Models for
Numerical Simulation, ESA CR(P) 4397, EMI report I 75/03, Freiburg October 31, 2003
[6] Soden P, Hinton M., Kaddour A. Failure criteria in fibre reinforced polymer composites, Composites
Science and Technology 1998, 58, Special Issue.
[7] Ryan S, Riedel W, Schfer F. Numerical Study of Hypervelocity Space Debris Impacts on CFRP/AL
Honeycomb Spacecraft Structures, International Astronautical Congress, Vancouver, 2004