3.1 Microstructural Evolutions Under Creep-fatigue Interaction
The changes in CF life as a function of dwell times (30, 120, 300, 600 and 900 s) for the welds are shown in Figure 5. All CF-tested specimens fractured in the BM region. Since it’s tedious and time consuming to perform nanoindentation for each specimen in Figure 5, the four typical specimens (whose CF lifetimes are close to the average) were selected for different dwell times (30, 120, 600 and 900 s). The CF samples with a dwell time of 300 s were not selected because the remarkable discreteness of their CF life. The CF lifetimes of four typical specimens were 1803 for dwell time of 30 s, 2312 for dwell time of 120 s, 967 for dwell time of 600 s and 417 for dwell time of 900 s, respectively. It is clear that the lifetime of CF decreased with increasing dwell time in the range of 120 to 900 s. This observation is consistent with the change in CF lifetime with increasing dwell time for P92 steel [15]. Nevertheless, the improved CF lifetime with increasing dwell time is observed in the range of 30 to 120 s dwell times. This phenomenon is not reported for the P92 steel welded joint due to the limited data for CF testing.
To investigate the change in CF life with different dwell times, the peak tensile stress per cycle was plotted as a function of normalized CF life (N / Nf ) for the four typical CF samples in Figure 6. N is the number of instantaneous cycles, and Nf represents the number of failure cycles. Figure 6 shows two different softening behaviors for long dwell times (600 and 900 s) and short dwell times (30 and 120 s). For the CF specimens with short dwell times, three levels can be seen in Figure 6. To illustrate the CF samples with a dwell time of 30 s: in the first stage (0–10% CF lifetime), the peak tensile stress decreases from about 255 to 200 MPa. Then, a stationary stage (10%–60%) occurs where the peak tensile stress decreases slightly from 200 to 155 MPa. The third stage of the CF specimen starts at 60% of the life, and finally the specimen breaks. This softening behavior was also observed by Zhang et al. [15] when working on P92 steel in the circumferential direction. It is also observed that the stress level of the CF specimen with 120 s dwell times is slightly lower than that of the specimen with 30 s dwell time in all phases. However, only two stages are observed for the CF specimens with long dwell times, as shown in Figure 6. For the sample CF with 900 s dwell, the peak tensile stress decreases rapidly from 270 to 225 MPa in first 10% of the lifetime. The decreasing rate becomes slow and remains almost constant, and the stress is then linearly reduced from 225 to 200 MPa during the remaining CF lifetime. It is also observed that the peak tensile stress level (from 250 to 200 MPa at the first stage, from 200 to 175 MPa at the second stage) is lower in CF with 600 s dwell time at both stages than with 900 s dwell time. The excessive stress level of CF specimen with long dwell times may exceed the capacity of cyclic resistance of P92 steel welded joint and eventually lead to premature fracture. The inelastic strains for the four typical specimens were estimated from their hysteresis loops at a 50% lifetime fraction. The inelastic strains of the different dwell times were 0.251% (30 s dwell), 0.247% (120 s dwell time), 0.281% (600 s dwell time) and 0.282% (900 s dwell time). The increased inelastic strain could be the reason that the long dwell time leads to high peak tensile stress. Moreover, the significantly increased inelastic strain from 120 s dwell time to 600 s dwell time could also lead to the two cyclic softening processes.
Stress relaxation is considered a process of creep damage accumulation in strain-controlled CF experiments. On the other hand, the stress-strain behavior response of the first cycle and the cyclic loading corresponding to 50% of the CF life are two representative behaviors to show the interaction between creep and fatigue in alloys [26,27,28]. Therefore, the stress relaxation histories of the first cycle and the 50% CF life (901 for a dwell time of 30 s, 1506 for a dwell time of 120 s, 483 for a dwell time of 600 s, and 208 for a dwell time of 900 s) are shown in Figure 7. For the first cycle of four typical specimens, a good overlap of the stress relaxation behavior can be seen in Figure 7(a). The stress decreases rapidly from 270 to 125 MPa in the first 40 s. After that, the stress slowly decreases and slightly decreases to 80 MPa with the dwell time. At the cyclic loading corresponding to 50% life (CF), the peak tensile stress values were different for the four specimens: 160 MPa at a dwell time of 30 s, 155 MPa at a dwell time of 120 s, 175 MPa at a dwell time of 600 s and 200 MPa at a dwell time of 900 s, as shown in Figure 7(b). At the end of the dwell time, the stress was 110 MPa at a dwell time of 30 s, 70 MPa at a dwell time of 120 s, 80 MPa at a dwell time of 600 s and 110 MPa at a dwell time of 900 s. Since the dwell time for a typical specimen is only 30 s, the stress is not completely relieved even in the first 40 s of stress relaxation. The stress level of CF specimen with a dwell time of 30 s is the highest among the four typical specimens in the first cycle and at 50% lifetime. For the remaining three samples, the stress level increases with increasing dwell time at 50% CF lifetime. This observation indicates that the higher stress level contributes to the lower CF lifetime at longer dwell time. In addition, the excessively high stress level in the CF sample with 30 s dwell time leads to a shorter CF lifetime than in the CF sample with 120 s dwell time.
Using scanning electron microscope (SEM), the fracture morphologies of CF samples with dwell time of 30 s, 120 s, 600 s and 900 s are shown in Figure 8(a)–(d). Figure 8(a) and (b) shows that the fracture surfaces of CF specimens with short dwell time (30 s and 120 s), have fatigue striations, while no creep voids are seen in the fracture morphology. This indicates that fatigue is the predominant damage mechanism for the CF short dwell time specimens. In contrast, creep voids were observed in the fracture surfaces of the CF short dwell time specimens (600 s and 900 s), as shown in Figure 8(c) and (d). The size of the creep voids increase with the increasing dwell time, about 20 μm at dwell time of 600 s and 40 μm at dwell time of 900 s. On the other hand, the creep voids were surrounded by fatigue striations, as shown in Figure 8(c) and (d). This observation suggests that the interaction between creep and fatigue is the dominant damage mechanism for CF specimens with long dwell time. Figure 9(a) and (b) shows the two fracture mechanisms schematically. In Figure 9(a), the fatigue cracks developed in the surface of the specimens with short dwell time and then propagated transgranularly until the specimen fractured. For the specimens with long dwell time, the fatigue cracks also formed at the surface, while the creep voids formed at the grain boundaries. During the creep-fatigue process, the fatigue cracks met the large creep voids and propagated until the specimen fractured, as shown in Figure 9(b). This interactive creep-fatigue fracture mechanism was able to accelerate the propagation rate of the cracks and led to premature failure of the P92 welded joint.
In retrospect, the specimens used in this work have all experienced prior CF loading before nanoindentation. Figure 10 shows the hardness and elastic modulus values measured at the fracture edges of PWHT and CF-tested specimen with 30, 120, 600 and 900 s dwell times. To reduce the effect of indentation size in this material [29], the hardness and elastic modulus were measured at 800 μm for all specimens. From Figure 10, the effects of spacing from the fracture edge (ranging from 50 to 350 μm) on the hardness and elastic modulus are negligible for all CF specimens. Compared with the BM for PWHT samples (3.05 GPa), the hardness measured near the fracture edge is slightly reduced for CF samples with 600 and 900 s dwell times (2.81 and 2.90 GPa), as shown in Figure 10(a). In contrast, the hardness at the fracture edge decreases significantly to 2.61 and 2.06 GPa for the CF samples with 30 and 120 s dwell times. A similar observation (Figure 10(b)) is also obtained for elastic the modulus, which is almost unchanged at long-term dwell times (208 GPa for PWHT, 202 GPa at 600 s dwell time and 206 GPa at 900 s dwell time) and significantly reduced at short-term dwell times (71 GPa at 30 s dwell time and 63 GPa at 120 s dwell time). Note that the CF specimens with short-term dwell times remain at the third softening stage before fracture, while the specimens with long -term dwell times are stay at the steady softening stage (Figure 6). The stress level at the third stage is much lower than at the stable stage (Figure 6), which could lead to the remarkable reduction in hardness in the CF samples with short-term dwells. Moreover, the two CF specimens with short dwell time were subjected to more cyclic loading (1803 and 2312 cycles) than those with long dwells (967 and 417 cycles). In the CF-tested specimens, microvoids form during cyclic loading at high temperatures, which then lead to nano- to micro-scale defects. The author speculates that these microvoids/defects could increase the pore volume fraction in the tiny nanoindentation test area, thus decreasing the elastic modulus. On the other hand, it is reported that the cyclic softening behavior of P92 steel at high temperatures is caused by the lower dislocation density and the decreasing ability of grain boundaries impede the dislocation motion [22,23,24]. Normally, this microstructural evolution should lead to a significant deterioration in deformation resistance [30, 31]. The reduced resistance to deformation would lead to increased deformation at the same stress level, thus deteriorating the elastic modulus. Based on the characteristics of the fracture surface (Figure 8), the significant reduction in hardness and elastic modulus could be caused by fatigue-induced damage.
3.2 Nanoindentation Characterization
The representative load-displacement (P-h) curves measured during nanoindentation creep near the fracture edge are shown in Figure 11(a) for the CF specimens with 30, 120, 600 and 900 s dwells. The initial holding displacements of the different typical P-h curves are 1003.5 nm (30 s dwell), 1011.6 nm (120 s dwell), 1004.2 nm (600 s dwell) and 1003.5 nm (900 s dwell). It is observed that the holding forces are higher for longer CF dwells except for 120 s dwell time where the holding force is lowest. The corresponding creep displacement determined from the nanoindentation depth during the dwell time is shown in Figure 11(b). Two common primary and secondary stages of nanoindentation are also observed. The creep displacement increases rapidly while the creep rate decreases rapidly at the primary stage. The creep displacement increases almost linearly with the holding time, and the creep rate remains constant at the secondary stage.
Figure 12 shows that the creep displacement as a function of the dwell times measured at different distances from the fracture edge for PWHT and CF-tested specimens. It can also be seen that the effects of the distances from the fracture edge (ranging from 50 to 350 μm) on the creep displacements are negligible for all CF-tested specimens. Compared to the BM in the PWHT specimen (29.3 nm), the creep displacements (36.8 nm at 30 s dwell, 36.9 nm at 600 s dwell and 36.4 nm at 900 s dwell) are slightly increased at longer CF dwell times except for the 120 s dwell time, which results in a significant creep displacement (98.4 nm), as shown in Figure 12. Hardness is generally considered as the ability of alloys to resist creep deformation [32]. From Figure 10(a) and Figure 12, it can be seen that the CF specimen with lower hardness have relatively high creep deformation. Although the CF specimens with 30, 600 and 900 s dwell times have almost the same creep deformation at room temperature, the decrease in their hardness is different. It can be seen that the creep resistance of CF specimens with long-term dwell times is lowered compared to 30 s dwell time. This conclusion is consistent with the high temperature creep-fatigue behavior that longer dwell time can lead to more creep damage which decreases the creep resistance. Since the hardness of the CF sample decreases significantly with a dwell time of 120 s, its creep resistance cannot be directly compared to the other three CF specimens. It should be noted that this sample was subjected to the most cyclic loading among the four typical samples. On the other hand, the dwell time of this CF sample also longer than that of another fatigue loaded CF sample with only 30 s dwell time per cycle (Figure 8(a) and (b)). During the dwell time, the grain boundary (GB) is in a pronounced relaxation process due to the lower stress level in the CF samples with 120 s dwell time. During this process, the dislocations at the GBs were annihilated [23, 24], reducing the stress fields generated by the dislocation-GBs interaction. This relaxation process on the GBs also led to a deterioration of the deformation resistance and interacted with a lower hardness, which promoted creep deformation at room temperature. Moreover, the decreased elastic modulus due to cyclic loading also leads to decreased deformation resistance of the CF specimens with short dwell times(Figure 10(b)). It is also possible to affect room temperature creep if the deformation mechanism is due to dislocation activities. To verify this, the deformation mechanism of room temperature creep should be investigated.
3.3 Strain Rate Sensitivity
Nanoindentation creep has been widely used to estimate strain rate sensitivity (SRS), which can be used to reveal the creep mechanism [33, 34]. The SRS value (m) under the self-similar Berkovich indenter can be approximately estimated by Eq. (1) [35].
$$ m = \frac{\partial \ln H}{{\partial \ln \dot{\varepsilon }}}. $$
(1)
The creep curve can be fitted with high reliability (R2 > 0.99) by an empirical equation:
$$ h(t) = h_{0} + A(t - t_{0} )^{B} + Kt, $$
(2)
where h0 and t0 are the initial holding depth and time at the beginning of the holding period, and A, B, and K are the fitting constants.
Based on the data of the fitting curves, the change of hardness H during the holding period can be estimated as follows:
$$ H = \frac{P}{{24.3h_{{\text{c}}}^{2} }}, $$
(3)
where P and hc are the holding load and contact displacement of Berkovich indenter. The hc measured by the Berkovich indenter can be expressed as follows:
$$ h_{c} = h - 0.72 \times P/S, $$
(4)
where h is the displacement of the indentation and the contact displacement, S is the stiffness deduced from the unloading curve. Furthermore, the strain rate \(\dot{\varepsilon }\) can be estimated as follows:
$$ \dot{\varepsilon } = \frac{{{\text{d}}h}}{{{\text{d}}t}} \cdot \frac{1}{h}. $$
(5)
As an example of the usefulness of this method, Figure 13(a) shows the fit (R2 > 0.99) obtained by this method to describe the creep curve at the fracture edge of the CF specimen with 120 s holding time. From Figure 13(b), the strain rates stabilize at 6.23×10−4 s−1 after decreasing significantly during the hold time of 0.012 s−1. Figure 13(c) shows the decrease in nanoindentation hardness as a function of holding time. A natural logarithmic correlation between the nanoindentation hardness and the creep strain rate is shown in Figure 13(d). Thus, the value of SRS could be estimated by a linear relationship in the secondary creep phase.
Figure 14 shows the variation of SRS (m) for the fracture edge in PWHT and CF specimens with different dwell times. The values of SRS estimated from the CF specimens with long-term dwells (0.0105 for 600 s dwell and 0.0108 for 900 s dwell) are slightly increased in comparison to the BM in PWHT specimen (0.0075) [23], as plotted in Figure 14. For the CF specimens with short-term dwells, a significant increase in SRS value is observed at 30 s dwell time (0.0278) and especially at 120 s dwell time (0.1709). In general, the values of m less than 0.3 indicate that the mechanism of creep at room temperature is dislocation activities. The m can be expressed as a function of dislocation density as suggested in Ref. [22].
$$ m = \frac{3\sqrt 3 kbT}{{a\mu bcV^{ * } }}\frac{1}{\sqrt \rho }, $$
(6)
where ρ is the dislocation; V* is the activation volume; kb is the Boltzmann constant; T is the temperature in Kelvin, and a, μ, b, c are constants for a given material.
Eq. (6) shows that the SRS is inversely proportional to the square root of the dislocation density when a, μ, b, c, and V* are positive for a given material. Under CF loading, the reduction in dislocation density was observed, and enhanced SRS was also estimated in Refs. [22, 23]. For coarse-grain materials, the deformation typically occurs within the grain, and grain boundary can act as sinks for dislocation motions [36, 37]. Therefore, the stress field is produced by dislocation-dislocation and dislocation-GB interactions during deformation. The annihilation of dislocation reduces the dislocation density during the dwell period and enhances SRS for the four typical specimens. As for long-term dwells (600 and 900 s), the relaxation process is saturated due to the stable stress level after the first 40 s (Figure 7). Furthermore, the decrease in dislocation density and increased SRS are also reported to occur after fatigue loading [38]. Compared with long-term dwells (600 and 900 s), the higher number of cyclic loading history could result in lower dislocation density for the CF specimens with short-term dwells (30 and 120 s). Consequently, the significantly enhanced SRS was estimated in CF specimens with short dwells.
Finally, it should be noticed that the fracture mechanisms are different between long-term dwell and short-term dwell, as shown in Figures 8 and 9. For CF specimens with short-term dwells, the fracture mechanism was fatigue-dominate. The softening behaviors were serious in short-term dwells, which caused a remarkable reduction in hardness and lower dislocation density, which could significantly increase local SRS. As for long-term dwells, the long dwell time led to long thermal-mechanical aging time, which could result in nucleation and growth of creep voids, increase of grain size [23, 24], and coarsening of precipitates. The growth of creep voids would reduce the effective load area of specimens, which eventually caused the specimen rupture, namely the creep-fatigue-dominated fracture mechanism. In this fracture mechanism, the grain size increased for a long thermal-mechanical aging time, which could induce the slight reduction of local hardness and enhancement of SRS [20].