Fatigue Characterization of SiC/SiC Ceramic Matrix Composites in Combustion Environment

Ceramic matrix composites (CMCs) are candidate materials for future propulsion systems due to their low weight, high temperature capability, and high strength at high temperatures [1]. Despite many advantages offered by CMCs, one disadvantage is that these materials are prone to oxidation and surface recession in a high water vapor containing combustion environment which limits their application at high temperatures. Over the years many studies [2–8] have been conducted to understand and characterize the oxidation behavior of CMCs under stress at high temperatures. These studies were performed under isothermal furnace conditions ether in dry or humid environment which cannot simulate the conditions experienced by the gas turbine hot section components. Only a few studies [9–14] have been conducted where simultaneous mechanical and combustion loading was applied to CMCs. Hot section components of the jet engines such as vanes, combustor liners, and shrouds experience some mechanical loading along with thermal gradient stress. To successfully implement CMCs in jet engines for improved thermal efficiency one must characterize these materials under similar engine conditions which requires the development of a special experimental facility.

Recent studies [10,12,14] in the combustion environment revealed that fatigue life of the CMCs was significantly lower in combustion environment in comparison to the isothermal furnace environment. The decrease in the fatigue life was attributed to the presence of high moisture content and thermal gradient stress.

Electrical resistance (ER) is a structural health monitoring technique to monitor the damage in the composite at room temperature and high temperature. Several researchers [15–21] have extensively investigated the ER behavior of the melt infiltrated (MI) SiC/SiC composites under various loading conditions at room and high temperatures to understand the damage initiation and propagation. ER in MI systems is more sensitive to matrix damage since the matrix contains free silicon—which is the majority current charge carrying phase in the composite.

The main aim of this study was to further a special experimental facility where simultaneous mechanical and combustion loading can be applied on coupon level CMCs more toward the conditions experienced by the hot section components of the jet engine. Two different MI SiC/SiC CMCs were investigated to understand the effect of the combustion environment on fatigue life and to understand the relevant damage mechanisms in the combustion environment. Several fatigue tests were conducted at a stress ratio of 0.1, frequency of 1 Hz, and a specimen surface temperature of 1200 °C at different stress levels. For comparison similar tests were performed in furnace environment. Specimens that achieved runout condition were subjected to room temperature monotonic tension testing to evaluate the residual properties. Post-test microscopy was performed under scanning electron microscope (SEM) to understand the oxidation embrittlement and damage mechanisms.

A special experimental facility was developed at The University of Akron where coupled mechanical and combustion conditions can be induced on the CMC coupons. A horizontal hydraulic MTS machine (MTS 810 systems) was used for the mechanical loading and a high velocity oxygen fuel (HVOF) gun (HP 2700, Plasma Powders and systems) which uses propane as fuel and oxygen as oxidizer was used for the combustion conditions. The experimental setup is shown in Fig. 1. This combustion facility is capable of generating a peak flame temperature of 2300 °C and a velocity of Mach 1 (930 m/s). Wide range of combustion conditions can be achieved by adjusting the flow rates of fuel and oxygen. For this study, the specimens were fixed in the MTS machine at an angle of 45 deg with respect to the flame. This was thought to be a worst-case scenario for a CMC as the as-machined edges are completely exposed to the high velocity high temperature flame. The specimen was heated on the front side, and the backside was left open for natural convection. A significant thermal gradient was observed which induces severe thermal gradient stresses. Two forward lean infrared cameras (FLIR A6700 SC) were used to monitor the front and back temperatures of the specimen. A representative temperature distribution of the specimen in the burner rig is shown in Fig. 2. The major combustion by products formed by burning of propane and oxygen are NO_x, H₂O, CO, CO₂, and excess O₂. The moisture content in the flame was estimated to be ∼30% by using chemical equilibrium with application software. Velocity of the flame was calculated using a high-speed camera by spraying chrome carbide particles with the flame. The entire experimental facility was setup in a soundproof room as the flame generates excessive noise. Compressed air was provided to the system to prevent the nozzle from overheating. To protect the grips and wedges of the MTS machine steel plates with thick ceramic insulation along with recirculated chilled water was used. A lean burn condition (equivalence ratio of 0.83) which would be similar to the operating environment of a turbine engine was selected for this study. To achieve this lean burn condition, the pressures and flowrates of fuel and oxidizer were selected as 80 psi, 55 slpm (standard liters per minute) for propane, 150 psi, 300 slpm for oxygen and 75 psi, 400 slpm of compressed air. Further details are provided in Panakarajupally et al. [14].

The ER leads were attached to the specimen by using a high temperature silver paste along with high temperature wires. Epoxy glass tabs were used on the ends of the specimen to prevent the current leaking to the grips. The leads are placed inside the grips to protect the wires and silver paste form excessive heat generated by the open flame heating.

The CMCs tested in this study were fabricated using melt infiltration process by Goodrich corporation [2]. Two versions of MI systems differed by the fiber type were tested. The first consists of Hi-Nicalon type S fibers (Nippon Carbon, Tokyo, Japan) and the second version consists of Tyranno SA fibers (Ube Corporation, Tokyo, Japan). The fibers were woven into a cloth with five-harness satin weave. The fibers were then coated with BN interphase and a chemically vapor infiltrated (CVI) SiC overcoat. A SiC-containing slurry was then infiltrated into the open porosity of the composite and dried followed by molten Si infiltration to densify the matrix. Microstructure of the two systems is shown in the Fig. 3. All the specimens considered had been machined into dog bone specimens (∼10 mm wide gauge section), the general properties are shown in Table 1. Proportional limit was determined based on 0.005% offset strain method (ASTM C1275-15).

The specimens considered in this study were subjected to tension–tension fatigue loading at a selected combustion condition (Table 2). Prior to heating, the specimens were fixed in a load-controlled mode with a minimum load of 50 N to prevent the specimen experiencing compression due to thermal expansion of the material. The flame was then started at a distance far away from the specimen to prevent the thermal shock. The flame was advanced toward the specimen slowly, and the specimen surface temperature was monitored with the FLIR camera. Once the target temperature was achieved the gun was locked at that position and the fatigue loading was initiated after 3 min to allow the temperature to stabilize. Fatigue loading was applied in steps. Initially, the specimens were ramped to mean load, fatigued at 0.1 Hz for five cycles, 0.5 Hz for five cycles, and 1 Hz for 2 to 4 h. The specimen was then unloaded and cooled to room temperature. The procedure was repeated until either failure occurred, or runout was attained. The runout condition in the burner rig was set as 24 H (86, 400 cycles). If a specimen survives 24 H, then it was subjected to uni-axial monotonic tension testing for residual properties at room temperature.

Fatigue tests were conducted in an isothermal furnace at a temperature of 1200 °C and fatigue conditions the same as those for the burner rig tests. A resistive furnace is used to apply the temperature condition (∼25 mm isothermal hot zone length). The specimens were ramped at a rate of 40 °C per minute. Runout condition in the furnace environment is set as 100 H (360,000 cycles), and the specimens achieved runout condition were subjected to monotonic tension testing.

Applied stress versus cycles to failure for the specimens fatigued in burner rig and furnace environment is shown in Fig. 4. Fatigue life increased with decreasing stress in both environments. The furnace environment had significantly higher fatigue life in comparison to the burner rig environment in both MI systems indicating the more deleterious environment present in the burner rig. Comparing the Hi–NiC S (HNS) and the Tyranno SA systems, the SA system exhibited significantly higher fatigue life in the burner rig and furnace environments. Presumably, this was at least in part due to the higher fiber volume content of the SA composites (Table 1) which also accounts for the higher PL and ultimate stress properties of this composite panel [22].

In addition to velocity and water vapor content, burner rig specimens experience significant axial and through-thickness thermal gradients. Figure 5 shows representative thermal gradients for the two specimens. A thermal gradient of 175 °C and 100 °C was observed for the HNS specimen and SA specimen, respectively. Since SA specimens considered in this study are thinner compared to the HNS, they experienced lower thermal gradients. A simplified approximation of through thickness thermal stress based on thermal gradients and modulus of these composites (Table 1) estimates a net stress of 79 MPa for the HNS specimen and 47 MPa for the SA specimen. Addition of this thermal gradient stress to the applied stress would increase the overall net stress at some locations which induces significant matrix cracks leading to faster degradation and rapid failure.

Runout condition of 24 H was achieved in burner rig at a stress of 85 MPa for Tyranno SA system and at a stress of 70 MPa for Hi–NiC S system. A 100 H runout was achieved for furnace tests at a stress of 125 MPa for Tyranno SA system and at a stress of 85 for Hi–NiC S system.

The ER behavior in MI systems are more sensitive to matrix damage, i.e., matrix cracking, since the free silicon in the matrix is the most conductive phase in the composite. Matrix cracking results in an increase in ER. Several studies [15–21] have been conducted to understand the ER behavior of MI SiC/SiC CMCs under various loading conditions at high temperatures and room temperature. High temperature ER behavior is complicated by the intrinsic temperature dependence of resistivity of SiC/SiC composites.

The measured resistance during burner rig fatigue is shown in Fig. 6 for Hi–NiC S composites. Resistance (R) increases with heating due to increasing resistivity of free silicon with temperature. For MI composites, R increases under isothermal heating up to about 900 °C [23]. In a burner rig environment, due to the large thermal gradient (Fig. 5), the volume of material that is at the peak temperature was much smaller compared to the isothermal condition and relegated to the surface side of the specimen. Upon heating, resistance continuously increases until the peak surface temperature is achieved which was different than what was measured for isothermal furnace conditions as will be shown below.

For the 100 MPa HNS fatigue test (Fig. 6(a)), the time for mechanical loading is indicated in the figure. It is not clear why the dip and then rapid increase occurred upon loading. However, following the initial loading, little increase in ER was observed until failure for this short time test. Presumably, the increase in ER with loading is due to initial matrix cracking exacerbated by the thermal gradient at this higher stress condition for HNS. For the 85 and 70 MPa, more significant increases in resistance were observed during fatigue cycling after some time (Figs. 6(b) and 6(c), respectively). For 85 MPa (Fig. 6(b)), there was a total change in resistance of over 2 Ω after loading with distinct increases in resistance occurring at about 13,000 and 25,000 s. After both distinct increases in resistance, the rate of increasing resistance markedly increased. This appears to indicate a progression of matrix damage leading up to failure. The 70 MPa test (Fig. 6(c)) had only about a total of 0.4 Ω increase in resistance over the entire test and resistance increased much more gradually. This specimen did not fail during burner rig exposure. However, the rate of increasing resistance does increase with time, especially near the end of the test indicating some degree of matrix cracking.

The measured resistance during burner rig fatigue for SA composites is shown in Fig. 7. The shortest time, highest stress level (150 MPa peak stress—Fig. 7(a)) had sharp increases in resistance approaching the end of the test. Interestingly, the lowest stress (85 MPa peak stress—Fig. 7(b)) run out test had multiple distinct increases in resistance followed by more gradual resistance increase. Again, there must have been significant matrix crack events to cause the increases in resistance. The resistance plot for the 100 and 125 MPa fatigue specimens are not shown due to erratic resistance measure from lead attachment issues.

Figures 8 and 9 show the measured resistance for isothermal furnace tests of HNS and SA composites, respectively. The overall trend was similar for both the systems. With heating, resistance increased up to 900 °C followed by decreasing resistance up to the 1200 C peak temperature. This is most noticeable for the shorter time tests in Figs. 8(a) and 9(a) and has been observed for MI in other studies [23]. The temperature dependence for ER has been explained by the change in charge carrier concentration from extrinsic to intrinsic in silicon at the saturation temperature [24]. The reduction in resistance with temperature did not occur in the burner rig because the temperature gradient created a smaller volume at the higher temperatures and a large volume of the specimen at intermediate temperatures. For the isothermal tests, an axial temperature gradient exists, albeit more gradual than the burner rig, but there is no through-thickness gradient. Therefore, at a given set temperature, the largest volume of material at a given temperature would be the highest temperature and would have more influence on the total measured resistance across the entire temperature gradient.

Unlike the burner rig resistance behavior, no sharp increases in ER were observed for the HNS or SA specimens in the furnace environment during fatigue at temperature. There was an increase in resistance an hour prior to failure for the 100 MPa HNS test (Fig. 8(a)). SA specimens fatigued at 175 and 150 MPa (Figs. 9(a) and 9(b)) showed a jump in resistance as soon as the loading initiated with gradual increasing ER after. Presumably matrix cracking occurred upon loading at the higher stresses.

Specimens that survived 24 H of testing in burner rig environment and 100 H in isothermal furnace environment were subjected to room temperature monotonic tension testing to evaluate the residual properties. The stress–strain plots are shown in Fig. 10 and residual strength values of runout and as-produced specimens are shown in Table 3. The strength and strain to failure of the runout specimens degraded significantly in both environments. The HNS specimen fatigued at 70 MPa peak stress in the burner rig retained a strength of 201 MPa. The strength degradation of the SA composite after burner rig runout was about 15% whereas the degradation in HNS composites after burner rig runout was 24%. The strength degradation of the HNS composite after furnace runout however was significantly greater, about 40%. The furnace runout condition achieved stresses only marginally above the proportional limit for the HNS composite. Most interesting, the HNS composites showed much lower strains above the proportional limit of the as-produced specimen (123 MPa) and failure at a relatively low strain. For these composites, matrix microcracks must have formed at the lower stresses and grew during high temperature fatigue resulting in the strong bonding of fibers to one another and the matrix. The large strain accumulation for the stress–strain curve of the as-produced HNS specimen implies very low interfacial shear strengths. This would encourage greater oxidation kinetics for within the crack due to relatively larger crack openings and an increased driving force for matrix crack growth under stressed-oxidation conditions. After thermomechanical exposure, room temperature mechanical behavior would be “stiffened” due to the increased bonding between fibers and matrix in the matrix crack as is evidenced in the fracture surface analysis below.

Specimens failed in the burner rig environment and isothermal furnace environment were observed under scanning electron microscope to understand the oxidation behavior and its role on the fatigue life. The SEM images of the burner rig fractured specimens are shown in Figs. 11 and 12. Two clear distinct regions are observed on the fracture surface oxidized and unoxidized (separated by a white dotted line). The oxidized regions revealed complete oxidation of matrix, interphase, and fibers cladding to the matrix whereas some fiber pullout is observed in the unoxidized regions.

Fracture surface of the Hi–NiC S specimen fatigued at 100 MPa stress level in burner rig (Fig. 11(a)) showed some fiber pullout whereas the specimens fatigued at lower stress levels (85 and 70 MPa (Figs. 11(b) and 11(c)) showed complete oxidation embrittlement on the fracture surface. Fatigue life increased with decreasing stress. This suggests that the specimen exposed to burner rig environment for longer durations experiences oxidation within at least one crack or multiple cracks that link up compared to specimens at higher stress and shorter failure times. Fiber pullout on one end and complete oxidation on the other end of the fracture surface (Fig. 11(a)) suggests that cracks may have originated from the as-machined edges as they were directly exposed to the high velocity combustion flame [14]. Additionally, HNS fracture surface revealed multiple large interior voids. Voids can act as stress concentrators where cracks may originate leading to much faster degradation of the composite.

Fracture surface of the SA specimens (Fig. 12) showed similar oxidative features as HNS specimens. Edges of the SA specimens were completely oxidized which indicates the effect of angular impingement of flame on the test coupons. The test coupons were mounted in the horizontal MTS at an angle of 45 deg where the top and bottom edges are completely exposed to high velocity high temperature flame making them more prone to oxidation. A macro-image of the SA specimen survived 24 H in burner rig environment is shown in Fig. 13.

Fracture surface of the SA and HNS specimens in furnace environment is shown in Figs. 14 and 15. SA specimen fatigued at higher stress level, 170 MPa revealed complete oxidation of fracture surface and a little fiber pullout was observed for the specimen fatigued at 150 MPa stress. For these composites, stress applied was beyond the matrix cracking stress which induces significant matrix cracks leading to oxidation and final fracture. Similar behavior was observed for the HNS specimens where little fiber pullout was visible for 100 MPa fatigued specimen and complete oxidation for 85 and 70 MPa stressed specimens. Evidently crack(s) formed and propagated at these low stresses for the runout specimens (including in the burner rig condition) so as to enable oxidation embrittlement through the cross section.

Applied stress versus cycles to failure for the Hi–NiC S and SA systems revealed significant reduction of the fatigue life in burner rig environment compared to the isothermal furnace environment when stressed at similar loading conditions and temperature. The decrease of fatigue life could be attributed to several factors present in the burner rig environment such as (a) thermal gradient stress, (b) moisture content, and (c) velocity of the high temperature gases which is especially deleterious to exposed cut edges of the composite. For both systems considered in this study, applied stress was well below the proportional limit stress in burner rig environment. One would expect a runout condition at these stress levels. However, runout was achieved only at 70 MPa for Hi–NiC S specimen and 85 MPa for SA specimen. Directional heating of burner rig specimens induces a significant thermal gradient stress. Steady-state calculations estimated a thermal stress of 79 MPa for Hi–NiC S specimen and 47 MPa for SA specimen. This thermal stress in combination with the applied fatigue stress would be expected to cause additional matrix cracking allowing oxidation of fibers, interphase, and matrix within a matrix crack resulting in strong bonds between the fibers and matrix resulting in a reduced failure stress.

Lower fatigue life in the burner rig environment could also be partly attributed to high water vapor content present in the burner rig. It was estimated from chemical equilibrium analysis software that approximately 30% of water content exists in the burner rig due to primary and secondary combustion of fuel and oxygen. Moisture content in the furnace environment was estimated to be approximately 2%. The presence of high moisture content in the burner rig may accelerates the oxidation of matrix and interphase in the presence of matrix cracks leads to faster degradation of the material and reduced fatigue life.

Although, it was not explored in this study, velocity might be an additional factor which makes the hot gases penetrate deep into the composite system causing degradation of the composite.

A special experimental facility is setup to test the MI woven SiC/SiC CMCs under representative turbine engine conditions. Two versions of the MI systems were considered in this study. First version consists of SiC matrix, BN interphase and reinforced with Hi–NiC S fibers. The second system consists of SiC matrix, BN interphase, and reinforced with Tyranno SA fibers. All the specimens were subjected to a tension–tension fatigue loading at different stress levels with a stress ratio of 0.1, frequency of 1 Hz, and a specimen surface temperature of 1200 °C. Similar tests were performed at the same temperature and fatigue conditions in isothermal furnace for comparison.

Stress versus cycles to failure indicates higher fatigue life for the furnace environment compared to burner rig environment at the same stress level indicating the drastic effect of the combustion environment. The SA composite had higher fatigue life at a given stress level than the HNS composite which was attributed to the higher fiber volume fraction. Residual strength results suggested significant strength reduction in both furnace and burner rig environments. Microscopy images revealed little fiber pullout and significant oxidation on the fracture surface for both material systems in burner rig and furnace environment. Directional heating of the burner rig specimens indicated significant thermal gradient stresses for both systems which causes significant damage to the material reducing fatigue life.

Electrical resistance showed some promising results to monitor the damage in the combustion environment. The results revealed that ER was very sensitive to temperature fluctuations. Most tests showed an increase in resistance during fatigue cycling which might be due to some damage development. Although, ER results were promising, more in-depth study needs be performed to isolate the individual effects of oxidation and fatigue loading in the burner rig.

The authors would like to thank Dr. David Shiffler, Office of Naval Research for funding this project under the Grant No. N00014-18-1-2646 (Funder ID: 10.13039/100000006).