Sei sulla pagina 1di 40

Modeling of High Strain Rate Axial Response and Shear Failure

of Thermoplastic Composites

Pierce D. Umberger

Dissertation submitted to the Faculty of


Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of

Doctor of Philosophy
in
Engineering Mechanics

Dr. Scott Case, Chair


Dr. Romesh Batra
Dr. Michael Hyer
Dr. Sunghwan Jung
Dr. Robert West

August 1, 2013
Blacksburg, Virginia

Keywords: UHMWPE, time-temperature superposition, composite materials, shear lag,


Monte Carlo, FEM, ABAQUS, punch shear

Copyright 2013 by Pierce D. Umberger

Contents
List of Figures

iv

1 Background

1.1

Motivation & Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

High Strain Rate Tensile Properties and tTSP . . . . . . . . . . . . . . . . .

1.3

Axial Progressive Damage Modeling . . . . . . . . . . . . . . . . . . . . . . .

1.4

Through Thickness Shear of UHMWPE Laminates . . . . . . . . . . . . . .

1.4.1

High SPR Investigations . . . . . . . . . . . . . . . . . . . . . . . . .

1.4.2

Low SPR Investigations . . . . . . . . . . . . . . . . . . . . . . . . .

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.5

2 Preliminary Results: Punch Shear

10

2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

2.2

Laboratory Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

2.2.1

Testing Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

12

2.2.1.1

Punch Shear Test Apparatus . . . . . . . . . . . . . . . . .

12

2.2.1.2

Punch Configurations . . . . . . . . . . . . . . . . . . . . .

13

2.2.1.3

Punch Shear Test Procedure . . . . . . . . . . . . . . . . . .

14

2.2.2

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

16

2.2.3

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

17

Finite Element Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

20

2.3.1

Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . .

20

2.3.2

Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .

21

2.3

ii

3 Preliminary Results: Axial Model

22

3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

3.2

Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

3.3

Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

24

3.4

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

4 Proposed Work

28

4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

4.2

Punch Shear Modeling and Analysis . . . . . . . . . . . . . . . . . . . . . . .

28

4.3

Axial Modeling and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

4.4

Project Time Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

4.5

Expected Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

4.6

Anticipated Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

31

Bibliography

32

iii

List of Figures
1.1

Tensile strength vs. strain rate for several temperatures in a PP tape. Tensile
strength increases with increasing strain rate and decreasing temperature [2].

Master curve developed from the data in Figure 1.1 with a reference temperature of 20 C [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Weibull distribution of S3000 fiber strength at three thermorheologically "equivalent" strain rates [6]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Axial and shear stresses associated with 1 and 2 fiber breaks, respectively.
Both stress components approach their far field value away from the fiber
break. [16] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.5

Schematic representation of the shear lag model of Okabe et al. [16] . . . . .

1.6

Node numbering convention for the shear lag model of Okabe et al. [16] . . .

1.7

Schematic of the fixture used in laboratory testing conducted by Xiao et al.


[27] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.8

Finite element model of high-SPR punch testing by Xiao et al. [27] . . . . .

2.1

SS1214 laminate samples before and after testing. . . . . . . . . . . . . . . .

11

2.2

Punch shear test apparatus exploded assembly.

. . . . . . . . . . . . . . . .

12

2.3

Side view of punch shear test apparatus exploded assembly. . . . . . . . . . .

13

2.4

Unsupported shear punch fabricated from A2 tool steel. . . . . . . . . . . . .

14

2.5

Supported shear punch fabricated from A2 tool steel. . . . . . . . . . . . . .

15

2.6

Assembled punch shear test apparatus in MTS hydraulic load frame. . . . .

16

2.7

Ultimate shear stress versus thickness. . . . . . . . . . . . . . . . . . . . . .

17

2.8

Load versus displacement for selected unsupported punch samples. . . . . . .

18

2.9

Load versus displacement for selected supported punch samples. . . . . . . .

18

2.10 Detail of 3D solid quarter model of punch shear test apparatus. . . . . . . .

20

1.2
1.3
1.4

iv

2.11 Boundary conditions of 3D solid quarter model of punch shear test apparatus. 21
3.1

Simulated stress-strain curve for SS3124 composite at 0.83s1 . . . . . . . . .

25

3.2

Simulated stress-strain curve for SS3124 composite at 8.3s1 . . . . . . . . . .

25

3.3

Simulated stress-strain curve for SS3124 composite at 6.4x105 s1 . . . . . . .

26

4.1

Expected project timetable. . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

1 | Background
1.1

Motivation & Aims

Ultra-high molecular weight polyethylene (UHMWPE) composites are used in many applications requiring light weight, high strength, and impact resistance, particularly armor applications [1]. The Spectra/SpectraShield line, manufactured by Honeywell, is a UHMWPE
material system commonly used in ballistic armors. SpectraShield consists of a cross-ply laminate of high molecular weight polyethylene fibers in a weak low molecular weight blended
polyethylene matrix, formed by hot pressing. As is the case in viscoelastic polymers, mechanical response of UHMWPE fibers and composites is a function of temperature and strain rate
[2, 3, 4, 5]. In particular, UHMWPE fibers and composites show an increase in stiffness and
strength with increased strain rate and/or decreased temperature [6, 7, 8, 9, 10]. Therefore,
composite properties are of particular interest at strain rates equivalent to those occurring
during ballistic events, which are typically 105 s1 to 106 s1 .

1.2

High Strain Rate Tensile Properties and tTSP

Laboratory testing of composites at strain rates relevant to ballistic events is difficult. Conventional tensile test methods such as servohydraulic and electromagnetic load frames cannot
come close to reaching the high strain rates needed. Some high strain rate methods exist,
such as Split Hopkinson Pressure Bar (SHPB) testing, but even these test methods can not
achieve all desired strain rates [5, 10]. Additionally, high strain rate tensile measurements
present additional sample gripping and sensing concerns [11]. Viscoelastic modeling using
the time-temperature superposition principle (tTSP) has been used to predict behavior of
viscoelastic materials at time scales or strain rates that are not physically achievable in a
laboratory [3, 2].
In 2007, Alcock et al. investigated the effects of temperature and strain rate on the mechanical properties of highly oriented polypropylene (PP) tapes and all-PP composites. The
authors analyzed strain rate and temperature effects on tensile modulus and strength and
developed master curves for each. These master curves were used to predict the constitutive
behavior of the tapes and composites at various strain rates, including those that are diffi1

2
cult to achieve with physical testing. Figure 1.1 shows strength vs. strain rate for several
different temperatures. Figure 1.2 shows the master curve developed from the data in Figure
1.1, covering a much wider range of strain rate [2].

Figure 1.1: Tensile strength vs. strain rate for several temperatures in a PP tape. Tensile
strength increases with increasing strain rate and decreasing temperature [2].

Figure 1.2: Master curve developed from the data in Figure 1.1 with a reference temperature
of 20 C [2].
Previous work with UHMWPE fibers attempted to predict fiber strength at high strain rates
using tTSP and subambient temperature testing. While only directly applicable to fully
amorphous polymers, tTSP has been applied to failure mechanisms of composite systems
and has been successfully used to predict tensile modulus and strength at shifted time scales

3
[12]. Honeywell Spectra S3000 UHMWPE fibers were tested in a TA instruments Q800 DMA
at a range of temperatures and strain rates in order to predict strength properties at ballistic
strain rates. Statistical tests of these strength distributions did not indicate problems with
these predictions with a P-value of 0.374, it was not shown that the Weibull modulus of
UHMWPE fibers tested at several thermorheologically equivalent temperature/strain rate
combinations was different; Weibull modulus is the slope of fit lines shown in Figure 1.3
[6]. This indicates that there is not a fundamental difference in the strength distribution of
UHMWPE fibers at thermorheologically equivalent temperature/strain rate combinations.
However, while no problems were found with predicting raw fiber strengths, high strain rate
predictions using tTSP did not appear to work for manufactured composite laminates, as
discovered by Cook in 2010 [7].

Figure 1.3: Weibull distribution of S3000 fiber strength at three thermorheologically "equivalent" strain rates [6].
The current alternative to a viable model for high strain rate material property prediction

4
is to conduct multiple batteries of ballistic tests on composite panels. Currently UHMWPE
prices far exceed those of many other conventional armor systems, making these tests very
expensive [1]. There is a need for UHMWPE composite strength predictions at a range of
elevated strain rates.

1.3

Axial Progressive Damage Modeling

The mechanical behavior of fiber reinforced composite materials is heavily dependent on the
nature of the imperfect fibers that comprise the composite. The evolution of the damage in
the composite is critical to understanding and predicting ultimate composite strength and
failure.
As a unidirectional composite material has an axial tensile load applied, individual fiber
failures are generated due to localized imperfections [13]. Axial stress in a fiber is necessarily
zero at the location of a break, and stress concentrations form at this location in nearby
fibers [14]. Load is gradually transferred back to the broken fiber via the matrix material,
until after some ineffective length, the fiber stress approaches its far-field value, as shown
in Figure 1.4. It is generally assumed that axial stresses are carried entirely by the fiber
strands, while shear stresses are carried entirely in the matrix between broken and unbroken
fibers. However, in composite materials where the matrix stiffness is appreciable compared
to that of the fibers, the contribution of the axial stress in the matrix material has been
considered by Beyerlein and Landis [15].

Figure 1.4: Axial and shear stresses associated with 1 and 2 fiber breaks, respectively. Both
stress components approach their far field value away from the fiber break. [16]

5
Landis et al. developed a finite element shear lag model which determined the stress profile
around a broken fiber segment. This stress profile was then used to repeatedly simulate the
damage progression in the form of a Monte Carlo method [17]. Several models have been
proposed to describe the nature of composite damage evolution, in both 2 and 3 dimensions [18, 17, 19, 16, 13, 15, 20, 21, 22, 23, 24, 25]. For some sets of composite constituent
properties, fiber/matrix debonding failure modes can also occur [20, 19, 16]. Most recent
work involving fiber debonding and/or slipping has been in 2d, which cannot fully characterize the behavior and formation of fiber clusters that ultimately lead to composite failure
[21, 25, 24, 22, 23], however, the most recent work on the topic by Okabe et al. considers
fiber slipping and debonding in three dimensions [16].
Okabe et al. make the typical simplifying shear lag assumptions that fibers carry only
axial load, while the matrix carries only shear. Furthermore, they assume that in an NxM
fiber array, fibers are round, of constant cross section, and evenly spaced [16]. The model
schematic is shown in Figure 1.5.

Figure 1.5: Schematic representation of the shear lag model of Okabe et al. [16]
Considering the composite fibers as axial springs and the matrix material as shear springs,
the equilibrium equation for the arrangement shown in Figure 1.5 can be written as
4

di,j,k
r X l
A
=

dz
2 l=1 i,j,k

(1.1)

6
where A is the cross section area of the fiber, r is fiber radius, and is the interfacial shear
stress, which can be written as
ui+1,j,k ui,j,k
d
ui,j+1,k ui,j,k
= Gm
d
ui1,j,k ui,j,k
= Gm
d
ui,j1,k ui,j,k
= Gm
d

1
i,j,k
= Gm

(1.2a)

2
i,j,k

(1.2b)

3
i,j,k
4
i,j,k

(1.2c)
(1.2d)

where d is the fiber spacing and the ui,j,k are the relative axial displacements of the individual
fiber elements. The numbering convention used by Okabe et al. is shown in Figure 1.6

Figure 1.6: Node numbering convention for the shear lag model of Okabe et al. [16]
Equation 1.1 is solved numerically in combination with applied boundary conditions for
composite ends and fiber breaks, resulting in the displacement field for a simulated composite
with an arbitrary number of broken fibers.

1.4

Through Thickness Shear of UHMWPE Laminates

In addition to axial properties, through-thickness shear response of UHMWPE composites


is of interest for modeling impact events. Particularly at high speeds where inertial effects
are relevant, the through-thickness shear properties play a significant role in overall laminate
behavior during an impact event [1, 26]. The most common means of characterization of

7
the through-thickness behavior of composite laminates is a punch test. In punch testing,
the span-to-punch ratio (SPR), that is, the diameter of the composite fixture relative to the
diameter of the punch implement, is a controlling factor in shear versus bending deformation
modes [27, 28]. SPR values near 1 lead to shear dominated deformation and failure, while
large unsupported spans (SPR 1) lead to bending dominated deformation and failure.

1.4.1

High SPR Investigations

Xiao et al. studied delamination damage in S-2 glass composites from punch-shear loading
[29]. A schematic of the test fixture is shown in Figure 1.7. Sun and Potti compared quasistatic punch shear experiments with ballistic penetration tests on AS4 GFRP composites
[28]. In both the work of Gama and Gillespie and Xiao et al., the SPR was much greater
than 1, meaning that bending mechanics were significant. In the work of Sun and Potti,
the SPR varied, but was between 3 and 12, indicating that bending behavior was likewise
present [28]. Additionally, Xiao et al. conducted finite element modeling of the laboratory
experiments. Their quarter-plate FE model is shown in Figure 1.8.

Figure 1.7: Schematic of the fixture used in laboratory testing conducted by Xiao et al. [27]

Figure 1.8: Finite element model of high-SPR punch testing by Xiao et al. [27]

1.4.2

Low SPR Investigations

ASTM D732 specifies a SPR 1 test of through thickness shear strength of plastics [30].
It is noted in the standard that the nature of the stress concentration between the fixture
and sample as well as between the punch and sample can have an appreciable effect on the
outcome of the testing. Liu and Piggott used ASTM D732 on epoxy matrix materials and
thermoplastic polymers. They noted that this method might not be well suited to testing
of composites due to the close clearances between the punch and fixture. It was suggested
that fiber pullout and friction between the punch, fixture, and sample would hamper results
[29]. This standard was successfully employed by Crescenzi et al. on S-2 glass composites
[31]. No problems were encountered with friction or fiber pullout/entanglement. To date,
there is no published work on the characterization of the through-thickness shear strength
of thermoplastic composite laminates using a test similar to ASTM D732.

1.5

Summary

High strain rate constitutive properties of UHMWPE composite materials are not predictable
using low strain rate testing accelerated by tTSP [7], however prior work indicates that

9
constituent properties are predictable using low strain rate testing accelerated by tTSP
[6, 12]. Failure models incorporating progressive damage are well established and validated
against laboratory data [18, 17, 19, 16, 13, 15, 20, 21, 22, 23, 24, 25].
To date, a progressive damage model that incorporates fiber debonding has not been applied to
predict the high strain rate response of UHMWPE composites. Additional studies are needed
to determine the efficacy of using such a model to predict composite strength at a range
of strain rates approaching 106 s1 . Furthermore, validation of these predictions against
existing experimental data needs to be performed. Finally, subject to model validation, a
study of composite failure mode, and other important parameters such as critical cluster
sizes, ineffective fiber lengths, and fiber/matrix debonding behavior needs to be conducted to
provide information that cannot be obtained through laboratory investigation.
Through-thickness shear properties of UHMWPE laminates have not been determined. A
punch test with a SPR 1 demonstrates primarily shear behavior. ASTM D732 has been
used successfully on neat matrix samples as well as rigid thermoplastic polymers and epoxy
matrix composites. Finite element simulation has been performed and validated against
laboratory testing at a variety of SPR values substantially greater than 1.
A punch-shear test similar to ASTM D732 is a viable means of characterizing the throughthickness shear strength of UHMWPE laminates. However, such testing has not been published to date. Characterization of the through-thickness behavior of UHMWPE laminates
needs to be performed. Due to the nature of stress concentrations inherent in this test, finite
element analysis needs to be conducted concurrent with the laboratory testing. Investigation
of the stress concentrations and resulting implications for the physical testing needs to be completed in order to estimate constitutive properties for the through-thickness shear strength of
these laminates.

2 | Preliminary Results: Punch Shear


A partial form of this investigation was presented at the SAMPE 2010 Technical Conference.

2.1

Introduction

Much work has been done to characterize the properties of laminates in terms of their energy
absorption via a punch type test. Gama and Gillespie and Sun and Potti studied the relation
between quasi-static punch-shear behavior and ballistic penetration models for thick section
S-2 glass composites [26, 28, 32]. They found that different span to punch ratios changed the
mechanics of failure. SPR close to 1 leads to shear dominated failure, while large support
spans lead to bending dominated failure. In 2005, Xiao et al. studied delamination damage in
S-2 glass composites due to punch-shear loading [27]. In both of these cases, the supported
diameter is substantially larger than the punch diameter, leading to a significant mix of
bending and shear.
In order to support modeling efforts, it is desired to understand the through-thickness shear
behavior of UHMWPE composites in terms of material properties. While any mechanical
test is always a measurement of a mechanical system property rather than a true material
property, it is desired to isolate the through-thickness shear behavior inasmuch as is possible.
In 1995, Liu and Piggott studied the shear properties of polymers and fiber composites [29].
Liu and Piggott used ASTM standard D732-85 as a basis for evaluating shear properties
of thermoplastic polymers as well as several epoxy matrix materials. They concluded that
the punch shear test may have problems associated with friction between the punch and
die, as well as problems developing a stress state close to pure shear, rather than a tensiondominated state. Lee and Sun and Potti and Sun performed punch testing on graphite fiber
reinforced composites in a similar manner and found that the primary damage modes were
delamination and plugging [33, 34]. Nemes et al. performed quasi-static as well as high strain
rate punch shear characterization of graphite/epoxy laminates using a SHPB apparatus.
They observed that the overall penetration behavior of the laminate was relatively insensitive
to construction parameters other than thickness [35]. Liu and Piggott also tested samples
using an Iosipescu type test geometry. They found that it provided unreliable results for
polymers in the rubbery state [29]. To date, literature has not been published to establish
constitutive property estimates for the shear behavior of UHMWPE or similar laminates.
10

11

2.2

Laboratory Testing

Samples tested were commercially available Honeywell SpectraShield 1214 product. Either
4 or 8 layers of SpectraShield 1214, each having a lay-up of [0,90]2 were combined to form
laminates of lay-up [0,90]4 or [0,90]8 . Sheets of the SpectraShield product were cut into
approximately 30.5 cm x 30.5 cm panels using a CNC cutting table. Sheets were stacked in
the aforementioned configuration. Multiple sheets were pressed in one cycle, with a sheet
of silicone coated paper between each. The stack of panels was placed between two 0.635
cm thick smooth aluminum sheets and placed in a hot press. Panels were pressed at a
temperature of 118 C and pressure of 19 MPa for 15 minutes. The pressure was removed
and the panel was allowed to cool in the press for approximately 15 minutes before being
removed. These pressed plates were clamped between rigid fiberglass plates and cut into
50 mm x 50 mm samples using a wet saw. Samples were sufficiently large that cut edges
were far away from the punch area. Figure 2.1 shows SpectraShield samples before and after
testing. Typical sample thicknesses were 0.53 mm and 1.04 mm, for 8 and 16 layer samples,
respectively. Typical fiber volume fractions range from 70 to 75%.

Figure 2.1: SS1214 laminate samples before and after testing.

12

2.2.1

Testing Procedure

Specimens were tested using an in-house punch shear assembly. The test setup is similar to
ASTM D732-85, but with smaller dimensions to achieve failure within a desired load range
based on load frame specifications and predicted strength.
2.2.1.1

Punch Shear Test Apparatus

The in-house punch-shear assembly consists of a two piece rigid steel die with a 19.05 mm diameter hole through which a 19.00 mm diameter punch is pressed. Thin samples are clamped
in between the two die halves. The punch to die fit is 0.025 mm nominal radius, effectively
minimizing the unsupported span of the sample during testing. Laminate thickness is 25 to
50 times this clearance. The goal is to create a region of intense through-thickness shear in
the test specimen at the punch/die interface, while minimizing other mechanics phenomena
such as bending. Figures 2.2 and 2.3 show the assembly. Up to eight 21 -20 inch bolts are
used to provide clamping pressure to the specimen. The entire assembly was manufactured
out of A2 tool steel and then air-hardened to approximately Rockwell C50 to minimize wear
on the edge of the hole at the punch-die-sample interface.

Figure 2.2: Punch shear test apparatus exploded assembly.

13

Figure 2.3: Side view of punch shear test apparatus exploded assembly.
2.2.1.2

Punch Configurations

Two main punch configurations were considered; one allowing and one preventing curvature
of the sample during testing. It was hypothesized that disallowing curvature of the sample
during testing would create a stress state closer to pure shear, and thereby more effectively
decouple tensile and shear modes of failure.
Unsupported Punch The unsupported punch was simply a right circular cylinder of
diameter 19.00 mm and a length of 50 mm. The punch was intended to create a region of
intense shear in the sample at the punch-die interface. The punch was manufactured from
A2 tool steel and then air-hardened to approximately Rockwell C50 to minimize wear of the
punch edge. Additionally, the face of the punch was surface ground to make the edge as
square as possible to reduce bending at the punch-die interface. Figure 2.4 shows the punch
used.
Supported Punch The supported punch was designed to prevent curvature of the sample
during testing. The main body of the punch was a right circular cylinder of diameter 19.00
mm and a length of 50 mm. One end of the punch was drilled and tapped with a 14 -20

14

Figure 2.4: Unsupported shear punch fabricated from A2 tool steel.


inch hole and surface ground to create a square edge. A second right circular cylinder was
machined with a 19.00 mm diameter and a 10 mm length with a 6.35 mm diameter hole
through the center to form the clamp. Both pieces were fabricated from A2 tool steel and
air-hardened to Rockwell C50, as before. Figure 2.5 shows the punch assembly.
2.2.1.3

Punch Shear Test Procedure

Samples of size 50 mm x 50 mm were centered on the bottom portion of the die assembly. A 5
mm diameter hole was cut using a razor blade for samples tested with the supported punch.
The top die section was installed. Proper alignment of the holes in the upper and lower
section of the die was achieved via pressed dowel pins. Four 12 -20 inch bolts were inserted
in the bolt holes located at the corners of the die halves and tightened to a torque of 68
N-m. The die assembly was designed with eight bolt holes due to anticipated problems with
satisfactorily gripping the samples, however four bolts proved to be sufficient to achieve the
necessary grip pressure on the sample to prevent slipping, based upon post-test inspection.
Punch shear testing was conducted in an MTS hydraulic load frame. An 88-kN capacity MTS
Force Transducer 661.20 load cell was used to measure load. Displacement measurements
were taken from the load frame stroke transducer. An aluminum block was clamped in the

15

Figure 2.5: Supported shear punch fabricated from A2 tool steel.


upper MTS 647 Hydraulic Wedge Grip and used to transfer load to the punch head. Control
and data collection was provided using an in-house National Instruments (NI) LabVIEW
program. The punch shear test assembly was rested on the lower grip frame, which was
verified to be parallel to the aluminum block in the upper grip to ensure load was applied
vertically and did not cause the punch to jam in the die assembly. Figure 2.6 shows the
assembled test apparatus in the load frame prior to a typical test.
Tests were conducted in load control at a load rate of 667 Ns . Load and displacement
channels were recorded at 50 Hz using the in-house NI LabVIEW program. After each test,
the punch was removed, die halves unbolted, and the sample removed for inspection of the
failure surface.

16

Figure 2.6: Assembled punch shear test apparatus in MTS hydraulic load frame.

2.2.2

Results

For this test geometry, through-thickness shear stress is ideally represented as

F
Ashear

(2.1)

where Ashear is defined as

Ashear = Dp t

(2.2)

where Dp is the average of the punch and die diameters and t is the sample thickness.
The results for ultimate shear stress versus sample thickness are shown in Figure 2.7. Average
ultimate shear stress for each thickness and punch type are given in Table 2.1. Additionally,
standard deviation and coefficient of variation (COV) for each data set is shown in Table
2.2.

17

Figure 2.7: Ultimate shear stress versus thickness.


Table 2.1: Average ultimate shear strength and deviation for supported and unsupported
punch types (MPa).
8-layer
16-layer
Absolute Difference
Percent Difference

Unsupported

Supported

266
242
24
9.0

196
179
17
8.7

Typical load-displacement curves for unsupported punch samples are given in Figure 2.8.
We note that there is a distinct slope change for both 8 and 16 layer samples. Similarly,
Figure 2.9 shows typical load-displacement curves for samples tested using the supported
punch.

2.2.3

Discussion

There is a trend of increased apparent shear strength with decreasing thickness. If the state
of stress was truly pure shear through the thickness of the sample, apparent shear strength

18

Figure 2.8: Load versus displacement for selected unsupported punch samples.

Figure 2.9: Load versus displacement for selected supported punch samples.

19
Table 2.2: Standard deviation and COV for supported and unsupported punch types.
8-layer
16-layer

Std. Dev. (MPa)


COV (%)
Std. Dev. (MPa)
COV (%)

Unsupported

Supported

12.1
4.6
17.1
7.1

15.1
7.7
12.2
6.8

would not be expected to vary with specimen thickness. Initially, when only unsupported
punch tests had been performed, it was hypothesized that this increase in apparent strength
was due to a membrane type behavior. Thinner 8-layer samples have less bending stiffness
and more ability to rotate when compared to 16-layer samples, given the constant clearance
in the punch and die geometry. It stands to reason that apparent shear strength would then
be higher for samples that are experiencing more of a membrane type state of stress, due to
the extreme anisotropy of UHMWPE fibers that comprise the SpectraShield laminates.
This notion is further supported by the distinct change in slope of the load-displacement
curves shown in Figures 2.8 and 2.9. It is possible that the increase in slope during the
course of the punch shear test is due to a nonlinear geometric effect that is governed by a
membrane type behavior. The same phenomenon that governs this increased mechanical
stiffness could be governing the increase in observed shear strength.
It was initially expected that the supported punch that prevented curvature of the sample
would reduce this hypothesized membrane behavior and would result in apparent shear
strengths that were more consistent versus sample thickness. However, given the results in
Figure 2.7 and Tables 2.1 and 2.2, this does not appear to be the case. For both supported
and unsupported punch types, an 8 to 9 percent decrease in apparent shear strength is
observed with a two fold increase in sample thickness. Finite element modeling is in progress
to investigate these phenomena.

20

2.3
2.3.1

Finite Element Modeling


Model Formulation

To better understand the results of the through thickness punch shear testing, a finite element model of the test apparatus was constructed, as shown in Figure 2.10. The model
currently treats the punch and fixture assembly as rigid, but has the ability to model them
as elastic structures. Composite elements are 3D C3D20 quadratic 20-node elements with
full integration. The bottom edge of the test fixture is fixed in space, and a load is applied to
the top half of the fixture to simulate clamping pressure. A second load is applied to the top
of the punch to simulate an applied load. A schematic of the boundary conditions is shown
in Figure 2.11. Interaction between the punch, fixture, and composite sample is modeled as
contact and friction loads.

Figure 2.10: Detail of 3D solid quarter model of punch shear test apparatus.
It is assumed that composite layers are perfectly bonded and that no slipping of the composite
occurs in the grip region of the fixture.

21

Figure 2.11: Boundary conditions of 3D solid quarter model of punch shear test apparatus.

2.3.2

Preliminary Results

The development of the aforementioned finite element model is complete, but substantial
results have not yet been produced and analyzed. Further work is proposed in Section 4.2.

3 | Preliminary Results: Axial Model


3.1

Introduction

3.2

Model Formulation

A shear-lag type model is being used to characterize the load transfer away from fiber breaks.
This load transfer is governed by several parameters, most important in this context are the
elastic modulus of the fiber and the shear modulus of the matrix. Both of these properties
change with strain rate, but the rate of change is profoundly different between the two.
Fiber modulus changes are small compared to the several orders of magnitude change that
the matrix material exhibits over 5 orders of magnitude of strain rate. Due to these changes,
the nature of the load transfer away from fiber breaks is expected to very with strain rate.
The notation used is the same as that of Okabe et al. [16]. The arrangement of elements in
the model is shown in Figure 1.5. The i and j indices indicate fiber number, and the k index
is the element number along the length of a given (i,j) fiber.
Fiber properties are randomly assigned based on the Weibull distribution. Previous work on
this project has demonstrated the Weibull distribution to be a suitable fit for this material
system [6]. Each fiber element is assigned an ultimate fiber tensile strength as shown in
Equation 3.1. Matrix properties and all other fiber properties are assumed to be deterministic. This is partly because the composite strength is largely governed by the fiber strength
distribution, and partly because an accurate distribution for other material properties is
difficult to obtain.

uti,j,k

 
= ln

1
1 i,j,k

L0
z

1
m

(3.1)

Following the aforementioned method of Okabe et al., we note that the equilibrium equation
for axial stress on a fiber element is
4

r X l
di,j,k
A
=

dz
2 l=1 i,j,k
22

(3.2)

23
And we assume that the shear stress at the fiber-matrix interface is a function of the displacements of the surrounding fibers.
ui+1,j,k ui,j,k
d
ui,j+1,k ui,j,k
= Gm
d
ui1,j,k ui,j,k
= Gm
d
ui,j1,k ui,j,k
= Gm
d

1
i,j,k
= Gm

(3.3a)

2
i,j,k

(3.3b)

3
i,j,k
4
i,j,k

(3.3c)
(3.3d)

We can then write a finite difference method equation to solve for the displacements of the
fiber nodes, using the approach of Oh [20].

4Ef A (i,j,k (ui,j,k+1 ui,j,k ) i,j,k1 (ui,j,k ui,j,k1 )


(2 + i,j,k1 + i,j,k ) z2
h


ui+1,j,k ui,j,k
+h Gm
1 P 1 i,j,k 1 D1 i,j,k
d
i

1
1
1
+1 P i,j,k 1 D i,j,k y + 1 D i,j,k s

h

ui,j+1,k ui,j,k
1 P 2 i,j,k 1 D2 i,j,k
+h Gm
d
i

+2 P 2 i,j,k 1 D2 i,j,k y + 2 D2 i,j,k s

h

ui1,j,k ui,j,k
3
3
1 P i,j,k 1 D i,j,k
+h Gm
d
i

+3 P 3 i,j,k 1 D3 i,j,k y + 3 D3 i,j,k s
h


ui,j1,k ui,j,k
4
4
+h Gm
1 P i,j,k 1 D i,j,k
d
i

+4 P 4 i,j,k 1 D4 i,j,k y + 4 D4 i,j,k s = 0

(3.4)

Where P and D are equal to 1 if matrix yielding or debonding has occurred, respectively, and
are zero otherwise. = 1 only if the indexed fiber element has broken and is zero otherwise.

24
The boundary conditions are

ui,j,1 = 0

(3.5a)

ui,j,K = c L

(3.5b)

We solve the above system of equations. From the computed displacements, the fiber element
stresses are computed, and fiber breaks are calculated. The above steps are then repeated
for the new composite configuration with fiber breaks. This continues until the composite
has failed.

3.3

Preliminary Results

Simulations have been completed for ultra-high molecular weight polyethylene (UHMWPE)
composites at a variety of strain rates, using fiber and matrix constitutive properties previously measured. As expected due to the very low matrix stiffness of these materials relative
to fiber stiffness, the low strain rate simulation results are very close to the bundle strength
result for the UHMWPE fibers, despite the presence of load transfer between fibers that
the model allows. Figure 3.1 shows the simulation result for 32 simulations of a SS3124
composite at a strain rate of 0.83s1 , while Figure 3.2 shows result of 32 simulations at a
strain rate 10 times faster, 8.3s1 .
At low strain rates results are within 2% of the bundle strength result. Monte Carlo simulation results have a relatively low amount of scatter, with a coefficient of variation of <2.5%.
High strain rate simulations exhibit a strength substantially higher than the bundle strength
result, as the matrix becomes relatively stiffer compared to the fibers with changes in strain
rate. The ultimate tensile strengths shown in Figures 7 and 8 indicate an increase of 7% with
one order of magnitude change in strain rate. When the strain rate is increased further to
6.4x105 s1 , the average ultimate tensile strength increases to 2.9 GPa, as shown in Figure
3.3.

25

Figure 3.1: Simulated stress-strain curve for SS3124 composite at 0.83s1 .

Figure 3.2: Simulated stress-strain curve for SS3124 composite at 8.3s1 .

26

Figure 3.3: Simulated stress-strain curve for SS3124 composite at 6.4x105 s1 .

3.4

Discussion

At high strain rate simulated strength tends to overestimate the experimentally measured
composite strength. The current model does not incorporate fiber-matrix debonding or
matrix yielding, although it has the ability to do so. Current efforts are focusing on obtaining
fiber-matrix bond strength data so as to enable the model to simulate fiber-matrix debonding.
Since the current model does not use any fit parameters, efforts are being made to obtain
an estimate of the bond strength, rather than treat the value as a fit parameter.
The results of the axial strength model agree well with both bundle strength estimates as
well as measured composite strength at low strain rates. However, at high strain rates, the
model predicts strengths substantially higher than measured values, while bundle strength
estimates are substantially lower. It was hypothesized that this was due to the fact that the
axial simulation model did not implement debonding at the fiber/matrix interface. This is
a planned feature of the model that has now been implemented. However, the interfacial
strength of the fiber/matrix bond is being estimated. Several techniques, including microbead debond, and fiber pushout were deemed to be impractical to perform on this composite.
Currently, work is in progress to determine an estimate of this property from two sources.
The first source is tensile tests of [(+/-45)N ]S composites that were previously conducted.
This method incorporates other modes of deformation than those of interest, but the test

27
data already exists. The second method involves estimating the debond strength from a
transverse tension test of a unidirectional specimen. This is a much simpler test, but the
debonding mode is not correct. In addition to obtaining estimates for the fiber/matrix
debonding strength, a sensitivity study is in progress to determine how much the model
strength estimates are affected by the debonding stress. The necessary precision of the
estimate is dependent on how sensitive the model is to the variable.
Further work is proposed in Section 4.3.

4 | Proposed Work
4.1

Introduction

The objective of the proposed studies is to develop models capable of producing reliable
estimates for constitutive properties of thermoplastic composites to enable a robust modeling
effort.
Finite element modeling of through-thickness punch shear testing in conjunction with analysis of stress concentrations will be used in conjunction with test data to predict values of
through-thickness shear strength and modulus. Additionally, the axial progressive damage
model will be used in conjunction with measured and predicted constituent properties to
predict composite response at strain rates not attainable in a laboratory setting.

4.2

Punch Shear Modeling and Analysis

Objectives: Validate and refine FE model to examine the mechanics governing the failure
of composite laminates as tested in Section 2.2.
Methodology: The finite element model is being refined to include a finer mesh with several
elements through the thickness of each composite layer.
Additional investigations will be performed as to the stress state at the interface of the punch
and sample:
Refine FE model to >1 element through the thickness of each composite layer.
Determine the ratio of in plane to out of plane stresses from the FE model as a function
of applied load, sample thickness, and punch configuration.
Determine the magnitude of stress concentration at punch-composite interface and the
implications of this stress concentration on measured strength.
Develop analytical half-space contact solution for punch-composite interface for comparison to FE model results.
Estimate through-thickness shear strength of UHMWPE composite laminates.
28

29

4.3

Axial Modeling and Analysis

Objectives: Expand the progressive damage model to include fiber/matrix debonding and
validate the model. Use the model to predict composite properties at high strain rates that
are not attainable in laboratory testing.
Methodology: The progressive damage model is currently being expanded to include matrix plasticity and fiber/matrix debonding. After determining interfacial bond strength,
implement debonding and simulate composite failure at a range of strain rates.
Estimate fiber/matrix bond strength from off-axis tensile test data
Implement fiber/matrix debonding and simulate composite failure at strain rates ranging from quasi-static to 106 s1
Examine and explain any changes in failure mode, fiber ineffective length, critical
cluster size, or other relevant failure characteristics
Estimate composite tensile performance at strain rates ranging from quasi-static to 106
s1

4.4

Project Time Table

Work on all major areas of proposed study is underway. An expected timetable of work is
shown in Figure 4.1.
Completion is expected in July 2013.

30

Figure 4.1: Expected project timetable.

4.5

Expected Contributions

To date, a prediction of through-thickness shear properties of UHMWPE composites has


not been realized. A method for their prediction is outlined in Section 2.3. This research
consisting of combined mechanical testing and simulation will produce an estimate of shear
modulus and strength along with a discussion of the confidence in this prediction and any
associated issues.
To date, a progressive damage tensile strength model that incorporates fiber debonding has
not been applied to predict the high strain rate response of UHMWPE composites. This
research will produce a model framework and a prediction of composite strength at a range
of strain rates from quasi-static to 106 s1 . These predictions will be validated against
existing experimental data.
To date, a study of UHMWPE composite failure including critical break cluster sizes, ineffective fiber lengths, and fiber/matrix debonding behavior has not been conducted. Using the
framework developed in this research, a study of these parameters will be performed.
The above outlined contributions constitute a significant addition to the body of knowledge
regarding thermoplastic composite laminates and their mechanical response.

31

4.6

Anticipated Publications

Umberger, P.D., Case, S.W., Tensile strength and modeling of UHMWPE composites at
high strain rates, Journal TBD
Umberger, P.D., Cook, F.P. Case, S.W., Characterization and response of UHMWPE reinforced thermoplastic composites and constituents using time-temperature superposition,
Journal of Thermoplastic Composite Materials
Umberger, P.D., Case, S.W., Through-thickness shear behavior of UHMWPE composites,
Applied Composite Materials

Bibliography
[1] Bhatnagar, A. Lightweight ballistic composites: military and law-enforcement applications. Woodhead, 2006.
[2] Alcock B., Cabrera N.O., Barkoula N.M., Reynolds C.T., Govaert L.E., Peijs T. The
effect of temperature and strain rate on the mechanical properties of highly oriented
polypropylene tapes and all-polypropylene composites. Composites Science and Technology, 67:20612070, 2007.
[3] Brinson H., Brinson L. Polymer engineering science and viscoelasticity. Springer, 2008.
[4] Ariyama T., Mori Y., Kaneko K. Tensile properties and stress relaxation of polypropylene at elevated temperatures. Polymer Engineering and Science, 37 Iss. 1:8190, 2004.
[5] Tan V.B.C., Zeng X.S. Characterization and constitutive modeling of aramid fibers at
high strain rates. International Journal of Impact Engineering, 32:17371751, 2006.
[6] Umberger, P.D. Characterization and response of thermoplastic composites and constituents. Virginia Polytechnic Institute and State University, 2010.
[7] Cook, F.P. Characterization of uhmwpe laminates for high strain rate applications.
Virginia Polytechnic Institute and State University, 2010.
[8] Peijs, T., et al. Role of interface and fibre anisotropy in controlling the performance of
polyethylene-fibre-reinforced composites. Composites Science and Technology, 52:449
466, 1994.
[9] Tan, V.B.C., Ching T.W. Computational simulation of fabric armour subjected to
ballistic impacts. International Journal of Impact Engineering, 32:17371751, 2006.
[10] Koh C.P., Shim V.P.W., Tan V.B.C. Response of a high-strength flexible laminate to
dynamic tension. International Journal of Impact Engineering, 35:559568, 2007.
[11] Chen, W.W., Lim, J., Zheng, J.Q. Dynamic small strain measurements of kevlar 129
single fibers with a miniaturized tension kolsky bar. Polymer Testing, 29:701705, 2010.
[12] Miyano Y., Nakada M., Naoyuki S. Accelerated testing for long-term durability of gfrp
laminates for marine use. Composites: Part B, 35:497502, 2004.
32

33
[13] Hedgepeth, J.M., vanDyke, P. Local stress concentrations in imperfect filamentary
composite materials. Journal of Composite Materials, pages 294309, 1967.
[14] Hedgepeth, J.M. Stress concentrations in filamentary structures. NASA TN D-882,
1961.
[15] Beyerlein, I.J., Landis, C.M. Shear-lag model for failure simulations of unidirectional
fiber composites including matrix stiness. Mechanics of Materials, 31:331350, 1999.
[16] Okabe, T., Takeda, N., Kamoshida, Y., Shimizu, M., Curtin, W.A. A 3d shear-lag
model considering micro-damage and statistical strength prediction of unidirectional
fiber-reinforced composites. Composites Science and Technology, 61:17731787, 2001.
[17] Landis, C.M., Beyerlein, I.J., McMeeking, R.M. Micromechanical simulation of the
failure of fiber reinforced composites. Journal of the Mechanics and Physics of Solids,
48:621648, 2000.
[18] Landis, C.M., McGlockton, M.A., McMeeking, R.M. An improved shear lag model for
broken fibers in composites. Journal of Composite Materials, 33:667680, 1999.
[19] Landis, C.M., McMeeking, R.M. Stress concentrations in composites with interface
sliding, matrix stiness, and uneven fiber spacing using shear lag theory. International
Journal of Solids and Structures, 36:43334361, 1999.
[20] Oh K.P. A monte carlo study of the strength of unidirectional fiber-reinforced composites. Journal of Composite Materials, 13:311328, 1979.
[21] Beyerlein I.J., Phoenix S.L. Stress concentrations around multiple fiber breaks in an
elastic matrix with local yielding or debonding using quadratic influence superposition.
Journal of Mechanics and Physics of Solids, 44:19972039, 1996.
[22] Beyerlein, I.J., Phoenix, S.L. Statistics of fracture for an elastic notched composite
lamina containing weibull fibers -part i. features from monte-carlo simulation. Eng.
Frac. Mech., 57:241265, 1997.
[23] Beyerlein, I.J., Phoenix, S.L. Statistics of fracture for an elastic notched composite
lamina containing weibull fibers - part ii. probability models of crack growth. Eng.
Frac. Mech., 57:267299, 1997.

34
[24] Ibnabdeljalil, M., Curtin, W.A. Strength and reliability of fiber-reinforced composites:
localized load-sharing and associated size effects. Int. J. Solids Structures, 34:26492668,
1997.
[25] Sastry, A.M., Phoenix, S.L. Load redistribution near non-aligned fiber breaks in a two
dimensional unidirectional composite using break influence superposition. J. Mater.
Sci. Lett., 12:15961599, 1993.
[26] Gama, B.A., Gillespie, J.W. Punch shear based penetration model of ballistic impact
of thick-section composites. Composite Structures, 86:356369, 2008.
[27] Xiao, J., Gama, B., Gillespie Jr., J. Progressive damage and delamination in plain
weave s-2 glass/sc-15 composites under quasi-static punch-shear loading. Composite
Structures, 78:182196, 2007.
[28] Sun C.T., Potti S.V. A simple model to predict residual velocities of thick composite
laminates subjected to high velocity impact. International Journal of Impact Engineering, 18:339353, 1996.
[29] Liu, K, Piggott, M. Shear strength of polymers and fibre composites: 1. thermoplastic
and thermoset polymers. Composites, 26(12):829840, 1995.
[30] ASTM D732. Standard test method for shear strength of plastics by punch tool. ASTM,
2009.
[31] Crescenzi, F., Marini, F., Nardi, C., Pizzuto, A., Rossi, P., Verdini, L., Rajainmaki, H.,
Knaster, J., Bettinali, L. Mechanical characterization of glass fibreepoxy composite
material for iter pre-compression rings. Proceedings of the 26th Symposium of Fusion
Technology, 86:25532556, 2011.
[32] BA Gama, M Rahman, and JW Gillespie Jr. Energy absorbing damage mechanisms in
quasi-static punch shear of thick-section composites. CD Proceedings, SAMPE, pages
90921, 2004.
[33] S-W. R. Lee and C.T. Sun. A Quasi-Static Penetration Model for Composite Laminates.
Journal of Composite Materials, 27(3):251271, January 1993.
[34] S.V. Potti and C.T. Sun. Prediction of impact induced penetration and delamination in
thick composite laminates. International Journal of Impact Engineering, 19(1):3148,
January 1997.

35
[35] JA Nemes, H Eskandari, and L Rakitch. Effect of laminate parameters on penetration
of graphite/epoxy composites. International journal of impact engineering, (97):97112,
1998.

Potrebbero piacerti anche