Short-Beam Shear Fatigue Behavior on Unidirectional GLARE: Mean Shear Stress Effect, Scatter, and Anisotropy

Abstract

This paper investigates the effect of mean shear stress on short-beam shear fatigue in a GLARE 1-3/2 commercial fiber–metal laminate (FML). This study explores three shear stress ratios (${\mathit{R}}_{\mathit{\tau }}$ 0.1, 0.3, and 0.5) and two material orientations (longitudinal and transversal) under constant amplitude fatigue. Different stress levels for each ${\mathit{R}}_{\mathit{\tau }}$ value were explored to obtain failures between 10³ and 10⁶ load cycles. The experimental results reveal anisotropy, with transversal specimens exhibiting lower performance and increased scatter. The mean shear stress effect is discussed herein, with insights into the critical role of mean shear of fatigue performance. ${\mathit{R}}_{\mathit{\tau }}$ 0.1 was the most severe condition and ${\mathit{R}}_{\mathit{\tau }}$ 0.5 was the least severe. The ${\mathit{R}}_{\mathit{\tau }}$ 0.3 condition produced steeper S-N curves, indicating that the combined effect of mean shear stress and shear stress amplitude led to a higher rate of damage accumulation. The fractographic analysis investigated the failure modes and confirmed the damage dominated by Mode II, supporting the test methodology employed.

1. Introduction

Fiber–metal laminates (FMLs), which were designed primarily for aeronautical applications, are more complex than typical fiber-reinforced composites. They offer advantages over monolithic materials, including higher specific strength and a slower crack growth rate in fatigue, resulting from their unique design and manufacturing process [1].

However, when it comes to interlaminar cracks, FMLs tend to exhibit relatively lower values of fracture toughness and fatigue crack growth resistance [2,3] compared to other crack orientations in the material [4,5]. This phenomenon is observed in both mode I and mode II loading conditions.

Notably, mode II loading has gained more research attention due to the potential occurrence of interlaminar shear stresses in various scenarios. These include bending, internal or external ply drop, straight free edges, corners, and crack bridging, all of which can contribute to mode II-driven delaminations [6,7,8,9,10,11,12].

The fatigue interlaminar crack growth in mode II has received significant attention in recent research. Researchers like Amaral et al. [13], Giallanza et al. [3], and Bieniaś and Dadej [14] have conducted experiments and developed models to assess mode II crack growth in various FML types. These investigations employed the end-notched flexure (ENF) test fixture, a reliable method for quantifying mode II fatigue-related interlaminar crack growth. However, it is important to note that this test does not provide insights into the stage of fatigue initiation. These previous studies highlighted the complexities associated with assessing interlaminar shear fatigue. Consequently, the literature contains several alternative tests that offer the means to approximate interlaminar shear fatigue, encompassing both the initiation and propagation stages, albeit with certain limitations. Lessard et al. [15] presented a list of 31 shear test methods and summarized the advantages and disadvantages of each test. The most commonly used methods were discussed: the rail shear test [16] presents a questionable region of pure shear due to the specimens being prone to twisting, low interlaboratory repeatability, and problems due to free edges [15]. The [±45]_ns test [17] is inexpensive; however, it requires superior specimen quality and does not produce reliable results [15]. The Iosipescu test [18] is considered the main choice to obtain static shear properties; however, there is a need to adapt the test rig for fatigue and its application in fatigue scenarios is not well established [15,19].

The short-beam shear (SBS) test [20] is widely employed in laminate composites to obtain the apparent interlaminar shear strength. Although this test method produces only an apparent measure of this shear property [15,20], the SBS test presents some advantages over the abovementioned methodologies, like the small specimens needed, simple test apparatus, and absence of problems related to grip between the test sample and the device [21].

This quasi-static test has been employed to investigate its applicability in FMLs [22,23] or to examine other factors related to SBS strength [24,25,26,27,28,29,30,31,32,33]. Regarding the application of cyclic loads, Kotik and Perez Ipiña [34] used this test in fatigue to study the behavior of commercial GLARE 1-3/2 laminate in two orientations for a shear stress ratio (${\mathit{R}}_{\mathit{\tau }}$) of 0.1. Additionally, they evaluated the failure mode, along with some aspects associated with the test limitations. Surowska et al. [35] evaluated the effect of thermal cycles on various carbon–aluminum laminates with protective glass interlayers in a 3/2 configuration, employing SBS fatigue tests. They observed the dependence of the fatigue response on thermal cycles, although they did not provide the shear stress ratio used. Both studies used only one ${\mathit{R}}_{\mathit{\tau }}$, not enabling conclusions about the possible effect of the mean shear stress.

The effect of mean stress on the fatigue life of laminate composites has been investigated under various cyclic loading conditions, including tension–tension, tension–compression, or compression–compression [36,37,38,39]. Vassilopoulos and Keller [40] explained that the influence of mean stress in normal stress fatigue can be accurately represented by the constant life diagram (CLD), which is constructed from classical S-N curves (also known as Wöhler curves).

The effect of mean shear stress on fatigue life has been explored in studies involving glass fiber- [19,41] and carbon fiber-reinforced [42,43] polymers, utilizing SBS fatigue results. However, this effect remains an open question in FML research regarding interlaminar shear stresses.

This paper explores the effect of mean shear stress fatigue on a commercial FML known as GLARE 1-3/2. The experimental program involved testing SBS fatigue specimens at three ${\mathit{R}}_{\mathit{\tau }}$ values in both the longitudinal and transversal orientations of the material, enabling an evaluation of the anisotropic effect. Basquin’s and Sendeckyj’s fitting models were adapted for shear stresses and then applied to the experimental results to build a partial piecewise CLD to illustrate the mean stress effect. A fractographic analysis of the failure mechanisms involved and a fractographic validation of the methodology applied were also included.

2. Materials and Methods

The material chosen in this investigation was a commercial GLARE ^® 1-3/2 laminate [1] produced by the Structural Laminates Company, Pittsburgh, PA, US. The lay-up of this fiber–metal laminate comprises three Al 7475-76 alloy layers interleaved with two S-glass-reinforced FM-94 epoxy prepreg layers. Some characteristics and mechanical properties obtained from the manufacturer data sheet [44] are shown in

Table 1. Characteristics and mechanical properties of GLARE 1-3/2 [44].

Material	GLARE 1
Lay-up (Al/prepreg)	3/2
Fibers	S-glass
Al alloy	7475-T76
Post-stretch (%)	0.4
$\mathrm{Longitudinal} \mathrm{yield} \mathrm{strength} ({\mathit{F}}_1^{\mathit{y}\mathit{s}}$) (MPa)	545
$\mathrm{Transversal} \mathrm{yield} \mathrm{strength} ({\mathit{F}}_2^{\mathit{y}\mathit{s}}$) (MPa)	338
Metallic percentage (Vol)	67.9
$\mathrm{Longitudinal} \mathrm{ultimate} \mathrm{tensile} \mathrm{strength} ({\mathit{F}}_1^{\mathit{u}\mathit{t}\mathit{s}}$) (MPa)	1282
$\mathrm{Transversal} \mathrm{ultimate} \mathrm{tensile} \mathrm{strength} ({\mathit{F}}_2^{\mathit{u}\mathit{t}\mathit{s}}$) (MPa)	352
Longitudinal Young’s modulus (E1) (GPa)	64
Transversal Young’s modulus (E2) (GPa)	49
Density (ρ) (kg/m3)	2520
Total thickness (t) (mm)	1.42

. The short-beam strength in both orientations was determined in a previous study [34] and were ${\mathit{F}}_1^{\mathit{s}\mathit{b}\mathit{s}}$ = 81.6 MPa and ${\mathit{F}}_2^{\mathit{s}\mathit{b}\mathit{s}}$ = 72.5 MPa.

The specimens were sliced in longitudinal and transversal directions with a circular blade saw, as schematically shown in Figure 1. Nominal specimen dimensions were thickness (t) 1.42 mm, width (b) 2.84 mm, and length (l) 9 mm. A total of 60 longitudinal specimens and 70 transversal specimens were tested. Sandpaper grit grades of 80, 240, 600, and 1200 were used to produce a smooth surface in planes 1–3 and 2–3. The top and bottom surfaces (planes 1–2) were left as manufactured.

The specimens were tested in an electrodynamic testing machine, Instron ElectroPuls^® E3000 (Instron, High Wycombe, UK), with a load cell of ±5 kN. The tests were monitored with a digital camera, Imperx IPX-2M30 (Imperx, Boca Raton, FL, USA), equipped with a 1:1.4/25 mm Fujifilm lens. The testing device, shown in Figure 2, used a span-to-thickness ratio of 4.0, supports of 1 mm in diameter, and a loading nose of 2 mm in diameter.

The SBS stress (${\mathit{\tau }}_{\left(\mathit{i}\right)}^{\mathit{s}\mathit{b}\mathit{s}}$) was calculated by employing the classical equation of beams in ASTM D2344-16 [20] for quasi-static SBS tests, as follows:

$${\mathit{\tau }}_{\left(\mathit{i}\right)}^{\mathit{s}\mathit{b}\mathit{s}}={\displaystyle \frac{3{\mathit{P}}_{\left(\mathit{i}\right)}}{4\mathit{b}\mathit{t}}}$$

(1)

where ${\mathit{P}}_{\left(\mathit{i}\right)}$ is the force at the i-th data point observed during the fatigue test.

Specimens were tested using values for ${\mathit{R}}_{\mathit{\tau }}$ = 0.1, 0.3, and 0.5, with the shear stress ratio (${\mathit{R}}_{\mathit{\tau }}$) defined as:

$${\mathit{R}}_{\mathit{\tau }}={\displaystyle \frac{{\mathit{\tau }}_{\mathit{m}\mathit{i}\mathit{n}}^{\mathit{s}\mathit{b}\mathit{s}}}{{\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}}}$$

(2)

where ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ and ${\mathit{\tau }}_{\mathit{m}\mathit{i}\mathit{n}}^{\mathit{s}\mathit{b}\mathit{s}}$ are the maximum and minimum SBS stress values, respectively, observed in one cycle (Figure 3).

The tests were performed at a loading frequency (f) of 5 Hz, constant amplitude (${\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}$), sinusoidal waveform, load control (see Figure 3), and room conditions (25 °C; HR 50 ± 10%). The tests ended when the specimens collapsed or a limit of 10⁶ cycles was reached. The load cycles were defined by the ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ and ${\mathit{R}}_{\mathit{\tau }}$ values. The ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ values were chosen as a function of the short-beam strength, to obtain failures between 10³; and 10⁶ cycles. The ${\mathit{\tau }}_{\mathit{m}\mathit{i}\mathit{n}}^{\mathit{s}\mathit{b}\mathit{s}}$ were calculated using Equation (2). The specimen orientation and the values of ${\mathit{R}}_{\mathit{\tau }}$, ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$, and the number of test replicates for each condition employed are detailed in

Table 2. Test program.

Orientation	${\mathit{R}}_{\mathit{\tau }}$	${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ (MPa)	Number of Replicates	Orientation	${\mathit{R}}_{\mathit{\tau }}$	${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ (MPa)	Number of Replicates
Longitudinal	0.1 [34]	60.2	4	Transversal	0.1 [34]	53.6	5
		56.0	6			50.0	4
		52.0	6			49.6	1
		48.0	5			46.4	5
		44.9	6			42.7	2
		41.8	3			42.4	3
	0.3	72.0	3			42.2	1
		66.9	3			41.3	1
		62.9	3			39.8	6
		57.7	4		0.3	66.9	4
		51.4	3			57.1	5
	0.5	72.0	4			51.4	5
		68.0	3			45.7	4
		64.0	3			42.9	5
		60.0	4		0.5	64.0	5
						60.0	5
						56.0	5
						48.0	4

.

Fatigue results are used to construct S-N curves and can be fitted using either deterministic or statistical models. Vassilopoulos and Keller [40] highlighted that the choice of model significantly affects the outcomes of the CLD [40], a factor that is not always taken into account in many publications. The deterministic model by Basquin [45] and the statistical model by Sendeckyj [46] were chosen, due to their good fit to the experimental data and reliable performance in both low- and high-cycle fatigue regimes in a previous analysis. In that analysis, presented in the Supplementary Materials, these models were compared with other models recommended by Vassilopoulos and Keller [40], such as NLD [40], ASTM [47], and Whitney [48], as well as the models proposed by Mandell-Meier [49], Appel-Olthoff [50], Bach [50], and Gao et al. [51]. A partial piecewise linear CLD formulation was chosen due to model simplicity.

Although all the models were originally proposed for normal stresses, they were adapted for application to the current situation of short-beam shear (SBS) stresses. In this way, values such as normal stress amplitude (${\mathit{\sigma }}_{\mathit{a}}$) are superseded by shear stress amplitude (${\mathit{\tau }}_{\mathit{a}}$), mean stress (${\mathit{\sigma }}_{\mathit{m}}),$ by mean shear stress (${\mathit{\tau }}_{\mathit{m}})$, and so on.

The deterministic Basquin’s relation adapted for shear stresses is presented in Equation (3):

$${\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}={\mathit{\tau }}_0{{\mathit{N}}_{\mathit{f}}}^{-1/\mathit{k}}$$

(3)

where ${\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}$ is the SBS cyclic stress amplitude, ${\mathit{N}}_{\mathit{f}}$ is the number of cycles at failure, $-1/\mathit{k}$ and ${\mathit{\tau }}_0$ are model parameters that were estimated by linear regression.

The statistical Sendeckyj’s wear-out model was implemented using Equation (4):

$${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}=\mathit{\beta }\left\{{\left[-\mathit{l}\mathit{n}\left({\mathit{P}}_{\mathit{s}}\left(\mathit{N}\right)\right)\right]}^{{\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{${\mathit{\alpha }}_{\mathit{f}}$}\right.}}\right\}{\left\{\left[\mathit{N}-\left({\displaystyle \frac{1-\mathit{C}}{\mathit{C}}}\right)\right]\mathit{C}\right\}}^{-\mathit{G}}$$

(4)

where ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ is the maximum SBS stress level, ${\mathit{P}}_{\mathit{s}}\left(\mathit{N}\right)$ is the probability of survival after N cycles, C is a constant that defines the shape of the low-cycle fatigue region in the S-N curve, and G is the slope of the S-N curve. $\mathit{\beta }$ and ${\mathit{\alpha }}_{\mathit{f}}$ are the scale and shape parameters of the 2P Weibull distribution, respectively. For a better comparison with the Basquin model fit, the S-N curves for the Sendeckyj model were plotted for a 50% reliability level.

The piecewise linear CLD [40] arises from the linear interpolation between known values in the (${\mathit{\tau }}_{\mathit{m}}^{\mathit{s}\mathit{b}\mathit{s}}-{\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}$) plane and the material’s ultimate strength. This CLD was built by connecting adjacent points with the same time-to-failure value from different S-N curves using straight lines (constant life lines). The first known R-ratio (counted counterclockwise from the x-axis) was connected to the ultimate strength.

Samples with similar ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ and different ${\mathit{R}}_{\mathit{\tau }}$ values were selected to evaluate the failure mechanisms by observing the exposed failure surfaces. The samples were then coated with gold using a sputter coater, the Denton Vacuum model Desk V, and were observed using the scanning electronic microscopes (SEM) Tescan VEGA 3. An example of a cut specimen with the identification of the layers observed in SEM is shown in Figure 4. The failure analysis was carried out on the interfaces between layers 2 and 3 or 3 and 4 (where the shear stresses are higher) and between layers 1 and 2 or 4 and 5 to compare the failure mechanisms of the different layers.

3. Results and Discussion

Plots of the SBS stress amplitude (${\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}$) and maximum SBS stress (${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$) versus the log of cycles to failure are presented in Figure 5 and Figure 6, respectively. a and Figure 6a show the results for longitudinal specimens and Figure 5b and Figure 6b for transversal specimens.

Table 3. Fitting parameters for the Basquin model.

		Basquin Curve Parameters
	${\mathit{R}}_{\mathit{\tau }}$	${\mathit{\tau }}_0$	1/k
Longitudinal	0.1	93.7	0.06
	0.3	124.2	0.07
	0.5	94.6	0.03
Transversal	0.1	82.0	0.06
	0.3	90.2	0.05
	0.5	84.7	0.04

presents the parameters of the Basquin model and

Table 4. Fitting parameters for the Sendeckyj model.

		Sendeckyj Curve Parameters
	${\mathit{R}}_{\mathit{\tau }}$	αf	β (MPa)	C	G
Longitudinal	0.1	28.5	84.8	0.485	0.052
	0.3	30.0	85.1	0.004	0.071
	0.5	28.0	84.7	0.052	0.034
Transversal	0.1	30.9	74.2	0.178	0.053
	0.3	18.4	72.4	0.003	0.064
	0.5	29.0	73.8	0.002	0.050

presents the parameters of the Sendeckyj model when fitted to the data. Figure 6 presents the plots of these adjustments.

3.1. Scatter and Anisotropy

The experimental results exhibited a larger scatter for transversal specimens than longitudinal ones. Specimens tested under an ${\mathit{R}}_{\mathit{\tau }}$ value of 0.1 exhibited significant scatter in the high-cycle regime for both orientations. In opposition to the typical behavior of metals [52], this FML displayed an increased scatter in the low cycle region (<10⁴ cycles) for ${\mathit{R}}_{\mathit{\tau }}$ values of 0.3 and 0.5 in transversal specimens. May and Hallett [19], who observed similar behavior in GFRP specimens, associated this phenomenon with variations in both material properties and specimen quality. It is worth clarifying that May and Hallett analyzed the initiation results, while this work evaluated both damage initiation and growth.

The material exhibited anisotropy for the studied situations, as was previously reported for an ${\mathit{R}}_{\mathit{\tau }}$ value of 0.1 [34]. This effect can also be observed in the CLD diagram (see Section 3.2). The transversal specimens presented lower performance than the longitudinal ones for all three ${\mathit{R}}_{\mathit{\tau }}$ values tested.

3.2. Mean Shear Stress Effect

A pronounced effect of mean shear stress on the SBS fatigue performance of GLARE 1-3/2 was observed. For instance, when ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ = 60 MPa in Figure 6a, specimens tested at an ${\mathit{R}}_{\mathit{\tau }}$ of 0.1 failed at around 2500 cycles, whereas failure occurred at around 50,000 cycles for ${\mathit{R}}_{\mathit{\tau }}$ 0.3 and 350,000 cycles for ${\mathit{R}}_{\mathit{\tau }}$ 0.5. This underscores the critical role of shear stress amplitude in determining fatigue life.

${\mathit{R}}_{\mathit{\tau }}$ 0.3 and 0.5 exhibited similar cycles to failure under the highest ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ condition (around 70 MPa$)$. For low ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}}$ conditions (around 40 MPa ${\mathit{\tau }}_{\mathit{m}\mathit{a}\mathit{x}}^{\mathit{s}\mathit{b}\mathit{s}})$, ${\mathit{R}}_{\mathit{\tau }}$ 0.3 appears to have a similar time-to-failure as ${\mathit{R}}_{\mathit{\tau }}$ 0.1, as seen in the transversal specimen data (Figure 6b). However, the ${\mathit{R}}_{\mathit{\tau }}$ 0.3 condition presented the highest exponential coefficients (curve slope) for both the Basquin and Sendeckyj models (refer to Table 3 and Table 4), except for the transversal dataset adjusted with Basquin’s relation. This effect is more prominent in longitudinal orientation but can also be observed in the transversal specimens (see Figure 6a,b). The high exponential coefficients of an ${\mathit{R}}_{\mathit{\tau }}$ value of 0.3 indicate a higher damage rate for this condition.

In contrast, the ${\mathit{R}}_{\mathit{\tau }}$ value of 0.5 exhibited the lowest exponential coefficients for both models, indicating that this condition is milder. ${\mathit{R}}_{\mathit{\tau }}$ 0.1 showed intermediate coefficient values, except for the Sendeckyj model applied for transversal specimen data. Despite that, this condition presented the lowest N_f.

The lowest S-N curve slope for the ${\mathit{R}}_{\mathit{\tau }}$ 0.5 condition was also reported for axial fatigue tests on E-glass/polyester composites, independent of specimen configurations (on- or off-axis angles) [40], and for ±45° angle-ply E-glass/epoxy composites [53]. May and Hallett [19] evaluated the failure initiation of IM7/8552 carbon/epoxy composite under shear stress fatigue, considering only the initiation stage. The S-N curves were fitted using the Basquin relation and presented a marked decrease in slope with increasing ${\mathit{R}}_{\mathit{\tau }}$ (0.5 < 0.3 < 0.1 for the double-notched shear test and 0.5 < 0.1 for the SBS test). They also observed that damage onset occurred later with high values of ${\mathit{R}}_{\mathit{\tau }}$, indicating that the shear stress amplitude has a more significant impact on initiation than the mean shear stress.

Figure 7a shows the partial piecewise linear CLD [40] built using the S-N curves fitted to Basquin’s relation for longitudinal and transversal data. In Figure 7b, the values of mean stress (${\mathit{\tau }}_{\mathit{m}}^{\mathit{s}\mathit{b}\mathit{s}}$) and stress amplitude (${\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}$) were normalized by the short-beam strength (${\mathit{F}}^{\mathit{s}\mathit{b}\mathit{s}}$) of the respective orientation to enable a better comparison between both datasets. The transversal data set seems to have a “smooth” constant life curve, while the longitudinal dataset presented a deviation for a ${\mathit{R}}_{\mathit{\tau }}$ value of 0.3. In this condition, the 10⁴ and 10⁶ points seem to be farther away when compared to ${\mathit{R}}_{\mathit{\tau }}$ 0.1 and ${\mathit{R}}_{\mathit{\tau }}$ 0.5 conditions. The CLDs in Figure 7b evidence the greater degradation with fatigue in the transverse direction. This combined effect of anisotropy and mean shear stress must be considered in models that aim to approximate fatigue properties from quasi-static strength values.

3.3. Fractographic Analysis

SEM images of specimens with typical failure modes are shown in Figure 8. Interlaminar failure was the preferential mode in the longitudinal samples, as can be seen in Figure 8a,c,e. Delamination was the first macroscopic damage observed during the failure process, as reported in Ref. [34], with later growth of a crack in the middle span of the external aluminum layer under tensile stresses. This can be observed in Figure 8a–e. Some specimens even exhibited a crack in the middle aluminum layer, as shown in Figure 8c. All these cracks corresponded to the final stage of life.

The transversal specimens showed interlaminar and intralaminar damage, as can be seen in Figure 8b,d,e. The first damage observed in these specimens was seen in the aluminum/GFRP interface of the central layers. The interlaminar crack exhibited ply splitting during its propagation. The crack generated visible intralaminar damage in the GFRP layer near the roller (Figure 8b). Another failure mode can be observed in Figure 8d. As observed and reported in Ref. [34], an interlaminar crack first originates due to shear stresses and grows at the mid aluminum/GFRP interface in the mid-span. The interlaminar crack was the first damage that was observed on the specimen. Then, the crack propagated in intralaminar mode approximately parallel to the thickness direction. This delamination promoted through-thickness crack growth in the aluminum layers and led to specimen collapse. The last failure mode (Figure 8f) included inter and translaminar cracks in the GFRP, without macroscopic cracks in the aluminum layer being observed.

The fracture surfaces of three longitudinal specimens tested at ${\mathit{R}}_{\mathit{\tau }}$ 0.1, 0.3, and 0.5 are shown in Figure 9, Figure 10 and Figure 11, respectively. Two different regions can be observed in the three figures at lower magnification (image a), with a more noticeable transition visible in Figure 9 and Figure 10 (highlighted with a dashed line). The final quasi-static fracture can be observed on the left of image a, while the fatigue-related surface is on the right side of the images. The quasi-static surface (images b and c) presented a relatively clean surface with shear cusps, indicated by the arrows in images c. Shear cusps are morphologies that are characteristic of a predominantly mode II quasi-static fracture [54]. The fracture surfaces in Figure 9 and Figure 10 were observed on the surface dominated by the matrix, while Figure 11 presents the fiber-dominated failure surface. The direction of crack propagation can be assessed by the shear cusps’ tilt direction. On the matrix-dominated face, the crack growth direction is the same as the cusp tilts, while on the fiber-dominated face, the crack growth direction is opposite to the cusp tilt direction. The analysis of these fractures indicated that the cracks propagated from the right side to the left, which is consistent with the digital microscope observations during the test. The images in (d) and (e) correspond to that part of the specimen under cyclic shear stresses. The surface is worn and has matrix debris and broken fibers. The arrows in (e) indicate some rudimentary rollers (rounded, deformed shafts of resin) that were observed in the mode II fatigue failure surfaces of fiber-reinforced polymer composites [55].

The fracture surface of a transversal specimen tested at ${\mathit{R}}_{\mathit{\tau }}$ 0.3 and ${\mathit{\tau }}_{\mathit{a}}^{\mathit{s}\mathit{b}\mathit{s}}$ = 18.0 MPa is shown in Figure 12. The fracture exhibited two zones with different damage morphologies. The final quasi-static fracture is shown at the left of Figure 12a and is enlarged in Figure 12b. The damage associated with the fatigue processes can be observed on the right of Figure 12a and in detail in Figure 12c. The fracture surface presented ply-splitting behavior coinciding with the intralaminar crack in Figure 8b,f. The final fracture (left area) showed some debris that was distributed randomly and fibers that were partially detached. None of the analyzed particles in this region presented a roller morphology. The fatigue zone (right area) showed a worn surface that included particles and broken fibers dispersed along the surface. Most of the matrix particles were longer than those observed in the longitudinal specimens. Various particles showed the morphology of rollers. The surface characteristics coincided with a predominantly mode II fracture [55].

There are differences in the appearance of the fibers on both sides of the analyzed area. Those at the side dominated by the fatigue process were cleaner than the fibers corresponding to the final fracture. Other researchers observed this behavior with compressive and tensile fatigue [56,57]. They explained that cyclic loading leads to preferential damage at the fiber/matrix interfaces, producing a cleaner surface on the fibers.

The initiation region of the dominant delamination behavior could not be found during our analysis of the fatigue surfaces. It is important to highlight that the SBS specimens showed damage morphologies that present the characteristics of mode II fatigue and fracture in composites. Although the SBS test presents several limitations [22,34], it does produce, at least predominantly, interlaminar shear fatigue damage.

4. Conclusions

The noticeable effects of anisotropy and mean shear stress in short-beam shear (SBS) fatigue were evidenced, as well as the dependence of scatter on loading conditions and specimen orientation. In comparison, transversal specimens exhibited a lower performance than longitudinal ones, accompanied by more significant scatter. These tendencies were confirmed in the constant life diagrams (CLDs) constructed for SBS stresses.

Regarding the mean stress effect, ${\mathit{R}}_{\mathit{\tau }}$ = 0.1 was the most severe condition, leading to the shortest time-to-failure. In contrast, ${\mathit{R}}_{\mathit{\tau }}$ = 0.5 was the least severe condition. ${\mathit{R}}_{\mathit{\tau }}$ = 0.3 presented intermediate values of fatigue life. Nonetheless, this condition produced steeper S-N curves, indicating that the combined effect of mean shear stress and shear stress amplitude led to a higher rate of damage accumulation.

The fractographic analysis confirmed the Mode II damage dominance hypothesis, validating the test methodology employed. Although it was not possible to identify the nucleation region, the delamination growth direction analysis matched the macroscopic observations on the specimen’s surface during the test.