Microstructural Effects on Thermal-Mechanical Alleviation of Cold Dwell Fatigue in Titanium Alloys

Abstract

Cold dwell fatigue is a well-known problem in the titanium components of aircraft engines. The high temperature and low dwell stress of in-service conditions have been reported to give rise to dwell fatigue resistance through a thermal-mechanical alleviation process. Here, dwell fatigue tests at room temperature and the component operating temperature were performed on IMI834 titanium alloy to assess the microstructural effects on thermal-mechanical alleviation of cold dwell fatigue while eliminating the effect of chemical composition. The ratcheting strain rates under different loading conditions were quantitatively investigated to aid the understanding of thermal-mechanical alleviation.

1. Introduction

Titanium alloys are widely used in manufacturing high-stressed components of aero-engines due to their high strength-to-weight ratio and good corrosion resistance [1,2,3,4,5]. These titanium components have been found to suffer from cold dwell fatigue since the 1970s [6]. Serious lifetime reduction is observed when loading cycles contain a dwell period where the stress is held at a high magnitude. The increasing peak stress at the soft-hard grain boundaries induced by ratcheting strain accumulation is argued to contribute to the facet crack nucleation under dwell fatigue loading and cause early failure of key gas turbines components [7,8,9,10,11,12]. The ratcheting behavior under cyclic loadings of titanium alloys is related to their strong strain rate sensitivity (SRS) at near-ambient temperatures [13].

Experimental [14,15,16,17] and analytical [18,19,20,21,22] observations have shown that the dwell and strain rate sensitivity is associated with the microstructure of titanium alloys. The microstructural heterogeneity in dual-phase titanium alloys arises from the anisotropy of α and β phase properties. The rate-dependent properties of the two phases have been found to be remarkably different at near-room temperatures [23]. As a result, alloys with different morphology and texture show different rate sensitivity and, hence, dwell sensitivity. Shen et al. [24,25,26] have shown that titanium alloys with different microstructures have remarkably different fatigue crack resistance. Experimental work on dwell fatigue sensitivities of Ti-6Al-2Sn-4Zr-xMo (Ti-624x) alloys by Qiu et al. [27] found that the Mo content apparently had a large influence on the dwell fatigue life debit. However, both chemical composition and microstructure were different in the Ti-624x series alloys considered, although the alloys were designed to have the same nominal type of microstructure, but the average grain sizes and phase volume fractions were different. In addition, the atomistic simulations suggest that Mo content does not give rise to the observed differences in SRS of Ti-6242 and Ti-6246 alloys [28]. Hence, for the purpose of investigating microstructural effects on cold dwell fatigue, it is important and necessary to eliminate the composition effects of alloys.

On the other hand, the temperature has been found to significantly affect dwell-sensitivity by altering the strain rate sensitivity [29,30]. At near-room temperature, near-α Ti alloys (such as Ti-6242 and Ti-6Al) are known to demonstrate strong rate and dwell sensitivities. With the temperature increasing, the load shedding is found to peak at about 120 °C and progressively diminishes when the temperature increases to ≥230 °C. Other types of Ti alloys also show a similar trend of dwell sensitivity with increasing temperatures, but the most significant temperature may be shifted (e.g., the most dwell sensitive temperature for Ti-6246 alloy has been suggested to be about 300 °C) [30]. The diminution of SRS at in-service temperatures has been reported to give rise to resistance to dwell fatigue [31]. In addition, the hoop stress (primary loading) under in-service conditions has been found to be much lower than the macroscopic yield strengths of the alloy. The combinations of high operating temperatures and low operating stresses are argued to resist the facet nucleation by inhibiting plastic strain ratcheting due to cyclic loading, which is termed as thermal-mechanical alleviation [31].

The thermal-mechanical alleviation in IMI834 alloy with a bimodal microstructure has been systematically investigated using a dual-phase crystal plasticity model with α-β morphology explicitly represented [31]. The reduction in dwell sensitivity has been clearly demonstrated, but the microstructural effects on thermal-mechanical alleviation have not yet been fully addressed. Thus, the motivation of the present work is to explore the thermal-mechanical alleviation in four commonly observed microstructures of titanium alloys and to evaluate the microstructural effects on dwell sensitivity.

2. Materials and Methods

The near-α titanium alloy IMI834 was supplied as an ingot consisting of mainly equiaxed α grains (with an average grain size ≈40 μm) and some lamellar structures with thin β laths located near grain boundaries. Specimens of approximately 20 mm × 20 mm × 100 mm cuboids were cut from the ingot for heat treatment. Four different heat treatment routes were used to achieve the commonly observed microstructures: an equiaxed structure, a Widmanstätten structure, a bi-modal structure, and a trimodal structure. The hot working processes and the corresponding phase contents and microstructures are presented in Figure 1. The heat treatment was conducted in the air atmosphere. The equiaxed microstructure was obtained by thermal processing at near-β temperatures (~10–20 °C below β-transus temperature, $T_{\beta }$ ≈ 1055 °C for IMI834) for 1 h and then furnace cooled (FC) to room temperature. The resulting average grain size was ~50 μm. The Widmanstätten microstructure was achieved by heating up to ~15–30 °C above $T_{\beta }$ for 1 h, followed by furnace cooling. The bimodal microstructure was obtained by a two-step heat treatment: specimens were first heated to near-β temperatures and then were annealed at ~800 °C. This microstructure consists of equiaxed α grains (${\alpha }_p$) and transformed β matrix (${\beta }_t$). For the trimodal microstructure, there is an extra step before annealing: the temperature was kept at ~30–50 °C below $T_{\beta }$ for 1 h, which allows lamellar α grains (${\alpha }_l$) to grow from the transformed β matrix. The cooling conditions for both bimodal and trimodal were air cooling (AC). Scanning electron microscopy (SEM) examinations were carried out to characterize the morphology of each microstructure. The phase volume fraction was then calculated using ImageJ software [32]. Electron Backscattering Diffraction (EBSD) characterizations were conducted on all the obtained microstructures using a scanning step of 0.3 mm and an acceleration voltage of 20 kV. EBSD results on the four microstructures demonstrated a similar strong texture with the basal poles tilted ±30° from the normal direction (ND).

A set of stress relaxation tests were first carried out at room temperature (e.g., 25 °C) and at in-service temperature (e.g., ~350 °C [31]) to investigate the effect of temperature and microstructures on the SRS of the alloy. Tensile specimens with a gauge length of 40 mm and a diameter of 6 mm were machined with the loading direction parallel to ND, as shown in Figure 2. Specimens with different microstructures were uniaxially stretched up to 2 mm within 5 s, which results in a constant strain rate of $\dot{\epsilon }=1\times {10}^{-2} {\mathrm{s}}^{-1}$, and then held at the maximum strain for 60 s. Substantial stress relaxation is expected during the hold if the specimen experiences strong rate-dependent plasticity. In order to obtain reproducible results, the stress relaxations of each microstructure were repeated three times.

3. Results and Discussion

In order to isolate the effect of microstructure from the effect of chemical composition on dwell sensitivity, energy dispersive spectroscopy (EDS) was performed to compare the element distribution of different microstructures. By analyzing five main alloying elements, Al, Sn, Zr, Nb, and Mo, although there is segregation between the α and β phases at the microstructural scale, the overall composition was not changed after different heat treatment processes.

Figure 3 shows the engineering stress-time curves obtained at 25 °C and 350 °C, respectively, for the four microstructures. From the first 5-s uniaxial tension part, the 0.2 pct proof stress ${\sigma }_{0.2}$ can be extracted. The strain rate sensitivity exponent $m$ can be determined from the stress-drop during the strain hold period given by [33]

$$m=\frac{\mathrm{d} \ln \left(\sigma \right)}{\mathrm{d} \ln \left(-\dot{\sigma }\right)}.$$

(1)

The first 1 s of the stress relaxation data, where the stress-drop is the most significant, was used to calculate the $m$ value for all the specimens. Resultant proof stresses and SRS exponents are summarized in Figure 4a,b, respectively. The error bars in Figure 4a represent the scatters of the repeated tests. Although a higher scatter in the proof stress can be observed at the room temperature compared to 350 °C, the general trend of the strength between microstructures is unchanged. In addition, due to the relatively good repeatability during the stress relaxation for all the repeated tests, there is no noticeable difference (<0.0002) in the measured strain rate sensitivity exponent. Overall, for all the considered microstructures, both proof stresses and the $m$ values are lower at the high temperature as expected. The bimodal and trimodal structures show a higher strength under both temperatures, while the Widmanstätten microstructure has the lowest strength. Material strength is one of the aspects that affect the in-service performance, but, more importantly, the strain-rate-dependent plasticity is fundamentally related to the dwell fatigue life. The equiaxed and Widmanstätten structures are the most rate-sensitive, while the bimodal and trimodal are relatively less sensitive to strain rate. However, at 350 °C, all the SRS exponents are reduced to about 0.003, which means the strain rate sensitivity of IMI834 alloy is negligible at this temperature. The stress relaxation curves in Figure 3 also suggest that there is almost no stress drop at the high temperature. Other evidence in the literature has demonstrated that the rate sensitivity diminishes progressively at temperatures above 230 °C for some Ti alloys [29,30]. The diminishment of SRS at the high temperature was attributed to the energy barrier of the pinned dislocation from the obstacle under the considered loading rates being easily overcome of. Since the low strain rate regime where the SRS is predominated by the thermal activation process [34], a high deformation temperature can lead to a high frequency of dislocation escape from the pinned obstacle. As a result, the average dislocation velocity is much higher than the value to accommodate the applied deformation rate; hence, the variation in strain rates does not alter the flow stress of the material at high temperatures.

In order to investigate the microstructural effect on dwell fatigue, low cycle fatigue (LCF) and low cycle dwell fatigue (LCDF) tests were performed at room (25 °C) temperature on specimens with a 15 × 6 mm diameter cylindrical gauge section, as shown in the inset of Figure 5. The peak stress for both loading modes was identical and chosen to be 95% of the proof stress ${\sigma }_{0.2}$, as determined in Figure 4a, i.e., 900 MPa, 845 MPa, 931 MPa, and 923 MPa for the equiaxed, Widmanstätten, bimodal, and trimodal microstructure, respectively. The rise and fall time for both LCF and LCDF were chosen to be identical as 1 s. The dwell period in the LCDF tests was 60 s, and the stress ratio R in both loading modes was 0.

The accumulated plastic strain at the end of each loading cycle was recorded, as shown in Figure 5. The differing microstructures under LCF showed different plastic strain accumulations at 25 °C, as shown in Figure 5a. The plastic strains-to-failure for the Widmanstätten and bimodal structures was about 3%, while those for the equiaxed and trimodal structures were about 2%. On the other hand, the plastic strain accumulation under LCDF at 25 °C in Figure 5b displays both higher ratcheting rates and strains-to-failure. The results clearly demonstrate that all the considered microstructures experience a strong dwell sensitivity at room temperature. In order to investigate this phenomenon quantitatively, the LCF and LCDF lives and the dwell debit ($N_{LCF}/N_{LCDF}$) are summarized in Figure 5c. The maximum dwell debit is 28.66 in the Widmanstätten structure, followed by a value of 24.76 in the equiaxed structure. Generally, the fatigue responses in the bimodal and trimodal structure are similar to each other, but that for the trimodal is slightly less sensitive to dwell fatigue. Lei et al. [35] have also reported that the fatigue life of the trimodal structure of TA-15 alloy is slightly longer than that of the bimodal structure. Stroh [36] proposed a dislocation pile-up model to explain fatigue crack nucleation. The progressive build-up of dislocations within a favorably oriented grain under cyclic loading is argued to be responsible for the crack nucleation at the boundary with the neighboring “hard” grain. The mean-free path of dislocations in the trimodal microstructure is effectively reduced by the thick lamellar ${\alpha }_l$ grains. Hence, the peak stress achieved at the grain boundaries is lower, which makes the fatigue life longer. Discrete dislocation plasticity modeling [37] has demonstrated that thick laths within a lamellar structure can reduce the size of the pile-ups and change the SRS of the material, which then leads to different dwell fatigue susceptibility. The Widmanstätten structure in the present study was designed to have thick α laths and very thin β laths (≤2 μm). The average size of dislocation pile-ups is larger in the microstructure with a larger α grain size and thinner β laths. Thus, the resistance to dislocation motions provided the grain and phase boundaries is lower. The local resolved shear stress acting on the leading dislocation of each pile-up group is higher, which promotes the thermally-activated escaping from the pinned obstacle. As a result, the average velocity of dislocations is higher, and the strain rate sensitivity and dwell-sensitivity of the Widmanstätten and equiaxed structures are stronger.

The analysis of the fatigue responses has so far been carried out at room temperature and at relatively high stress. In order to understand the behavior of the different microstructures under in-service loading conditions, the LCDF tests were performed at a high temperature and low maximum stress. The thermal-mechanical loading histories of IMI834 rig components indicate that the first dwell period of a long haul flight consists of a stress hold at magnitude very much below the macroscopic yield stress and at a temperature of about 340–370 °C [31]. Hence, the temperature for the LCDF tests was chosen to be 350 °C, and the maximum stress chosen was 85% of the proof stress ${\sigma }_{0.2}$ at the corresponding temperature. The evolutions of the plastic strains are shown in Figure 5d. No specimen at 350 °C was found to fail under LCDF loadings; hence, the tests were terminated after 5000 cycles because of the very low strain accumulations. The dwell fatigue responses were totally different compared to the low-temperature scenario. The accumulated plastic strain was found barely to increase after the first loading cycle, and the maximum residual plastic strain after 5000 loading cycles was found to remain below 1.5% for all the specimens. The resistance to dwell fatigue through the thermal-mechanical alleviation process was clearly demonstrated, for the first time, in four different types of microstructure. The observed reduction in the dwell sensitivity under in-service loading conditions can be ascribed to two aspects: on the thermal side, the strain rate sensitivity was substantially reduced at the high temperature, while, on the mechanical side, the lower hold-stress made the amount of plastic deformation much lower. The former is argued to be the main reason for the observed thermal-mechanical alleviation in dwell fatigue. In addition, the maximum in-service stress can even be lower than 0.85 ${\sigma }_{0.2}$, but, due to the difference in the stress-multiaxiality and the temperature histories between the laboratory fatigue tests and the in-service rig spin tests, lower stress may not be able to trigger plasticity in the laboratory LCDF tests. Although low applied stress can effectively reduce the amount of plastic deformation and prolong the dwell fatigue life of the material, fundamentally, it does not change the sensitivity to rate or stress-hold, and the residual plasticity strain is still accumulated with the cycles. Hence, the reduction in the maximum applied stress cannot be ignored in the thermal-mechanical alleviation observed here but only plays a secondary role.

Comparing the accumulated plastic strain evolution curves under differing loading conditions in Figure 5, there are three main ratcheting behaviors observed: (1) a relatively low ratcheting strain rate under low-temperature LCF; (2) a high ratcheting strain rate under low-temperature LCDF; and (3) a very low ratcheting strain rate under high-temperature LCDF. These three types of ratcheting behaviors in fatigue have been recently reviewed [38]. The ratcheting strain rates for different microstructures under different loading conditions are extracted from the steady stage of the plastic strain accumulation curves and are summarized in Figure 6. Broadly, the ratcheting strain rates show similar trends to the fatigue life of the specimen: a higher ratcheting rate leads to faster strain accumulation, which eventually causes plastic instability and shorter fatigue life. In particular, the equiaxed structure has an LCDF ratcheting rate about 110 times higher than that under LCF. Nevertheless, the ratcheting rates under in-service (elevated temperature) conditions for all the considered microstructures are reduced to a negligible level where no obvious additional ratcheting strain accumulation was noticed after the initial few cycles. Since the energy barrier to thermally activated dislocation escape is significantly reduced with higher temperature, a stable dislocation structure is established at the early stage of the loading. In addition, the applied stress is lower compare to the room temperature LCDF, and there is no driving force for new dislocations to be nucleated, so the alloy is no longer sensitive to strain rate or dwell loading, regardless of the microstructure.

4. Conclusions

In summary, stress relaxation tests were carried out on IMI834 alloys with four commonly observed microstructures at room temperature and 350 °C. The strain rate sensitivity at room temperature is highly microstructure-dependent but is reduced significantly at elevated temperature and becomes nearly independent of microstructure in this alloy. The low cycle fatigue and low cycle dwell fatigue results at room temperature suggest that all the microstructures experience a strong dwell sensitivity. The LCDF tests under in-service loading conditions demonstrate the thermal-mechanical alleviation in all the considered microstructures. The dwell fatigue lives under in-service loading conditions are much longer than that at room temperature, indicating temperature and the applied stress play a key role in cold dwell fatigue.