JOURNAL OF

COMPOSITE
Article M AT E R I A L S

Journal of Composite Materials

2021, Vol. 55(9) 1231–1246
Measurement of in-plane and ! The Author(s) 2020
Article reuse guidelines:
out-of-plane elastic properties of sagepub.com/journals-permissions
DOI: 10.1177/0021998320967073
woven fabric composites using journals.sagepub.com/home/jcm

digital image correlation

opez-Santos1,2,3, A May-Pat2, ER Ledesma-Orozco1,

F L
erez4 and F Avil
A Hernández-P es2

Abstract
The digital image correlation (DIC) technique is employed to measure the strain fields and the five independent elastic
material constants of a glass/vinyl ester plain weave laminated composite. Transverse (through-thickness) tensile and
Iosipescu shear properties were measured using thick (
19 mm) laminates, and the strain fields were compared against
those predicted by a homogenized linear elastic finite element analysis. DIC reveals that the strain fields of the laminates
may present heterogeneous zones, which are attributed to the heterogeneous nature of the woven fabric architecture.
Homogenization of the DIC strain fields to produce two-dimensional stress-strain plots yields average strains and elastic
properties of the woven composite which can be similar to those measured by conventional (localized) strain gages.
Important differences were found for the transverse tensile and Iosipescu shear tests, where differences of 26.4%, 28.6%
and 11.9% were obtained for the measured transverse elastic modulus (Ez), in-plane shear modulus (Gxy) and transverse
shear modulus (Gxz), respectively.

Keywords
Digital image correlation, plain weave fabric, elastic properties, Iosipescu shear, out-of-plane, strain

Introduction
properties is an important but complex task, because
Polymeric composites are widely employed in the air- the undulation of the yarns, type of woven fabric, het-
craft, automotive, civil, energy and many other indus- erogeneity of the architecture, and size of the represen-
tries. They are still the focus of scientific investigations tative unit cell (RUC) may influence the effective elastic
because their outstanding properties, such as low properties of the composite.5–10 The in-plane elastic
weight, high specific stiffness and strength, toughness, properties of PWCs have been extensively investigat-
durability, chemical resistance, and workability. ed,11–13 and there exist standard test methods to
Unidirectional fiber reinforced composites have excel-
lent mechanical properties only along one direction,
which motivates the use of woven fabrics with more 1
Departamento de Ingenierıa Mecánica, Universidad de Guanajuato,
balanced properties in structural applications.1,2 The Mexico
2
added benefit of more balanced mechanical properties Centro de Investigacion Cientıfica de Yucatán AC, Unidad de Materiales,
when using woven fabrics comes with the price of Mexico
3
Tecnol
ogico Nacional de Mexico/I.T. Superior de Coatzacoalcos, Mexico
architectural complexity.1–3 A plain weave composite 4
Departamento de Ciencias de la Tierra, Tecnol ogico Nacional de
(PWC) consists of yarns interlaced orthogonally (at Mexico/I.T. Merida, Mexico
90 ) having only one undulation along the thickness
as a result of the yarn interlacing; this architecture Corresponding author:
F Aviles, Centro de Investigaci
on Cientıfica de Yucatán AC, Unidad de
allows balanced in-plane properties, as well as higher Materiales, Calle 43 # 130 x 32 y 34, Col. Chuburná de Hidalgo, CP
in-plane impact and transverse resistance due to the 97205, Merida, Yucatán, Mexico.
yarn interlacing.4 Measurement of the PWC elastic Email: [email protected]
1232 Journal of Composite Materials 55(9)

measure in-plane tension (ASTM D303914) and shear Iosipescu shear tests, to determine out-of-plane (trans-
properties (ASTM D537915). Despite their importance, verse) elastic properties. The DIC-measured strain
however, the out-of-plane properties of PWCs have fields are compared against homogenized finite element
been significantly less studied, likely because of the modeling. Special interest is placed on the strain gra-
complexity of their three-dimensional architecture.16 dients due to the woven fabric geometry, comparing
Several analytical (laminated theory)4,11,12,17–20 and the average strains and elastic constants measured by
numerical21–30 research works have been devoted to DIC to those measured by conventional strain gages.
study the elastic response of PWC laminates.
Experimental research on PWCs seems to be more lim-
ited, and published research has investigated the effect
Material and experimental methods
of the type of woven fabric, thickness of the laminate, Materials
undulation, gaps and continuity of the yarns, the geom-
etry of the weave, the stacking sequence, the relative PWCs were fabricated using a Derakane Momentum
position between plies and off-axial loading.4,29,31–37 470–300 vinyl ester resin reinforced with several layers
Simulation of textile composites such as plain weave of 0/90 plain weaves made of E-glass yarns. Plain
and twill or satin weaves is challenging, because the weaves were acquired from “Poliformas Plásticas S.A
simplified assumptions of common micromechanical de C.V.” (Merida, Mexico), whereas the vinyl ester
methods remain no longer valid.20 In order to take resin (Derakane Momentum 470–300) was acquired
advantage of both analytical and numerical methods, from Ashland Inc. (Convington, Kentucky, USA).
some semi-analytical approaches have been proposed The vinyl ester resin has a density of 1.08 g/mL,
to improve the efficiency over purely numerical meth- 33 wt.% of styrene monomer, and is typically employed
ods.38–42 For example, a two-step homogenization pro- for manufacturing laminated composites using infusion
cedure to predict the elastic properties of woven processes due to its low viscosity and room tempera-
polymer matrix composites using the generalized ture curing.
method of cells has showed good agreement with
experimental results.38 An issue related to mechanical Manufacturing of laminated composites
testing of PWCs is the accuracy of the (local) strain
measurements due to the application of stresses given Laminated composite plates were manufactured by
the possibility of strain gradients. Strain gradients in vacuum-assisted resin transfer molding. Three sets of
PWCs are caused by the inherent woven fabric archi- flat plates with different thickness were manufactured,
tecture, whose RUC is typically of the order of a few viz.
8 mm thick (20 layers),
14 mm thick (32 layers)
millimeters, see e.g. Adumitroaie and Barbero.13 and
19 mm thick (44 layers), all with volume fractions
Progress on this issue may be achieved by using the between 55% and 57%, measured according to the
digital image correlation (DIC) technique. DIC is a “Test Method I” and “Procedure G” from the ASTM
relatively new optical method that allows direct non- standard D3171.46 To obtain thick laminates (14 and
contact measurements of strain fields, which can be 19 mm thick) with uniform resin infiltration, the infu-
conducted at length scales significantly larger than sev- sion process needed a few modifications. To this aim,
eral RUCs of a PWC. For example, Lomov et al.5 used four 6.3 mm diameter resin inlet tubes were used, which
DIC on two PWCs and a braided composite to validate were end-connected to a stack of two 19 mm diameter
their finite element model (FEM), identifying damage spiral-wrap tubes for resin distribution and wetting. To
initiation. Brynk et al.43 employed DIC to measure the delay the curing time and promote uniform resin infil-
longitudinal and transverse strains of unidirectional, tration into the weaves, a concentration of 0.4 wt.% of
plain weave and twill weave carbon reinforced compos- both, cobalt naphthenate (promoter) and methyl ethyl
ite plates loaded in tension, in order to determine their ketone peroxide (Norox, MEKP-925, initiator) were
elastic properties. Qin et al.44 used DIC to measure the used. Such a selected concentration was obtained
in-plane and out-of-plane shear properties of a PWC, after conducting a program of several gelation tests
whereas Pollock et al.45 measured the in-plane elastic at different concentrations of promoter and initiator.
properties of a PWC under off-axis tension. Given this Once the plain weaves were infiltrated, the composite
background, this work uses DIC as a central tool to material was cured for 24 h at room temperature and
investigate the strain fields and the five independent post-cured for 4 h at 100  C. Two flat plates of
elastic constants of an E-glass/vinyl ester PWC lami- 270 mm  150 mm  8 mm were used to manufacture
nate. Uniaxial and V-notched (Iosipescu) shear tests seven tensile specimens (replicates) for in-plane
are conducted to determine in-plane properties, where- elastic modulus (Ex), in-plane Poisson’s ratio ( xy)
as dedicated thick PWC laminates are manufactured to and transverse Poisson’s ratio ( xz) measurements.
conduct through-thickness tensile tests and transverse Flat plates of 270 mm  150 mm  14 mm,
Lopez-Santos et al. 1233

270 mm  150 mm  19 mm and 90 mm  90 mm  according to the ASTM standard D3039 procedure,14

19 mm were used to manufacture seven replicates for as illustrated in Figure 1. An inset with a photograph of
in-plane shear modulus (Gxy), transverse shear modu- the RUC is included in Figure 1(a). The RUC has an
lus (Gxz) and transverse elastic modulus (Ez) tests, average size of 4.97 mm  4.95 mm along the warp and
respectively. Because the information on the 0/90 weft (fill) directions, respectively.
plain weave architecture suggests that is balanced in To prevent premature failure, tabs were bonded to
the 0 and 90 directions, only five independent elastic the specimens ends by using Hysol EA9309 aerospace
constants were considered for the PWC (Ex  Ey,  xz  grade adhesive (Henkel, CA, USA). End tabs were
 yz and Gxz  Gyz). This consideration, however, will be made of a 25 mm  25 mm glass/polyester composite
further examined in the forthcoming sections. with 1.5 mm thickness, see Figure 1(a). One surface of
the specimen in the xy plane (front, see Figure 1(b))
was prepared with the speckle pattern needed to con-
Preparation of specimens and tests duct DIC measurements, while strain gages were
In-plane tensile testing. In order to determine Ex,  xy and bonded to the back xy surface and to a lateral surface
 xz, flat rectangular specimens of 250 mm length, (thickness) in the xz plane. An extensometer (not
25 mm width and 8 mm thickness were employed employed herein) might be used as an alternative that

(b) Speckle paern

(a)

y In-plane
z strain gage

z
8 Transverse
1.5 strain gage
x
x
y

25 x

250
Dimensions in mm

(c)

Specimen
DIC cameras

Strain
indicator

Figure 1. In-plane tensile test. (a) Dimensions of test specimens and photograph of a glass fiber weave showing the unit cell, (b)
tensile specimen showing the DIC speckle pattern and strain gages, (c) full test setup.
1234 Journal of Composite Materials 55(9)

covers more unit cells, but its accuracy depends on Transverse tensile testing. In order to determine Ez, an I-
its gripping action, and would not provide strain shaped profile was chosen for the tensile specimen. The
field measurements. For the DIC speckle, a white back- selection of this geometry for through-thickness tensile
ground was painted and small black dots were random- testing was motivated by previous attempts docu-
ly sprayed on top of the white background, to achieve mented in the literature for this complex test.48,49 The
high contrast and fine facets. The strain gages used specimen was obtained by first machining a 19 mm
were EA-13-062AQ-350/LE from Vishay Micro- thick plate into 18 mm cubes, and then machining
Measurements (Raleigh, NC, USA), with overall those 18 mm cubes to the desired “I” profile, as
dimensions of 6.60 mm  3.80 mm, active (sensing) shown in Figure 2(a) and (b). Although the ASTM
gage length of 2.90 mm and electrical resistance of standard D729150 recommends the use of a specimen
350 X. Two mutually perpendicular strain gages were with flat ends bonded to a metallic test fixture, a mod-
attached to the back xy surface of the specimen to ification was employed herein to provide mechanical
measure Ex and  xy, whereas one strain gage was assembly between the specimen flange and its specially
adhered to the lateral xz plane to measure  xz, as designed fixture, avoiding the use of adhesive bonding.
shown in Figure 1(b). The DIC measurements were For this test, tensile loading is introduced to the speci-
conducted using a GOM ARAMIS 5 M (Optical men by loading both flanges of the specimen, which are
Measuring Techniques, Braunschweig, Germany)47 firmly fit to rabbets machined on the steel test rig
with a shutter time of 6 ms and sampling speed of 1 blocks, as indicated in the photograph of Figure 2(c).
image per second. Further details about the DIC tech- As shown in Figure 2(b), the frontal xz surface of
nique are included in ‘Digital image correlation setup’ the specimen was prepared with the DIC speckle pat-
section. Tension tests were conducted using a tern, while a strain gage (EA-13-062AQ-350/LE) was
Shimadzu Autograph AG-I (Kyoto, Japan) universal bonded to the back xz surface (the opposite face to the
testing machine (see Figure 1(c)), with a load cell of DIC-prepared surface), orientated along the loading
100 kN, cross-head speed of 1 mm/min and recording direction. The same GOM ARAMIS 5 M equipment
1 data per second. The full stress-strain response up to was employed with a shutter time of 90 ms, and sam-
failure of the PWC was firstly determined by conduct- pling speed of 1 image per second. The tests were con-
ing two tensile tests until specimen failure. Because this ducted in a Shimadzu Autograph AG-I universal
work focuses on elastic properties, and in order to testing machine with a load cell of 5 kN, cross-head
acquire more data, subsequent testing of the in-plane speed of 1 mm/min and recording 1 data per second.
tensile specimens included two initial loading- Because the laminate is highly sensitive to load changes
unloading cycles well within the linear elastic regime, along the ply direction, and has a relatively low
followed by a third loading excursion up to failure, through-thickness strength, all seven specimens of the
obtaining 3 elastic property data per specimen. through-thickness testes were tested until failure at
Loading-unloading cycles were applied to each of the their first load excursion. Therefore, average and stan-
seven replicates tested within the linear elastic zone dard deviation values reported for Ez were obtained
(axial strain ex  0.06%, axial stress rx  15 MPa), from seven data. The out-of-plane strain (ez) used to
and finally they were loaded once more until specimen generate the reported transverse stress (rz) vs. strain
failure. It was visually verified the lack of any damage or (ez) plots was calculated as an average along a 12 mm
tendency to diminish the elastic modulus (or strength) vertical long line centered between the specimen
during the first two loading-unloading cycles. A similar flanges. Ez was determined from the slope (least
methodology was applied for the other tests (discussed squares linear regression) of the rz vs. ez curves
in the forthcoming sections) with the exception of the within 0.0025%  ez  0.03% region.
transverse tensile test (for Ez determination). Thus, the
average and standard deviation values of each elastic In-plane shear and transverse shear testing. The V-notched
property were calculated from 21 data (7 samples and (Iosipescu) shear test method was employed to measure
3 data per sample). Ex,  xy and  xz were determined Gxy and Gxz according to the ASTM D5379 standard.15
from a least squares linear fit from the stress-strain The specimen consists of a rectangular plate with
curve. This fitting was conducted within 0  ex  0.06% V-shaped notches at the mid-section, as indicated in
for the two initial cycles, and within 0.05%  ex  0.20% Figure 3. Each notch forms a 90 angle with a radius
for the final loading excursion up to failure. Because the of 1.3 mm, and is oriented at 45 with respect to the
laminate configuration indicated by the ASTM standard longitudinal axis, having a 4 mm depth (
20% of the
is thin for this test, it was not possible to measure the specimen width). All Iosipescu shear specimens were
through-thickness strain by DIC in the tensile in-plane 76 mm long, 19 mm wide and 4 mm thick, as indicated
test. Therefore  xz was solely measured by using strain in Figure 3(a). Specimens for Gxz were machined from
gages. the 19 mm thick plates, see Figure 3(a), while specimens
Lopez-Santos et al. 1235

(a) (b)

z
Surface with
strain gage
x Speckle paern

P
Dimensions in mm
(c)

Grip

Specimen Test fixture

Grip

Figure 2. Through-thickness tensile test. (a) Specimen dimensions, (b) speckle pattern used for DIC (front) and strain gages (back)
xz surfaces, (c) specimen mounted on the test rig (front surface).

DIC face
(b)
(a)

z z z

D=19 e=11 x y x

r= 1.3 t=4
L=76

Dimensions in mm Strain gage face

(c)
Load
Introducon
Fixture

Specimen

Figure 3. Transverse Iosipescu shear test (xz plane). (a) Specimen dimensions, (b) speckle pattern used for DIC (front) and strain
gages (back) xz surfaces, (c) specimen mounted on the test fixture (showing front surface).
1236 Journal of Composite Materials 55(9)

for Gxy (not shown) were machined from the 14 mm Finite element analysis
thick plates.
Model for transverse tension. A homogenized linear elastic
For Gxz and Gxy measurements, the speckle pattern
FEM of the through-thickness specimen depicted in
was sprayed on the front surface while strain gages
Figure 2 was conducted using the ANSYSVR 52 software.
were bonded to the back surface (Figure 3(b)), in a
Modeling was conducted in order to assess the suitabil-
similar manner to the tensile tests. Strain gage rosettes ity of the proposed I-shaped specimen for the trans-
at  45 (FCA-2-11-3L) with an active gage length of verse tensile test, the homogeneity of the strain fields,
2 mm from Tokyo measuring Instruments Laboratory and as an approximation to compare against DIC
Co (Shinagawa, Tokyo, Japan) were employed. The measurements. Smaller scale FEMs which include the
GOM ARAMIS 5 M equipment was employed with a detailed architecture of the reinforcement may certainly
shutter time of 72 ms and sampling speed of 1 image improve the FEM predictions,53,54 but this is out of the
per second. The tests were conducted in a Shimadzu scope of this work.
Autograph AG-I universal testing machine with a The model is depicted in Figure 4, and was con-
load cell of 5 kN, cross-head speed of 1 mm/min and structed using 20-node brick elements with quadratic
recording 1 data per second. A V-notched (Iosipescu) interpolation (“SOLID 186”) with 3 degrees of freedom
test fixture standard model WTF-IO from Wyoming at each node (translations in the x, y and z directions,
Test Fixtures Inc (Salt Lake City, UT, USA) was represented by the standard elasticity variables u, v and
used to load the specimen, see Figure 3(c). The average w, respectively). The nodes located at the bottom sur-
shear stress (s) was calculated by dividing the applied face (z ¼ C/2 ¼ 9 mm) were fixed (i.e. u ¼ v ¼ w ¼ 0),
compressive force between the cross-section area locat- as indicated in Figure 4. A nodal displacement of
ed between the V-shaped notches. The shear strain (c) w ¼ wApp ¼ 5.90  106 m was applied to all nodes
was simultaneously measured by the strain gage belonging to the top surface (z ¼ C/2). This applied dis-
rosettes and DIC. The average DIC shear strain used placement corresponds to an applied load (P) of 736 N
to produce s vs. c plots was computed along a 11 mm and was used in order to match an specific DIC load
long vertical line centered between the notches, by step. The elastic properties used in this model were
using the conventional mathematical (integral) defini- Ex ¼ Ey ¼ 28.6 GPa, Ez ¼ 12.7 GPa,  xy ¼ 0.35,  xz ¼
tion of the average of a spatially varying function.51  yz ¼ 0.36, Gxy ¼ 4.33 GPa, Gxz ¼ Gyz ¼ 3.77 GPa,
For the strain gages, c was calculated as the absolute which correspond to the average DIC measurements,
value of the sum of the strains measured by both strain as it will be further discussed. After a strict convergence
gages at  45 . Gxz and Gxy were determined from the analysis, the final (convergent) mesh used approximate-
slope (calculated by a least squares linear regression) of ly 21, 000 elements, with a nominal (average) element
the s vs. c curves within the 0.2%  c  0.6% region size of 0.9 mm  0.6 mm  0.19 mm at the gage length.
(cxy or cxz, depending on the case).

Digital image correlation setup

Full field strains were measured by DIC using a GOM z
ARAMIS 5 M instrument (Optical Measuring y
Techniques, Braunschweig, Germany). Measurements
utilized two cameras of 2448  2050 pixels resolution,
lenses with a focal length of 35 mm and a camera angle
of 25 . Images were acquired at a speed of 1 image per
w(x,y,C/2)= wApp
second, using a 15  15 pixels facet size, facet step of 13
pixels and an overlap area of 2 pixels. Measurements of
Ex and  xy used a vision field (measurement area) of
150 mm  130 mm (length  width), distance between C
cameras of 272 mm, and average measuring distance
(distance between the sensor and specimen) of x
700 mm; a CP20/MV 175  140 calibration panel
from GOM was used, as recommend by the
ARAMIS user’s manual.47 Measurements of Gxy, Gxz
and Ez used a vision field of 35 mm  29 mm with u,v,w (x,y,-C/2)=0
directly abutting cameras, measuring distance of
215 mm, and a CQ/CP20 30  24 GOM calibration Figure 4. Finite element model of the through-thickness tensile
panel. specimen. C ¼ 18 mm, wApp ¼ 5.9 mm.
Lopez-Santos et al. 1237

In-plane and transverse iosipescu shear models. Two- with a specific load step of the DIC results. The input
dimensional (plane stress) linear elastic FEMs were elastic properties were the same ones used in the trans-
constructed in ANSYS.52 A two dimensional analysis verse tensile model described in the previous section.
is deemed sufficient to simulate the shear strain distri-
butions of the in-plane and transverse Iosipescu shear Results and discussion
specimens, as has been previously stated by other
authors.55–57 Mesh refinements were gradually con- In-plane elastic modulus and Poisson’s ratio
ducted at the vicinity of the V-shaped notches, as indi-
cated in Figure 5. For efficiency, a hybrid mesh was Figure 6 shows the axial stress (rx Þ vs. strain ðex Þ curve
employed by using linear interpolation 4-node quadri- of a representative in-plane tensile specimen tested until
failure. The figure contains three insets of the DIC full
lateral elements (“PLANE 182”) for the zones far away
field strain corresponding to average strains
of the (central) notched region ( L/18  x   L/2)
ex ¼ 0.24%, 0.58% and 0.88%, by using an inspection
and 6-node quadratic interpolation triangular elements
zone of 55 mm  16 mm centered at the gage section.
(“PLANE 183”) at the vicinity of the notches (L/
Dashed lines forming a grid of 4.97 mm  4.95 mm are
18  x  þL/18); both element types have two degrees
included within those insets to facilitate the interpreta-
of freedom per node (translations along x and y direc-
tion of results. Each 4.97 mm  4.95 mm rectangle of
tions). After performing a convergence analysis to
such a grid corresponds to the average RUC size of
ensure a size-independent mesh, the convergent mesh
the PWC, see Figure 1(a). All three DIC strain fields
used comprised
23,000 elements of three element sizes show local variations (gradients) of ex, due to the yarn
with gradual variations of size, as shown in Figure 5. architecture. These gradients seem larger at the bound-
The smallest (quadratic interpolation) elements are ary of the yarns which are perpendicular to the load
located in the zone between notches, having an approx- direction. Because the mean RUC length of the PWC is
imate size of 0.1 mm for each side. In Figure 5 the
4.96 mm and the active length of the strain gage used
dimensions are L ¼ 76 mm and D ¼ 19 mm, and the is only 2.90 mm, the strain gage is limited to represent
origin of the coordinate system is centered between only localized values of strain, and is not able to cap-
both notches. The nodes located at the zones (L/ ture gradients of ex within the RUC size in zones larger
6  x  L/2, z ¼ D/2) and (L/12  x  L/2, than the active strain gage size. These strain gradients
z ¼ D/2) were constrained in the z direction (w ¼ 0), yield variability in the full strain fields, which are not
whereas corner nodes located at (x ¼ L/2, z¼D/2) possible to detect with the use of an individual strain
were constrained in the x direction (u ¼ 0). A displace- gage. Nevertheless, the average Ex calculated from the
ment w ¼ wApp ¼ 0.119 mm (corresponding to a force DIC strain fields along a 19.5 mm long vertical line
of 993 N, s ¼ 18 MPa) was applied to nodes located at centered at the specimen gage section yields similar
the zones (L/12  x  L/2, z ¼ D/2) and (L/6  x  L/2, values than those measured by the strain gage. As
z ¼ D/2), as indicated in Figure 5. The applied dis- reported in Table 1 (averages and one standard devia-
placement (w ¼ wApp ¼ 0.119 mm) simulates a specific tion), for this test, differences in Ex measured by DIC
force (993 N) which will be subsequently compared and by a strain gage are
1.05%. Similar trends and

Figure 5. Finite element model of the through thickness (xz) Iosipescu shear specimen. L ¼ 76 mm, D ¼ 19 mm, wApp ¼ –0.119 mm.
Analogous conditions were used for the in-plane (xy) Iosipescu model.
1238 Journal of Composite Materials 55(9)

Figure 6. Axial stress-strain curves for the in-plane tensile specimen.

Table 1. Average and one standard deviation measurements for

Ex,  xy, and  xz by DIC and strain gages. Out-of-plane elastic modulus
Property DIC Strain gage The out-of-plane stress (rz Þ - strain ðez ) response of a
representative I-shaped tensile specimen is shown in
Ex (GPa) 28.6  3.0 28.3  1.9 Figure 8. A photograph of a fractured specimen in
 xy 0.135  0.018 0.139  0.006 the test rig is shown in the inset, with the DIC strain
 xz – 0.368  0.024
field of ez superimposed in the lower part of the
specimen.
behaviors occurred for the in-plane Poisson’s ratio The stress-strain response is fairly linear until fail-
( xy), as seen in Figure 7 and Table 1. ure, given that Ez is an interlaminar, matrix dominated,
Figure 7 displays the negative of the transverse property.59 Failure occurred by a crack perpendicular
strain (–ey) as function of the axial strain (ex) of a rep- to the loading direction in the central zone of the speci-
resentative in-plane tensile specimen tested until failure, men (see inset), reaching a strength of 4.3 MPa
to calculate the Poisson’s ratio. The insets shown in (
539 N), corresponding to a strain of
0.033%. It
Figure 7 correspond to DIC strain field snapshots at should be noted that this strength is significantly
an average ex ¼ 0.10%. Both DIC strain fields (ex and lower than the one measured along the x direction
ey) present spatial gradients which are more abrupt for (see Figure 6), as is usual with the interlaminar/trans-
ex, probably because its elastic response is dominated verse properties of in-plane weaved composites without
by the yarn architecture. The regions that undergo z-reinforcement. A few scattered points in the DIC
larger strains are those with higher content of resin curves were present in this transverse tensile test,
between the yarns; since lateral deformations due to which is due to the small values of ez achieved.
Poisson’s effect are dominated by the matrix,58 ey Nevertheless, the trend of the DIC curve is clear and
presents a more uniform strain distribution. As seen an average Ez can be obtained from the slope of the
in Table 1, similar to Ex, the difference between the rz  ez response. The average (and one standard devi-
average  xy measured by DIC and by strain gages is ation) elastic modulus measured by DIC is Ez ¼ 12.7
only 2.9%. Notice in Table 1 that  xz could not be ( 0.94) GPa, while the one measured by using strain
measured by DIC, because the ASTM standard gages is 9.33 ( 2.59) GPa. Notice that the elastic mod-
D303914 requires a thin laminate for this test. ulus measured by the strain gages displays considerable
Lopez-Santos et al. 1239

Figure 7. Axial (ex) versus transverse (ey) strain for the in-plane tensile specimen.

Figure 8. Through-thickness stress (rz) vs. strain (ez) of an I-shaped tensile specimen loaded in the z-direction.

scattering (coefficient of variation of 28%). It is A difference of 26.4% is found for Ez measured by

believed that this stems from the heterogeneity of the the DIC and the strain gage. The reason for this dif-
material in the through-thickness direction, and the ference relies on the strain field/data that is considered
limited size of the strain gage. The influence of a by each instrument for rendering the average values
small bending moment produced by adjustment of used in the stress-strain curves. The strain gage covers
the test fixture can also not be fully discarded. only a length of 2.90 mm (
4 layers), while DIC
1240 Journal of Composite Materials 55(9)

average values are obtained from a 6 mm  12 mm b ¼ 6 mm (Figure 9(d)). FEM (blue circles) indicates
mask, averaged over a 12 mm straight line (
20 uniform (nearly constant) transverse strain distribution
layers). Although at the mesoscale the PWC may be along both, vertical (Figure 9(c)) and horizontal
considered homogeneous, at the microscale its complex (Figure 9(d)) central regions of the specimen. The
weaved architecture renders strain field variations, in DIC measurements agree fairly well with FEM, but
length scales of the order of millimeters. While the the DIC measurements present some oscillations
average strain measured by DIC averages those strain around the FEM predictions, as well as some small
gradients, the strain gages capture only localized negative values near the flanges (z/a   0.5 in Figure
values. Another limitation of strain gages is their lim- 9(c)). These small compressive strains at the flanges of
ited capacity to stay perfectly bonded to the PWC sur- the specimen (see also Figure 9(a)) may be an indica-
face, which may be of particular importance for tion that, for a few specimens, the flanges may be under
through-thickness tests which are prone to suffer flexural loading. Although these negative values of
delamination.60 strain were not observed in all specimens tested, this
To further investigate this specimen, Figure 9 shows specific contour plot showing the potential occurrence
the DIC strain field of the transverse tension specimen of flexural loading at the flanges was selected here for
(Figure 9(a)) and the corresponding FEM predictions presentation as an indication that future improvements
(Figure 9(b)). Both, DIC and FEM, indicate that ez is on the specimen are possible. Other geometries pro-
uniform within the specimen gage zone (web), support- posed for through-thickness tensile tests of laminated
ing the adequacy of the proposed specimen for out-of- composites are tapered cylinders60 and circular-waisted
blocks.60,61
plane tensile testing.
To further investigate the distribution of the strain
field in the specimen gage zone (web), ez was mapped In-plane and out-of-plane shear moduli
onto centered vertical (Figure 9(c)) and horizontal Figure 10 shows the in-plane shear stress (sxy ) vs. strain
(Figure 9(d)) lines, which are indicated in the inset as (cxy ) response of a representative V-notched (Iosipescu)
dashed lines of length a ¼ 12 mm (Figure 9(c)) and shear specimen. A small linear zone followed by a

Figure 9. ez strain field for the through-thickness tension specimen. (a) DIC strain field, (b) FEM strain field, (c) distribution of ez over
a vertical center line (a ¼ 12 mm) on the specimen web, (d) distribution of ez over a horizontal center line (b ¼ 6 mm) on the specimen
web.
Lopez-Santos et al. 1241

nonlinear response is observed, attaining large shear along the notches, Figure 11(c) and (d) map cxy along
strains, which indicates large influence of the matrix a 11 mm long vertical (labeled “e” in Figure 11(c)) and
in the shear response. 8 mm long horizontal (labeled “f” in Figure 11(d)) cen-
Two insets of DIC strain fields at selected values of tral lines. According to Figure 11(c), FEM predicts a
cxy are included in Figure 10. Dashed lines on the speci- uniform distribution of cxy around cxy  0.25%, with
men shown at cxy ¼ 2.9% (sxy ¼ 53.4 MPa) correspond strain singularities (peak values) only near the notch
to the RUC size (4.97 mm  4.95 mm), used as a refer- roots. The DIC measurements agree very well with
ence for analysis. The inset at the bottom of Figure 10 the FEM predictions, slightly oscillating around
corresponds to a specimen subjected to overly large cxy ¼ 0.25%, and varying between 0.19%  cxy 
strains (cxy ¼ 13%, sxy ¼ 73.2 MPa), for a failed PWC. 0.27%. These small oscillations may be caused by the
This value is not plotted in Figure 10 but the deformed heterogeneous architecture of the PWC. This is because
shape of the specimen and corresponding strain field the mean cxy determined by DIC include regions of
are shown only to provide an idea of the shear behavior only yarns (with and without undulation), only
at large strains. Notice that the ASTM D5379 standard matrix, and regions with a combination of both mate-
procedure,15 suggests to stop the test at cxy ¼ 5%. Both rials. On the other hand, the FEM carried out here is
insets show that DIC rightfully captures the zone of an homogenized model which does not include micro-
pure shear between the notches, which is depicted by structural details of the fiber and yarn architectures,
an elliptical (blue) zone where the maximum shear which makes its length-scale unsuitable to predict
strains occur, comprising approximately two RUCs such small oscillations and gradients in cxy . In Figure
between the notches. The ellipse has major and minor 11(d), both FEM and DIC indicate symmetric cxy dis-
axes of
11 mm and 5 mm, which are bigger than the tributions with maximum at the center, although FEM
2 mm active length of the strain gage. A more detailed slightly overpredicts the DIC measurements.
analysis of the distribution of cxy between the notches Analogous plots for the out-of-plane (transverse, xz)
of the Iosipescu shear specimen can be obtained from Iosipescu specimen are shown in Figures 12 and 13.
examination of Figure 11. Figure 11(a) and (b) are DIC The DIC inset figures of the specimen in Figure 12
and FEM in-plane shear strain distributions for an correspond to cxz ¼ 2.93% (sxz ¼ 45.6 MPa) and cxz ¼
average cxy ¼ 0.23% (sxy ¼ 9.84 MPa), showing good 13% (sxz ¼ 38.4 MPa, failed specimen). Interlaminar
agreement between the DIC and FEM. To extract shear failure (delamination) is identified as the failure
more information about the shear strain distribution mechanism for this specimen. According to the inset at

Figure 10. Shear stress (sxy) versus shear strain (cxy) plot of an Iosipescu shear in-plane specimen.
1242 Journal of Composite Materials 55(9)

Figure 11. In-plane shear strain (cxy) fields of an Iosipescu specimen. (a) Shear strain field measured by DIC, (b) shear strain field
predicted by FEM, (c) distribution of cxy on a vertical line (e ¼ 11 mm) centered between the notches, (d) distribution of cxy on a
horizontal line (f ¼ 8 mm) centered between the notches.

Figure 12. Shear stress (sxz) versus shear strain (cxz) plot for an Iosipescu shear transverse (xz) specimen.
Lopez-Santos et al. 1243

Figure 13. Transverse shear strain (cxz) fields of an Iosipescu specimen. (a) Shear strain field measured by DIC, (b) shear strain field
predicted by FEM, (c) distribution of cxz on a vertical line (e ¼ 11 mm) centered between the notches, (d) distribution of cxz on a
horizontal line (f ¼ 8 mm) centered between the notches.

cxz ¼ 13%, a sliding shear pattern at
45 is formed for Table 2. Average and one standard deviation values of Gxy and
the fibers between the notches. Gxz measured by DIC and strain gages.
As for cxy, at the macroscopic level, the cxz strain Property DIC Strain gages
field is quite uniform in the zone between the notches.
At the mesoscale (
RUC size), however, there are gra- Gxy (GPa) 4.70  1.16 3.24  0.74
dients in the shear strain field, as indicated in Figure 13. Gxz (GPa) 3.77  0.33 3.32  0.19
Figure 13(a) and (b) are DIC and FEM transverse
shear strain distributions for an average cxz ¼ 0.38% architecture is not fully isotropic, symmetric and bal-
(sxz ¼ 19.9 MPa). The distribution of cxz is mapped
anced in such a plane. This discrepancy with the iso-
along vertical (e) and horizontal (f) lines centered
tropic relationship for laminated composites using
between the notches. Given the increased influence of
plain weaves that are expected to be symmetric and
the polymeric matrix in the xz plane and the important
balanced has been previously observed by other
presence of yarn undulations in this plane, the DIC
measurements present higher strain gradients in this authors.25,28
plane, with values that vary between The shear moduli measured by the strain gages are
0.28%  cxz  0.43%. lower than those determined by DIC. This is deemed to
By comparing Figures 10 and 12 is evident that the be due to the small active length of the strain gage
in-plane shear modulus (slope) and in-plane strength (
2 mm) which does not cover a whole RUC of the
are higher than those corresponding to the transverse weave (
4.97 mm  4.95 mm). Using carbon/carbon
(xz) shear. The average (and one standard deviation) woven composites, Qin et al.44 have reported that
values of shear moduli are listed in Table 2. measurements of shear strain strongly depend on the
A difference of 31.1% between the average measure- strain gage size and its position in the woven compos-
ments of Gxy using DIC and strain gages is obtained, ite. This means that measurements by small strain
whereas for Gxz the difference is 11.9%. It should be gages is localized and its consideration as an average
noted that Gxy does not fulfill the isotropic relationship value of shear strain in the material might yield some
for the xy plane, pointing out that the PWC inaccuracies. Larger strain gages are commercially
1244 Journal of Composite Materials 55(9)

available, but strain gages with active lengths larger (PRODEP) for his PhD studies. Technical assistance of
than 5 mm are not common, and may be expensive. Miguel Angel Rivero Ayala (CICY) is appreciated.
Extensometers also do not map strain fields, and may
suffer from slippery at high loads. Thus, DIC arises as Declaration of Conflicting Interests
an ideal tool for this task. Furthermore, DIC measure- The author(s) declared no potential conflicts of interest with
ments may be used as an experimental tool to obtain respect to the research, authorship, and/or publication of this
calibration factors for average strains, especially in sit- article.
uations of materials with complex or heterogeneous
architecture, see e.g. Qin et al.44 Funding
The author(s) disclosed receipt of the following financial sup-
Conclusions port for the research, authorship, and/or publication of this
article: This research was supported by the U.S. Office of
The digital image correlation (DIC) technique was used
Naval Research Global (ONRG) under award number
to measure strain fields and the five independent elastic
N62909-19-1-2119.
constants of a glass/vinyl ester plain weave laminated
composite (PWC). The woven textile used consisted of
wrap and weft E-glass yarns interlaced orthogonally (0/ ORCID iD
90 ) with a representative unit cell of approximately F Aviles https://orcid.org/0000-0002-0227-1758
4.97 mm  4.95 mm. Because of its architecture, the
PWC exhibited strain field gradients and DIC proved References
to render high resolution to capture those strain field 1. Choi J and Tamma KK. Woven fabric composites – part
gradients, in fair agreement with homogenized finite I: predictions of homogenized elastic properties and
element analysis. DIC measurements not only provided micromechanical damage analysis. Int J Numer Meth
full field strains under different loading scenarios, but Engng 2001; 50: 2285–2298.
also improved the determination of the elastic con- 2. Bannister MK. Development and application of
stants of the PWC. This is because DIC is able to advanced textile composites. Proc Imeche 2004; 218:
yield stress-strain curves whose average strain includes 253–260.
3. Lawrence CA. High performance textiles for geotechnical
several representative unit cells, and thus better cap-
engineering: geotextiles and related materials. In:
tures the heterogeneity of the material. The DIC meas-
Lawrence CA (ed.) High performance textiles and their
urements showed that the PWC laminate exhibits applications. Holland: Woodhead Publishing, 2014,
gradients of strain fields at the mesoscale (i.e. at the pp.256–350.
scale of the representative unit cell), attributed to 4. Naik NK and Shembekar PS. Elastic behavior of woven
the architecture of the plain weave. Despite of this, fabric composites: I – lamina analysis. J Compos Mater
the averaged DIC strain fields yielded similar (differ- 1992; 26: 2196–2225.
ences lower than 3%) elastic constants than those cal- 5. Lomov SV, Ivanov DS, Verpoest I, et al. Full-field strain
culated by strain gages (
2–3 mm active gage length) measurements for validation of meso-FE analysis of tex-
for Ex and  xy. For Ez, Gxy, and Gxz, differences tile composites. Compos, Part A 2008; 39: 1218–1231.
between both techniques were found as 26.4%, 6. Lomov SV, Boisse P, Deluycker E, et al. Full-field strain
measurements in textile deformability studies. Compos
26.8%, and 11.9%, respectively. This is due to the
Part A 2008; 39: 1232–1244.
scale where the plain weave architecture can be consid-
7. Nicoletto G, Anzelotti G and Riva E. Mesoscopic strain
ered homogeneous or repetitive, and the presence of fields in woven composites: experiments vs. finite element
yarn undulations through the laminate thickness. The modeling. Opt Lasers Eng 2009; 47: 352–359.
plain weave architecture yields strain gradients at the 8. Tekieli M, De Santis S, De Felice G, et al. Application of
mesoscale, which affect the average values of the strain digital image correlation to composite reinforcements
and consequently those of the elastic constants, espe- testing. Compos Struct 2017; 160: 670–688.
cially when the spatial extension of these gradients is 9. Jiang Y, Tabiei A and Simitses GJ. A novel
comparable to the RUC size. Therefore, DIC arises as micromechanics-based approach to the derivation of con-
an ideal optical tool for capturing strain gradients in stitutive equations for local/global analysis of a plain-
woven composites, and their result could guide the weave fabric composite. Compos Sci Technol 2000; 60:
1825–1833.
determination of calibration factors for measurements
10. Bejan L, Axinte A and Taranu N. Influence of the geo-
of elastic constants using conventional strain gages. metric parameters on the elastic properties of textile poly-
meric composites. Mat Plast 2016; 53: 264–268.
Acknowledgements 11. Ishikawa T and Chou TW. Stiffness and strength behav-
Francisco L
opez acknowledges the economic support of iour of woven fabric composites. J Mater Sci 1982; 17:
“Programa para el Desarrollo Profesional Docente” 3211–3220.
Lopez-Santos et al. 1245

12. Naik NK and Ganesh VK. Prediction of on-axes elastic fabric composites under off-axial loading. Int J Mech
properties of plain weave fabric composites. Compos Sci Sci 2018; 141: 157–167.
Technol 1992; 45: 135–152. 30. Hayat K, Lei X and Ali HT. Prediction of elastic behav-
13. Adumitroaie A and Barbero EJ. Stiffness and strength ior of woven fabric reinforced plastics composites using
prediction for plain weave textile reinforced composites. two-step homogenization. In: 2018 15th International
Mech Adv Mater Struct 2012; 19: 169–183. Bhurban conference on applied sciences and technology
14. ASTM D3039. Standard test method for tensile properties (IBCAST), Islamabad, Pakistan, 9–13 January 2018,
of polymer matrix composite materials. West pp.36–42. New York: IEEE.
Conshohocken, PA: ASTM International, 2006. 31. Dow NF, Ramnath V and Rosen BW. Analysis of woven
15. ASTM D5379. Standard test method for shear fabrics for reinforced composite materials. Report NASA-
properties of composite materials by the V-notched beam CR-178275. National Aeronautics and Space
method. West Conshohocken, PA: ASTM International, Administration, Hampton VA, August 1987.
2005. 32. Ishikawa T, Matsushima M, Hayashi Y, et al.
16. Mespoulet S, Hodgkinson JM, Matthews FL, et al. Experimental confirmation of the theory of elastic
Design, development, and implementation of test meth- moduli of fabric composites. J Compos Mater 1985; 19:
ods for determination of through thickness properties of 443–458.
laminated composites. Plast Rubber Compos 2000; 29: 33. Ishikawa T, Matsushima M and Hayashi Y. Geometrical
496–502. and material nonlinear properties of two-dimensional
17. Ishikawa T. Anti-symmetric elastic properties of compos- fabric composites. AIAA Journal 1987; 25: 107–113.
ite plates of satin weave cloth. Fibre Sci Technol 1981; 15: 34. Shembekar PS and Naik NK. Elastic behavior of woven
127–145. fabric composites: II – laminate analysis. J Compos
18. Ishikawa T and Chou TW. One-dimensional microme- Mater 1992; 26: 2226–2246.
chanical analysis of woven fabric composites. AIAA J 35. Naik NK and Ganesh VK. Thermo-mechanical behav-
1983; 21: 1714–1721. iour of plain weave fabric composites: experimental
19. Ito M and Chou TW. Elastic moduli and stress field of investigations. J Mater Sci 1997; 32: 267–277.
plain-weave composites under tensile loading. Compos 36. Cai D, Zhou G, Wang X, et al. Experimental investiga-
Sci Technol 1997; 57: 787–800. tion on mechanical properties of unidirectional and
20. Hallal A, Younes R and Fardoun F. Review and com- woven fabric glass/epoxy composites under off-axis ten-
parative study of analytical modeling for the elastic prop- sile loading. Polym Test 2017; 58: 142–152.
erties of textile composites. Compos B Eng 2013; 50: 37. Khatkar V, Vijayalakshmi AGS, Manjunath RN, et al.
22–31. Experimental investigation into the mechanical behavior
21. Naik NK. Numerical modelling of woven fabric compos- of textile composites with various fiber reinforcement
ite materials. In: Bull JW (ed.) Numerical analysis and architectures. Mech Compos Mater 2020; 56: 367–378.
modelling of composite materials. UK: Springer, 1996, 38. Bednarcyk BA and Arnold SM. Micromechanics-based
pp.400–438. modeling of woven polymer matrix composites. AIAA
22. Roze AV and Zhigun IG. Three-dimensional reinforced Journal 2003; 41: 1788–1796.
fabric materials. Polym Mech 1973; 6: 272–278. 39. Prodromou AG, Lomov SV and Verpoest I. The method
23. Johnson WS, Masters JE, O’Brien TK, et al. Modeling of cells and the mechanical properties of textile compo-
damage in a plain weave fabric-reinforced composite sites. Compos Struct 2011; 93: 1290–1299.
material. J Compos Technol Res 1993; 15: 136–142. 40. Kirane K, Salviato M and Bazant ZP. Microplane triad
24. Scida D, Aboura Z, Benzeggagh ML, et al. Prediction of model for simple and accurate prediction of orthotropic
the elastic behaviour of hybrid and non-hybrid woven elastic constants of woven fabric composites. J Compos
composites. Compos Sci Technol 1998; 57: 1727–1740. Mater 2016; 50: 1247–1260.
25. Scida D, Aboura Z and Benzeggagh ML. A microme- 41. Liu X, Rouf K, Peng B, et al. Two-step homogenization
chanics model for 3D elasticity and failure of woven- of textile composites using mechanics of structure
fibre composite materials. Compos Sci Technol 1999; 59: genome. Compos Struct 2017; 171: 252–262.
505–517. 42. Rouf K, Liu X and Yu W. Multiscale structural analysis
26. Aitharaju VR and Averill RC. Three-dimensional prop- of textile composites using mechanics of structure
erties of woven-fabric composites. Compos Sci Technol genome. Int J Solids Struct 2017; 136: 89–102.
1999; 59: 1901–1911. 43. Brynk T, Molak RM, Janiszewska M, et al. Digital image
27. Kollegal MG and Sridharan S. A simplified model for correlation measurements as a tool of composites defor-
plain woven fabrics. J Compos Mater 2000; 34: mation description. Comput Mater Sci 2012; 64: 157–161.
1756–1786. 44. Qin L, Zhang Z, Li X, et al. Full-field analysis of shear
28. Donadon MV, Falzon BG, Iannucci L, et al. A 3-D test on 3D orthogonal woven C/C composites. Compos,
micromechanical model for predicting the elastic behav- Part A 2012; 43: 310–316.
iour of woven laminates. Compos Sci Technol 2007; 67: 45. Pollock P, Yu L, Sutton MA, et al. Full-field measure-
2467–2477. ments for determining orthotropic elastic parameters of
29. Wang L, Zhao B, Wu J, et al. Experimental and numer- woven glass-epoxy composites using off-axis tensile speci-
ical investigation on mechanical behaviors of woven mens. Expe Tech 2014; 38: 61–71.
1246 Journal of Composite Materials 55(9)

46. D3171 ASTM. Standard test methods for constituent con- 55. Broughton WR, Kumosa M and Hull D. Analysis of the
tent of composite materials. West Conshohocken, PA: Iosipescu shear test as applied to unidirectional carbon-
ASTM International, 2006. fibre reinforced composites. Compos Sci Technol 1990;
47. GOM Optical Measuring Techniques. ARAMIS user’s 38: 299–325.
information and hardware manual: sensor configuration 56. Grediac M, Pierron F and Vautrin A. The Iosipescu in-
examples. Braunschweig, Germany: GOM GmbH, 2013. plane shear test applied to composites: a new approach
48. Broughton WR and Sims GD. An overview of through- based on displacement field processing. Compos Sci
thickness test methods for polymer matrix composites. Technol 1994; 51: 409–417.
NPL report DMM(A) 148. Middlesex, UK: National 57. Chiang MY and He J. An analytical assessment of using
Physical Laboratory, 1994. the losipescu shear test for hybrid composites. Compos B
49. Ferguson RF, Hinton MJ and Hiley MJ. Determining Eng 2002; 33: 461–470.
the through-thickness properties of FRP materials. 58. Lee JR, Molimard J, Vautrin A, et al. Digital phase-
Compos Sci Technol 1998; 58: 1411–1420. shifting grating shearography for experimental analysis
50. ASTM D7291Through-thickness “flatwise” tensile of fabric composites under tension. Compos, Part A
strength and elastic modulus of a fiber-reinforced polymer 2004; 35: 849–859.
matrix composite material. West Conshohocken, PA: 59. Avva VS, Allen HG and Shivakumar KN. Through-the-
ASTM International, 2015. thickness tension strength of 3-D braided composites.
51. Ho H, Tsai MY, Morton J, et al. Numerical analysis of J Compos Mater 1996; 30: 51–68.
the Iosipescu specimen for composite materials. Compos 60. Karkkainen RL, Moy P and Tzeng JT. Through-thickness
Sci Technol 1993; 46: 115–128. property measurement of three-dimensional textile compo-
52. ANSYS 12.0. ANSYS user’s manual: theory reference. sites. Report ARL-TR-4765. US Army Research
Canonsburg, PA: Swanson Analysis Systems, Inc., 2009. Laboratory, Aberdeen Proving Ground, MD, April 2009.
53. Barbero EJ, Trovillion J, Mayugo JA, et al. Finite ele- 61. Abot JL and Daniel IM. Through-thickness mechanical
ment modeling of plain weave fabrics from photomicro- characterization of woven fabric composites. J Compos
graph measurements. Compos Struct 2006; 73: 41–52. Mater 2004; 38: 543–553.
54. Li S, Zhou C, Yu H, et al. Formulation of a unit cell of a
reduced size for plain weave textile composites. Comput
Mater Sci 2011; 50: 1770–1780.