3.1. Mechanical Response under Static Tension
The stress–strain response of notched and dogbone SiC/BMAS specimens under cyclic and monotonic tension, respectively, is presented in
Figure 1. Un-notched specimens exhibited a triple regime behavior consisting of a linear initial part followed by a regime of gradually decreasing tangent modulus and a final regime of apparent stiffening. The second regime (regime “II”) is associated mainly with interfacial damage, most importantly interfacial debonding, but also with progressive matrix cracking evidenced as decreasing material stiffness (average slope of unloading/reloading loops). In the third regime (regime “III”) an increase in material stiffness and tangent modulus coupled with an almost linear stress–strain relationship are apparent. In this ultimate regime, the mechanisms of interfacial debonding and matrix cracking have reached a saturated state, hence material damage is not governed by the interface or matrix anymore, but by a mechanism of superior strength, essentially load bearing by intact fibers [
15]. Similar triple regime phenomena with prefailure macroscopic stiffening and linear stress–strain relationships have been encountered before [
16]. As observed in the curves of the bottom graph of
Figure 1, Regime III is absent from the mechanical behaviors of notched specimens. This is probably due to premature fiber—hence also composite—failure stemming from stress concentration gradients in the vicinity of the notch roots. In notched specimens instead, the material failed soon after the maximum load was attained, giving minimal “tail” effects.
Most importantly, the stress–strain curves of un-notched specimens defined with unique precision have a common intersection point (CIP) of unloading–reloading curves in the first quadrant of the stress–strain curve in the tension domain. The coordinates of the CIP, 0.001 strain and 90 MPa stress, are directly related to the axial residual stress state of the composite [
17,
18]. While a thorough analysis of the CIP feature for the particular composite has been the subject of a previous work [
19], it is interesting to repeat here that a self-assembled CIP has never before been encountered experimentally.
Comparing the monotonic and cyclic tension curves for the SiC/BMAS composite (top graph of
Figure 1), it can be concluded that cyclic loading results in an increase by 20% in attainable material stress, calculated at fracture. If this increase is due to higher amounts of energy dissipated at damage mechanisms such as interfacial debonding, matrix cracking, and load bearing by intact fibers [
20], it is then suggested that cyclic loading by itself improves the energy dissipation capacity of the material. The existence of another energy dissipation mechanism, pull-out, anticipated by the weak interfacial bond discussed in the experimental section, was verified after the end of the tests: Failed specimens had not separated in two pieces after removal from the grips, with the frame still indicating small load values of the order of a few Newtons. This meant that fibers had failed within the matrix environment and had pulled-out noncompletely before removal of the specimens from the grips [
21].
Composite strength and modulus appeared to increase with decreasing notch length. Un-notched specimens enjoyed average strengths and moduli of 355 MPa and 151 GPa, respectively. The corresponding values for the 0.2 and 0.35 notched-to-width length were 280 MPa/119 GPa and 270 MPa/108 GPa, respectively.
3.2. Static/Cyclic Loading
Temperature variation, ΔT, measured by IRT and instant load are plotted in
Figure 2i as a function of time for a DEN specimen with a 0.35 notch-to-width ratio. Indices “a” through “e” denote the instances of the thermographs shown in
Figure 2ii, collected at the notched ligaments of the composites. It is observed that peak ΔT location coincides with the location of maximum load for every cycle, whereas peak ΔT magnitude increases with progressing loading, hence also material damage. At the ultimate cycle, the ΔT trace appears to follow a completely different pattern to those in previous cycles, wherein temperature appears to drastically increase, indicating that the specimen is heading for catastrophic fracture.
In
Figure 2ii, the locations of crack initiation, as identified by IRT, are indicated by a red circle mark. The apparent high-temperature area located outside and to the right of the circle mark is a baseline pattern that exists even before load application and remains constant until catastrophic failure. It is associated with the specimen’s surface emissivity, not with material damage. It should not be ignored that IRT is concerned with temperature variations as a result of progressive damage, not with absolute values. Under this rationale, no noticeable change in temperature is seen up to 32 s experimental time (
Figure 2ii.a). Within 73 s of testing (
Figure 2ii.b), very small temperature variations can be observed within the marked (circle) area. Temperature increases become more obvious in
Figure 2ii.c,d, 135 s and 209 s into loading, respectively. It is indicated that damage is extending in area and magnitude. In the last loading cycle, 285 s in the test, a dramatic increase in temperature throughout the whole notched ligament signifies that material failure is imminent.
The thermographic behavior within the ultimate loading cycle of the same 0.35 notch-to-width ratio specimen is demonstrated in
Figure 3 and analyzed in the following. Indices “a” through “h” presented in
Figure 3ii define the instances of the thermographs in
Figure 3i (note the notch roots). In
Figure 3i.a, no heat appears to exist within the notched ligament. In the next instance,
Figure 3i.b, a red arrow indicates what appears to be crack initiation. At 285 s,
Figure 3i.c, a significant temperature difference is observed, which coincides with the change in the slope of the mechanical response curve. It is believed that from this instance on, subsurface cracks start propagating from the left notch root with direction to the right. It is important to establish this instance as precisely as possible, as this will facilitate early prediction of the final fracture. It is noted that the associated time (285 s) corresponds to 73% of the total duration of this ultimate loading cycle. Five seconds later (
Figure 3i.d at 290 s) the subsurface crack appears to span half of the notched ligament, while only another 300 ms later (
Figure 3i.e) it propagates abruptly and unstably towards the right notch. The maximum temperature is attained (
Figure 3i.f) at an instance that coincides with the maximum load of the final loading cycle. This temperature is associated with the matrix cracking saturation and the load-bearing of the reinforcing fibers. Temperature starts decreasing at the left notch in the next
Figure 3i.g, while it increases above and under the subsurface crack, as indicated by black arrows. This increase is due to the fiber failure under the critical level of applied load. Failed fibers pull-out, giving rise to the frictional thermal energy [
20,
22] evidenced in
Figure 3i.h.
Similar trends were observed for DEN composites with smaller notches, as in specimens with 0.2 notch-to-width ratios; the thermographic behavior within the ultimate loading cycle of such a specimen is demonstrated in
Figure 4 and analyzed in the following. Again, indices “a” through “h” presented in
Figure 4ii define the instances of the thermographs in
Figure 4i. After 600 s of testing time,
Figure 4i.b, no “warm” damage areas are seen in the thermographs. Crack initiation appears at 680 s,
Figure 4i.c, which compares favorably with the instance of slope change in mechanical behavior of the material. It is hence possible to foresee early fracturing at 80% of the final cycle duration. In the subsequent thermographs
Figure 4i.d,e, the subsurface crack propagates from the left notch towards the middle of the notched ligament. The crack then propagates abruptly and unstably towards the right notch. A temperature variation profile compatible with pull-out is seen again in the last thermograph,
Figure 4i.h, where the specimen has failed completely.
Variation of ΔΤ is shown in
Figure 5 for the two notch lengths used in the current study. It is observed that temperature maximizes, from the baseline value, in a vertical manner and then gradually decreases. The steep maximization is due to heat release from the sudden ultimate failure of the specimen, which is the most prominent damage mechanism hence emitting the maximum heat, while the gradual decrease is due to natural dissipation of the released heat to the environment. Specimens with shorter initial notches, hence higher gauge length volumes, exhibit peak ΔΤ (increase of ΔΤ taken from the baseline value) of 12 °C at fracture, while the ones with longer notches exhibit lower peak ΔT values around 5 °C. It is believed that in specimens with larger notched ligaments (smaller notches), damage evolves over a wider material region throughout testing, hence peak temperature at the critical load is not high. On the other hand, damage is accumulated and relieved not so drastically in a specimen with less material available within the notched region.
3.3. Fatigue
Lock-in thermography was applied during fatigue loading of SiC/BMAS dogbone specimens. The intrinsically dissipated energy, as monitored by the IR camera for ten different stress levels ranging from 30% to 90% σ
UTS, is plotted as a function of % σ
UTS in
Figure 6.
Therein, the data can be visually categorized into two distinct groups depending on the intensity of the energy and the slope of the linear regressions which best approximate the data, as demonstrated in
Figure 6. In the first regime, pertaining to stress levels from 30% to 60% σ
ULS, the dissipated energy increases at a very low, almost constant rate, while from 70% σ
UTS and upwards, energy rises considerably more rapidly. Given the fact that thermal energy is a direct measure of damage accumulation in the material, the constant rate in the first regime signifies a minimal damage accumulation rate therein. Accordingly, the steep slope of data in the second regime indicates a high damage accumulation rate due to fatigue. Under the effect of cyclic stresses of the amplitudes pertaining to the second regime, the material is prone to fatigue failure. Hence, the transition from one behavior to the other defines the shift of damage accumulation from stable to fatigue failure. Correspondingly, the intersection point of the two regression lines defines the fatigue limit of the material [
9]. The value of fatigue limit calculated for the cross-ply SiC/BMAS through the thermographic approach of this study is 70% σ
UTS, or 205 MPa.
An examination of the thermographic pattern of cross-ply SiC/BMAS at different stress levels is of particular interest in view of the established fatigue limit value. This information is presented in
Figure 7a–j, wherein the 70% σ
UTS fatigue level which corresponds to thermograph 7f. Two distinct cases are made obvious by examination of this Figure: (i) Thermographs 7a–d depict low energies in cold (blue) color coding in the initial loading stages associated with minimal material damage and (ii) thermographs 7e–j capture progressive damage accumulation is captured by increasingly warmer colors (high energy). In the first four thermographs, up to 176 MPa applies stress, there is practically no appreciable change in the dissipated energy. In the fourth thermograph,
Figure 7, a slight change in color can be attributed to the saturation of elastic energy accumulation on the onset of appearance of fatigue. A totally dissimilar energy distribution pattern appears in thermograph 7f due to the unfolding of internal energy dissipation phenomena such as interfacial damage, delamination, and fiber sliding across the debonded interface [
23]. In thermographs 7g–7j (75–90% of σ
UTS), a raise in energy can be noticed, as indicated by the increment of the magenta spots until fracture. At a particular loading level, this energy can be attributed to fiber bridging, fiber failure, and pull-out.
3.4. Crack Growth
Measurement of crack growth in SiC/BMAS composites was made possible by analyzing the thermal energy dissipated by the material during cyclic loading with unloading/reloading loops, in terms of temperature variation. Information originating only from the area in front of the notch root, termed the “reference zone”, where crack growth and damage were anticipated to concentrate, was considered. For reference purposes in the forthcoming analysis, the mechanical response of CMCs under cyclic loading is presented in
Figure 8.
Initially, the maximum temperatures (relating to the maximum damage) were identified and plotted as a function of time as per
Figure 9, wherein crack opening displacement (right axis) and cycle count (number in red) are included to facilitate perception of the loading instance. It is observed that temperature variation exhibits periodical local maxima which are prominent after 1000 s of testing time. Using the thermal camera’s native analysis software, the reference zone was divided into a large number of 0.226 mm wide subzones by introduction of neighboring vertical lines, called “control lines”. Knowledge of the exact time instance where these peak temperatures occur, from
Figure 9, and realizing that these maxima originate from crack growth arriving at a specific control line, one can identify the exact corresponding control line, and hence derive the crack length.
Crack lengths determined based on the aforementioned IRT approach are plotted versus testing time in
Figure 10 (hollow cycle symbols). It is therein observed that the initial five loading/unloading cycles do not induce crack formation or growth. A 0.45 mm long crack is eventually formed at the 6
th cycle, approximately 600 s into loading. In the subsequent 500 s, cycles 7 and 8, the crack does not appear to propagate further. Extensive but stable crack growth appears to commence after the eighth cycle, ca 1000 s into loading. Within the subsequent 1000 s, or four cycles, crack length eventually reaches 9.3 mm, or
ca. 37% of the specimen’s un-notched ligament, before ultimate failure. It is noted that in the particular specimen configuration crack growth can only span a fraction of the un-notched ligament of the material as a result on the compressive forces that develop at the specimen’s back face. Plotted alongside the experimentally measured (IRT) crack growth in
Figure 10 is the theoretically expected counterpart based on the conventional elastic compliance technique [
24]. A wide discrepancy between experimental values and theoretical predictions is observed, especially within the initial nine cycles where actual crack growth is not significant. This discrepancy is due to the shortcoming of the analytical method to approach compliance changes independently of crack growth. Contrary to the theoretical approach, where compliance drops are considered only as results of crack growth, in a CMC, such decreases can also be promoted by mechanisms such as high stress concentration at the crack tip. Moreover, in view of the brittle nature of the glass–ceramic matrix, crack growth is not expected to follow the highly stable, almost linear, trend predicted by theory. Under the same rationale, within the last three unloading/reloading stages related to stable crack growth, the experimental values appear to follow the theoretically predicted trend, albeit with an overestimation in actual crack length by 25%.
Figure 11 represents typical thermographs collected at indicative successive stages of crack growth. Warmer colors correspond to higher amounts of thermally dissipated energy, representing areas of higher damage accumulation as opposed to colder colors. Violet color, corresponding to the maximum temperature, is attributed to crack growth. The onset of cracking is barely seen in
Figure 11a, 1100 s into testing. As the crack propagates, the damage zone expands radially due to stress transfer from the high stress area to the surrounding material but also due to the unfolding of complementary damage mechanisms such as interfacial damage, fiber sliding along the debonded interface, fiber bridging, and pull-out [
14].
Figure 11e, collected just before composite failure, represents the maximum damage span on the specimen. Using image analysis, the contours to the maximum temperature zones found at the crack tip (violet color) can be reconstructed as shown in
Figure 11f. In this manner, the shape and size of the maximum-damage zone and its evolution with time during composite fracture can be visualized.