Composites Science and Technology 178 (2019) 11–25

Contents lists available at ScienceDirect

Composites Science and Technology

journal homepage: www.elsevier.com/locate/compscitech

On the optimal choice of fibre orientation angle in off-axis tensile test using T
oblique end-tabs: Theoretical and experimental studies
J.C. Marín, J. Justo, A. Barroso∗, J. Cañas, F. París
E.T.S. Ingeniería, Universidad de Sevilla, Camino de los Descubrimientos s/n, 41092, Sevilla, Spain

ARTICLE INFO ABSTRACT

Keywords: Shear strength evaluation in composite materials still remains a not fully solved problem. One of most promising
B. Shear strength proposals to characterise a material in shear by means of a simple tensile test is the off-axis tension test with
D. Off-axis tensile test oblique end-tabs whose inclination coincides with that of the longitudinal isodisplacement lines. This end-tab
C. Stress concentrations configuration presents an essential difficulty since the tab angle depends on the material properties including the
shear modulus G12, which is one of the values to be obtained from off-axis tension tests.
In the present work, the study of the most suitable fibre orientation for the performance of off-axis tests under
an oblique tabs configuration has been addressed. An analytical study on the dependence of the tab angle ϕ on
both the fibre orientation θ and the material properties E11, E22, ν12 and G12 has been carried out, the quotient
G12/E11 being identified as the parameter that controls the evolution of the tab angle.
The influence of deviations in the tab angle on the stress state near the corners of specimen has been analysed
by means of a novel study of the nominal singular stress state at these points. The results of this study show that
negative deviations in the angle of the tabs contribute to decrease the severity of the stress state at the corner,
whereas positive deviations contribute to increase the order of the stress singularity. Additionally, the range of
fibre orientations in which the evaluation of shear strength S is feasible has been determined, i.e., the effect of
σ12 in the failure is clearly dominant versus the effect of the rest of the stress-state components.
In the experimental part of this work, four different fibre orientation angles (5°, 10°, 15° and 20°) have been
considered to perform the off-axis tension test under oblique end-tabs configuration. The experimental results of
these tests show that only for 10° fibre orientation angle the majority of specimens fail at the central zone under
a uniform stress state. Finally, for a fibre orientation angle θ = 10°, specimens with induced positive deviations
in tab angle of +3° and +7° have been tested, it being observed that all the failures appears at the end of
specimens.
On the basis of both theoretic and experimental results, the optimal fibre orientation angle to use in the tests
is θ = 10°. The tab angle must be evaluated on the basis of an estimation of the material properties (quotient
G12/E11), it being recommendable, in case of ignoring the value of G12, to use a tab angle lesser than the
theoretical value to avoid positive deviations.

1. Introduction must be as simple as a tensile test, and mainly for this reason, the most
successful approach for shear characterization has been the ± 45° ten-
Shear characterization of unidirectional layers for composite ma- sion test [23]. Although is notorious [26] that the strength value ob-
terials continues being a not fully solved problem. In the last fifty years tained from this test is not the intralaminar shear strength S, because
many different approaches have been proposed to solve this problem the ± 45° layers fail due to a dominant normal transversal stress com-
(rail shear test [1–7], thin-walled tube torsion test [8,9], off-axis tensile ponent σ22.
test [10–22], ± 45° tensile test [23–26], Arcan test [27–30], Iosipescu One of most common tests for composite materials is the off-axis
test [31–33], asymmetric four-point bend test [34,35], plate twist test tension test. This test is frequently used in shear characterization of
[36,37], v-notched rail shear test [38–40], off-axis flexural test [41–43] unidirectional layer specimens and in the study of the off-axis beha-
and shear frame test [44]), although anyone of these approaches has viour of different fibre orientation angles under both static and fatigue
achieved an unquestionable support. In general, a characterization test loads. However, it is well known that the difficulties in the

∗
Corresponding author.
E-mail address: abc@us.es (A. Barroso).

https://doi.org/10.1016/j.compscitech.2019.04.022
Received 10 April 2019; Accepted 16 April 2019
Available online 19 April 2019
0266-3538/ © 2019 Elsevier Ltd. All rights reserved.
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 1. (a) Nominal off-axis tension test configuration. (b) Oblique end-tab specimen.

experimental performance of off-axis tension tests are associated with • The associated oblique angle of the tabs should present as little
the coupling effect (Pagano & Halpin [10]). Two types of approaches variation as possible with material properties.
have been proposed to solve these difficulties: the application of cor- • The fibre orientation should be adequate for shear modulus eva-
rection factors to the direct measurement of the test (Pindera & Her- luation G . 12
akovich [17], Marín et al. [21]), and the modification of the test fixture • The fibre orientation should be adequate for shear strength eva-
to attain a situation that best approximates a pure tension stress field. luation S.
For the latter, several different approaches have been proposed (Pin-
dera & Herakovich [17], Cron et al. [18], Sun & Berreth [19], Sun & To consider the first condition, an analytical study has been per-
Chung [20], Xiao et al. [22]). Among these studies, the approach pro- formed regarding the variation in the angle of the oblique end-tab ϕ
posed by Sun & Chung [20] has achieved a certain remarkable re- with the fibre orientation θ and material properties E11, E22, ν12 and
levance. This approach involves the use of oblique end-tabs whose G12. The second condition is not as restrictive as the others, because
angle agrees with isodisplacement lines for an ideal test configuration. accurate G12 values can be obtained by different fibre orientation angles
Pierron & Vautrin [45] have shown, by numerical study (finite element from 10° to 45°. Former studies have proved that the optimal orienta-
method (FEM)) of a 10-degree off-axis configuration, that specimens tion to produce maximum values of γ12 is around 10° (Chamis & Sinclair
with oblique end-tabs present a more uniform stress field than speci- [15]). To consider the last condition, a novel study of the singularities
mens with regular end-tabs. Additionally, the stress concentration that of the nominal stress state at the corner, to evaluate the changes in
appears at the corners of the specimen can be reduced by using an stress state at the ends of a specimen due to deviations in the tab angles,
oblique end-tab configuration. This fact, which was experimentally has been carried out. Additionally, the adequate orientation interval to
verified by Kawai et al. [46] by means of strain gages and by Pierron evaluate shear strength S has been determined. Several failure criteria
et al. [47] by means of a grid technique, constitutes the great advantage have been employed for this purpose.
that this configuration has in connection with strength characterization. Then, to support the theoretical results, a campaign of off-axis
Nevertheless, as pointed out by Pierron & Vautrin [45], the oblique end- tensile tests under oblique tab configuration has been carried out.
tab configuration does not produce a pure shear stress field; hence, a Different fibre orientations have been considered (5°, 10°, 15° and 20°)
failure criterion is needed to evaluate in-plane shear strength S. The in these tests. The corresponding tab angles for these orientations have
same authors, in a later work [48], show that coherent values of S can been estimated in the basis of the manufacturer's values of material
be obtained by applying the Tsai-Wu criterion to the results of both a properties. The experimental results and the failures observed for the
10-degree off-axis tension test and a 0-degree Iosipescu test for the same different orientations have been considered separately.
material.
However, the oblique end-tab configuration presents an essential 2. Oblique end-tab setup for off-axis tension tests
difficulty. The angle of the tabs ϕ depends on the material properties
including the shear modulus G12, which is one of the values that is The nominal off-axis tension test configuration consists of a rec-
determined from off-axis tension tests. Pierron et al. [47], by means of a tangular unidirectional layer specimen whose fibres are oriented at an
sensitivity analysis, show that the most influential parameters for the angle θ with respect to the longitudinal direction x, as shown in
angle of the tabs are E11 and G12. Additionally, based on numerical Fig. 1(a). This unidirectional layer specimen is ideally subjected to a
results (FEM) for a 10-degree fibre orientation angle, they concluded uniform distribution of longitudinal stress σx = σ. Under these
that a difference of 3° in the oblique angle resulted in non-significant boundary conditions, the lines of longitudinal isodisplacement are
stress concentrations near the tabs; however, a difference of 7° can lead straight lines that lie at an angle ϕ with respect to the x-axis; this angle
to significant stress concentrations near the tabs. can be derived from the analytical solution of the problem, and it can be
In the present work, the most adequate fibre orientation angle to expressed as [20]:
perform an off-axis tensile test under an oblique end-tabs configuration
S16
is theoretically and experimentally studied; this orientation angle cot = ,
S11 (1)
should meet the following three conditions:
where S11 and S16 are the compliance coefficients. These compliance

12
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

coefficients can be evaluated as follows: which ϕ = 32.4242° over the entire range of μ considered, i.e., for this
orientation, the tab angle becomes nonsensitive to the variation in μ.
S11 =
cos4
E11
+ ( 2 12
E11
+
1
G12 ) sin 2 cos2 +
sin4
E22 Likewise, the variation in the tab angle ϕ with the Poisson's ratio ν12 in
the range of properties considered (0.28, 0.34) can be evaluated; this
S16 = ( 2
E11
+
2 12
E11
1
G12 ) sin cos3 + ( 2 12
E11
2
E22
+
1
G12 ) sin 3 cos
shows that the Poisson's ratio has a negligible effect on the tab angle ϕ,
(2) as seen in Fig. 3.
The variation in quotient E11/E22 produces a variation in ϕ, which
Sun & Chung [20] propose the use of oblique end-tabs with an angle
affects the fibre orientation angle θϕ defined previously. Table 1 shows
ϕ (Fig. 1(b)), which produces a constant displacement along the lines of
the values of these orientations θϕ for the central values and the ex-
longitudinal isodisplacement at the ends of the specimen.
treme values of quotient E11/E22 in the range of properties considered.
Equations (1) and (2) indicate that the angle of the end-tab ϕ de-
The tab angles ϕ that correspond to these orientations are also pre-
pends on both material properties (E11, E22, G12 and ν12) and fibre or-
sented in Table 1.
ientation angle θ. The fibre orientation angle is known before the spe-
As seen from the values in Table 1, fibre orientations θϕ in the range
cimen manufacturing, but the material properties are generally
of 24.1–28.8° led to minimal variation in the tab angle ϕ. To estimate
unknown. Therefore, the practical performance of a test requires a
the error of ϕ that we could commit by choosing a specific orientation θ,
previous assumption regarding the material property values. Obviously,
we have used an orientation θ = 26°, which is approximately in the
the validity of the G12 and S values obtained from this test depends on
centre of the optimal interval. Fig. 4 shows the evolution of the tab
the validity of the assumption realized regarding the material property
angle ϕ that corresponds to an orientation θ = 26° for the extreme and
values. The feasibility of the testing procedure would thus be con-
central values of the quotient E11/E22. As can be observed, the greatest
strained by the sensitivity of the tab angle ϕ to variations in material
angular deviations occur at the highest μ values. Thus, for E11/
properties.
E22 = 10.9, a maximum negative angular deviation of −3.80° is ob-
tained, and for E11/E22 = 25, a maximum positive angular deviation of
3. Study of the tab angle ϕ as a function of material properties +2.75° is obtained. These would be the maximum angular errors in the
case of choosing ϕ = 32.5°, which is the value associated with θ = 26°
Given the great diversity of properties (E11, E22, G12 and ν12) in for the central value of the quotient E11/E22 = 18.
composite materials, it seems unlikely that a unique tab angle could be After evaluating the possible maximum angular deviations of ϕ that
found to be adequate for all materials. However, it is reasonable to find could appear for a fibre orientation θ = 26°, it would be necessary to
a unique orientation of tabs that is applicable to one material type, analyse if these deviations could be considered acceptable, i.e., the
generating acceptable errors in the evaluation of G12 and S. The prop- stress concentration generated by the angular deviation would not
erties for graphite-epoxy composites, according to literature references cause specimen failure. Pierron et al. [47], based on numerical studies,
(Tsai [49], Mil-HDBK-17 [50], Schwartz [51], Herakovich [52]), are in estimated that this error was ± 3° for an orientation θ = 10°; however,
the following range: the permissible error for different orientations is still unknown. For this
120 GPa ≤ E11 ≤ 200 GPa; 8 GPa ≤ E22 ≤ 11 GPa; 3.5 GPa ≤ G12 purpose, it is necessary to analyse how the stress concentration at the
≤ 8.5 GPa; 0.28 ≤ ν12 ≤ 0.34. This range is used as a basis to develop corners of the specimen varies when there is a deviation in the angle of
the present analytical study. tabs. A FEM 2D model of the problem has been made using the program
An expression for the reciprocal of the tangent of angle ϕ (cot ϕ) can ANSYS [53]; the model consists of 40000 SHELL63-type elements and
be derived using equations (1) and (2); multiplying both the numerator 40501 nodes. The mesh, as seen in Fig. 5, has been conveniently refined
and denominator of this expression by E11 leads to the following in the area close to the specimen ends to better analyse the effect of the
equation: stress concentration in this area. To take into consideration how the
cot stress concentration varies, a point (highlighted with a red circle in
Fig. 5(B)) sufficiently close to the corner (0.2 mm) has been chosen as a

=
(2 + 2 12
E11
G12 ) sin cos + ( 2 3
12
2E11
E22
+
E11
G12 ) sin 3 cos preliminary indicator for comparison. The discretisation has been ver-
ified by comparing the results for a mesh with an element size that has
+( 2 + ) sin E11 E11
cos4 12
2 cos2 + sin4
G12 E22 been reduced by half; the obtained results were equivalent at the con-
(3) sidered point.
Let us define a variable μ as the quotient G12/E11 to simplify the For graphite-epoxy composites, the following average values for
notation. From equation (3), we can derive an explicit formulation for each property (i.e., the average value of the entire range of considered
the tab angle ϕ: values) were used for the FEM: E11 = 160 GPa; E22 = 9.5 GPa;
G12 = 6 GPa; and ν12 = 0.31. Different orientations of the fibres have

= arctan
1
tan2 (2 12
1
µ )+ E11
E22
tan2 been considered, including θ = 5°, 10°, 15°, 20° and 26°. Fig. 6(a) and
(b) and 6(c) show how the concentrations of stress components (σ11,
(2 + 2 12
1
µ ) 1
tan
+ ( 2E11
E22
+2 12
1
µ ) tan (4)
σ22, σ12) vary with the deviation in the tab angle Δϕ at the selected
point. Only positive deviations have been represented since they were
As can be observed in equation (4), the parameters that ϕ depends the most unfavourable results. These concentrations have been eval-
on are the quotient μ = G12/E11, the fibre orientation angle θ, the uated as the quotient between the value of a particular stress compo-
Poisson's ratio ν12, and the quotient E11/E22. In Fig. 2, the variation in nent at the specified point and the value of the same stress component
tab angle ϕ as a function of μ is represented for three fibre orientation in the ideal configuration (identified by id in the figures) for the same
angles (10°, 30°, 45°). For this graphical representation, the average load level. For all the orientation angles considered, we observed that
values were used for both parameters ν12 (0.31) and quotient E11/E22 the stress concentrations increase with the angular deviation Δϕ. For a
(18) in the range of material properties considered. particular deviation, the stress concentrations generally increase with
As can be observed in Fig. 2, the evolution of the graphs is linear for the orientation angle of the fibres.
fibre orientation angles less than 30°, and it is nonlinear for angles To take into account the contribution of these stress components to
greater than 30°. In view of the evolution of ϕ for fibre orientation the failure of the specimen, it is necessary to use a failure criterion. In
angles less than 30° (which has an almost linear evolution), it can be this study, we have used the theories of Tsai-Wu [54] and Puck [55],
deduced that an intermediate angle near 30° having a constant evolu- whose equations ((8) and (6), respectively) will be detailed in section 5.
tion of ϕ should exist. This intermediate angle is θϕ = 25.8962°, in In each case, for an orientation θ and a deviation Δϕ, the stress state has

13
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 2. Variation in tab angle ϕ with μ for different fibre orientation angles θ (10°, 30°, 45°).

Fig. 3. Variation in ϕ versus μ for θ = 25.8962° and for the extreme values of ν12 (0.28, 0.31).

Table 1 criterion generally increases with the orientation of the fibres θ.

Fibre orientation angles θϕ for which the evolution of ϕ is constant with μ. If we assume that the maximum acceptable level of the criterion is
E11/E22 θϕ (°) ϕ (°)
reached for an orientation θ = 10° and a deviation Δϕ = +3° (as de-
clared admissible by Pierron et al. [47]), this value of the criterion
10.9 28.8214 38.2797 corresponds to approximately 1.34, as seen in Fig. 7. This assumption
18 25.8962 32.4242 would imply that the allowable deviations would be less than +3° at
25 24.0948 29.2059
orientations greater than 10°. Thus, for the cases where θ = 15°, 20°,
and 26°, the failure criterion value would be 1.34 for deviations Δϕ
been forced to take the values that would indicate failure in the ideal around +1.7°, +1.1° and +0.8°, respectively. The allowable deviations
configuration (i.e., the failure criterion equations are set equal to would increase only in the case of fibre orientations less than 10°. Thus,
unity). Then, the value of each failure criterion equation at the selected for θ = 5°, the acceptable level of the criterion would be reached with a
point (near the corner) has been represented in the following figures. deviation Δϕ of approximately +9°. Therefore, for fibre orientations
The variation in the value of the Tsai-Wu and Puck failure criteria greater than 10°, the amount of potential angular deviation decreases,
with the deviation in tab angle Δϕ is shown in Fig. 7, the results ob- but the maximum acceptable deviations drastically reduce. In contrast,
tained with both theories being very similar. As with the stress com- for fibre orientations of 10° and lower, the potential angular deviation
ponents, it is observed that the value of the criteria increases with the may be greater, but the maximum acceptable deviation is also greater.
angular deviation of tabs, and for each deviation Δϕ, the value of the Table 2 shows the potential deviations and the estimated acceptable

14
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 4. Variation of ϕ versus μ for θ = 26° and for different values of E11/E22 (10.9, 18, 25).

deviations for the different orientations analysed. It should be men- 4. Study of the singular stress state at the corners of the specimen
tioned that negative acceptable deviations have been recorded along-
side the positive deviations, but in this particular case, we can say that Experimental evidence shows that premature undesirable failures
these values would represent a minimum of the acceptable deviations. near the tabs of the specimen typically occur in configurations with
As previously mentioned, negative deviations with respect to the angle oblique tabs, as seen in Fig. 8.
of the tabs are less unfavourable than the positive deviations, and To analyse the role of the singular stress states induced by the
therefore, they lead to lower stress concentrations. presence of a corner at the end-tabs of the specimen, a specific study of
It should be noted that the consequences obtained are based on the the stress singularities has been carried out.
numerical results of the model studied, and, as previously mentioned, Considering a planar model of the specimen, the orders of stress
the stress state at the corner is singular. A singular stress state analysis singularities, δ in (r , ) K r k f (k ) ( ) (with a polar coordinate
k k
is carried out in the next section to give more representativeness to the system centred at the corner tip, and where K and f ( ) are the gen-
previously obtained results. eralized stress intensity factor and the characteristic angular function,

Fig. 5. View of the mesh: (A) general view; (B) detailed view of the corner.

15
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 6. Variation in the stress concentration of: (a) σ11, (b) σ22 and (c) σ12 with the angular deviation of the tabs Δϕ.

respectively [56]), have been evaluated for different combinations of Table 2

the fibre and tab angles, θ and ϕ. The calculations have been carried out Potential deviations in tab angle ϕ and acceptable deviations Δϕ.
using the semi-analytical procedure developed by Barroso et al. [56]. θ (°) Theoretical ϕ (°) Potential deviation in ϕ Acceptable deviation Δϕ
The procedure is fully analytical with the exception of a final numerical (°) (°)
search for the roots (the orders of stress singularities) of the char-
5 29.6 ± 15.4 ±9
acteristic equation; the accuracy of this procedure is extremely good
10 23.1 ± 9.1 ±3
(up to 13 digits), which has been demonstrated in a previous work by 15 24.2 ± 5.9 ± 1.7
Mantic et al. [57]. 20 27.5 ±4 ± 1.1
The tool used for the stress singularities at multimaterial corners 26 32.5 +2.8–3.8 ± 0.8
allows any number of elastic wedges converging at the same corner and

Fig. 7. Variation in the value of the criteria with the angular deviation of the tabs Δϕ: (a) Tsai-Wu C., (b) Puck C.

16
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

straight configuration (ϕ = 90°) to that (15°) giving a negative order of

stress singularity (i.e.: no unbounded stresses).
It can be observed in Fig. 10 that the behaviour for all five fibre
angles is similar in the sense that the higher the tab angle is, the higher
the order of stress singularity. Additionally, it can be observed that the
order of stress singularity associated with the theoretical ϕ (see Table 2)
for each orientation θ increases with the angle of the fibres. Conse-
quently, fibre orientation angles of 5° and 10° would be more appro-
priate in this sense, as they result in smaller orders of stress singularity
Fig. 8. Premature failures in off-axis samples with oblique tabs. at corner A.
Assuming the maximum acceptable level in the order of stress sin-
any type of linear elastic constitutive law, any perfect adhesion or gularity to be the one that is reached at a fibre orientation θ = 10° and
sliding friction contact between material wedges, and any general a tab angle deviation Δϕ = +3°, as has been done previously in the
homogeneous boundary conditions imposed along the external faces (in stress results from the numerical analysis, we can determine, based on a
case of open corners) or closed corners (all material wedges being nominal singular stress state, the acceptable deviations Δϕ for different
connected), to be taken into consideration [56,57]. The tool also allows fibre orientations. Table 3 shows the estimated values of the acceptable
the presence of the so-called mathematically degenerate materials in deviations for the different orientations θ considered.
the framework of the Stroh formalism of anisotropic elasticity in com- As can be observed, the values of the allowable deviations estimated
plex variable, which is the base of the code. With such characteristics, on the basis of the analytical study of the order of stress singularity
the present tool is, to the best authors' knowledge, one of the few (Table 3) are very similar to those obtained on the basis of numerical
available codes to determine the order of stress singularities with these models (Table 2), i.e., both showed the same trend. It should be noted
types of materials and boundary conditions. that the results of the study of stress singularity are more restrictive, as
The procedure is able to evaluate complex-valued orders of stress they estimate lower values of deviations. In the particular case where
singularities, which typically appear in interface cracks in bimaterial the fibre orientation angle θ is 26°, no positive deviation would be al-
configurations. This fact implies that the generalized stress intensity lowable according to the study of stress singularity.
factors are also complex number, which is not the case for the problem Another important consequence that can be observed in Fig. 10 is
under analysis. The program also incorporates the argument principle, that negative Δϕ deviations contribute to a decrease in the order of
which is an excellent tool for identifying the number of roots of a ho- stress singularity at the corner. For each orientation θ, there will be a
lomorphic function in a particular region of the complex plane, which certain negative deviation that eliminates the stress singularity (δ = 0).
allows all orders of stress singularities to be perfectly identified and These points have been highlighted by a red rectangle in Fig. 10. From
evaluated. As a final consideration of the program characteristics, it can the fibre orientations considered, θ = 10° is the one in which the the-
evaluate a singular stress field with multiple terms, all of them being oretical angle of the tabs (ϕ = 23.1°) is closest to the angle that pro-
singular. In the problem under analysis, only one singular term appears, duces δ = 0 (21.8°). For that reason, the 10° orientation could be con-
allowing a straightforward interpretation of the results. sidered optimal in terms of the singularity of the nominal stress state at
In the particular problem under analysis, where the free edge of the the corner (in the neighbourhood of the corner).
sample is used as the horizontal reference direction, the corner con- Another fact to be emphasised, which can be deduced from the
figuration consists of a single orthotropic material wedge that is free results presented in Fig. 10, is that the stress singularity orders that
along the lateral side of the specimen and clamped along the side of the correspond to the straight tab configuration (ϕ = 90°) are greater than
specimen inclined an angle ϕ; the fibre orientation θ is measured those of the oblique tab configuration for all the fibre orientations
counterclockwise from the free edge. considered. Therefore, the configuration of oblique tabs would be more
A detailed view of the left end of an off-axis specimen is shown in favourable than that of straight tabs in terms of the stress severity levels
Fig. 9, where the two corner configurations that generate stress singu- that appear in the corners.
larities are identified as corner A and corner B. Due to the properties in
the homogeneous 2D model associated with the internal structure (or- 5. Study of the optimal fibre orientation θ for the determination of
ientation of the fibres) at both corners, only corner A will be considered shear strength
hereafter. A failure that initiates at corner A via a combination of shear
and circumferential (normal) stresses can progress throughout the As previously mentioned in the introduction section, the stress state
matrix, parallel to the fibre direction, all along the specimen, thereby in the off-axis test is not a pure shear stress field; therefore, these tests
generating a catastrophic failure. However, a failure that initiates at require the application of a failure criterion for the indirect determi-
corner B via a combination of shear and circumferential (normal) nation of shear strength S. Chamis & Sinclair [15], using the Chamis
stresses can only progress by means of fibre failure, which is, of course, criterion, identified that orienting the fibres at 10° is suitable because
less likely to occur than the matrix failure at corner A. the term associated with the shear stress σ12 is dominant in comparison
Fig. 10 shows the orders of stress singularities for five different fibre to those terms associated with the normal stresses. Following this re-
angles (θ = 5°, 10°, 15°, 20°, 26°) and various tab angles from the commendation, the orientation θ = 10° is the most used configuration
in the off-axis test with regular tabs. In the recent work by Vargas &
Mujika [42], we can find similar discussions that propose this particular
fibre orientation above others. In particular, regarding the oblique tab
configuration, the 10° orientation has been one of the most frequently
used [45,47,48,58,59]. Nevertheless, other orientations have also been
used with the oblique tab configuration as for example in the original
proposal by Sun & Chung [20] where a 20° orientation is used, or in the
works by Xiao et al. [22] and Xavier et al. [60] where a 15° orientation
is employed.
To determine which fibre orientation angle is best suited for shear
Fig. 9. Detailed view of the corners at the specimen end, which generate stress strength determination under an oblique tabs configuration, it is ne-
singularities. cessary to analyse when the failure will take place, and under which

17
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 10. Orders of stress singularity δ at corner A for various combinations of fibre angles (θ = 5°, 10°, 15°, 20°, 26°) and tab angles (20° < ϕ < 90°).

Table 3 stress term showed in (5). Using these values of σx, the difference be-
Estimated values of the acceptable. tween the term associated with σ12 and the sum of the rest of the terms
θ (°) Theoretical ϕ (°) Acceptable deviation Δϕ (°)
can be evaluated. When this difference is positive, it means that the
term associated with σ12 is dominant. The optimum orientation corre-
5 29.6 +8 sponds to the angle for which this difference is maximum.
10 23.1 +3 Taking various sources for the mechanical properties of graphite-
15 24.2 +1.6
epoxy composites [49–52], the ratio YT/S typically varies between
20 27.5 +0.75
26 32.5 +0 0.4227 and 0.7282, and the tensile strength in the fibre direction XT
varies between 1200 MPa and 2900 MPa. After making the aforemen-
tioned evaluations and calculations for the distinct combinations of the
stress state it will occur. For the failure determination, it is necessary to mechanical properties of the materials (YT, XT, S) for the previously
select a failure criterion; two of the proposals that have had the best mentioned range of values, the intervals and optimum orientations are
results in the world wide failure exercise [61], Puck's criterion [55] and obtained and summarized in Table 4.
Tsai-Wu's criterion [62], were selected in the present work. With the results of Table 4, the interval of orientations in which it
Regarding the stress state at the instant that failure occurs in the can be assumed that the shear stress term is dominant, i.e., the inter-
oblique tabs configuration, the one appearing in the central area of the section of all intervals, would lie between 4° and 21°. The maximum
specimen, which will be considered to be approximately equal to the values of the computed differences lie between 3° and 8°. Discarding the
one in the ideal off-axis configuration, will be assumed. This hypothesis angle of 3°, which is outside the allowable interval, it can be stated that
is supported by the numerical results [20,45] and experimental results the optimum interval lies between 4° and 8°. Thus, according to Puck's
[46,47] of previous works in the literature. Thus, we can assume that criterion, the previous intervals ((4°, 21°) and (4°, 8°)) would be the
the stress state under which the failure takes place can be approximated allowable and optimum orientation intervals, respectively, for the de-
as follows: termination of shear strength by means of off-axis tension tests.
With reference to Tsai-Wu's criterion [54], it is given by the fol-
11 = x cos2 ; 22 = x sin2 ; 12 = x sin cos , (5)
lowing equation:
where σx is the applied longitudinal stress.
With the stress state given by (5), the activated failure mode, ac- 2 2 2
12 11 22 1 1 1 1
+ + + 11 + 22
cording to Puck's criterion [55], would be the “inter-fibre failure mode S XT XC YT YC XT XC YT YC
A”, whose equation is given by: 1
11 22 = 1.
2 2 2 6 XT XC YT YC (8)
12 (+) YT 22 (+) 22 11
+ 1 P +P =1 ,
S S YT S 1.1XT (6) Following the same procedure previously outlined for Puck's cri-
where P is 0.35 for graphite composites, as recommended by the
(+) terion, and assuming the same variation ranges of the mechanical
author. Through a convenient manipulation, equation (6) is trans- properties (XT, XC, YT, YC, S), the orientation interval for which the
formed into: shear stress term is dominant can be determined. In this case, besides
the variation ranges of XT (1200 MPa, 2900 MPa) and the ratio YT/S
2 2
12
+ 1 2P (+)
YT 22
+ 2P (+) 22 (0.4227, 0.7282) previously considered, it is necessary to take into
S S YT S account the variation in XC/XT (0.58, 0.89) and YC/YT (3.37, 5.31). In
6 6
22 11 11
2P (+) 2+ = 1, Table 4
S 1.1XT 1.1XT (7) Allowable angular intervals and optimum orientations according to Puck's
criterion.
where the term associated with σ12 can be isolated from the rest, which
makes it easier to evaluate its contribution to the criterion apart from YT/S XT = 1200 MPa XT = 1800 MPa XT = 2900 MPa
the contributions of the other stress terms. For that aim, all possible
0.4227 (4°, 21°) 8° (3°, 21°) 6° (2°, 21°) 4°
values of angular orientations have been covered, from 0° to 90°, at
0.6667 (4°, 30°) 8° (3°, 30°) 6° (2°, 30°) 4°
intervals of 1°. Moreover, for each angle, the value of σx that gives rise 0.7282 (3°, 32°) 7° (2°, 32°) 5° (2°, 32°) 3°
to the failure was obtained by means of solving equation (7) for the

18
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Table 5
Allowable angular intervals and optimum orientations according to the Tsai-Wu criterion.
YT/S XT (MPa) XC/XT = 0.58 YC/YT = 3.37 XC/XT = 0.58 YC/YT = 5.31 XC/XT = 0.89 YC/YT = 3.37 XC/XT = 0.89 YC/YT = 5.31

0.4227 1200 (5°, 22°) 10° (5°, 21°) 9° (5°, 20°) 10° (5°, 19°) 9°
2900 (2°, 20°) 5° (2°, 19°) 5° (2°, 19°) 6° (2°, 18°) 5°
0.7282 1200 (4°, 33°) 9° (4°, 32°) 9° (4°, 31°) 10° (4°, 30°) 10°
2900 (2°, 31°) 5° (2°, 30°) 5° (2°, 31°) 6° (2°, 29°) 5°

Table 5, the allowable intervals and optimum orientations obtained for and the failure strain in fibre direction XεT have been obtained. These
different combinations of the mechanical properties are shown. average values with their corresponding standard deviations (STD) are:
The interval of allowable orientations will be the intersection of all
obtained intervals for the different combinations of mechanical prop- E11 = 125.159 GPa (STD: 2.615 GPa);
erties, which in this case is (5°, 18°). In each one of the evaluated cases, XT = 1746.458 MPa (STD: 36.079 MPa),
the optimum orientation inside the corresponding interval was de- X T = 1.394%(STD: 0.029 %)
termined, which showed that all these orientations occurred in the
range of (5°, 10°). As results of the 90° tension tests the average values for both the
With the results of the different failure criteria considered, an in- elastic modulus perpendicular to the fibre direction E22, the tensile
terval of allowable orientations, which is evaluated as the intersection strength perpendicular to the fibre direction YT and the failure strain
of the intervals obtained for the different criteria, can be determined; perpendicular to the fibre direction YεT have been obtained. These
this interval is (5°, 18°). For all orientations inside this interval, it can be average values with their corresponding standard deviations are:
stated that the shear stress term σ12 is dominant according to the two E22 = 8.112 GPa (STD: 0.232 GPa);
considered failure criteria. The range where the optimum orientations
are located could be defined as the combination of the optimum in- YT = 53.666 MPa (STD: 4.875 MPa),
tervals, which would correspond to the interval (4°, 10°), where the Y T = 0.662%(STD: 0.060 %)
maximum of the differences are located for all failure criteria. As done
Four rectangular panels with fibre orientation angles 5°, 10°, 15°
previously for Puck's criterion, the lowest part of the interval (4°) will
and 20° have been manufactured by laying up 4 pre-preg layers, the
be discarded, as it falls outside of the allowable interval; therefore, the
dimensions of these panels are respectively: 215 × 123 mm2,
optimum orientation interval is reduced to (5°, 10°).
207 × 115 mm2, 208 × 97 mm2, and 199 × 88 mm2. From the 5°
panel, 8 specimens with dimensions 202 × 10 mm2 have been ob-
6. Experimental study
tained; 7 specimens from the 10° panel with dimensions 200 × 10 mm2;
6 specimens from the 15° panel with dimensions 205 × 10 mm2; and 6
A graphite-epoxy composite denominated AS4/8552 and manu-
specimens from the 20° panel with dimensions 198 × 10 mm2.
factured by HEXCEL Composites has been considered at the present
On the basis of the average values of material properties (9), and
work. The average values of material properties, obtained from the
using equation (4), the tab angle ϕ for every different fibre orientation θ
available manufacturer's data, are:
can be estimated. These angles are shown in Table 6.
E11 = 135 GPa , E22 = 8.75 GPa , G12 = 4.75 GPa , 12 = 0.3. (9) Due to the difficulty in precisely cut these estimated angles, prac-
tical tab angles ϕp have been selected, as shown in Table 6. Using these
These values allow the tab angles for the different fibre orientation
practical angles ϕp woven fabric glass fibre tabs have been cut with a
in the off-axis tests to be estimated. These values of the tab angle can
length of 40 mm approximately, 10 mm width and trapezoidal shape.
present certain deviation due to the uncertainty in the values of ma-
These oblique tabs have been bonded to specimens by means of SA-
terial properties. To check how correct the estimated tab angles are, 0°
FRAN adhesive.
and 90° tension tests have been carried out additionally. For this pur-
The off-axis tensile tests have been carried out in an INSTRON 4483
pose, two unidirectional panels with fibre orientation at 0° and 90° have
universal testing machine, with a 150 KN load cell, and a 50 mm length
been manufactured to determine the elastic modulus E11 and E22, and
extensometer. The room temperature for the tests, the relative humidity
the allowable values XT and YT for the material considered. The panel
and the displacement velocity have been respectively, 22 °C, 50% and
with fibre orientation 0° has been prepared by laying up 4 pre-preg
1 mm/min. Mechanical grips have been used in the tests, and all the
layers with dimensions 220 × 150 mm2 with a resultant thickness of
surface of the tab was griped to try to reproduce the displacement
0.85 mm approximately. The panel with fibre orientation 90° has been
boundary condition at the end of specimen. Experimental curves of
prepared by laying up 6 pre-preg layers with dimensions 200 × 150
longitudinal stress σx versus longitudinal strain εx are shown in Fig. 11.
mm2 with a resultant thickness of 1.18 mm approximately. Both panels
Due to the uniformity of stress field at the central zone of specimen, it
have been cured in autoclave, and then woven fabric glass fibre tabs
must be mentioned that measured values (which are average values) of
have been joined to the ends of the panels. Rectangular specimens have
both stress and strain are representatives of the values at the points in
been obtained from the panels by cut-out. From the 0° panel, 9 speci-
this central zone.
mens with dimensions 200 × 10 mm2 have been obtained, presenting
As is shown in Fig. 11, the scattering in the behaviour of specimens
regular tabs with 40 mm length approximately, and a specimen ratio L/
decreases with the increase in the fibre orientation angle θ, it being
w > 10 (L = free length, w = width). From the 90° panel, 8 specimens
maximum for 5° and almost negligible for 20°. The average values of the
with dimensions 200 × 15 mm2 have been obtained, presenting regular
tensile strength σxu and the failure strain εxu with their corresponding
tabs with 50 mm length approximately, and a specimen ratio L/
w = 6.67.
Table 6
The tensile tests have been carried out in an INSTRON 4483 uni-
Estimated and practical tab angles for the different fibre orientation.
versal testing machine, with a 150 KN load cell, and a 50 mm length
extensometer. The room temperature for tests was 22 °C, the relative θ (°) 5 10 15 20
humidity was 50%, and the displacement velocity was 1 mm/min. As
ϕ (°) 28.23 22.36 23.8 27.44
results of the 0° tension tests the average values for both the elastic ϕp (°) 28 22 24 27
modulus in fibre direction E11, the tensile strength in fibre direction XT

19
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 11. Stress-strain curves (σx vs. εx) for the different fibre orientations considered.

Table 7
Failure stress and strain, and elastic modulus.
θ (°) σxu (MPa) STD (MPa) εxu (%) STD (%) Ex (GPa) Exsec (GPa)

5 753.539 53.792 0.976 0.119 92.574 76.043

10 414.945 17.850 0.989 0.198 70.928 43.053
15 296.842 10.063 1.034 0.150 50.204 29.284
20 200.602 12.585 1.140 0.231 33.889 18.122

standard deviations are shown in Table 7. The elastic modulus Ex (slope

at the beginning of the curves) and the secant value of the modulus Exsec
(Exsec = σxu/εxu) are also shown in Table 7.
The failures observed in specimens with fibre directions 5°, 10°, 15°
and 20° are shown in Figs. 12–15 respectively. As can be seen in Fig. 12,
all specimens with fibre orientation angle θ = 5° present failures near
the corners. As can be seen in Fig. 13, the specimens with fibre or-
ientation 10° fail mainly at the central zone (5 failures at central zone, 1
failure at lateral zone, 1 failure at the corner). As can be seen in Fig. 14, Fig. 13. View of the failure specimens for fibre orientation angle 10°.
the majority of specimens with fibre orientation 15° fail at the lateral
zone (5 failures at lateral zone, 1 failure at central zone). Finally, as can 7. Analysis of experimental results and discussion
be seen in Fig. 15, the specimens with fibre orientation 20° fail mainly
at the lateral zone (5 failures at lateral zone, 1 failure at central zone). The shear properties of the material considered (modulus G12 and
strength S) are the values that are wanted to be obtained from the direct
results of tests (longitudinal modulus Ex and failure stress σxu). In what
concerns to the shear modulus G12, this property can be evaluated by
means of the following expression:

1 cos4 2 12 1 sin4
= + + sin2 cos2 +
Ex E11 E11 G12 E22 (10)

Substituting the Ex values (see Table 7) in equation (10) and the

experimental values of both E11 and E22 (E11 = 125.159 GPa,
E22 = 8.112 GPa), with the Poisson's ratio of the material (ν12 = 0.3),
the shear modulus G12 for every fibre orientation θ can be evaluated.
The G12 values for the different orientations θ are shown in Table 8. The
values of the secant shear modulus G12sec, which can be obtained from
equation (10) using the secant longitudinal modulus Exsec (see Table 7),
are shown also in Table 8.
As can be observed from the values in Table 8, a high difference
exists between the value of the shear modulus G12 (slope at the be-
ginning) and the value of the secant shear modulus G12sec. The meaning
of this difference is obviously the non-linear character of the shear
behaviour. Therefore, to characterise the material behaviour the
Fig. 12. View of the failure specimens for fibre orientation angle 5°. knowledge of shear curve (σ12 vs. γ12) is needed. For this purpose, the

20
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

shear stress component σ12 can be evaluated from the longitudinal

stress σx using equation (5) corresponding to the stress field of the ideal
off-axis tension test configuration (Fig. 1(a)). The shear strain γ12, ac-
cording to the material constitutive law in orthotropic axes, can be
expressed as the quotient σ12/G12. In this way, using equation (5) and
the constitutive law in geometric axes of specimen (σx = Ex · εx), the
following expression is obtained:

x sin cos Ex x sin2 cos2

= =
12
G12 sin cos G12 (11)

Using equation (10), the above expression becomes:

Ex x 1 cos4 2 12 sin4
12 = + sin2 cos2
sin cos Ex E11 E11 E22 (12)

Manipulating adequately equation (12) we can obtain:

x cos3 2 12 sin3
12 = + x sin cos +
sin cos sin E11 E11 cos E22
(13)
As can be observed in equation (13), the shear strain γ12 is expressed
as a function of both the measured values in the tests (σx, εx) and
constant values that depend on the fibre orientation angle θ and the
material properties (E11, E22, ν12). A direct measurement of γ12 can be
obtained by instrumentation of specimens with 3 strain gages. This
procedure is independent of material properties but is more expensive.
Fig. 14. View of the failure specimens for fibre orientation angle 15°. Thus, using equations (5) and (13) the curves σ12 vs. γ12 for the dif-
ferent orientations considered can be evaluated. These curves are re-
presented in Fig. 16.
As can be seen in Fig. 16, the scattering in the shear behaviour
decreases with the increase in the fibre orientation angle θ, it being
higher for 5° and smaller for 20°. Theoretically it might be expected that
all fibre orientations presented the same shear behaviour, but it can be
observed that the curves for θ = 5° shows a more flexible behaviour
than the other orientations, leading to G12 values smaller than for the
other cases. A possible explanation of this fact can be that for small fibre
orientation angles, as θ = 5°, the material behaviour is dominated by
the fibre. For that reason, variables associated with the matrix, as the
shear strain γ12, can present significant errors. Therefore, the fibre or-
ientation angle θ = 5° can not be considered adequate to evaluate the
shear modulus by means of the off-axis tension test with oblique end-
tabs.
With reference to the other orientations (10°, 15°, 20°), the differ-
ences noticed in shear behaviour are small, it being observed that 10°
and 20° orientation angles show a similar behaviour, while the speci-
mens for θ = 15° present a slightly stiffer behaviour than the others.
In what concerns to the evaluation of shear strength S, taking into
account that the stress field in the specimens is a full stress state, with
three components (σ1, σ2, σ12), the use of a failure criterion is needed to
evaluate S in an inverse manner. For this purpose, in this work, the
theories of Puck [55] and Tsai-Wu [54] will be used again.
Theoretically, in the off-axis tension test with oblique end-tabs it
can be supposed that the failure appears in the central zone of spe-
cimen, under a uniform stress field very approximated to that of the
Fig. 15. View of the failure specimens for fibre orientation angle 20°. ideal test configuration, which can be expressed by equation (5).
Therefore, the stress component values (σ1u, σ2u, σ12u) in the central
zone of specimen at the failure instant can be calculated substituting the
experimental values of σxu (see Table 7) in expression (5). These
Table 8 stresses values for the different orientations are shown in Table 9.
Shear modulus G12 and G12sec. As can be observed from the values in Table 9, the stresses in the
θ (°) G12 (GPa) G12sec (GPa) central zone of specimens at the failure instant are different for the
different orientations, and then, every orientation must be analysed
5 2.545 1.419 separately. Previously, to address the analysis of the strength behaviour
10 4.423 1.858
for every orientation, the angle of the tabs employed ϕp (estimated on
15 4.917 2.320
20 4.677 2.163 the basis of the average values of material properties (9)) must be
checked to verify that agrees with the theoretical values associated with

21
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 16. Curves σ12 vs. γ12 for different orientations.

Table 9 equation (8). The shear strength value obtained from equation (8), for
Stress components in the regularized zone at failure. the stress components corresponding to θ = 10°, is S = 77.887 MPa. As
θ (°) σ1u (MPa) σ2u (MPa) σ12u (MPa)
can be observed, the values evaluated by both criteria are very similar,
then the average from these values, S = 77.353 MPa, can be taken as
5 747.815 5.724 65.425 representative of the shear strength.
10 402.433 12.512 70.960 In the case of θ = 15°, as can be seen in Fig. 14, the majority of
15 276.957 19.885 74.210
20 177.136 23.466 64.472
specimens fail at a lateral zone near to the end, but at a certain distance
from the corner. In order to evaluate the stress field that appears in such
zone a FEM model has been developed in commercial code ANSYS [53],
Table 10 consisting of 30150 elements type SHELL63, with 40501 nodes. The
Theoretical and practical tab angles for the different fibre orientation. mesh (the same used in section 3), as can be seen in Fig. 5, has been
refined conveniently in the area close to the ends to collect the effect of
θ (°) 5 10 15 20
the stress concentration. The discretisation has been verified by com-
ϕ (°) 27.74 22.10 23.63 27.35 paring the results for a mesh with an element size reduced by half,
ϕp (°) 28 22 24 27 obtaining the same results. To simulate the effect of the non-linear
shear behaviour (produced at the failure instant), by means of a linear
elastic analysis, the secant shear modulus (G12sec = 2.32 GPa) has been
the material properties (experimental values: E11 = 125.159 GPa, used. As boundary condition, a uniform displacement is applied to
E22 = 8.112 GPa, G12 = 4.6 GPa, ν12 = 0.3). These theoretical values produce a longitudinal resultant load Fx = 2970 N, which corresponds
can be evaluated substituting the material properties in expression (4), to an average stress σx = 297 MPa (value of σxu for θ = 15°).
these values together with the values of ϕp (for comparison purposes) The distribution of stress components (σ1, σ2, σ12) in the specimen is
are shown in Table 10. shown in Fig. 17, where a stress concentration (formally a singularity)
As can be seen from the values in Table 10, the tab angles ϕp used in in σ2 around the upper corner (left end) can be observed, while σ12
specimens present deviations lesser than 0.5°, and therefore, they can presents both a stress concentration (formally a singularity) around the
be acceptable. This fact shows that the tab angles can be precisely lower corner (left end) and a slight concentration at a certain distance
evaluated on the basis of an estimation of material properties. away from the end of specimen. The central zone of specimen is under a
As is shown in Fig. 12, the majority of specimens for θ = 5° fails uniform stress state, whose values are approximately those showed in
near the corner, where there is a singularity in the stress state (as can be Table 9 for θ = 15°. Additionally, in Fig. 17, the value of Tsai-Wu's
seen in section 4). For that reason, this orientation is not suitable for the criterion has been represented, it being observed that a lateral zone
evaluation of shear strength S. (near the extremes of the specimen) appears with values of the criterion
For θ = 10°, as can be observed in Fig. 13, the majority of specimens higher than one, while in the central zone the values of the criterion are
fail at the central zone, and therefore it can be assumed that this failure lesser than one. The values of the criterion achieved in this lateral zone
has been generated under the regularized stress field, whose values are (coloured in light blue in Fig. 17) are produced by the interaction of
shown in Table 9. Replacing these stress components in the considered both stress components σ2 and σ12. The failures observed in specimens
failure criteria with the allowable of material (XT = 1746.458 MPa, (see Fig. 14) at this lateral zone are consequently associated with the
YT = 53.666 MPa, XC = 1300 MPa, YC = 245 MPa) the shear strength S high value of the criterion at this zone. Due to the non-uniform stress
can be evaluated in an indirect way. field observed at this lateral zone, where the failures are located, the
In the case of Puck's criterion [55], the failure mode which would be shear strength S cannot be evaluated from the measurements of the
activated is the so-called “inter-fiber failure mode A”, whose expression tests for this fibre orientation.
is given by equation (6), where P (+) is 0.35 for graphite composites, as In the case of θ = 20°, as can be seen in Fig. 15, the majority of
recommended by the author. The shear strength value obtained from specimens fail at a lateral zone near to the end, with a similar config-
equation (6), for the stress components corresponding to θ = 10°, is uration to the above analysed for θ = 15°, therefore the explanation
S = 76.819 MPa. made for θ = 15° can be extended to interpret the results for θ = 20°.
In the case of Tsai-Wu's criterion [54], the expression is given by

22
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Fig. 17. Numerical results for θ = 15°: stress components σ1, σ2, σ12, and value of Tsai-Wu's criterion.

Fig. 18. View of the failure specimens with fibre orientation θ = 10° and tab angles: (a) ϕ = 25°, (b) ϕ = 29°.

Once all considered orientations have been analysed and the 10° those specimens were configured with oblique tabs at ϕ = 25°, and in
orientation is shown to be the most suitable for test execution, the effect the others three specimens an angle of ϕ = 29° is used for tabs. These
of deviation in tab angle for this orientation (θ = 10°) has been ex- tab angles, 25° and 29°, produce deviations Δϕ, with respect to the
perimentally studied. For this purpose, from a panel with fibre or- angle corresponding to θ = 10° (ϕ = 22°), of +3° and +7°, respec-
ientation angle θ = 10° six specimens have been obtained, three of tively. These deviations Δϕ have been selected to experimentally check

23
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

the numerical estimations carried out in previous works [47], which taken as representative of the shear strength S value. The other or-
suppose that 3° deviations could not affect the strength behaviour but 7° ientations considered (5°, 15° and 20°) fail in the majority of the cases
deviations could produce premature failures. The failures observed in near the extreme of specimen where the stress state is non uniform.
specimens of fibre orientation θ = 10° for both tab angles ϕ = 25° and Additionally, for fibre orientation angle θ = 10°, specimens with
ϕ = 29° are shown in Fig. 18 (a) and (b) respectively. deviations in tab angle ϕ of +3° and +7° have been tested, it being
As can be observed in Fig. 18, in both cases ((a) and (b)) all the observed that all the failures appear at the end of specimens.
failures occur at the ends of specimens, therefore both deviations +3° Consequently, deviations in tab angle must be avoided, they having to
and +7° generate changes in the stress field around the corner of be lesser than +3°.
specimens that increase it severity and produces premature failures. As In summary, the most suitable fibre orientation angle to perform the
a consequence, the maximum positive deviation in tab angle must be off-axis tensile test under oblique end-tabs configuration in graphite-
less than +3° for fibre orientation θ = 10°. epoxy composites is θ = 10°. Based on an estimation of material
In summary, it would be recommendable to use a tab angle lesser properties (quotient G12/E11) the angle of tabs ϕ can be determined, to
than the theoretic value of ϕ. The reasons behind this recommendation use a tab angle lesser than the theoretic one to avoid positive deviations
are two: being recommendable.

- Negative deviations in tab angle are favourable in order to decrease Acknowledgement

the severity of the stress field around the corner of specimen (as can
be seen in section 4). This work was supported by the Spanish Ministry of Education,
- Positive deviations in tab angle must be avoided because the max- Culture and Sports under Grant MAT2016-80879-P.
imum allowable deviation is lesser than +3°.
Appendix A. Supplementary data
8. Conclusions
Supplementary data to this article can be found online at https://
In the present work, the most adequate fibre orientation angle to doi.org/10.1016/j.compscitech.2019.04.022.
perform off-axis tensile tests under an oblique end-tabs configuration is
studied. This study is appropriate because, in this configuration, the References
angle of the tabs depends on both the fibre orientation angle and the
properties of the material. For this purpose, an analytical equation for [1] L.H. Floeter, K.H. Boller, Use of Experimental Rails to Evaluate Edgewise Shear
the tab angle ϕ has been obtained, showing that the parameters that ϕ Properties of Glass-Reinforced Plastic Laminates, U.S. Forest Products Laboratory,
Madison, Wisconsin, Task Force for Test Methods vol. 17, Industry Advisory Group,
depends on are the fibre orientation angle θ, the quotient G12/E11, the MIL-HDBK, 1967 28 abril.
quotient E11/E22 and the Poisson's ratio ν12. On the basis of this ana- [2] J.M. Whitney, D.L. Stansbarger, H.B. Howell, Analysis of the rail shear test – ap-
lytical equation, it is shown that the quotient G12/E11 is the most im- plications and limitations, J. Compos. Mater. 5 (1) (1971) 24–34.
[3] A.K. Hussain, D.F. Adams, The Wyoming-modified two-rail shear test fixture for
portant parameter in the value of the tab angle for a fixed fibre or- composite materials, J. Compos. Technol. Res. 21 (4) (1999) 215–223.
ientation angle. The identification of this main parameter is [4] A.K. Hussain, D.F. Adams, Analytical evaluation of the two-rail shear test method
fundamental to control the deviations in the tab angle due to the un- for composite materials, Compos. Sci. Technol. 64 (2) (2004) 221–238.
[5] A.K. Hussain, D.F. Adams, Experimental evaluation of the Wyoming-modified two-
certainty in the value of the material properties. rail shear test method for composite materials, Exp. Mech. 44 (4) (2004) 354–364.
Additionally, to study how the deviations in the tab angle affect the [6] I. De Baere, W. Van Paepegem, J. Degrieck, Design of a modified three-rail shear
stress state at the corners of the specimen, a novel study of the singu- test for shear fatigue of composites, Polym. Test. 27 (3) (2008) 346–359.
[7] S.M. Mohseni Shakib, S. Li, Modified three rail shear fixture (ASTM D4255/D
larity of the nominal stress state at the corner where the failure starts
4255M) and an experimental study of non-linear in-plane shear behaviour of FRC,
has been developed. The results of this study show that positive de- Compos. Sci. Technol. 69 (11–12) (2009) 1854–1866.
viations in the angle of the tabs ϕ increases the order of the stress [8] J.M. Whitney, J.C. Halpin, Analysis of laminated anisotropic tubes under combined
singularity, leading to premature failures in the neighbourhood of the loading, J. Compos. Mater. 2 (1968) 360–367.
[9] N.J. Pagano, J.M. Whitney, Geometric design of composite cylindrical character-
corner as is observed in specimens with induced deviations. In contrast, ization specimens, J. Compos. Mater. 4 (1970) 360–378.
negative deviations in the angle of the tabs ϕ contribute to decrease the [10] N.J. Pagano, J.C. Halpin, Influence of end constraint in the testing of anisotropic
severity of the stress state at the corner. From this point of view, as has bodies, J. Compos. Mater. 2 (1) (1968) 18–31.
[11] E.M. Wu, R.L. Thomas, Off-axis test of a composite, J. Compos. Mater. 2 (4) (1968)
been shown by this study of stress singularity at the corner of the 523–526.
specimen, the fibre orientation θ = 10° is optimal in terms of having a [12] R.R. Rizzo, More on the influence of end constraint on off-axis tensile test, J.
tab angle ϕ closest to the angle that would eliminate the singularity Compos. Mater. 3 (1969) 202–219.
[13] G.L. Richards, T.P. Airhart, J.E. Ashton, Off-axis tensile coupon testing, J. Compos.
(δ = 0). Mater. 3 (1969) 586–589.
As has been shown in section 5, a range of suitable fibre orientations [14] R.B. Pipes, B.W. Cole, On the off-axis strength test for anisotropic materials, J.
θ for the determination of the shear strength S, i.e., those in which the Compos. Mater. 7 (1973) 246–256.
[15] C.C. Chamis, J.H. Sinclair, Ten-deg off-axis test for shear properties in fiber com-
contribution of the shear stress σ12 to the failure of the layer is domi- posites, Exp. Mech. 17 (9) (1977) 339–346.
nant, are located between 5° and 18°. The optimal orientations (where [16] B.W. Chang, P.H. Huang, D.G. Smith, A pinned-end fixture for off-axis testing, Exp.
the contribution of σ12 to failure is maximal) are located at the lower Tech. 8 (6) (1984) 28–30.
[17] M.J. Pindera, C.T. Herakovich, Shear characterization of unidirectional composites
end of the range, i.e., between 5° and 10°.
with the off-axis tension test, Exp. Mech. 26 (1) (1986) 103–112.
In the experimental part of the present work, off-axis tensile tests of [18] S.M. Cron, A.N. Palazotto, R.S. Sandhu, The improvement of end-boundary condi-
graphite-epoxy (AS4/8552) specimens with oblique tabs for four dif- tions for off-axis tension specimen use, Exp. Mech. 28 (1) (1988) 14–19.
ferent fibre orientations (5°, 10°, 15° and 20°) have been carried out. [19] C.T. Sun, S.P. Berreth, A new end tab design for off-axis tension test of composite
materials, J. Compos. Mater. 22 (8) (1988) 766–779.
Experimental results show that only 10° fibre orientation angle is sui- [20] C.T. Sun, I. Chung, An oblique end-tab design for testing off-axis composite speci-
table for the evaluation of shear strength S, because for this orientation mens, Composites 24 (8) (1993) 619–623.
the majority of specimens fail at the central zone, the predictions of the [21] J.C. Marín, J. Cañas, F. París, J. Morton, Determination of G12 by means of the off-
axis tension test. Part II: a self-consistent approach to the application of correction
study of stress singularities being then confirmed. Thus, for this or- factors, Compos. Part A Appl. Sci. Manuf. 33 (1) (2002) 101–111.
ientation (θ = 10°) the shear strength S can be evaluated from the [22] Y. Xiao, M. Kawai, H. Hatta, An integrated method for off-axis tension and com-
regularized stress field by means of a failure criterion. In the present pression testing of unidirectional composites, J. Compos. Mater. 45 (6) (2011)
657–669.
study Puck's and Tsai-Wu's criteria have been used for this purpose [23] P.H. Petit, A Simplified Method of Determining the in Plane Shear Stress-Strain
obtaining similar values, the average value (77.353 MPa) having been

24
J.C. Marín, et al. Composites Science and Technology 178 (2019) 11–25

Response of Unidirectional Composites vol. 460, ASTM STP, 1969, pp. 83–93. [43] J.M. Cabrero, G. Vargas, Analysis of the validity of the three-point off-axis bending
[24] B.W. Rosen, A simple procedure for experimental determination of the longitudinal method, Appl. Math. Model. 39 (2015) 5265–5277.
shear modulus of unidirectional composites, J. Compos. Mater. 6 (3) (1972) [44] V. Trappe, R. Basan, F. Grasse, Stiffness and Fracture of Shear Loaded Laminates
552–554. with Unidirectional and Biaxial Fibre Orientation Investigated with a Picture Frame
[25] C.C. Chiao, R.L. Moore, T.T. Chiao, Measurement of shear properties of fibre Test, 16th European Conference on Composite Materials vol. 16, ECCM, 2014.
composites. Part 1. Evaluation of test methods, Composites 8 (3) (1977) 161–169. [45] F. Pierron, A. Vautrin, 10° off-axis tensile test: a critical approach, Compos. Sci.
[26] S. Kellas, J. Morton, K.E. Jackson, Damage and Failure Mechanisms in Scaled Angle- Technol. 56 (4) (Apr 1996) 483–488.
Ply Laminates vol. 1156, ASTM STP, 1993, pp. 257–280. [46] M. Kawai, M. Morishita, H. Satoh, S. Tomura, K. Kemmochi, Effects of end-tab
[27] M. Arcan, Z. Hashin, A. Voloshin, A method to produce uniform plane-stress states shape on strain field of unidirectional carbon/epoxy composite specimens subjected
with applications to fiber-reinforced materials, Exp. Mech. 18 (4) (1978) 141–146. to off-axis tension, Compos. Part A Appl. Sci. Manuf. 28 (3) (1997) 267–275.
[28] S.C. Yen, J.N. Craddok, K.T. Teh, Evaluation of a modified Arcan fixture for the in- [47] F. Pierron, E. Alloba, Y. Surrel, A. Vautrin, Whole-field assessment of the effects of
plane shear test of materials, Exp. Tech. 12 (12) (1988) 22–25. boundary conditions on the strain field in off-axis tensile testing of unidirectional
[29] S.C. Hung, K.M. Liechti, An evaluation of the Arcan specimen for determining the composites, Compos. Sci. Technol. 58 (12) (Dec 1998) 1939–1947.
shear moduli of fiber-reinforced composites, Exp. Mech. 37 (4) (1997) 460–468. [48] F. Pierron, A. Vautrin, New ideas on the measurement of the in-plane shear strength
[30] K.W. Gan, T. Laux, S.T. Taher, J.M. Dulieu-Barton, O.T. Thomsen, A novel fixture of unidirectional composites, J. Compos. Mater. 31 (9) (1997) 889–895.
for determining the tension/compression-shear failure envelope of multidirectional [49] S.W. Tsai, Composites Design, Think Composites, 1988.
composite laminates, Compos. Struct. 184 (2018) 662–673. [50] MIL-HDBK-17-2E, Polymer Matrix Composites. Materials Properties, Department of
[31] D.E. Walrath, D.F. Adams, The Iosipescu shear test as applied to composite mate- Defense Handbook, 1998.
rials, Exp. Mech. 23 (1) (1983) 105–110. [51] M.M. Schwartz, Composite Materials Handbook, Mc Graw-Hill Book Company,
[32] D.F. Adams, D.E. Walrath, Further development of the Iosipescu shear test method, 1984.
Exp. Mech. 27 (2) (1987) 113–119. [52] C.T. Herakovich, Mechanics of Fibrous Composites, John Wiley & Sons, Inc., 1998.
[33] D.F. Adams, D.E. Walrath, Current status of the Iosipescu shear test method, J. [53] A.N.S.Y.S. Swanson, Analysis System, Inc., 2011.
Compos. Mater. 21 (6) (1987) 494–507. [54] S.W. Tsai, E.M. Wu, A general theory of strength for anisotropic materials, J.
[34] R. Prabhakaran, W. Sawyer, A photoelastic investigation of asymmetric four point Compos. Mater. 5 (1971) 58–80.
bend shear test for composite materials, Compos. Struct. 5 (3) (1986) 217–231. [55] A. Puck, H. Schurmann, Failure analysis of frp laminates by means of physically
[35] H. Inone, M. Takanashi, T. Nakamura, T. Aoki, T. Ogasawara, Modified asymmetric based phenomenological models, Compos. Sci. Technol. 58 (7) (1998) 1045–1067.
four-point bend test method for in-plane shear properties of ceramic matrix com- [56] A. Barroso, V. Mantic, F. París, Singularity analysis of anisotropic multimaterial
posites at elevated temperatures, Ceram. Eng. Sci. Proc. 37 (2) (2017) 53–60. corners, Int. J. Fract. 119 (2003) 1–23.
[36] G.D. Sims, W. Nimmo, A.F. Johnson, D.H. Ferriss, Analysis of Plate-Twist Test for [57] V. Mantic, A. Barroso, F. París, Singular elastic solutions in anisotropic multi-
In-Plane Shear Modulus of Composite Materials (Revised), Tech. Rep. 54 National material corners. Application to composites, in: V. Mantic (Ed.), Chapter 11 of the
Physical Laboratory Report DMM(A), 1994. Book “Mathematical Methods and Models in Composites”, Imperial College Press,
[37] B. Gommers, I. Verpoest, P. Van Houtte, Further developments in testing and 2014, pp. 425–495.
analysis of the plate twist test for in-plane shear modulus measurements, Compos. [58] G. Odegard, M. Kumosa, Determination of shear strength of unidirectional com-
Part A Appl. Sci. Manuf. 27 (11) (1996) 1085–1087. posite materials with the iosipescu and 10° off-axis shear tests, Compos. Sci.
[38] D.O. Adams, J.M. Moriarty, A.M. Gallegos, D.F. Adams, Development and Technol. 60 (16) (2000) 2917–2943.
Evaluation of the V-notched Rail Shear Test for Composite Laminates, Tech. Rep. [59] G. Odegard, M. Kumosa, Elastic-plastic and failure properties of a unidirectional
Federal Aviation Administration Report DOT/FAA/AR-03/63 FAA Office of carbon/PMR-15 composite at room and elevated temperatures, Compos. Sci.
Aviation Research, Washington, DC, 2003. Technol. 60 (16) (2000) 2979–2988.
[39] D.O. Adams, J.M. Moriarty, A.M. Gallegos, D.F. Adams, The V-notched rail shear [60] J.C. Xavier, N.M. Garrido, M. Oliveira, J.L. Morais, P.P. Camanho, F. Pierron, A
test, J. Compos. Mater. 41 (3) (2007) 281–297. comparison between the Iosipescu and off-axis shear test methods for the char-
[40] M. Gude, W. Hufenbach, M. Andrich, A. Mertel, R. Schirner, Modified V-notched acterization of Pinus Pinaster Ait, Compos. Part A Appl. Sci. Manuf. 35A (7–8)
rail shear test for shear characterisation of textile-reinforced composite materials, (2004) 827–840.
Polym. Test. 43 (2015) 147–153. [61] M.J. Hinton, A.S. Kaddour, P.D. Soden, A comparison of the predictive capabilities
[41] F. Mujika, A. Valea, P. Gañan, I. Mondragon, Off-axis flexure test: a new method for of current failure theories for composite laminates, judged against experimental
obtaining in-plane shear properties, J. Compos. Mater. 39 (11) (2005) 953–980. evidence, Compos. Sci. Technol. 62 (12–13) (2002) 1725–1797.
[42] G. Vargas, F. Mujika, Determination of in-plane shear strength of unidirectional [62] K.-S. Liu, S.W. Tsai, A progressive quadratic failure criterion for a laminate,
composite materials using the off-axis three-point flexure and off-axis tensile tests, Compos. Sci. Technol. 58 (7) (1998) 1023–1032.
J. Compos. Mater. 44 (21) (2010) 2487–2507.

25