Composite Structures 200 (2018) 173–186

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Finite element analysis of pre-stretch effects on ballistic impact performance T

of woven fabrics
Guodong Guo
Department of Mechanical and Aerospace Engineering, North Carolina State University, Raleigh, NC 27695, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Pre-stretch effects on ballistic impact performance of woven fabrics that made of Kevlar KM2 is investigated with
Ballistic impact finite element simulation. Individual yarns of the fabrics are modeled with truss elements incorporated with
Pre-stretch equivalent material properties of real fiber yarns. Before the projectile is issued, a fabric panel that has a rec-
Fabrics tangular configuration is pre-stretched by directly applying displacement boundary conditions. This two-step
Finite element analysis (FEA)
loading condition is realized through results transferring capability offered by the commercial software
ABAQUS. Simulation results reveal that pre-stretch can significantly influence the fabric’s ballistic response such
as ballistic limit, energy absorption and wave propagation. The fabric with higher pre-stretch absorbs more
energy, however fails earlier than the fabric with lower pre-stretch. Parametric studies show that with the
increase of pre-stretch, the deformation contour evolves from a pyramid shape that is conventionally observed in
non-pre-stretched fabrics into a conical shape.

1. Introduction
Woven fabrics that made of high strength aramid fibers such as
Kevlar have attracted great attention in applications such as soft body
armor and bulletproof vest. In these applications, the fabrics are re-
quired to provide ballistic resistance against the incident high speed
projectiles. To continuously improving the protection performance and
comfortability of the protective apparels, extensive studies have been
conducted on the ballistic impact behavior of woven fabrics, from ex-
perimental characterization to numerical simulation, in the past several
decades [1–3]. In Cheeseman’s [4] review paper, seven factors were
outlined that influence the ballistic performance of woven fabrics:
material properties, fabric structure, projectile geometry, impact velo-
city, multiple ply interaction, far field boundary conditions and friction
[5,6]. These factors can be further categorized into intrinsic factors like
material properties and extrinsic factors like impact velocity. Extrinsic
factors are often related to experimental method used, such as projectile
geometry and clamp designing. Another extrinsic factor that exerts a
great influence on the fabric response however has rarely been studied
is fabric pre-stretch that arises from clamping during experiment [7].
For example, in Ref. [8], it was reported that the fabric was slightly pre-
stretched before been fired upon. However, it is not known how much
pre-stretch was applied and it is not well understood how the pre-
stretch precisely affects the fabric deformation and its impact perfor-
mance.
Although many different clamping configurations have been studied
[9,10], it is tacitly assumed that the pre-stretch is zero in these con-
figurations. In reality, pre-stretch can be generated intentionally or
unintentionally when griping the fabric into a test machine. One of the
most obvious effects of pre-stretch is to diminish or eliminate initial
fiber crimp that arises from the elegant woven structures of fabrics. For
a single yarn, de-crimping generates a zero force with no initial strain,
while further stretch induces a pre-strain in fiber yarns. Shin et al. [11]
observed that yarn pre-tension in a fabric serves to induce failure at an
earlier instance and also reduce the cut energy of the fabric. Further-
more, while it is well known that friction improves the ballistic re-
sponse by maintaining the integrity of woven pattern, pretension has a
significant effect on friction. For example, in a quasi-static test, it was
found that the peak pull-out force of a fabric yarn increases with the
increase of pre-load magnitudes [12]. Under impact, to eliminate crimp
and consequently keep the fiber yarn straight, a slight pre-load is in-
evitably introduced [13]. On the other hand, even though for a fiber
reinforced composites rather than a dry fabric, many studies have been
conducted on pre-load effects and it is found that pre-load has sig-
nificant effects on their deformation, energy absorption and damage
property [14,15], there is actually very few literature reports on such
effects on woven fabrics under impact [16]. Since different researchers
stretch their fabric targets to a different degree before shooting, it is
highly critical to investigate the pre-stretch effect when comparing the
experimental results from different studies.
The challenge in studying the pre-stretch effects lies in the precise
controlling of the clamp to produce a determined pre-stretch. Also, the

E-mail address: gguo2@ncsu.edu.

https://doi.org/10.1016/j.compstruct.2018.05.067
Received 2 January 2018; Received in revised form 27 April 2018; Accepted 14 May 2018
Available online 19 May 2018
0263-8223/ © 2018 Elsevier Ltd. All rights reserved.
G. Guo Composite Structures 200 (2018) 173–186

fabric targets usually slip at its edges when projected upon [17], Table 1
making it extremely difficult to isolate the individual effect of these Material properties of fabric yarns.
factors. In contrast to experimental study, numerical simulation has Density (g/cm3) Elastic modulus (GPa) Failure strain (%)
substantial advantage in overcoming these deficiencies. The most pro-
minent advantage of numerical simulation is it can isolate each in- Kevlar KM2 1.44 82.6 4.7
Truss element 9.355 536.6 4.7
dividual mechanism and eliminate unwanted variables that would
couple to affect the impact behavior. It also allows parametric studies of
the individual effect where the dominant parameter can be meticu-
0.00785 mm2. This means it loses its accuracy in modeling the area of a
lously controlled. Over the past two decades, finite element has gained
real yarn. Thus the mechanical response of a single yarn, and therefore
great attention in modeling fabric impact behavior due to the excep-
a woven fabric, would not have been captured if the physical Kevlar
tional advantages. Finite element models across different scales require
KM2 properties were used. To account for this cross-sectional area ef-
different computational efforts. In the continuum level, the entire fabric
fect, an equivalent elastic modulus is introduced for the truss element
is modeled as a shell or membrane and anisotropic behavior is in-
such that it is equivalent to its fiber yarn counterpart in the sense of
corporated with material constitutive models [18,19]. This approach is
strain energy. In this way, the equivalent elastic modulus for the truss
computationally efficient and can model the global behavior of fabric
model is obtained through scaling that of the fiber yarn by its area ratio.
response, however ignores important fabric behavior such as yarn
To conserve the mass of a truss element with respect to a real yarn, the
sliding and yarn-yarn friction. To capture the individual yarn behavior,
volumetric density of which is scaled in the same manner. This process
meso-scale model has been developed [5,6,9,20,21]. In the meso-scale
gives out an equivalent elastic modulus of 536.6 GPa and an equivalent
approach, fabric yarns are explicitly modeled as solid elements and the
density of 9.355 g/cm3. In this manner, the sound speed of the material
material properties are set to dominant in a principal direction. This
is also retained and calculated as E / ρ = 7.57 km/s . Furthermore, in
approach has been widely used in the study of the effect of friction,
accordance with experimental results as reported in Ref. [25], a tensile
material properties, clamping configurations. However, the computa-
failure strain, implemented through the user subroutine VUSDFLD in
tion cost is extremely high. Another disadvantage is that stability be-
ABAQUS, is set to be 4.7% for a single yarn, which corresponds to a
comes an important consideration when setting the orthotropic mate-
failure stress of 3.88 GPa. These equivalent material properties are
rial properties to a single yarn, i.e., the transverse stiffness should never
listed in Table 1 and will be used in the simulations hereafter. In ad-
be too small as required to mimic an ideal yarn, which leads to ex-
dition, the yarn path is determined by following the procedure de-
cessive bending stiffness for a fabric thus distracts the impact behavior.
scribed in Ref. [5]. A single yarn geometry created in this way is shown
A more simplified technique is to use truss elements to model fabric
in Fig. 1a. The mechanical response of a single yarn predicted by the FE
yarns. Tan et al. [1] and Sanhita et al. [22] fall into this category. Truss
model is shown in Fig. 1b. Here stretch ratio, which is defined as the
element model significantly reduces the element numbers and is thus
ratio between applied displacement and the initial distance between the
more computationally efficient. However, it loses important fabric in-
two fiber ends, is used rather than strain because initially the yarn
formation such as yarn cross-sectional area.
strain is zero due to crimp. As can be readily seen from Fig. 1b, the force
The goal of this paper is to investigate pre-stretch effects on ballistic
is initially very small and close to zero, which represents the crimp
impact performance of woven fabrics. To achieve the goal, a finite
stage that is usually found in a yarn directly extracted from a fabric. The
element model is constructed at the yarn level with truss element. It
amount of yarn crimp accounts for approximately 1.2% of its stretch
differs from the conventional truss element method in that equivalent
ratio. After the yarn straightens out, a force is developed in the yarn and
material properties are used thus single yarn behavior is well captured.
increases linearly until it reaches the peak value determined by its
Pre-stretch effects on a single yarn impact is first studied and validated
failure strain. This exactly reproduce the behavior of a physical yarn
against analytical solutions. Fabric response is then validated against
that has an elastic modulus of 82.6 GPa.
experimental results. The pre-stretch effect is studied for rectangular
shaped woven fabrics with four edges fixed boundary conditions. The
fabrics were first stretched by directly applying displacement boundary
conditions at its edges and then shot by a spherical projectile. This two-
step simulation is achieved by results transferring capability provided 1.494 mm
by the commercial software ABAQUS [23]. The effect on ballistic per-
formance such as V50 velocity, energy dissipating mechanism and de-
formation shape is studied and compared against that of non-pre-
stretched fabrics. (a)
200
2. Pre-stretch effects on a single yarn impact

2.1. Yarn model and material properties 150

Force (N)

In reality, a single fabric yarn consists of hundreds of fiber filaments

and thus is a very complex material [24]. The fiber material used in this 100
study is Kevlar KM2, which has a Young’s modulus of 82.6 GPa and a
fiber volumetric density of 1.44 g/cm3 [25]. Since a single yarn has
50
little resistance to bending, it is modeled with truss element provided in
commercial software ABAQUS. The truss element assumes a homo-
genous material and a linear elastic material model is used to char- 0
acterize its behavior. An important consideration in modelling a single 0 1 2 3 4 5 6 7
yarn is its cross-sectional areal, which is usually set to be 0.051 mm2, Stretch-ratio (%)
including the area of Kevlar fiber and void space in a 600 denier yarn (b)
[26]. In the truss model, the radius of the truss element is set to be
0.05 mm to preserve a thickness of 0.2 mm for a single-ply plain woven Fig. 1. Single yarn, (a) a path diagram and (b) its force-stretch ratio relation-
fabric. However, this radius gives out a cross-sectional area of ship.

174
G. Guo Composite Structures 200 (2018) 173–186

2.2. Single yarn impact 0.25

0.2
Impact on a single yarn was studied by Smith et al. [27], it was
found that when a straight yarn with an infinitely length is impacted 0.15

Strain (%)
transversely, two waves start propagating along the yarn. One is long-
0.1
itudinal wave, which propagates outward along the yarn from the im-
Crimp1
pact point at the speed of sound. The material in the wake of the 0.05 Crimp2
Crimp3
longitudinal wave front is strained and set into motion toward the Straight1
0
impact point. The other one is transverse wave which propagates to- Straight2
Straight3
ward the direction of impact at a much lower speed. Smith et al. [27] -0.05 Strained1
Strained2
found that the tensile strain ε in the wake of the longitudinal wave front Strained3
is given by -0.1
0 0.5 1 1.5 2
time (μs)
V = c ε (2 ε (1 + ε ) −ε ) (1)
Fig. 2. Strain time history extracted from three locations on the yarns from the
where V is the impact velocity and c is the sound speed of the material. three models.
After the tensile strain is determined, the transverse wave speed U
with respect to laboratory can be obtained by
diminishes quickly and oscillate between positive and negative when
U = c ( ε (1 + ε ) −ε ) (2) the longitudinal wave passes through. Also, the peak strain value de-
grades gradually as the strain wave propagates along the yarn. Since the
For a tensile failure strain of 4.7%, Eq. (1) predicts that an impact material properties are all the same, these effects are all attributed to
velocity of 1033 m/s is needed to break the fiber yarn. However, it is the crimped geometry effect. For the ‘strained’ model, 2% pre-stretch
reported from literature that the actual velocity needed to break the generates a pre-strain of 0.8% in the yarn (Fig. 1b) and Eq. (3) yields a
yarn is significantly lower than this prediction [28]. strain of 0.093% due to impact in additional to pre-strain. In Fig. 2, the
When pre-strain is considered, an analytical solution that implicitly pre-strain value is subtracted from the final strain results to show the
gives out strain versus impact velocity relationship is derived as wave propagation. It can be seen that, due to pre-stretch, an additional
strain of approximately 0.092% is generated, which is very close to
V = c ε (2 (ε + ε0)(1 + ε ) −ε ) (3)
what predicted by Eq. (3). In addition, similar to the ‘straight’ model,
where ε0 is pre-strain within the yarn due to pre-tension. The transverse the strain magnitude is well maintained throughout the wave propa-
wave speed in the pre-tension model can be calculated as gation history.
Next we study the transverse wave propagation. The transverse
U = c ( (ε + ε0)(1 + ε ) −ε ) (4) wave speed calculated by the three FE models as well as by Eqs. (2) and
In reality, the yarns in woven fabrics are often crimped due to (4), respectively for ‘straight’ model and ‘strained’ model is summarized
permanent deformation or possible damages during manufacturing in Table 2. It can be seen that for both ‘straight’ model and ‘strained’
process. Therefore, a crimped yarn directly extracted from a fabric is a model, FE results is in good agreements with analytical results. On the
better representation of what is used in real applications. Fig. 1b shows other hand, compared to longitudinal wave, transverse wave propa-
that a crimped yarn undergoes three deformation stages when stret- gates at a much lower speed. However, as opposed to longitudinal
ched: (1) de-crimping with zero strain, (2) straight yarn with zero strain wave, which propagates at the same speed in all the three models,
and (3) straight yarn with non-zero strain. Based on these outlined transverse wave speed shows a high sensitivity to pre-stretch magni-
deformation stages, single yarn impact simulations were conducted for tude. It travels fastest in the ‘strained’ model, followed by ‘straight’
three sets of configurations: in the first simulation, impact was con- model and travels slowest in the ‘crimped’ model. This phenomenon is
ducted on a crimped yarn. This model is denoted as ‘crimped’ model; in shown evidently in Fig. 3, where the yarn deformation profile is plotted
the second simulation, impact was conducted on a straight yarn with at 8 μs after impact. It should be noted that in the ‘crimped’ model,
zero strain, which represents the yarn condition at the time instant it material in the wake of longitudinal wave not only moves along the
just straightens out. This model is denoted as ‘straight’ model; in the yarn direction, but also moves vertically. This explains why the strain
third simulation, before impact was conducted on the yarn, it was ap- magnitude diminishes as shown in Fig. 2.
plied with a stretch with a stretch ratio of 2%. It can be readily seen As a conclusion, it is found that the single yarn impact performance
from Fig. 1b that this pre-stretch magnitude generates an initial strain is significantly affected by pre-stretch. Strain in the yarn is determined
in the fiber yarn. This model is denoted as ‘strained’ model. In the si- by longitudinal wave propagation, while the deformation profile is
mulations, three yarns were applied with a constant impact velocity of determined by transverse wave propagation. These results provide
100 m/s to its midpoint. According to Eqs. (1) and (2), for a straight important insight into the impact performance of woven fabrics.
yarn, this impact velocity generates a constant tensile strain of 0.2% in
the wake of longitudinal wave front and a transverse wave with a speed
of 323.46 m/s. 3. Fabric model and validation
To study the longitudinal wave propagation of these three models,
the strain time history is tracked at three different points, of which the 3.1. Fabric impact model
positions are roughly the same for each model, on the yarn as shown in
Fig. 2. It can be readily see that in all these three models, the long- The FE model in the current work is comprised of a balanced plain-
itudinal wave propagates at a speed of 7.56 km/s, which is very close to
what predicted by the analytical model. It comes as no surprise because Table 2
Transverse wave speed (m/s).
the same material properties are applied for all the models. In the
‘straight’ model, a strain with a magnitude of approximately 0.2% is Analytical FE
generated, equals to what gives out by Eq. (1). This strain magnitude is
‘Crimped’ model 217.8
maintained in the wake of the longitudinal wave front as it propagates.
‘Straight’ model 323.5 330.2
For the ‘crimped’ yarn model, however, instead of maintaining a con- ‘Strained’ model 708.9 709.4
stant strain in the wake of the longitudinal wave front, the strain

175
G. Guo Composite Structures 200 (2018) 173–186

Transverse wave front

Crimped
Impact location

Straight

Strained

Fig. 3. Single yarn deformation 8 μs after impact.

woven Kevlar fabric with 34 yarns per-inch arrayed in the warp di-
rection and 34 yarns per-inch in the weft direction. This scheme creates Fig. 4. A comparison between FE simulation and experimental results on the
a yarn span of 0.747 mm. The projectile is assigned with a spherical deflection time history of the impact site at four different projectile velocities.
shape and modeled as a rigid body with a diameter of 5.56 mm and a
mass of 0.692 g. This configuration is chosen to be the same as what smaller elastic modulus leads to a larger peak deflection and a delay on
used in Ref. [2] for validating purpose. The general contact option in the stop of the projectile.
the commercial software ABAQUS is employed to determine the inter-
action between the fabric yarns as well as between fabric and the
4. Methodology of modelling pre-stretched fabrics
projectile. The friction coefficient is set to be 0.2 for both yarn-yarn
contact and fabric-projectile contact as found in Ref. [22]. During si-
So far the FE model for both a single yarn and a fabric has been
mulation process, the elements that meet the aforementioned strain
validated. Next the ballistic impact of fabrics that are under pre-stretch
failure criterion will lose their load bearing capacity and consequently
is illustrated.
will be removed from the model.
The fabric under study has a configuration as displayed in Fig. 5a. A
single layer of 55.278 mm * 55.278 mm rectangular fabric is clamped at
its four edges. The fabric is first stretched simultaneously in its two
3.2. Model validation
perpendicular directions and then impacted by the spherical projectile
at various speeds. The uniaxial force vs stretch ratio relationship of the
The constructed FE model is first validated against available ex-
fabric is shown in Fig. 5b. It can be seen that the force-stretch curve
perimental results. Experimental impact tests result reported in Ref. [2]
consists of two phases. The first phase is from zero to 0.22% stretch
are used to validate the FE model in the current study. In their ex-
ratio. The force applied in this phase is almost zero, so the fabric re-
periments, a comprehensive set of impact tests were performed at
sponse is characterized by de-crimping. However, it should be noted
various impact velocities on a Kevlar KM2 fabric. Both a cylindrical and
that compared to a single fiber, the de-crimping phase of a fabric is
a spherical projectile were used to shoot the fabric target. However,
significantly lower due to yarn-yarn interlacing which potentially
only the spherical projectile is exploited here for validation purpose. A
constraint the yarn mobility. The second phase is from 0.22% stretch to
singly-ply fabric was clamped to a steel frame that has a circular
2.35% stretch. The fabric response in this phase is dominated by yarn
aperture with a diameter of 50.8 mm. In simulation, this configuration
elongation as indicated by the linear force vs stretch ratio curve. It also
is accomplished through fixing all the degrees of freedom of the nodes
should be noted that the strain field generated due to stretch in the
along a circular boundary with the same diameter.
fabric is quite non-uniform. That is why at the instant of roughly 2.35%
The first quantity to validate is V50 velocity, which represents the
stretch (below the maximum strain failure criterion), rupture occurs to
projectile speed at which the fabric has a 50% possibility of penetrating.
some of the yarns. During simulations, the pre-stretch level is carefully
The FE simulation predicts a V50 velocity of 103.75 m/s, which is in
controlled by directly applying a displacement boundary condition. The
good agreement with experimental results, where it was found that the
same amount of displacement is applied in the two opposite directions.
V50 velocity is bounded between 83 m/s and 105 m/s. The second
The specific pre-stretch level that are studied include 0 (referred to as
quantity to validate is deflection time history of the impact site, which
baseline model), 0.15%, 0.4%, 0.8% and 1.4%. The corresponding
is compared against the experimental results at four different velocities
magnitude of force needed to achieve such stretch ratios can be readily
(three velocities below V50 velocity and one velocities above V50 ve-
read out from Fig. 5b. For each of the pre-stretch levels, approximately
locity). The results are shown in Fig. 4. For the non-penetrating cases,
ten simulations are conducted at various impact velocities. Such
the peak deflection corresponds to the instant of projectile arresting.
quantities as residual velocities, fabric deformation and energy dis-
For the penetrating cases, the peak deflection corresponds to the time
sipations are recorded and analyzed. In addition, it is assumed that the
instant the first yarn ruptures. It can be seen that the FE results agrees
kinetic energy of the projectile is dissipated by fabric absorption in such
reasonably well with the experimental results. The deflection rates of
mechanisms as fabric strain energy, fabric kinetic energy and friction
the fabric are exactly reproduced by the FE model among all the con-
energy. However, it is known that in a fabric configuration that has four
sidered velocities. In addition, the FE model slightly underestimate the
edges clamped, the friction energy only accounts for a very small per-
peak deflections and the peak deflections occurs at an earlier time than
centage of dissipation thus will not be analyzed further [9].
the experimental results. This is attributed to fabric slippage at the
fixture [17]. The slippage won’t happen in FE simulation. On the other
hand, we note that a truss element based FE model is studied in the Ref. 5. Results and discussion
[22]. However, the material properties are different here, i.e., an
equivalent material property is used to account for yarn volume effect. Upon shooting a fabric, depends on the projectile speed, the out-
It also should be noted that in Ref. [22], an overestimate of peak de- come is either non-penetrating or penetrating. The watershed that
flection was predicted. This might due the material properties that was characterize the critical speed of the dividing point is called ballistic
used. Even a slight difference in material properties would result in a limit or V50 velocity as stated before. It plays such an important role in
different impact response. For example, for a non-penetrating event, a ballistic impact performance thus will be studied first. During

176
G. Guo Composite Structures 200 (2018) 173–186

55.278 mm degrades when pre-stretched. Another important observation is that the

V50 velocity even drops as the pre-stretch increase from 0 to 0.15%, at
which point the applied force is still zero. This phenomenon again
highlights the superiority of stretch controlled fabric boundary condi-
tions over load controlled ones where this phenomenon might be
skipped over.
Displacement boundaries

Fig. 6 indicates that for an intrusion speed that is lower than 96.
25 m/s, the projectile will be stopped in all these pre-stretch levels,
while for an intrusion speed that is greater than 116.25 m/s, all fabrics

55.278 mm
will be penetrated. To fully understand the energy dissipating me-
chanisms, it is advantageous to first study the non-penetrating impact
where any premature failure interference is prevented. In the following
Weft sections, a 80 m/s impact velocity is first chosen to study the non-pe-
netrating impact, followed by a penetrating impact of which the impact
velocity is 120 m/s.
Warp
5.1. Pre-stretch effects on a non-penetrating impact

Fig. 7a–d show the projectile velocity history, fabric deflection

history (from impact site), fabric strain energy history and fabric kinetic
(a) energy history for the pre-stretched fabrics at the impact velocity of
80 m/s. At this speed, the projectile in all cases are arrested as indicated
8
by the negative residual velocities in Fig. 7a. By observing the projectile
velocity evolution, it can be seen that the pre-stretch affects the velocity
6 history in a very consistent manner. With the increase of fabric pre-
stretch, the projectile velocity slows down more quickly, which means
Force (kN )

projectile kinetic energy dissipates in a faster way. As a consequence,

4 the projectile is arrested in a shorter time with the increase of fabric
pre-stretch. For non-penetrated impact, the maximum fabric deflection
is an important parameter because which is usually related to the blunt
2 trauma [29]. From Fig. 7b, it can be observed that the more pre-stretch,
the lower the intrusion depth. The difference between the 1.4% pre-
stretched case and the baseline model is about 2 mm. Although it looks
0 like small, it leads to significantly different consequences for applica-
0 0.5 1 1.5 2 2.5 3 tions like a body protection.
Stretch-ratio (%)
(b) 5.1.1. Strain energy
Fig. 5. (a) Fabric model and (b) its uniaxial force-stretch ratio relationship.
For the baseline model and the 0.15% pre-stretched model, there is
no strain energy in the fabric before shooting. After shooting, strain
energies in this two cases remain zero for a few micro-seconds due to
120
crimp. The strain energy in the 0.15% pre-stretched case develops
slightly earlier than which in the baseline model since a portion of
crimp has been eliminated due to pre-stretch. After that, the strain
110
V50 velocity (m/s)

energy in both of this two cases gradually increase. In the situations

100
90
80
0 0.5 1 1.5
Pre-stretch ratio (%)
Fig. 6. V50 velocity vs pre-stretch relationship.
simulation, various projectile velocities were applied, when an upper
and lower bound speed limit that result in a penetrating and a non-
penetrating is found, a speed increment of 2.5 m/s is applied gradually
to find out the V50 velocity. The V50 velocities for all the considered
pre-stretching levels are plotted in Fig. 6. They are 116.25 m/s,
113.75 m/s, 111.25 m/s, 108 m/s and 96.25 m/s respectively for the
pre-stretch level of 0, 0.15%, 0.4%, 0.8% and 1.4%. It indicates that the
V50 velocity decreases gradually with the increase of fabric pre-stretch.
In the other words, the ballistic load bearing capacity of fabrics
that the pre-stretch is greater than 0.15%, initial strain energies had
been developed in the fabric due to yarn tension before shooting. A
close observation shows that immediately after impact, these is a
sudden tiny drop in the strain energy for these cases. This is because the
strain field generated in the pre-stretch stage is non-uniform due to
yarn-yarn friction in some regions as stated earlier. Those regions that
subjected to a larger strain relax to some extent due to material flow in
the wake of longitudinal strain waves. In Fig. 7c, for comparison pur-
pose, the strain energy due to pre-stretch is subtracted from the total
strain energy and the lowest point is assumed to be zero for those case
that the pre-stretch is beyond 0.15%. The strain energies generated
after impact are attributed to longitudinal straining of fabric yarns due
to longitudinal strain wave, which will be analyzed further later. Strains
in the yarn increase each time the wave is reflected off the fabric edges,
gradually building up the strain energy. It can be observed from Fig. 7c
that the strain energy develops faster and faster with the increase of
pre-stretch.
5.1.2. Fabric kinetic energy
Another important energy dissipating mechanism is kinetic energy
of fabrics, which is closely related to transverse wave propagation.
Upon impact, besides longitudinal wave, a transverse displacement

177
G. Guo Composite Structures 200 (2018) 173–186

(a) 80 (b) 7

60 6

Displacement (mm)
40 5

Velocity (m/s)
20 4

0 3
Pre-stretch 0% Pre-stretch 0%
-20 0.15% 2 0.15%
0.4% 0.4%
-40 0.8% 0.8%
1
1.4% 1.4%
-60 0
0 20 40 60 80 100 120 0 50 100 150
Time (μs) Time (μs)
(c) 2000 (d) 700
Pre-stretch 0% Pre-stretch 0%
0.15% 600 0.15%
0.4%

Kinetic energy (mJ)

0.4%
1500
Strain energy (mJ)

0.8% 500 0.8%

1.4% 1.4%
400
1000
300

200
500
100

0
0 0 20 40 60 80 100 120
0 20 40 60 80 100 120
Time (μs) Time (μs)

Fig. 7. Ballistic impact responses of fabrics at the projectile speed of 80 m/s. Time history of (a) projectile velocity, (b) impact site deflection, (c) fabric strain energy,
(d) fabric kinetic energy.

wave is initiated in the direction of projectile and propagates outward.
The fabrics that involved in the transverse wave propagation will be
accelerated. The kinetic energy arises from the motion of the ac-
celerated fabrics accounts for most of the kinetic energy in the fabric.
Thus the kinetic energy is determined by the fabric area that involved in
the transverse wave as well as transverse velocity of those fabric.
Initially, with the increase of the area of the accelerated fabric, the
kinetic energy increases. As shown in Fig. 7d, it achieves its maximum
value at the time instants of 86.4 μs, 83.6 μs, 76.4 μs, 65.6 μs and 52.4 μs
respectively for the pre-stretch magnitude of 0, 0.015%, 0.04%, 0.08%
and 0.14%. For a square fabric that under four edge clamped boundary
conditions, it generally achieves its maximum kinetic energy after the
transverse wave reaches the fabric edges and reflects back. So these
time instants are indications of the transverse wave propagation speed,
which will be discussed further later. In addition, the maximum kinetic
energy absorbed shows a decreasing tendency with the increase of pre-
stretch magnitude.
5.1.3. Deflection shape
Figs. 8 and 9 show deflection contours of the fabrics under 80 m/s
impact for the baseline model and the 0.15% pre-stretched model, re-
spectively at different time instants. For these two cases, the fabrics are
initially in their de-creeping stage. The transverse wave propagation
can be clearly seen from these deflection contour plots. It should be
noted that the transverse displacement due to pre-stretch and initial
contact adjustment is negligible. The deflection contour edges show the
transverse wave front. For the baseline model, the transverse wave
speed is faster in its principal axis than in the other directions, and it
assumes its lowest speed in the 45-degree direction. Consequently, the
transverse wave propagation generates an approximately pyramid
shape with the deflection edges in 45-degree concave slightly inward.
The two diagonals of the square base of the pyramid lies in the two
perpendicular central-most principal yarn directions. This phenomenon
has been confirmed by many experimental and numerical studies. The
pyramid shape is maintained until it hits the fabric edges at 68 μs. At
this time instant, those transverse displacement waves that hit the
fabric edges will reflect back toward the impact site, and the rest of the
waves continues to propagate until reach the fabric edges. The trans-
verse wave in 45-degree direction is the last one that reaches the fabric
edges since it has a longer distance to go as well as propagates in a
relatively slow speed. Finally, the transverse wave will spread out over
the entire fabric. Interestingly, the transverse wave spread over the
entire fabric at 128 μs, which is approximately the time instant the
projectile is completely stopped. The deflection contours for the 0.15%
pre-stretched case show a same trend with that of the baseline model.
The transverse wave propagation generates a pyramid deflection shape
and then spreads out over the entire fabric. However, a close observa-
tion reveals that the transverse wave in principal yarns arrives the
fabric edges at an earlier time than the baseline model. Also, the
transverse wave spreads out over the entire fabric at a slightly earlier
time than when the projectile is stopped. Indicating a faster transverse
wave speed.
Figs. 10 and 11 display the deflection contours for the case of 0.8%
and 1.4% pre-stretch, respectively at the impact speed of 80 m/s. For
these two cases, tensile strain had been developed due to yarn tension
before impact, as analyzed earlier. By looking at Fig. 11, a more
strained case, it can be seen that the deflection contours due to trans-
verse wave propagation displays a more conical shape with a circular
base, dramatically different from the pyramid shape in the baseline
model. The deflection edges in the 45 degree bulges outward as op-
posed to the concave shape in the baseline model, which indicates that
the transverse wave speed in the 45-degree direction is accelerated as
compared to that in the principal yarn direction. The transverse dis-
placement wave hits the fabric edges at the time instant of 43 µs. Again,
those wave that arrived at the fabric edges will get reflected backward
and the rest of which will continue to propagate toward the fabric

178
G. Guo
Fig. 8. Deflection contours of baseline model at four different time instants.
edges. The transverse displacement waves spread out the entire fabric
at the time instant of 65 µs, which is drastically earlier than the time
instant when the projectile is arrested. For the case of 0.8% pre-stretch,
the deflection contour shows an intermediate shape between a cone and
a pyramid. It firstly arrives at the fabric edges at the time instant of
51 µs and spread over the entire fabric at the time instant of 85 µs.
Further study shows that for the other pre-stretch levels (not shown in
here), the deflection contours and wave propagation displays a very
consistent manner, i.e., the deflection contour becomes more conical
and transverse wave spread out more quickly with the increasing of pre-
stretch levels. A summarization of these important time instants is given
in Table 3, including the time instants when I) transverse wave first hits
fabric edges; II) maximum kinetic energy (KE) is achieved III) trans-
verse wave spread over entire fabric IV) projective speed goes to zero.
A close correlation can be observed between these deflection con-
tour plots (or transverse wave propagation) and the fabric kinetic
179
Composite Structures 200 (2018) 173–186
energy histories. First, by comparing Figs. 8a, 9a, 10a, and 11a, which
display the deflection contours at the same time instant of 35 µs, it can
be seen that the transverse deflection involves the largest portion of
fabric area in the 1.4% pre-stretched case, followed by 0.8% and 0.15%
pre-stretched case while smallest in the baseline model. This indicates
an increased transverse wave speed with the increase of pre-stretching
magnitude. The maximum kinetic energy for all the cases occur at the
time instant after transverse wave arrives at the fabric edges. This ex-
plains why the time instants at which maximum kinetic energy occurs
show a decreasing tendency in Fig. 7d. Also, for the strained fabric, (the
pre-stretch level above 0.4%), at the first few micro seconds, the kinetic
energy in the more strained cases are higher than less strained cases.
5.1.4. Fabric strain
Strain developed in the fabrics is mainly due to longitudinal wave
propagation. Longitudinal wave initiates at the impact site and travels
G. Guo
Fig. 9. Deflection contours of 0.15% pre-stretched model at four different time instants.
in a much faster speed than the transverse wave. The secondary yarns
involve in longitudinal wave propagation due to interaction with
principal yarns although they are not directly in contact with the pro-
jectile. So the longitudinal wave quickly spread out the entire fabric.
After arrives at the fabric edges, it reflects back, each reflection in-
creases the strain magnitude. Consequently, the strain in the fabric
builds up gradually during impact. On the other hand, for different pre-
stretch levels, although longitudinal waves travel at the same speed, the
strain magnitude is different due to crimp and pre-stretch as seen from
Fig. 2. The strain magnitude will also be affected by friction at cross-
overs within the fabric. For pre-strained fabrics, a pre-stress is also
generated before impact. These factors work simultaneously during the
impact process, the outcome of which is a faster strain energy devel-
oping speed in more pre-stretched fabrics as shown in Fig. 7c.
Fig. 12 displays the contours of tensile strain for the baseline model,
0.4% pre-stretch, 0.8% pre-stretch and 1.4% pre-stretched cases at the
180
Composite Structures 200 (2018) 173–186
time instant of 80 µs for the 80 m/s impact. For the baseline model, the
maximum tensile strain occurs to the principal yarns in the two vertical
directions, while decreases to the neighboring areas. For the other three
cases, a strain field is already generated before impact was taking place.
Thus the strain in the pre-strained case consists of two parts: pre-strain
due to fabric stretch and strain due to impact. The pre-strain field is
quite non-uniform due to yarn-yarn interactions as stated earlier. For
example, the largest strain value always occurs in the four corner areas
after stretch. Nevertheless, after impact, the maximum strain again
acquires its maximum value along the principal yarns as shown in
Fig. 12b–d and strain spread out the entire fabric due to fast long-
itudinal wave propagation. It also shows that the maximum strain in a
fabric increase with the pre-stretch level. For a tensile strain based
failure criterion, this means the fabric with more pre-stretch fails at an
earlier time than less pre-stretched fabric.
G. Guo
Fig. 10. Deflection contours of 0.8% pre-stretched model at three different time instants.
5.2. Pre-stretch effects on a penetrating impact
In this section, effects of pre-stretch on a penetrating impact is in-
vestigated. Two major aspects are highlighted. First, what is the po-
tential indication of non-penetrating impact on penetrating impact?
Secondly, how pre-stretch may affect the mechanism that governs pe-
netration of the projectile into fabrics.
The impact velocity chosen for the penetrating simulation is 120 m/
s. As indicated by Fig. 6, at this velocity, fabric targets at all pre-
stretched levels will be penetrated through eventually. Nevertheless,
different energy dissipating mechanisms arise due to different pre-
stretch levels. Fig. 13 displays the velocity history, strain energy history
181
Composite Structures 200 (2018) 173–186
and kinetic energy history respectively for the baseline model, 0.15%,
0.4%, 0.8% and 1.4% pre-stretched model. The projectile first decele-
rates and finally acquires a constant residual velocity. Based on the
velocity history, the impact process can be divided into two phases:
before penetration and after penetration. The time instants when the
fabric is immediately penetrated are indicated by arrows in Fig. 13. It
can be observed that, before penetration, the variation of velocity his-
tory at different pre-stretch levels shows a very consistent manner. The
1.4% pre-stretched fabric decelerates more quickly than other cases
therefore absorbs energy faster. The projectile energy is dissipated
mainly through fabric strain energy and fabric kinetic energy. The
strain energy also builds up more quickly in a more pre-stretched fabric.
G. Guo Composite Structures 200 (2018) 173–186

Fig. 11. Deflection contours of 1.4% pre-stretched model at three different time instants.

Table 3 where the deflection contour is plotted 35 µs after impact, for baseline
Important time instants for different pre-stretch levels (µs). model and 1.4% pre-stretched model, respectively. It can be readily
Pre-stretch First hit Maximum KE Spread over Projectile
seen that due to pre-stretch, the transverse wave not only accelerated in
(%) fabric edges achieved entire fabric stopped principal yarn direction, but also in other directions. From these char-
acteristics, it can be concluded that before penetration, the energy
0 68 86.4 127 127.2 dissipating mechanism and wave propagation characteristics are ana-
0.15 66 83.6 120 123.6
0.4 60 76.4 107 118.5
logous to the non-penetrating cases analyzed earlier.
0.8 51 65.6 85 110.1 Fabric penetration (failure) is determined by the prescribed strain
1.4 43 52.4 65 98.5 criterion. Strain in each fiber bundle is determined by initial strain and
longitudinal wave propagation after impact. Since fabric initial strain
increases with the increase of pre-stretch value, the fabric with higher
The kinetic energy acquires its maximum value earliest in the most pre- pre-stretch will meet the strain criterion earlier than those with lower
stretched case, indicating a faster transverse wave propagation. The pre-stretch. Fig. 13a shows that the penetration instant for baseline
transverse wave propagation characteristic is evidently seen in Fig. 14 model, 0.015%, 0.4%, 0.8% and 1.4% pre-stretched model are 95 µs,

182
G. Guo
Fig. 12. Longitudinal strain at 80 µs at the impact speed of 80 m/s of (a) baseline model, (b) 0.4% pre-stretch model, (c) 0.8% pre-stretch model, (d) 1.4% pre-stretch
model.
92 µs, 84 µs, 77 µs and 64 µs, respectively. Due to premature failure,
energy of the projectile cannot be fully dissipated by the fabric target. A
sharp peak in strain energy history (Fig. 13b) manifest the time when
the fabric fails. A sharp peak indicating fabric failure also appears in
kinetic energy history plot as shown in Fig. 13c. However, it should be
noted that the peak in Fig. 13c might occur even before the kinetic
energy first achieves its maximum. For a specified strain criterion, this
could happen when the impact velocity is high enough or the pre-
stretch is high enough. In both of these two cases, it is possible that the
fabric fails before transverse wave starts to hits the fabric edge.
Fig. 15 shows deformed configuration of the baseline model and
1.4% pre-stretched model at the time instants immediately after the
fabrics are penetrated. It can be seen that the fabric penetration in all
the two cases are characterized by principal yarn failure right beneath
the projectile. The reason is that the yarns at the impact site build up
strain faster than other area due to longitudinal wave propagation in
crimped fabric structure.
6. Conclusions
This study investigates pre-stretch effects on ballistic impact per-
formance of woven fabrics using truss elements model incorporated
183
Composite Structures 200 (2018) 173–186
with equivalent material properties of real Kevlar yarns. Important
findings drawn from this research is summarized as follows.
First, the equivalent material properties based truss FE model suc-
cessfully captures the mechanical behavior of woven fabrics and is
computationally very efficient. It makes the study of pre-stretch effects
on the ballistic impact performance of woven fabrics possible and in an
acceptable amount of time, which would be otherwise super time
consuming if solid element is used.
Secondly, the ballistic impact performance of woven fabrics is sig-
nificantly affected by pre-stretch in the following aspects: 1) the bal-
listic load bearing capacity degrades and thus the ballistic limit is de-
creased due to pre-stretch; 2) after impact, the projectile decelerates
more quickly in a more pre-stretched fabric because it absorbs energies
more quickly; 3) although longitudinal waves travel at the same speed
in all pre-stretch levels, the magnitude of strain developed is dramati-
cally different, leading to different degrees of strain energy in each
cases, i.e., the higher the pre-stretch level, the more strain energy is
generated; 4) In a more pre-stretched fabric, the deformed shape dis-
plays a more conical shape rather than a diamond shape as is commonly
seen, which means transverse wave speed increases in 45-degree di-
rection relative to that in the principal yarn direction as compared to a
non-pre-stretched fabric; 5) transverse wave speed increases in all
G. Guo Composite Structures 200 (2018) 173–186

120 a 5000
Pre-stretch 0% 77µs 84µs92µs
64µs 95µs
100 0.15%
4000

Strain energy (mJ)

0.4%

Velocity (m/s)
80 0.8%
3000 1.4%
60
Pre-stretch 0%
2000
40 0.15%
64µs
0.4%
77µs
20 0.8% 84µs 1000
1.4% 92µs
95µs
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Time (μs) Time (Ps)
2000 c
Pre-stretch 0%
Kinetic energy (mJ) 0.15%
1500 0.4%
0.8%
1.4%

1000

95µs
500
64µs
84µs 92µs
77µs
0
0 20 40 60 80 100
Time (μs)
Fig. 13. Ballistic impact response of fabrics at the projectile speed of 120 m/s. Time history of (a) projectile velocity, (b) fabric strain energy, (c) fabric kinetic energy.
Arrows indicate when the fabric is penetrated.

directions as pre-stretch magnitude increases, leading to an earlier Acknowledgements

achievement of maximum fabric kinetic energy; 6) for a non-pene-
trating impact, the intrusion depth is decreased due to fabric pre- The author would like to acknowledge the China Scholarship
stretch; 7) for a penetrating impact, more pre-stretched fabrics pene- Council for the financial supporting of a scholarship. The discussions
trate at an earlier time instant than a less pre-stretched fabric. contributed from Prof. Kara Peters and Prof. Mark Pankow from NCSU
are greatly appreciated.

Fig. 14. Deflection contour plots after 35 µs impact of (a) baseline model and (b) 1.4% pre-stretched model.

184
G. Guo Composite Structures 200 (2018) 173–186

Fig. 15. Deformed fabric configuration immediately after penetration, at side view and bottom view. (a) Baseline model. (b) 1.4% pre-stretched model. Bottom views
are enlarged for easier visualization.

Appendix A

A.1. Derivations of Eqs. (3) and (4)

For a single yarn impact, after solving differential equations of motion, the transverse wave velocity, with respect to a Lagrangian coordinate
system, is solved by Smith as (Ref. [27], Eq. (23))
T
ULag =
M (1 + ε ) (A.1)
where M is the mass per unit length of the fiber yarn. ε is the strain generated after impact. T is the yarn tension. In a preloaded yarn, T consists of the
pretension and the force due to impact, which can be written as
T = T0 + F (A.2)
where T0 denotes the pretension and F denotes the force due to impact. Substituting Eq. (A.2) into Eq. (A.1) yields
T0 + F
ULag =
M (1 + ε ) (A.3)
In the laboratory coordinate system, the transverse wave velocity is given as (Ref. [27])
U = (1 + ε ) ULag−W (A.4)
where W is the velocity of material flow in the wake of longitudinal strain wave. Substituting Eq. (A.3) into Eq. (A.4) yields
T0 + F
U = (1 + ε ) −W
M (1 + ε ) (A.5)
For a linear elastic fiber yarn that has a cross section area A and Young’s modulus E. The following equations are satisfied.
T0 = Eε0 A (A.6)
F = EεA (A.7)
M = ρA (A.8)
W = cε (A.9)

E
c=
ρ (A.10)
Substituting Eqs. (A.6)–(A.10) into Eq. (A.5) yields the expression given in Eq. (4). Note when pretension is not considered this yields Eq. (2).
The impact velocity in terms of ε , ULag and W is given as (Ref. [27], Eq. (24))
2
V= (1 + ε )2ULag −[(1 + ε ) ULag−W ]2 (A.11)
Substituting Eq. (A.3) and Eq. (A.9) into Eq. (A.11) yields the expression given in Eq. (3). When pretension is not considered, this yields Eq. (1).

References ballistic impacts. Int J Impact Eng 2006;32(11):1737–51.

[2] Yu JH, Dehmer PG, Yen CF. High-speed photogrammetric analysis on the ballistic
behavior of Kevlar fabrics impacted by various projectiles. Army research labora-
[1] Tan VBC, Ching TW. Computational simulation of fabric armour subjected to tory report ARL-TR-5333; 2010.

185
G. Guo
[3] Seretis GV, Kostazos PK, Manolakos DE, Provatidis CG. On the mechanical response
of woven para-aramid protection fabrics. Compos Part B Eng 2015;79:67–73.
[4] Cheeseman BA, Bogetti TA. Ballistic impact into fabric and compliant composite
laminates. Compos Struct 2003;61(1):161–73.
[5] Wang Y, Chen X, Young R, Kinloch I. Finite element analysis of effect of inter-yarn
friction on ballistic impact response of woven fabrics. Compos Struct
2016;135:8–16.
[6] Chu T-L, Ha-Minh C, Imad A. A numerical investigation of the influence of yarn
mechanical and physical properties on the ballistic impact behavior of a Kevlar KM2
woven fabric. Compos Part B Eng 2016;95:144–54.
[7] Tabiei A, Nilakantan G. Ballistic impact of dry woven fabric composites: a review.
Appl Mech Rev 2008;61(1):010801–10813.
[8] Tan V, Lim C, Cheong C. Perforation of high-strength fabric by projectiles of dif-
ferent geometry. Int J Impact Eng 2003;28(2):207–22.
[9] Nilakantan G, Nutt S. Effects of clamping design on the ballistic impact response of
soft body armor. Compos Struct 2014;108:137–50.
[10] Zhou Y, Chen X, Wells G. Influence of yarn gripping on the ballistic performance of
woven fabrics from ultra-high molecular weight polyethylene fibre. Compos Part B
Eng 2014;62:198–204.
[11] Shin HS, Erlich DC, Shockey DA. Test for measuring cut resistance of yarns. J Mater
Sci 2003;38:3603–10.
[12] Tapie E, Guo YB, Shim VPW. Yarn mobility in woven fabrics-a computational and
experimental study. Int J Solids Struct 2016;80:212–26.
[13] Song B, Park H, Lu W, Chen W. Transverse impact response of linear elastic ballistic
fiber yarn. J Appl Mech 2011;78:0510231–510239.
[14] Garcia-Castillo SK, Sanchez-Saez S, Lopez-Puente J, Barbero E, Navarro C. Impact
behaviour of preloaded glass/polyester woven plates. Compos Sci Technol
2009;69(6):711–7.
[15] Schueler D, Pentecote NT, Voggenreiter H. Effects of static preloads on the high
velocity impact response of composite plates. Compos Struct 2016;153:549–56.
[16] Shim VPW, Guo YB, Tan XY. Effect of pre-tension on ballistic resistance of woven
fabric. Appl Mech Mater 2014;566:325–31.
[17] Nilakantan G, Wetzel ED, Merrill R, Bogetti TA, Adkinson R, Keefe M, et al.
186
Composite Structures 200 (2018) 173–186
Experimental and numerical testing of the V50 impact response of flexible fabrics:
addressing the effects of fabric boundary slippage. In: 11th International LS-DYNA
users conference, Dearborn, MI, USA June 6–8; 2010.
[18] Phoenix SL, Porwal PK. A new membrane model for the ballistic impact response
and V50 performance of multi-ply fibrous systems. Int J Solid Struct
2003;40:6723–65.
[19] Erol O, Powers B, Keefe M. Development of a non-orthogonal macroscale material
model for advanced woven fabrics based on mesoscale structure. Compos Part B
Eng 2017;110:497–510.
[20] Shockey DA, Erlich DC, Simons JW. Improved barriers to turbine engine fragments:
interim report III. DOT/FAA/AR-99/8, III; 2001.
[21] Chocron S, Figueroa E, King N, Kirchdoerfer T, Nicholls AE, Sagebiel E, et al.
Modeling and validation of full fabric targets under ballistic impact. Compos Sci
Technol 2010;70(13):2012–22.
[22] Das S, Jagan S, Shaw A, Pal A. Determination of inter-yarn friction and its effect on
ballistic response of para-aramid woven fabric under low velocity impact. Compos
Struct 2015;120:129–40.
[23] ABAQUS Documentation, Version 6. 14, Dassault Systems Simulia Corp.; 2014.
[24] Bandaru AK, Sachan Y, Ahmad S, Alagirusamy R, Bhatnagar N. On the mechanical
response of 2D plain woven and 3D angle-interlock fabrics. Compos Part B Eng
2017;118:135–48.
[25] Cheng M, Chen W, Weerasooriya T. Mechanical properties of Kevlar KM2 single
fiber. J Eng Mater Technol 2005;127(2):197–203.
[26] Mckee PJ, Sokolow AC, Yu JH, Long LL, Wetzel ED. Finite element simulation of
ballistic impact on single jersey knit fabric. Compos Struct 2017;162:98–107.
[27] Smith JC, McCrackin FL, Schiefer HF. Stress-strain relationships in yarns subjected
to rapid impact loading. Part V: Wave propagation long textile yarns impacted
transversely. Textile Res J 1958;28(4):288–302.
[28] Walker JD, Chocron S. Why impacted yarns break at lower speed than classical
theory predicts. J Appl Mech 2011;78(5):051021.
[29] Kedzierski P, Poplawski A, Gieleta R, Slawinski G. Experimental and numerical
investigation of fabric impact behavior. Compos Part B Eng 2015;69:452–9.