You are on page 1of 11

Instrumented drop weight

impact testing of cross-ply and


fabric composites
J.D. WINKEL * and D.F. ADAMS **
(* EDO Corporation/** University of Wyoming. USA)
A state-of-the-art instrumented drop wei ght impact test system
developed at the University of Wyomi ng was used to investigate the
impact performance of thin, simply-supported composite laminates.
System calibration, data acquisition and data reduction techniques
developed for this impact test system, which makes use of a
piezoelectric force transducer, are presented, along wi th insights into
system resonance characteristics. Composite material plates were
tested to identify performance differences between cross-ply and fabric
material forms. The six composite material systems investigated
included cross-ply and fabric laminates of Hercules AS4 graphite,
DuPont Kevlar 49 and Owens-Coming E-glass fibres impregnated wi th
a Hercules 3501- 6 epoxy resin. Test results are presented along wi th
results of a literature review in this area.
Key words: composite materials; laminates; i nstrumented i mpact
drop wei ght i mpact E-glass fibres; graphite fibres; Kev/ar fibres;
fabric; epoxy resins
During the past two decades, composite materials have
been used in an increasing number of hardware
applications that are susceptible to impact damage.
The realization that some of the current advanced
composite materials, particularly those incorporating
graphite fibres, are far more susceptible to low-velocity
impact damage than glass fibre composites has spurred
the desire for a better understanding of this critical
performance parameter. Future applications of
advanced composites in components that will be
susceptible to impact damage, such as aircraft wing
primary structures, ensure that interest in the impact
properties of composites will remain high for some
time to come.
Most of the prior effort aimed at understanding
composite impact damage mechanisms has centred
around the use of Charpy and lzod tests, and
particularly instrumented versions of these tests. 1
Although widely used, these tests are not necessarily
suited to an understanding of composite material
response since the test geometry often does not
represent the end-use application of the composite. 2
Instrumented drop weight impact test (DWIT) systems
have the advantage of more closely approximating the
impacted plate configuration typical of a variety of
composite material applications.
The present research had three objectives. First, an
instrumented DWIT system was designed and put into
operation at the University of Wyomin~ Second, this
instrumented DWIT system was used to explore impact
response differences between balanced-weave fabric
laminates and equivalent cross-ply tape laminates.
Finally, a comprehensive literature survey of ballistic
and instrumented DWIT impact results for fabric
laminates was performed. Only the first two objectives
are addressed here, additional details being given in
Reference 3.
Many advantages have been cited in the literature (see,
lot example, Reterences 4-7) in support of both fabric
and cross-ply laminates. However, relatively little
attention has been focussed on a comparison of these
two material forms with respect to their impact
properties. Consequently, the present study was
initiated to address this lack of data. As noted in
Reference 1, much of the prior impact testing of
composites has been performed using hybrids. The
present results are based only on non-hybridized
laminates, in the hope of providing a simpler view of
impact phenomena.
SYSTEM DESIGN AND CALIBRATION
A variety of possibilities can be explored when
designing an instrumented drop weight impact" tester.
Drop towers such as a guide-rod configurations and
0010-4361/ 85/ 040268-11 $03.00 1985 Butterworth Et Co (Publishers) Ltd
268 COMPOSITES. VOLUME 16 . NO 4 . OCTOBER 1985
Al umi ni um F ~ Pulley assembly
U | S.~r~
mechanism ~ , , , steel I
. . . . I
Guide ro Nicolet Explorer III
storage
o ~ i l l o ~ c o p e
wi t h f l oppy disc
j ? ; : 2 , ., ..o..,
Slotted Model 208A05 PCB force transducer
Specimen
I I Anvil
infrared I - ~ - - - - = Motor
transistor
!
Model 464A PCB [
charge amp
I
Fig. 1 University of Wyoming drop weight impact test (DWIT) tower
Fig. 2 University of Wyoming instrumented drop weight impact syst em
Gardner-style drop tubes 8,9 may be used. Instru-
mentation may vary from impact tups fitted with strain
gauge bridges to accelerometers attached to the cross-
head at a location removed from the tup. However, any
system will typically consist of three basic components:
1) a drop tower that guides a falling weight towards the
specimen;
2) hardware and software dedicated to data acquisition
and reduction; and
3) a force transducer.
Drop t ower
The drop tower in the University of Wyoming system is
generally patterned after commercially available, guide-
rod style drop towers, for two reasons. First, much of
the reported literature available for comparison with
the present results is based on testing conducted using
drop towers manufact ured by Effects Technology, Inc
(see, for example, References 8, 10 and 11). Second, the
Mechanical Engineering Depart ment at the University
of Wyomi ng already owned a Barry Controls model
VP-35 Shock Machi ne which could be modified to be
roughly equivalent to an Effects Technology, Inc model
8000 system. A drawing of the modified tower is
presented in Fig 1. The overall test system is shown in
Fig 2.
Modifications to the model VP-35 Shock Machine
consisted of replacing the guide-rods to extend the
drop height, adding a vertical support and base plate to
increase stability and rigidity of the drop tower and
mi nor modifications to the pulley and cross-head
assemblies. The addition of the base plate to the
original model VP-35 base resulted in a base weight of
320 kg The rated maxi mum cross-head weight for the
system is approximately 32 kg~ In the present study, use
was made of a 16 kg cross-head.
The University of Wyoming' s instrumented DW|T
system was designed to incorporate various tup and
anvil configurations. This flexibility allows the drop
Plan view
/
I
1- - -
~ radius chamfer typical
Anvil
Mountin 9 plate
Side view
Fig. 3 Anvil configuration, University of Wyoming DWlT system
COMPOSITES. OCTOBER 1985 269
tower to be used for non-st andard tests, plus Charpy,
Izod and three-point bend tests in addition to normal
plate i mpact tests. The only constraints i mposed on test
type by the drop tower are that the anvil and tup fit
within the 250 250 mm space available between the
guide rods Since the thrust of the present work was to
investigate the normal i mpact behavi our of plates, a
152 152 105 mm anvil speci men support was bui l t
This fixture, providing a 127 127 mm simply-
support ed edge condition for the speci men plates, is
shown in Fi g 3.
It should be noted that cross-head friction was not
negligible despite the fact the two of the model VP-35
cross-head bushi ngs floated on ' O-ring' seals. Both
bushi ng rod clearances and rod perpendi cul ari t y were
unsuccessfully investigated in an effort to further
reduce cross-head friction. The velocimeter instru-
mentation, however, allowed this friction to be taken
into account when the cross-head velocity at i mpact
was calculated. Consequently, the presence of friction
did not affect the validity of test results. The most
probabl e cause of this friction (bi ndi ng due to rotation
of the cross-head about a perpendi cul ar to the guide-
rod length) could be eliminated in future cross-head
designs by the use of kinematic design principles. For
example, reducing the number of contact points
between the guide rods an~l the cross-head and the use
of roller bearings would al most certainly i mprove the
friction performance of the DWIT system.
Data acquisition and reducti on
A Nicolet Expl orer III model 206-2 digital oscilloscope
was used to record and store the data output by the
velocimeter and the piezoelectric force transducer.
Dat a were then reduced using a Hewl et t -Packard
21MX E-series mi ni comput er to preprocess the data
and create a data file. This data file was then
transferred to a Tekt roni x 4027 Interactive Termi nal
which was linked to the University of Wyomi ng' s CDC
Cyber 760 computer. Dat a were then reduced to
provi de the maxi mum i mpact load (Pm) and the
correspondi ng energy (Em), as well as the total energy
absorbed by the speci men during i mpact (ET). In
addition, Fouri er filtering techniques could be imple-
ment ed to smoot h out put data curves These values,
along with the complete force vs time curve, were then
plotted using a Textronix 4662 Interactive Digital
Plotter.
The Nicolet Expl orer III model 206-2 digital oscillo-
scope is capabl e of storing up to 4096 data points per
test in digital form on a small magnetic disc. The scope
was triggered by the velocimeter, which consisted of a
slotted steel flag of known di mensi ons that cut an
infrared transistor beam. The slotted flag provided data
from which pre- and post -i mpact velocities could be
reduced. The velocimeter was equivalent to commer-
cially available units that typically have an accuracy of
1% or better, ~2 with the exception that the velocimeter
flag used here allowed det ermi nat i on of post -i mpact
cross-head speeds. Since the velocimeter data were
necessary for force/ time data reduction, one channel of
the Nicolet was dedicated to velocimeter out put The
other channel was dedicated to the force t ransducer
out put When operat ed in this mode, the Nicolet stored
2048 data points for each channel, at typical time
intervals of 20 to 200 ms.
Stored data were reduced according to the following
procedure. First, i nst ant aneous cross-head velocities
were calculated using the relation:
v(i) = v o - Av (1)
where
v(i) ---- cross-head velocity at time i;
v o = cross-head velocity at impact;
Av = F( i ) +F( i +I ) At
2 m'
m = cross-head mass;
F(z) = force at time i; and
At = time interval between data points.
Second, the i nst ant aneous velocity was used to
calculate the energy absorbed by the specimen up to
time i using the relation:
E(/) ~- E( i - 1) + F( i ) +F( i +I )
2 " vi" At (2)
Third, the maxi mum i mpact load sustained by the
speci men was comput ed using a sorting routine in the
data reduction program. Finally, the reduced data were
plotted.
Force transducer i nstrumentati on
The basic geomet ry of the i mpact tup was selected to
be a 12.7 mm di amet er steel rod with a hemi spheri cal
penetrating end. This geometry was chosen due to its
earlier use by Wardl e and Tokarsky? Takeda et aP 3
concl uded from ballistic data that a hemi spheri cal
shape produced local crushing of the t opmost l ami na
rather t han dissipating the i mpact energy over a larger
area as was characteristic with a bl unt -nosed impacter.
The material system they used was cross-plied
Scotchply 1003 glass/ epoxy. Al t hough this work is
arguabl y irrelevant in the non-ballistic context of the
present study, it does suggest that i mpact data
generated with a certain tup geometry should be used
cautiously in appl i cat i ons involving different i mpact
geometries.
Various options were considered in developing
sufficiently sensitiwt tup instrumentation. This review
resulted in the selection of a piezoelectric force
transducer, mount ed between the tup and the cross-
head. This option was selected for the following
reasons:
1) commerci al l y available piezoelectric force trans-
ducers offered the desired sensitivity;
2) piezoelectric force transducers offered off-the-shelf
reliability;
3) similar instrumentation (piezoelectric accelerometers)
had been successfully used by Zoller t4 and Stellbrink
and Aoki; 15 and finally
4) commerci al l y available force transducers permi t a
conveni ent tup construction in which the gauge is
connect ed in series with the tup and the tup base-
plate, which is in turn attached to the cross-head.
This configuration keeps the t ransducer close to the
poi nt of i mpact
270 COMPOSITES. OCTOBER 1985
Fig. 4 Photograph of tup assembly (anvil and v el oci met er ar e al so
shown)
The specific force t ransducer selected was a PCB
Piezotronics, Inc model 208A05 unit with a rated
nomi nal l oad ~ of 22 kN. The assembl ed unit is shown
in Fig, 4_
System calibration
The University of Wyomi ng' s DWlT system was
carefully calibrated in bot h static and dynami c modes.
Since i mpact energy is calculated using Equat i on (2),
the accuracy of the test results depends on valid
measurement s of i nst ant aneous force ( F ' O , i ncrement al
time (At), and the cross-head mass. These quantities
are needed to det ermi ne the initial i mpact velocity, vi,
indicated in Equat i on (2). Consequently, system
calibration effects centred on these quantities.
The weight of the cross-head was precisely measured
using a calibrated Ohaus 20 kg bal ance and was
det ermi ned to be 15.88 +__0.0014 kg. The accuracy of
velocities calculated from the velocimeter output trace
was highly dependent upon the precise measur ement
of the slot and space di mensi ons of the velocimeter
flag, The velocimeter flag was precision-milled and
measured to an accuracy of_--t-0.025 mm. The
velocimeter was calibrated by compar i ng velocities
calculated from the velocimeter data with three
different known cross-head di spl acement rates in an
MTS model 810 servo-hydraulic test frame. The
maxi mum mean error between the DWIT vel oci met er
and the MTS test frame data was 1.08%. Although the
calibration was limited by the range of cross-head
di spl acement rates available on the MTS, there was
excellent agreement between the measured velocimeter
and MTS velocities.
Electrical component s such as the Nicolet Expl orer III
oscilloscope and the PCB 464A amplifier played a role
in det ermi ni ng F i, v i and At Consequently, these two
component s were calibrated using known outputs of a
Hewl et t -Packard model 6204B power supply and a
Hewl et t -Packard model 202C low-frequency oscillator.
Although the model 208A05 piezoelectric force
t ransducer was designed for i mpact applications, the
literature review revealed that it had not been used in
i mpact of composi t e materials prior to the present
application, Early moment um bal ance (dynamic)
calibrations run with the model 208A05 on an
"inelastic' medi um provided unexpected results and
further i mpet us for a t horough calibration of the
system. Static calibration data were provided with the
gauge by PCB Piezotronics, Inc. This calibration was
verified by compressively loading the entire tup
assembly in an Inst ron model 1125 test frame.
Dynami c calibration was performed with the gauge
mount ed both in the drop weight system and isolated
from it. The literature reviewed did not report results of
dynami c calibration of i nst rument ed drop weight
i mpact systems. Consequently, efforts to dynami cal l y
calibrate the University of Wyomi ng' s DWI T system
were deemed to be useful in demonst rat i ng the general
validity of i nst rument ed i mpact testing. Initial dynami c
calibration efforts centred on confi rmi ng the
conservation of moment um when the tup i mpact ed an
assumed inelastic medi um (clay). Forces generated by
the t ransducer for this test configuration were
apparent l y too high, and moment um appeared not to
be conserved. Similar force discrepancies were
obt ai ned in low-energy tests of eight-ply, quasi-
isotropic and cross-plied graphite/ epoxy plates.
The above tests were repeated with a variety of other
gauges in an effort to determine whet her moment um
discrepancies resulted from a bad gauge. From these
tests it was concl uded that the calibration procedure
was fundament al l y flawed for two reasons. Fi rst the
assumpt i on of an inelastic medi um was not valid.
Second, the additional mass of clay adhering to the tup
needed to be taken into account for moment um to be
conserved during rebounc[
Some additional work was then done to isolate the
cause of the low-energy graphite/ epoxy plate results.
This work consisted of subjecting plates of various
materials to very low i mpact energies. Impact energies
were low enough so that the plates were undamaged,
and simply caused the cross-head to rebound. This
rebound phenomenon was assumed to be a perfectly
elastic collision, ie the cross-head velocity was expected
to go to zero at the location of maxi mum force, and
the maxi mum i mpact energy was expected to
correspond to the potential energy of the cross-head.
Experi ment al drops verified bot h of these expectations.
Therefore, it was concl uded that the i nst rument at i on
was not responsible for the noticed force discrepancies.
These discrepancies appeared, instead, to be due to an
undet ermi ned plate/ tup interaction phenomenon. It
further appeared that discrepancies were present only
when the i mpact energy required for plate penetration
COMPOSITES. OCTOBER 1985 271
became a very significant portion (90% or more) of the
cross-head potential energy. Such a test condition was
not representative of the present testing; therefore no
further efforts were made to isolate the observed
phenomenon.
The University of Wyomi ng DWIT system was
dynami cal l y calibrated by abandoni ng the ' inelastic'
test and instead correlating the loss of cross-head
kinetic energy with energy absorbed by plate
specimens. This calibration met hod was verified using
thin Plexiglass plates, and became a st andard feature
of subsequent testing
SPECIMEN PREPARATION
All specimens were prepared from prepregs produced
by the present authors, in order to remove material
processing as a variable which mi ght effect the final
results. Balanced, plain-weave fabric l ami nat es and
equivalent l ami nat es made from cross-plied, uni-
directional tapes of Kevlar 49/3501-6 epoxy, AS4
graphite/ 3501-6 epoxy and E-glass/ 3501-6 epoxy were
chosen for the current testing~ Both filament tow and
fabric material forms were i mpregnat ed with Hercules
Tabl e 1. Av er age pl at e t hi ck nesses and f i bre
v ol umes
Materi al Lay-up Average Average
thickness* fibre
( r am) vol ume**
( 9 0 )
AS4 graphite
Cross-ply [0/9015s 3.00 (0.05) 42.8 (0.02)
Fabric [0/9011o 2.16 (0.02) 49.7 (0.02)
Kevlar 49
Cross-ply [0/9015s 2.72 (0.03) 42.9 (0.02)
Fabric [0/9011o 2.34 (0.02) 46.2 (0.01)
E-glass
Cross-ply [0/9013s 2.54 (0.06) 45.0 (0.04)
Fabric [0/90110 1.85 (0.03) 36.6 (0.07)
Coefficients of variation are given in parentheses
*Plate thickness is an average of 16 measurements (four per specimen)
**Fi bre vol ume is an average of three measurements, per ASTM
D3171- 76
3501-6 epoxy resin at the University of Wyomi ng
Unidirectional fibres were wet filament wound on a
drum wi nder which had been previously developed at
Wyomi ng Fabri c material was i mpregnat ed using a
cold slurry process which was developed as part of the
present effort
Plates were laid up by hand and cured using the
prepreg described above. All plates were carefully
compact ed during lay-up and were then cured in a
bl anket press This press consists of a pressurized
silicone rubber bl anket in a heated cavity which
applies the desired level of pressure and heat to the
laminate. Steel caul plates were used when curing the
laminates in the bl anket press, to ensure even
compaction.
All l ami nat ed plates were approxi mat el y 380 380 mm
in size. These large plates were C-scanned and then
cut into four 152 mm square i mpact specimens, three
tensile speci mens and three losipescu shear specimens.
All plates were nomi nal l y 2.5 mm t hi ck Actual plate
thicknesses and fibre vol umes are presented in Tabl e 1.
I mpact speci mens were cut using an abrasive cutting
wheel ( LECO type 811-035) mount ed on a radial arm
saw. Tensile and shear specimens were cut using the
same type of abrasive cutting wheel mount ed on a
surface grinder. The grinder provided a firm surface on
which to cut the specimens, and thereby mi ni mi zed
edge delaminations. Based on C-scans of scrap E-glass,
Kevlar and graphite composite specimens (which were
used to verify test procedures prior to performi ng the
present testing), i mpact specimens were expected to
exhibit little sensitivity to the quality of specimen edge
preparation. Consequently, the radial arm saw
provided the most efficient means to cut i mpact
specimens.
C-scans were used as the basis for quantitative
compari sons of i mpact damage area in the fabric and
tape specimens. All scans were performed with a Sonic
Inst rument s Mar k IV Ultrasonic Flaw Detector.
TEST RESULTS
Static material properties
Tensile and shear static tests were first performed using
material cut from the same l ami nat ed plates as the
i mpact specimens. Results of these tests are presented,
along with predicted values, in Tables 2 and 3.
Tabl e 2. T ensi l e propert i es of t he vari ous mat er i al s t ast ed
Composite Number Meas ur ed Reported Number Me a s u r e d Reported
material of modul us modul us of strength strength
system specimens (GPa) (GPa) specimens (MPa) (MPa)
[ Reference] [ Reference]
AS4 graphite/epoxy
Cross-ply 3 52.6 (0.056) 51.2 [18] 3
Fabric 3 65.2 (0.018) 59.4 [18] 3
Kevlar 49/epoxy
Cross-ply 1 26.1 (--) 30.1 [19] 2
Fabric 3 29.5 (0.054) 32.2 [19] 3
E-glass/epoxy
Cross-ply 3 22.2 (0.039) 19.6 [20] 3
Fabric 3 20.8 (0.018) 16.9 [20] 3
535 (0.072) 738 [18]
697 (0.056) 855 [18]
470 (0.117) 496 [19]
437 (0.117) 531 [19]
340 (0.383) 490 [20]
157 (0.082) 421 [20]
Coefficients of variation are gi ven in parentheses
272 COMPOSITES. OCTOBER 1 985
Table 3. Iosipescu shear properties of the various materials tested
Composite Number Measured Reported
material of modulus modulus
system specimens (GPa) (GPa)
[Reference]
Number Measured Reported
of strength strength
specimens (MPa) (MPa)
[Reference]
AS4 graphite/epoxy
Cross-ply 2 3.31 (0.073) *
Fabric 3 3.86 (0.105) *
Kevlar 49/ epoxy
Cross-ply 2 1.79 (0.347) 1.52 [19]
Fabric 3 1.59 (0.102) 1.52 [19]
E-glass/epoxy
Cross-ply 3 4.41 (0.177) *
Fabric 2 4.28 (0.057) *
2 157 (0.18) *
3 150 (0.03) *
2 9 5 ( 0 . 15) 60 [19 ]
3 99 (0.17) 60 [19]
2 108 (0.07) *
2 124 (0.04) *
Coefficients of variation are given in parentheses
"Values not available
Tensile tests were performed according to ASTM
St andard D3039-76, except that r ecommended width-
to-length ratios could not be mai nt ai ned due to the
limited amount of material available for specimens.
The r ecommended width-to-length ratio given in
ASTM D3039-76 is 1:5; actual speci mens ranged from
approxi mat el y 1:2.5 to the full 1:5 ratio. Some of the
di screpancy between strength predictions and test data
can probabl y be attributed to stress concent rat i on
effects at the grips i nduced in specimens havi ng shorter
lengths t han those specified by ASTM D3039-76.
Shear tests were performed using the Iosipescu shear
test methodY ~-23 The test met hod was limited by the
fact that the strain gauges used to measure shear strain
on the Iosipescu speci men had a linearity limit of 3%
strain. Thus, the stress/ strain curves obt ai ned probabl y
reflected bot h material and strain gauge non-linearities
above 3% strain. Shear strength values obt ai ned (see
Tabl e 3) were higher t han those listed in the literature
for unidirectional material. Although little shear
strength difference was expected between unidirec-
tional and cross-plied laminates, this result is likely to
be valid. One possible expl anat i on for the higher shear
strengths exhibited by the cross-plied composi t es is
that perpendi cul arl y oriented fibres tend to bl unt
cracks that at t empt to propagat e between plies. Since
shear failure modes are currently not well understood,
future work in this area would be quite worthwhile.
Both tensile and shear properties obt ai ned
experimentally were compar ed with values available in
the literature. In those cases where adequate data were
not found, the fibre supplier data sheet values were
used, along with a rule-of-mixtures estimation, to
generate values for compari son. Matrix property values
used in the rule-of-mixtures estimations were
previously measured at the University of Wyomi ng ~7
Static test results indicated that the materials used in
the i mpact test speci mens were of sufficient quality to
remove material processing as a variable that would
affect the i mpact results.
I mpact test resul ts
I mpact tests were carried out on 24 specimens, four
each for the cross-ply and fabric forms of the three
selected material systems. Specimens were chosen to be
152 mm square plates approxi mat el y 2.5 mm thick~ so
that the results could be compar ed with those obt ai ned
by other investigators (see below). Speci men and
support (anvil) geometries resulted in a simply-
Table 4. Impact properties of t he various materials tested
Composite Number Impact Specimen Peak force, Energy to Total energy, Energy loss,
material of velocity thickness Pm peak, E m E T AKE
system specimens ( ms -1) (mm) (kN) (J) (J) (J)
ASS graphite/epoxy
l O-ply cross-ply 3 3.725 (0.02) 2.997 (0.02) 3.22 (0.04) 8.76 (0.18) 21.65 (0.07) 20.65 (0.06)
l O-pl y fabric 4 3.645 (0.03) 2.159 (0.01) 2.60 (0.01) 8.51 (0.13) 15.55 (0.05) 15.00 (0.07)
Kevlar 49/ epoxy
l O-pl y cross-ply 4 3.636 (0.03) 2.718 (0.01) 3.15 (0.10) 14.57 (0.12) 21.76 (0.02) 21. 20 (0.02)
l O-pl yfabri c 4 3.725 (0.02) 2.337 (0.01) 2.14 (0.02) 10.26 (0.05) 17.92 (0.04) 17.71 (0.04)
E-glass/epoxy
6-ply cross-ply 4 3.609 (0.03) 2. 540 (0.03) 7.19 (0.08) 35.43 (0.12) 46.73 (0.01) 47. 70 (0.02)
l O-pl ycross-pl y 2 3.243 (0.20) 3.785 (0.01) 10.97 (0.10) 56.35 (0.16) 81.77 (0.10) *
l O-ply fabric 4 3.71 2 (0.02) 1.854 (0.01) 1.83 (0.02) 7.21 (0.08) 13.33 (0.02) 11.88 (0.07)
Coefficients of variation are given in parentheses
* T hese data wer e obtained before mi nor modifications to the vel oci meter wer e made whi ch permi tted post-impact velocities to be measured wi th a high degree of
accuracy
COMPOSITES. OCTOBER 1985 273
supported area that was 127 mm square.
All specimens were impacted by a 12.7 mm diameter
tup mount ed on a failing cross-head that weighed
15.88 kg. The nomi nal impact velocity was 3.66 ms -l,
corresponding to a drop height of approximately
876 mm. The cross-head was manually released after
the PCB model 464A amplifier and Nicolet
oscilloscope settings were verified. Typical settings
consisted of runni ng the amplifier in the short time-
constant mode with a full-range output of 8.90 kN. The
oscilloscope was typically set to sample data at 20 ms
intervals, with a full scale range of__+10 V. The guide
rods were liberally lubricated to minimize friction and
thus reduce scatter in the resulting impact velocities.
Impact velocity, average specimen thickness, peak
impact force (Pm) and corresponding impact energy
(E,n), and total impact energy (ET) is given in Table 4
for each specimen. Cross-head kinetic energy losses are
also presented. Compari son of these data with total
impact energy data supports the previously stated
conclusion that the University of Wyomi ng DWlT
system exhibits good dynami c calibration.
Instrumented impact testing represents a powerful tool
for underst andi ng materials behavi our during impact,
and can be correlated with more conventional concepts
of impact performance. This is illustrated particularly
well when Fig: 5 is contrasted with Figs 6-11. Fig. 5
shows the brittle behavi our exhibited by one of the
Plexiglass calibration specimens; Figs 6-11 illustrate
considerably tougher impact behavi our and are
representative of the six composite material
configurations tested. Additional plots, along with
photographs of failed specimens, are included in
Reference 3. The results indicate that each of the
composite materials tested continued to carry some
load beyond the maximum. Brittle materials such as
the Plexiglass achieve a maxi mum impact loading
beyond which the force rapidly drops to zero. In Fi g 5,
the sudden drop in the force/ time curve corresponded
to shattering of the specimen.
" 3
_ 3
|
o
= = = =
i I Prom 2.28 kN
Era= = 5. 0 5 J
E t m t 5 . 0 5 J
5 . 0
0 5 tO 15
Ti me (ms)
Fig. 5 Typical impact response of a 6. 10 mm thick Plexiglass specimen
A
2 . 5 " ' ~o
. . J
7.5
Pro= = 3.32 kN
Em I = 7.44 d
Etot, I = 2102 J
5.0
_ J
2.5
0 , I . . . . 0
0 2.5 5.0 Z5
Time (ms)
Fig. 6 Representative impact response of an AS4/ 3501- 6 [0/9015 s
cross-ply laminate
30
" - 3
20
U.I
I0
4
Pro== = 2.58 kN
E m, x = 9 . 5 8 J
Eto, = = = 1 5. 58 d
15
5
A
z
. . J
I
0 , I , i i i 0
0 2.5 5.0 7.5
Time (me)
Fig. 7 Representative impact response of an AS4/ 3501--6 [0/90110
fabric laminate
I0 ~-
b J
7.5
'max = 3.17 kN
Emo = = 14.04 J
E~t,t = 22.40 d
5.0
25
L U
2,5
I I i I ~ 0
O0 2.5 5. 0 7.
Time (ms)
Fig. 8 Representative impact response of a Kevlar 49/ 3501- 6 [0/9015S
cross-ply laminate
2 74 COMPOSI TES . OCTOBER 1 9 8 5
IOO
DISCUSSION
It was observed in the present i mpact testing that
results were dependent upon plate thickness. The force
and energy data of Tabl e 4 have consequently been
normal i zed by dividing t hrough by the appropri at e
plate thicknesses. The normal i zed values are given in
Tabl e 5. These normal i zed data will be used when
discussing the present results.
The data of Figs 6-11 were processed to increase their
usefulness and clarity. For example, the data were run
through a Fourier t ransform filtering process to remove
resonance oscillations from the data signal. Fig. 12 is a
non-filtered data trace correspondi ng to F'ig. 11. Data
were t ransformed with the goal of achieving as "true" a
data signal as possible by limiting use of the filtering
process. The filter threshold was set to smoot h out only
high frequency oscillations. Typical oscillation
frequencies, such as those shown in Fig. 12, were of
two types. A low frequency oscillation was present in
all data; this is discussed in more detail below. A
second, higher frequency oscillation with a range of
2 5
'max = 6. 49 kN
Ema x = 90. 50 J 50
Etot= I = 47. 20 J
7.5
5 ~
~ v
50
" 25
2.5
0 ' ' ' ~ 0
0 25 5.0 7.5
Ti me ( ms)
F ig. 1 0 Re p r e s e n t a t i v e i mp a c t r e s p o n s e of an E- gl as s / 3501- 6 [ 0 / 9 0 ] 3 s
c r os s - pl y l ami nate
2.~
1.5
1.0
0 5
o
0 2.5
Ti me (ms)
'max = 1 . 8 2 k N
Ema x = 6. 94 J
Etot= I = 13. 46 J
I 0
uJ
5
I , i i i
Fig.11 Represent at i ve i mpact r e s p o n s e of an E- gl as s / 3501- 6 [ 0 / 9 0 ] 1 0
f abr i c l ami nate
T a b l e 5. N o r ma l i z e d i mp a c t f o r c e a nd e n e r g y
v a l u e s *
Composite Normalized Normalized
material system force, maximum energy
Pm/t (kN mm -1) Em/t (J mm -1)
AS4 graphite/epoxy
1 O-ply cross-ply 1.07 2.92
1 O-ply fabric 1.21 3.94
Kevlar 49/ epoxy
1 O-ply cross-ply 1.16 5.36
1 O-ply fabric 0.91 4.39
E-glass/epoxy
6-ply cross-ply 2.83 13.95
1 0 - p l y cross-ply 2.90 14.89
1 O-ply fabric 0.99 3.89
*Dat a of Tabl e 4 normal i zed by di vi di ng by t he cor r espondi ng pl ate
t hi cknesses
2 . 0
1.5
z
1.0
0 . 5
i
2 _ 5
T i m e ( m s )
/~m== = 182 kN
Emax : 6.58 J
Etota I = 1 2 . 7 7 J
Pmax = ?-.16 kN
Ema x = 10.18 J
Etoro I = 16.83 d 20
3
z .-o
g ~
_.1
10
I
[ I I f I
o 5 0
Ti me ( ms)
Fig. 9 Representative impact response of a Kevlar 49 / 35 0 1- 6 [ 0/ 9011o
fabri c l ami nate
I i i i
5. 0
IO
Ld
5
1
Fig. 12 Non-f i l t ered E- gl as s / 3501- 6 [ 0/ 9011 o f abri c l ami nate i mpact
dat a c o r r e s p o n d i n g t o Fig. 11
C O M P O S I T E S . O C T O B E R 1 9 8 5 2 7 5
10 to 20 kHz was also present in all tests and was
smoot hed by means of the Fourier transform filtering
process. The source of these high frequency oscillations
was not clearly pinpointed. The electrical charac-
teristics of the data acquisition hardware at rest were
investigated but revealed no periodic frequency of the
above magnitude; thus, electrical oscillation did not
appear to be a realistic source. A more likely source
may be related to the fact that the resonant frequency
of the gauge and the observed high frequency
oscillations have the same order of magnitude.
Each of Figs 6-12 exhibit a relatively low frequency
oscillation during the initial portion of the impact
event This oscillation has a typical frequency of
~1300-2500 Hz, and is similar to apparent signal
oscillations in data traces reported in the literature. H,24
Since it has a periodic character, this signal oscillation
does not affect impact energy test data. Because impact
energy is an integral of the force/ deflection data, the
oscillatory nature of the signal will be removed when
impact energies are calculated. Peak impact force
similarly does not appear to be affected since the
magnitude of these signal oscillations appears to be
damped down to a negligible amount at the peak
condition.
It was initially assumed that the observed 1300-
2500 Hz oscillation was due to vibration of the simply-
supported specimen prior to damage inducement.
However, the calculated 462 Hz natural frequency of a
5.94 mm thick Plexiglass plate that had been tested
earlier was considerably lower t han the observed
frequency. This disagreement was further investigated
by affixing a PCB model 308B10 accelerometer to the
centre of the Plexiglass plate. The mass of the
accelerometer was included in the above calculation of
plate natural frequency. When the centre of the plate
was tapped, a vibration frequency of 429 Hz resulted.
This mat ched closely the predicted natural frequency
of 462 Hz.
Subsequent investigation (dropping a weight ont o the
cross-head-tup assembly while it was resting on an
unbroken specimen) revealed transducer output signal
oscillations of 1670-2860 Hz. It was concluded that the
observed low frequency oscillation resulted from
resonance inherent in the conventional DWIT design.
This resonance could possibly be avoided by removing
the gauge/ tup assembly from the cross-head and
mount i ng it in such a way as to maximize its vibration
isolation from the guide-rods. This would necessitate
mount i ng the specimen on either the falling cross-head
or directly above the tup and striking its free edges.
In addition to the above conclusions, a look at the data
in Tabl e 5 confirms the expectation that E-glass/ epoxy
should out perform Kevlar 49/ epoxy which, in turn,
should out perform the AS4 graphite/ epoxy. The E-glass
fabric/ epoxy performance is regarded as anomal ous
due to its low fibre vol ume (see following discussion).
A review of representative pre- and post-impact
C-scans, along with visual inspection of the failed
specimens, led to the following conclusions, First,
impact of the fabric specimens resulted in smaller
damage areas t han were present in the cross-ply
specimens. Second, cross-ply specimens exhibited
back-face fibre splitting with the exception of the
Kevlar/ epoxy specimens. Third, the Kevlar/ epoxy cross-
ply specimens exhibited only slightly larger damage
areas t han did the fabric specimens. Consequently, it is
felt that reduced damage area might be a justification
for selecting fabric material forms in the case of
E-glass and graphite, but not in the case of Kevlar 49.
The remaining discussion will focus on impact force as
an indicator of specimen performance. Although
impact energies can be expected to mirror the trends
indicated by impact force data, they will not be used
in the following discussion for two reasons. First, the
use of impact energy obscures small magnitude or
short duration differences in comparabl e data traces
since it is an integrated quantity. Second, there is a
tendency on the part of some designers to assume that
comparabl e applied energies will yield comparable
impact performance. By discussing energies rather than
forces it is too easy to disregard factors (such as impact
velocity) that play a definite role in impact
performance of composites.
AS4 graphite~3501-6 epoxy
Little difference in performance was expected between
the AS4/3501-6 cross-ply and fabric laminates as a
result of the literature review? This conclusion was
particularly supported by Miller et al.X~ Their work also
suggested that incipient damage forces were slightly
higher for fabric laminates, and that peak force varied
linearly with thickness for thin specimens. The results
of the present testing generally support this conclusion,
and that there is little difference between peak impact
force (Pm) performance of AS4/3501-06 fabric and cross-
ply laminates when normalized with respect to
thickness. Thus, the peak force data shown in Table 5
are directly comparabl e since they have been norma-
lized with respect to thickness. After normalization, the
fabric specimens exhibited peak forces about 11%
higher t han the cross-ply specimens.
Kevlar 49/3501-6 epoxy
Work by Mi ner et aP suggests that plain-weave Kevlar
49 fabric absorbs more energy prior to damage t han
eight-harness satin-weave material, due to the greater
straightening effect possible in the plain-weave.
Consequently, it was expected that Kevlar fabric
specimens would provide better impact performance
than the comparabl e cross-ply specimens. Just the
opposite proved to be true. Comparisons of the peak
force data of Table 5 indicate a force 27% higher lor
the cross-ply laminate t han for the fabric, after
normalization with respect to thickness. A linear force-
to-thickness relationship was again assumed. Since this
result was unexpected, a closer look was taken at the
other available data.
Further discussion 25 with one of the authors of
Reference 10 revealed results similar to those reported
above. Thickness and lay-up data made available by
Wardle a5 allowed a more direct compari son of the
quasi-isotropic fabric and cross-ply data presented in
Reference 10, by suggesting the existence of a linear
force-to-thickness relationship. A direct compari son of
the reported fabric and cross-ply data remains difficult,
however, due to the use of somewhat different resin
systems and lay-ups between the fabric and tape
laminates. While the comparisons of present results
with those given by Wardl d ,25 should not be viewed
as conclusive, they do suggest that the present results
are plausible. Possible explanations for the superior
perl ormance of Kevlar 49 cross-ply laminates will be
discussed later.
276 COMPOSITES. OCTOBER 1985
E-glass/3501-6 epoxy
No literature was available that suggested an expected
behaviour for the E-glass fabric and cross-ply laminate
specimens. The applicability of linear force-to-
thickness scaling in the case of the graphite/ epoxy and
Kevlar/ epoxy specimens suggested, however, that a
similar relationship might exist for the E-glass/ epoxy
specimens.
A limited amount of testing was performed that tended
to confirm the existence of a linear relationship for the
specimen thicknesses being investigated. Demonstra-
tion of this relationship was particularly important
since it allowed compari son of disparate cross-ply and
fabric specimens. The loose weave of the E-glass cloth
used created difficulties in matching both the fibre
volume and thickness of the comparabl e cross-ply
specimens. Lami nat e processing samples, for example,
indicated that a fibre volume of 39% or more
corresponded to a poor quality, porous laminate. When
fibre volumes exceeded the above value, the laminate
lacked enough resin to physically fill the spaces
between fibres in the loosely woven fabric used in the
present testing Consequently, the fibre volume
mismatch between E-glass cross-ply and fabric
specimens was greater than that for either the graphite/
epoxy or Kevlar/ epoxy specimens. For the cross-ply
data (see Table 4), the thickness relationship was
linear, however.
After scaling the cross-ply laminate specimen data to
the corresponding fabric specimen thicknesses, average
peak forces of 5.25 kN and 1.83 kN were noted for the
cross-ply and fabric specimens, respectively. The fact
that the fabric performance was lower t han that of the
cross-ply material was consistent with the observed
failures. While all other through-penetrated specimens
clung to the tup after impact, the E-glass fabric/ epoxy
composites specimens hung very loosely from it. Most
displaced material also tended to move back into place
when the other specimens were removed from the tup;
this was not the case with the E-glass fabric composite.
Because of the differences in apparent failure modes
noted above, it is surmised that the poor impact
performance of the E-glass fabric laminate was an
artefact of the loose weave of the fabric style used in
these specimens. Tighter weaves with higher fibre
volumes would likely provide better impact
performance.
Correlations with theory
The complex nature of composite material impact
response has limited the development of a generalized
theoretical model. Much of the existing correlation
work is empirical in nature. Although empirical
approaches do provide general guidelines, it should be
noted that results obtained may not be reliable, and
there are bound to be many exceptions to the general
approximations obtained.
In the case of normal plate impacts, Coppa et aF 6
suggest that, as a first approximation, plate impact
energy is related to quasi-static flexural failure energy.
For a beam-type structure, this suggests that the impact
energy should be proportional to the quantity
tTe (or o2/E), where cr and e are the maxi mum stress
and strain of the beam at failure and E is the modulus.
Wardle and Tokarsky ~ have expanded this approach
and suggest that the impact energy should be propor-
tional to to2/E, where t is the specimen thickness and
and E are the specimen extensional (rather than
flexural) strength and stiffness.
Although this is a simple approach, it does predict the
general trends observed in the present testing Values
of the parameter to~/E have been calculated for each
material, using average specimen thickness values
given in Table 1 and the strength and modulus values
given in Table 2. These calculated energy parameter
values are listed in Table 6. The energy paramet er to~/E
is regarded here as only a qualitative measure of data
trends since it is derived from an approximate
formulation that will be particularly dependent on
specimen span-to-depth ratios, specimen edge
conditions, etc. The data trends which do appear to be
modelled by this simple energy parameter are:
1) similar maxi mum AS4/3501-6 cross-ply and fabric
impact forces. Fabric impact data exhibited peak
forces about 11% higher than cross-ply data. Use of
the energy parameter toa/E suggests cross-ply
laminate performance 1.6% higher than fabric
performance: ie a negligible difference;
2) maxi mum Kevlar 49/3501-6 cross-ply laminate
impact forces higher than fabric values. Experi-
mentally measured maxi mum impact forces were
43% higher for the cross-ply than for the fabric
laminates. A cross-ply laminate performance 52%
higher t han for the fabric laminate is indicated by
the energy parameter to~/E;
3) the wide disparity between E-glass fabric and cross-
ply laminate data. The energy parameter to~/E
clearly models this disparity, and supports the
T abl e 6. La mi na t e e ne r gy pa r a me t e r v al ues 1 f or t he v ar i ous mat er i al s t est ed
Composite Specimen Tensile Tensile Energy
material thickness, strength, modulus, parameter,
system t (mm) a( M Pa) E (GPa) t ~ / E (kPa- m)
AS4 graphite/epoxy
Cross-ply 3.00 535 52.6 16.3
Fabric 2.16 697 65.2 16.1
Kevlar 49/ epoxy
Cross-ply 2.72 470 26.1 23.0
Fabric 2.34 437 29.5 15.1
E-glass/epoxy
Cross-ply 2.54 340 22.2 13.2
Fabric 1.85 157 20.8 2.2
COMPOSITES . OCTOBER 1985 277
present experi ment al fi ndi ng of poor fabric
performance.
In summary, the si mpl e energy par amet er model
appears to provi de a good means of relating tensile
coupon dat a to relative i mpact per f or mance of t hi n
composi t e plates. It is clear, however, t hat the insights
gai ned by appl yi ng this model will be dependent on
the quality of the strength and modul us dat a used.
CONCLUSIONS
The Uni versi t y of Wyomi ng' s drop weight i mpact test
(DWIT) system was used to provi de new insights i nt o
low-leveL nor mal i mpact s of composi t e plates. The
present effort provi ded a variety of insights into drop
weight i mpact testing whi ch may be useful to future
investigators. Test results showed a st rong dependence
on speci men thickness, whi ch agreed with results of
previ ous work by ot her investigators? TM That is, the
existence of a l i near force-to-thickness rel at i onshi p for
t hi n graphi t e/ epoxy and Kevl ar/ epoxy l ami nat es was
verified. Thi s rel at i onshi p allows compar i sons of dat a
from speci mens of varyi ng thickness.
A number of concl usi ons can be drawn from the
present effort These include:
1) i nst rument ed drop weight i mpact testing can provi de
useful insights into i mpact behavi our of composi t e
materials. It is part i cul arl y useful in closely repli-
cat i ng i mpact condi t i ons t hat are typical of actual
appl i cat i ons;
2 piezoelectric force t ransducers provi de a pract i cal
alternative to more convent i onal st rai n-gauge
i nst rument at i on;
3) i nst rument ed drop wei ght i mpact test results are
strongly dependent upon changes in speci men
thickness. A test confi gurat i on such as t hat used in
the present testing shoul d not be regarded as a
pot ent i al st andar d unl ess special concer n is given to
assuri ng repeat abl e speci men thicknesses;
4) results of the present testing general l y conf i r med
previ ous observat i ons of i ncreasi ng i mpact
per f or mance when progressi ng from graphi t e/ epoxy
to Kevl ar 49/ epoxy to E-gl ass/ epoxy composites. The
present testing also demonst r at ed smal l er post-
i mpact damage areas for graphi t e and E-glass fabric
l ami nat es t han for compar abl e cross-ply laminates.
There were no significant damage area differences
between Kevl ar fabric and cross-ply l ami nat e forms;
and
5) the present results demonst r at e little difference
between the i mpact per f or mances of AS4 graphi t e
fabric and cross-ply forms, superi or per f or mance for
the Kevl ar 49 cross-ply form and i nconcl usi ve
results for the E-glass fabric vs cross-ply forms.
REFERENCES
1 Adams, D.F. "Impact response of polymer-matrix composite
materials" Composite Materials: Testing and Design (Fourth
Conference). ASTM STP 617 (American Society for Testing
and Materials, 1977) pp 409-426
2 Ireland, D.R. "Instrumented impact testing for evaluating
end-use performance" Physical Testing of Plastics-Correlation
with End-Use Performanc~ ASTM STP 736 (American Society
for Testing and Materials, 1981) pp 45-58
3 Winkel, .I.D. and Adams, D.F. "Instrumented drop weight
impact testing of composite materials" Report UWME- DR-
301-108-0 (Department of Mechanical Engineering.
University of Wyoming, USA. December 1983)
4 "Weaving engineered fabrics" Textile Industries(February
1975) pp 88-89
5 Van Hammersf el d, J. "Extensive cost reduction studies --
composite empennage component -- L-1011 commercial
airliner" SAMPE Journal (May/June 1978) pp 25-33
6 Wood, S.A. 'Structural composites find big new market:
jumbo jetliners" Modern Plastics (November 1981) pp 57-59
7 LeBlanc, D.J. et al' Advanced composite cost estimating
manual -- volume I" Technical Report AFFDLL TR-76-87
(Northrop Corporation, Aircraft Division, Hawthorne, CA.
USA, August 1976)
8 "Advanced impact resistant multidimensional composites"
'Final Technical Report (submitted to Naval Air Systems
Command by Fibre Materials, lnc under NASC contract no
N00019-77-C-0430, January 1979)
9 Miner, LH., Wolffe, R.A. and Zweben, C.H, "Fatigue. creep,
and impact resistance of aramid fibre reinforced composites"
Composite Reliabili~ ASTM STSP 580 (American Society for
Testing and Materials, 1975) pp 549-559
10 Wardle, M.W. and Tokarsky, E.W. "Drop weight impact
testing of laminates reinforced with Kevlar aramid fibres,
E-glass and graphite" Composites Technology Review 5 No I
(Spring 1983) pp 4-10
11 Miller, A.G., Hertzberg, P. E. and Rantala, V.W. "Toughness
testing of composite materials" SAMPE Quarterly (January
1981) pp 36-42
12 Telecommunication with John RuybalL Effects Technology,
Inc, Santa Barbara. CA, USA` July 1983
13 Takeda, N., Sierakowski, R,L. and Malvern, L.E. "Studies of
impacted glass fibre-reinforced composite laminates' SAMPE
Quarterly (January 1981) pp 9-16
14 Zoller, P. "Instrumentation for impact testing of plastics"
Polymer Testing 3 (1983) pp 197-208
15 Stellbrink, ICK. and Aoki, ll.M. 'Effect of defect on the
behaviour of composites' Proc 4th Int Conf on Composite
Materials Tokyo. Japan 1982 pp 853-860
16 Transducer instrumentation, Model 208 series force
transducer" (PBC Piezotronics, lnc, Buflhlo, NY. USA)
17 Cairns, D.S. and Adams, D.F. "Moisture and thermal
expansion properties of unidirectional composite materials
and the epoxy matrix' J Reinforced Plastics and Compo.~ites 2
No 4 (October 1983) pp 239-255
18 Hercules Product Data Sheet No 843-1 (Hercules, Inc.
Wilmington, DE, USA, August 1981)
19 Kevlar49 Data Manual (E. l . DuPont de Nemours and
Company, Inc, Wilmington, DE, USA, March 1978)
20 "Comparative data E, S and $2 glass" (Owens-Corning
Fiberglas Corp, Toledo, OH, USA` May 1975)
21 losipesett, N. "New accurate procedure for single shear
testing of metals' JMat er 2 No 3 (September 1967) pp 537-
666
22 Walrath, D.E. and Adams, D.F. 'The iosipescu shear test as
applied to composite materials" Experimental Mechanics 23
No 1 (March 1983) pp 105-110
23 Adams, D.F. and Walrath, D.E. "losipescu shear properties of
SMC composite materials' Composite Materials: Testing and
Design (Sixth Conference). ASTM STP 787 (American Society
for Testing and Materials, 1982) pp 19-33
24 Wardle, M.W. "Impact damage tolerance of composites
reinforced with Kevlar aramid fibres' Proc 4th lnt Conf on
Composite Materials op cit pp 837-844
25 Personal communication with M.W. Wardle, E.I. DuPont de
Nemours and Company, Inc, Wilmington, DE, USA.
October 1983
26 Coppa, A.P., Zweben, C.H. and Mirandy, L. 'Flywheel
containment technology assessment' Report No UCRL-15261
(Lawrence Livermore Laboratory, Livermore, CA. USA. July
1980)
AUTHORS
Mr Winkei is current l y a Project Engi neer at the Fibre
Sciences Division, EDO Corporat i on, Salt Lake City.
UT, USA. Professor Adams, to whom inquiries shoul d
be addressed, is with the Depar t ment of Mechani cal
Engineering, University of Wyomi ng, Box 3295,
Laramie, WY 82071, USA.
278 COMPOSITES. OCTOBER 1985

You might also like