Experimental investigation of fatigue crack growth behavior of GH2036 under combined high and low cycle fatigue

Fatigue crack growth rates have been experimentally determined for the superalloy GH2036 (in Chinese series) at an elevated temperature of 550 (cid:1) C under pure low cycle fatigue (LCF) and combined high and low cycle fatigue (CCF) loading conditions by establishing a CCF test rig and using corner-notched spec-imens. These studies reveal decelerated crack growth rates under CCF loading compared to pure LCF load-ing, and crack propagation accelerates as the dwell time prolongs. Then the mechanism of fatigue crack growth at different loadings has been discussed by using scanning electron microscope (SEM) analyses of the fracture surface.


Introduction
Turbine discs are important load-bearing components in modern aero-engines, whose structural integrity is critical to engine performance. Generally a turbine disc is connected with turbine blades by means of a fir-tree design [1], i.e., a typical multiple load path structure, to allow for different rates of expansion between the disc and blade while still holding the blade firmly against centrifugal loads. During operation, turbine discs experience significant tensile stresses due to centrifugal loading and thermal loading. In addition, they may also experience aerodynamically induced small-amplitude and high-frequency vibrations. The centrifugal and thermal stresses can cause turbine discs to fail by low-cycle fatigue (LCF), and meanwhile the aerodynamically induced vibration stresses can lead to high-cycle fatigue (HCF) failures [2][3][4]. This combined loading on the gas turbine disc posts serious reliability concerns and thus requires careful investigation.
Damage-tolerant design approach has been successfully used to reduce fatigue failures and to extend lifetimes of critical components [5]. In this approach, the predicted lifetime is calculated by integrating the da/dN versus DK curve. This approach is also shown to work well under combined high and low cycle fatigue (CCF) loading [6]. Fatigue crack propagation behavior due to the combined action of LCF and HCF is therefore seen to be of broad interest [4]. Several studies on titanium alloys and steels stated that the HCF cycles lead to accelerated crack growth rates in comparison to pure LCF loading as soon as the total stress intensity DK range overcomes a certain threshold DK onset [4,[6][7][8][9][10]. Meanwhile, the influence of the number of HCF cycles per LCF cycle on fatigue crack growth was investigated systematically. While the conclusion that the superimposed HCF loadings accelerate the growth of fatigue crack and thus decrease the fatigue life has been well drawn in literature, several studies on some certain superalloys, which were tested under CCF loading, showed that HCF cycles with small amplitudes increase the number of cycles to failure instead [11][12][13][14]. The available literature data in [11,13] indicated that the superposition of high frequency low amplitude vibrations on LCF loading may be both negative and positive as regards the number of LCF cycles to failure. It was dependent on the ratio of the amplitudes of HCF to LCF loading. When the ratio was large enough, shorter lifetime was observed as expected. However, for small ratio the effect of HCF amplitudes resulted in an increase of LCF lifetime.
The material of interest in the present study is the precipitation hardened iron-based superalloy GH2036, which is optimized for good mechanical properties at high temperatures, and good resistance to high temperature corrosion and oxidation [15]. It has been widely used for rotating structural components, including engine turbine discs and blades constantly subjected to large stress ratios and high temperatures [16][17][18]. In this paper, the fatigue crack propagation behaviors of GH2036 under CCF loading at 550°C corresponding to the turbine disc's working temperature are studied. Targeted experiments which achieved a synchronous treatment of the high cycle and low cycle loading are designed and performed to investigate the mechanism underlying the fatigue crack propagation of GH2036 under CCF loading. Three specimens are conducted under pure LCF loading, and the other three specimens were tested under CCF loading. The influence of the HCF/LCF cycle ratio (i.e. the number of HCF cycles per LCF cycle) on fatigue crack growth was investigated, two group tests of different number of HCF cycles are conducted which reveals the influence of the amount of HCF cycles. The fatigue life of GH2036   under both the LCF and CCF loadings at 550°C in air is experimentally determined. In the end, the mechanisms of fatigue crack growth of GH2036 superalloy under different loadings are discussed and our results are summarized.

Material
The experiments were performed on turbine discs made of GH2036 superalloy, the chemical composition of which is listed in Table 1. Fig. 1(a) shows the microstructure of the material revealed by scanning electron microscope (SEM). An energy dispersive spectroscopy (EDS) analysis of the microstructure indicates that GH2036 superalloy is strengthened by a random distribution of NbC and VC precipitates ( Fig. 1(b)-(d)).

Specimen geometry
The specimen geometry for fatigue testing is shown in Fig. 2. The corner-notched specimen has a rectangular cross section with a thickness and width of 5 mm and 10 mm respectively. The specimen orientation, as defined by ASTM notation, was longitudinalshort transverse, which corresponds to the growth of the corner through the length of the fir-tree mortise in response to centrifugally induced stresses. Each specimen was pre-cracked at room temperature, at a stress ratio of 0.01 and a frequency of 5.0 Hz in order to generate a near quarter-circular crack with a radius of 0.2 mm. To minimize the effects of pre-cracking at room temperature on the test at 550°C, we ensure that the plastic zone sizes at the end of the precracking stage were the same as the start of the crack growth rate testing [7]. The size and shape of the developed crack with this specimen are more representative of turbine disc's operation conditions than those reproducted in a compact tension or bend specimen [6,8,[19][20]. All the test pieces were cut from a forged turbine disc of GH2036 given in Fig. 3. The turbine disc is solution heat treated at a temperature of 1120°C, and held for 1 h, followed by an air cool. Following the solution heat treatment, the disc undergoes a stabilization heat treatment at 870°C for 1 h, followed by an air cool. Finally, the disc is heated to 650°C for 24 h, followed by an air cool.

Procedures for fatigue crack growth test
Although single actuator test machines could generate the required waveforms, their ability to do so is severely limited. In this regard, we built a new test rig, illustrated in Fig. 4(a), in order to achieve a rational and noninterfering CCF loading [21]. In this test rig, particular fixture has been designed to avoid the interactions between the control systems for the two loadings by employing a load train assembly with three movable joints in Fig. 4(b) to allow the specimen to bend. A pair of wedge-shape clamping  device was used to hold the specimen when LCF loading was applied by using a servo-hydraulic actuator. In addition, four smaller bolts shown in Fig. 4(b) were employed to maximize the damping of the load train by increasing the surface contact stiffness between the clamping device and specimen. The HCF loading with a 200 N electrodynamic shaker to excite up to 5000 Hz and an 8 mm amplitude was transferred to the clamping device through the vibration exciting rod. It is worth pointing out that with the structure of three movable joints, the interference between HCF loading and LCF loading can be eliminated and the loading efficiency of HCF is greatly improved because of the decrease of friction. Thus, a rational and noninterfering CCF loading is achieved.
The specimen was heated by a high-frequency induction heater. The temperature of the specimen was monitored by a K-type thermocouple placed in the center of the gauge length. Test temperature was 550°C, equal to the maximum continuous working temperature of the turbine disc. All the experiments were run in laboratory air. There were also validation and control measures regarding to LCF and HCF loading. For LCF loading, four strain gauges were stuck on the key section. For the HCF loading, test results manifested that the vibration stress of the key section had a linear relation with vibration amplitude of the excitation   point [22]. Accordingly, the excitation point was monitored and controlled by a vibrometer in order to insure the accuracy of the HCF loading. Crack lengths were measured by using a Questar telescope with a 2 lm resolution, which was verified to have a high degree of confidence [23]. Stress intensity factors were calculated by the following relationship developed by Pickard [24] for the free surface position of a quarter-circular crack growing under tension shown in Eq. (1): where M G is the comprehensive correction coefficient and is related to the type and position of crack, M B is back-face correction factor, and M S is side-face correction factor. F(r) is stress function representing the effect of the complicated stress field on the stress intensity factor. U is correction factor of ellipse, and is equal to 2/p in this paper. c represents the distance between the crack center and the crack front.
Crack growth rate da/dN was determined by the 3-point secant method of analysis. Fig. 5 shows the CCF experiment being conducted.
The waveform of the LCF loading was simplified and represented by a trapezoidal stress wave with 1 s rise and fall times with a stress ratio (R LCF ) of 0.1, as shown in Fig. 6(a). In the test programme two LCF loading blocks tests were undertaken using different dwell time in order to assess the importance of time dependent fatigue crack growth with 9 s and 108 s dwells at the maximum load, which results in LCF loading frequencies of 0.091 and 0.0091 Hz, respectively. A loading frequency of 91 Hz was superimposed on the LCF cycle dwell period at the maximum load, the conjoint action of which is referred to as CCF, as illustrated in Fig. 6(b). The value of 91 Hz was the resonant frequency of the test facility. The stress ratio of HCF cycle (R HCF ) was 0.95. The stress amplitude ratio of LCF to HCF cycles, Q, is defined as:    In the CCF tests, the use of a 9 s dwell at maximum load resulted in a cycle ratio (n) of 1000:1 HCF cycle for every LCF cycle, whilst the 108 s dwell at maximum load gave a cycle ratio (n) of 10,000:1. An overview over all LCF and CCF tests is given in Table 2. Traverses were made across each specimen surface, one from the crack origin to the diagonally opposite corner and another along one edge. By correcting the orientation of the microscope and eliminating tilting of the specimen, photographs were taken at approximately 0.5 mm intervals across the fracture surface. The crack length associated with each photograph was measured using the machine's traversing micro-meters.

Crack growth rate under pure LCF loading
The crack length evolution of three specimens under pure LCF loading at the 9 s and 108 s dwell time is displayed in Fig. 7. Fig. 8 demonstrates da/dN obtained from three pure LCF loading tests with respect to the stress intensity factor range DK defined by Eq. (1) on logarithmic axes. It is shown that crack growth data under LCF loading reveals scatter due to the random distribution in the microstructure of the test material. Thus it is convenient to replace the scatter of data points by the construction of the band, within which all crack growth data falls (see Fig. 8). The slopes of the least squares lines are 2.4 and 2.0, respectively. The effect of increasing dwell time is to markedly increase the fatigue crack growth rates. Consequently, there is time dependent crack growth for GH2036 at 550°C. It is observed from Fig. 8 that there is a strong stepwise increase in da/dN for a long dwell time of 108 s, which is different from that at the shorter dwell time of 9 s.

Effects of HCF cycles on crack growth rate of CCF loading
The evolution of crack length of the four specimens under CCF and LCF loadings at a 9 s dwell is shown in Fig. 9. The crack growth rate da/dN data obtained from two CCF tests at a cycle ratio of 1000:1 and the two pure LCF tests are compared in Fig. 10. The da/dN data falls outside the bottom band of the pure LCF band and as DK increases, crack growth rate da/dN moves progressively away from the band. The superimposed HCF loadings lead to decreased crack growth rate and increased lifetime in contrast to pure LCF loading, which demonstrates that the superimposed HCF cycles delay the crack growth rate for GH2036 superalloy at 550°C.
In addition, another two tests under pure LCF loading and CCF loading with a cycle ratio of 10,000:1 and at a dwell time of 108 s were carried out. The corresponding data are presented in Figs. 11 and 12, also indicating that the crack growth rate is less under CCF loading than that under pure LCF loading. Most crack growth rate data also falls outside the bottom band (see two solid black lines in Fig. 12) of the pure LCF band. Compared to the data in Fig. 10, the fatigue crack growth rate drops slowly because of superimposed HCF cycles. The stepwise increase in da/dN appears similar to the crack growth rate under pure LCF loading with a long dwell time (see Fig. 12). Fig. 13 shows the crack growth data using CCF tests with two cycle ratios. In contrast to CCF loading with a 10,000:1 cycle ratio (i.e. 10,000 HCF cycles are superimposed on every LCF cycle at a 108 s dwell), the crack growth rate decreases at a cycle ratio of 1000:1 (i.e. 1000 HCF cycles are superimposed on every LCF cycle at a 9 s dwell). Fig. 8 shows that long dwell time significantly    accelerates the fatigue crack growth rate, i.e. the crack under LCF loading with a 108 s dwell grows faster than with a 9 s dwell. Also, it is demonstrated that the more HCF cycles are superimposed on LCF loading, the less the crack growth rate decreases. This means that the crack growth rate with CCF loading with a 10,000:1 cycle ratio slightly decreases compared with LCF loading at a 108 s dwell. Therefore, as the cycle ratio increases, the fatigue crack growth accelerates.

Fractography
A systematic fractographic analysis were performed by SEM on the fractured specimens (see Fig. 14) from the above tests, in which #1 and #2 were tested under LCF loading at a 9 s dwell time while #3 under LCF loading at a 108 s dwell time, #4 and #5 were tested under CCF loading with a cycle ratio of 1000:1, and #6 was tested under CCF loading with a cycle ratio of 10,000:1. No significant differences were discovered between the fracture surface along the diagonal and along the edge of the specimens, corresponding to the plane strain and plane stress regions respectively. All the specimens initiated and grew the crack from the starter notches as shown in Fig. 15.

The effect of dwell time under pure LCF loading
The examination of fracture surfaces by SEM shows that fatigue cracks originate at the surface. Fig. 16(a) clearly shows the specimen #1, with a 9 s dwell, has several crack initiation points while  specimen #3, with long dwell time, has only one crack source as illustrated in Fig. 16(b). This situation consistent with the long dwell time at elevated temperature causes the superalloy GH2036 defect sensitivity. As noted in Fig. 17(a), a large amount of oxide forms on the crack face in addition to fatigue striations. Also of significance is that the crack surface is highly oxidized for the specimen at pure LCF loading with a long dwell time in Fig. 17(c).
At a higher DK of 8.3 MPa ffiffiffiffi ffi m p , the fracture surface of specimen #1, with a 9 s dwell time, is a mixture of intergranular and transgranular failure (Fig. 17(b)). However, for the specimen #3, with a long dwell time at a higher DK, an intergranular fracture appears dominant (Fig. 17(d)). A change in the failure mode from transgranular fatigue crack propagation to intergranular fatigue propagation causes the stepwise increase in higher crack growth rates (see Fig. 8).

The difference between pure LCF and CCF tests
In the case of CCF loading with short dwell time, SEM observation shows that multiple crack initiation sites for specimen #4 are at the surface (Fig. 18), similar to pure LCF loading ( Fig. 16(a)). Fig. 19(a) shows the enlarged image of Fig. 17(b). In the high magnification SEM image, the fatigue crack growth of specimen #1, under pure LCF loading at higher DK, reveals dimples. In contrast, obvious fatigue striations are observed in the enlarged image of Fig. 19(b) for specimen #4, under CCF loading at higher DK, as shown in Fig. 19 (c). These fatigue striations induced by superimposed HCF loading are important for decreasing the fatigue crack growth rate. Fig. 20 depicts that the crack source for specimen #6, under CCF loading with a cycle ratio of 10,000:1, is different from specimen #4 with a cycle ratio of 1000:1 in Fig. 18. It is well known that superalloy GH2036 is defect sensitivity with a long dwell time at elevated temperature, which eventually leads to the observed only one crack initiation zone. At a higher DK, the fracture surface of specimen #6 shows intergranular failure (Fig. 21) instead of a mixed transgranular and intergranular failure for specimen #4. Consequently, a stepwise increase in crack growth rate occurs at a higher DK under CCF loading with a long dwell time (see Fig. 12).

The effect of cycle ratio under CCF tests
In addition, oxide thicknesses of the specimens #4 and #6 under CCF loading are compared as shown in Fig. 22. The maximum oxide thickness of specimen #4 is 4.147 lm (Fig. 22(a)), while the minimum and maximum values are 5.340 lm and 15.780 lm for specimen #6, respectively ( Fig. 22(b)). The oxidation thickness of specimen #6 with a high cycle ratio (n = 10,000:1) is considerably greater than that of specimen #4, resulting in a faster crack growth rate.

Conclusions
Overall our studies establish an experimental system at elevated temperature with a new CCF loading scheme and a design of fixture to achieve a noninterfering treatment of the HCF and LCF loading by employing a load train assembly with three movable joints. Then pure LCF and CCF tests on specimens cut from an actual forged turbine disc of GH2036 superalloy were performed in laboratory air at 550°C. Compared with previous work, our results confirm that superimposed small high frequency HCF amplitudes delay the crack growth rate using LCF tests for GH2036 superalloy. In addition, as the cycle ratio increases under CCF loading, the fatigue crack growth rate accelerates. From SEM observations, all of the cracks initiate from the surface of the starter notches. A change in the failure mode from transgranular to intergranular causes the stepwise increase in fatigue crack growth rates under a long dwell time. Furthermore, the oxidation thickness of the specimen under CCF loading with a larger cycle ratio is significantly greater than a small cycle ratio. In future research, crack growth rate corresponding to the creep damage should be considered to investigate crack growth behavior at elevated temperature.