Jtam-A4.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 56, 2, pp. 497-510, Warsaw 2018 DOI: 10.15632/jtam-pl.56.2.497 EFFECT OF CYCLIC HARDENING ON STRESS RELAXATION IN SUS316HTP UNDER CREEP-FATIGUE LOADING AT 700◦C: EXPERIMENTS AND SIMULATIONS NOBUTADA OHNO Nagoya Industrial Science Research Institute, Nagoya, Japan e-mail: nobuohno@nagoya-u.jp; nobuohno@nifty.com Tatsuya Sasaki Department of Computational Science and Engineering, Nagoya University, Nagoya, Japan e-mail: t.sasaki@mml.mech.nagoya-u.ac.jp Takehiro Shimada, Kenji Tokuda, Kimiaki Yoshida IHI Corporation, Yokohama, Japan e-mail: takehiro shimada@ihi.co.jp; kenji tokuda@ihi.co.jp; kimiaki yoshida@ihi.co.jp Dai Okumura Department of Mechanical Engineering, Osaka University, Osaka, Japan e-mail: okumura@mech.eng.osaka-u.ac.jp Cyclic hardening and stress relaxation experiments of SUS316HTP were performed under creep-fatigue loading with tensile strain holding at 700◦C. Experiments revealed that un- der strain holding, the slow stress-relaxation stage satisfying Norton’s lawwith slight cyclic hardening followed a rapid stress-relaxation stage that was noticeably affected by cyclic hardening. This suggests that in the slow stress-relaxation stage, inelastic deformation me- chanisms different from that of viscoplasticity occurred. Experiments were simulated using a cyclic viscoplastic-creep model in which the inelastic strain-rate was decomposed into viscoplastic and creep components that were affected differently by cyclic hardening. The simulation accurately reproduced the experiments. Keywords: creep-fatigue loading, cyclic hardening, stress relaxation, SUS316, constitutive model 1. Introduction Creep-fatigue tests with tensile and/or compressive strain holding at high temperatures have been performed to investigate the effect of creep damage on the fatigue lives of materials. For polycrystalline metals, creep damage under creep-fatigue loading is caused by grain boundary cavitation that develops with the accumulation of creep strain under strain holding (e.g., Hales, 1980; Priest and Ellison, 1981; Nam, 2002), and has beenmacroscopically evaluated in terms of the changes in stress and creep strain under strain holding (e.g., Inoue et al., 1989; Takahashi et al., 2008; Yan et al., 2015). To numerically evaluate creep damage in structural components, it is necessary to use a constitutive model that can accurately simulate the stress-strain behavior under cyclic loading with strain holding. Theductility exhaustionmethodproposed byPriest andEllison (1981) andHales (1983) is a well-knownmethod to evaluate creep damage under creep-fatigue loading (Ainsworth, 2006;Yan et al., 2015). This method assumes that creep damage develops with the accumulation of creep or inelastic strain under strain holding. Priest and Ellison (1981) proposed that creep damage develops when the inelastic strain-rate under strain holding is smaller than the transition rate 498 N. Ohno et al. below which the diffusion creep and grain boundary sliding become important, whereas Hales (1983) considered that the development of creep damage depends on the variations in inela- stic strain in three periods under strain holding. Takahashi (1998) and Takahashi et al. (2008) adopted the Priest and Ellison (1981) model and decomposed the inelastic strain-rate under strain holding into viscoplastic and creep components occurring at high and low inelastic-strain rates, respectively, and assumed that only the creep component contributes to the development of creep damage. They thus accurately predicted the creep-fatigue lives of 316 stainless steel at 550◦C and 600◦C. The decomposition of inelastic strain-rate is physically valid in the presence of dislocation viscoplasticity at high inelastic strain-rates and diffusion creep at low inelastic strain-rates. In the constitutive modeling of cyclic plasticity, however, the decomposition of inelastic strain into viscoplastic and creep strains has been regarded as a conventional assumption. The work has been focusedon thedevelopment of unifiedconstitutivemodels, inwhichbothviscoplasticity and creep are considered to be caused by dislocation movements (Miller, 1976; Krausz and Krausz, 1996; Chaboche, 2008). It is, therefore, worthwhile to investigate the stress relaxation behavior in creep-fatigue tests to examine the appropriateness of the inelastic strain-rate decomposition. This point of view was not taken by Nouailhas (1989) for using a unifiedmodel to simulate the creep-fatigue tests of 316 stainless steel at 600◦C performed byGoodall et al. (1981). In this study, the stress relaxation behavior under tensile strain holding was measured in creep-fatigue tests of SUS316HTP at 700◦C to examine the decomposition of inelastic strain- -rate. It was assumed that the increase in dislocation density, which occurs in grains and is observed as cyclic hardening, affected viscoplasticity significantlymore than diffusion creep and grain boundary sliding. It was thus suggested that inelastic deformationmechanisms other than viscoplasticity started to operate shortly after the onset of strain holding, and consequently that the inelastic strain-rate consisted of viscoplastic and creep components under strain holding in the creep-fatigue tests. The experiments were then simulated using a cyclic viscoplastic-creep model in which cyclic hardening was assumed to have different effects on the viscoplastic and creep strain-rates. Throughout this paper, a superposed dot indicates differentiation with respect to time t, a colon represents the inner product between tensors (e.g., σ : ε= σijεij and D : ε =Dijklεkl), ‖·‖ denotes theEuclidean normof second rank tensors (e.g., ‖σ‖=(σ :σ)1/2), and 〈·〉 indicates theMacaulay brackets (i.e., 〈x〉=x if x> 0 and 〈x〉=0 if x¬ 0). 2. Experiments 2.1. Material tested and cyclic loading conditions Uniaxial creep-fatigue tests with tensile strain holding were performed at 700◦C using an electric-hydraulic servo-type material testing machine MTS810. The material tested was SUS316HTP (a 316 stainless steel), which had the chemical composition andmechanical proper- ties at room temperature given inTables 1 and 2. Solid bar specimenswith the shape illustrated in Fig. 1 were used. The tests performed are listed in Table 3. Hereafter, ∆ε, ε̇, and th indicate the strain range, strain-rate, and strain hold time, respectively, in the creep-fatigue tests. Table 1.Chemical composition of SUS316HTP bymass percent C Si Mn P S Ni Cr Mo 0.07 0.28 1.75 0.27 0.0 11.2 16.6 2.12 Effect of cyclic hardening on stress relaxation in SUS316HTP... 499 Table 2.Mechanical properties at room temperature 0.2% proof stress Tensile strength Tensile rupture strain [MPa] [MPa] [%] 267 553 68 Fig. 1. Shape of the experimental specimens; dimensions in mm Table 3.Tests performed Strain rate Strain range Tensile strain hold time [1/s] [–] [min] 10−3 0.010 0, 10, 60 10−3 0.007 0, 10, 60 10−3 0.004 0 10−4 0.010 0 2.2. Effect of cyclic hardening on stress relaxation The effect of cyclic hardening on the stress relaxation behavior under strain holding was investigated using the experimental data obtained in the creep-fatigue tests with th =60min. Under stain holding, we have ε̇= ε̇e+ ε̇in =0 (2.1) where ε̇e and ε̇in are the elastic and inelastic parts of ε̇, respectively. Assuming isothermal Hooke’s law for ε̇e in Eq. (2.1) gives ε̇in =− σ̇ E (2.2) Here, σ and E indicate the uniaxial tensile stress and Young’s modulus, respectively. Using a difference approximation for σ̇ in the above equation, ε̇in at time t is represented as ε̇in(t)=− σ(ti+1)−σ(ti) (ti+1− ti)E (2.3) where ti and ti+1 are times slightly before and after t, respectively. ApplyingEq. (2.3) to the experimental data at ε̇=10−3 s−1 with th =60min at∆ε=0.007 and 0.01 provided the relationships between logσ(t) and log ε̇in(t) under strain holding (Figs. 2a and2b). Stress increasedwith the increasingnumberof cyclesN, especially in the stage justafter the onset of strain holding. This is the phenomenon known as cyclic hardening. The relationship became linear to satisfy Norton’s law shortly after the onset of tensile strain holding. Hereafter, the stage of stress relaxation satisfying Norton’s law is referred to as the Norton stage, and is distinguished from the transient stage in which stress rapidly relaxes just after the onset of strain holding. Figure 2 shows that the Norton stage had much less cyclic hardening than the transient stage. 500 N. Ohno et al. Fig. 2. Relationship between ε̇in and σ under tensile strain holding in the creep-fatigue tests at ε̇=10−3 s−1 with th =60min at: (a)∆ε=0.01 and (b)∆ε=0.007 To discuss the effect of cyclic hardening on the stress relaxation in more detail, the tensile peak stress σ+peak and a representative stress in theNorton stage, σ10E−7, are plotted againstN in Fig. 3. Here, σ10E−7 denotes the stress at which ε̇ in became equal to 10−7 s−1 under strain holding. As seen in the figure, σ10E−7 increased slightly with N compared to σ+peak, which increased noticeably fromN =1 toN ≈ 20. This confirms that theNorton stage hadmuch less cyclic hardening than the transient stage. It is physically valid to assume that cyclic hardening is primarily caused by an increase in dislocation density, which occurs in grains and is responsible for viscoplasticity. It is thus suggested that the Norton stage is rate-controlled by inelastic deformationmechanisms such as diffusion creep and grain boundary sliding, which are different from viscoplasticity. Therefore, to simulate the creep-fatigue tests performed in this study, the inelastic strain-rate under strain holding should be decomposed into a viscoplastic component responsible for the transient stage and a creep component responsible for the Norton stage. Fig. 3. Variations in σ+peak and σ10E−7 with N and∆ε in the creep-fatigue tests at ε̇=10 −3 s−1 with th =60min 3. Constitutive model Because the inelastic-strain rate under strain holding is decomposed into viscoplastic and cre- ep components (Section 2.2), it is assumed that the strain-rate ε̇ is decomposed into an ela- stic component ε̇e obeying Hooke’s law, a viscoplastic component ε̇p associated with combined Effect of cyclic hardening on stress relaxation in SUS316HTP... 501 isotropic-kinematic hardening, and a creep component ε̇c satisfying Norton’s law1 ε̇= ε̇e+ ε̇p+ ε̇c σ=De : εe (3.1) and ε̇ p = 3 2 ε̇ p 0 [ yeq (1+θp)σ p 0 ]ms−a yeq ε̇ c = 3 2 ε̇c0 [ σeq (1+θc)σc0 ]n s σeq (3.2) whereσ is the stress,De is the isotropic elastic stiffness, ε̇ p 0, σ p 0, andm are viscoplastic parame- ters, s is the deviatoric stress, a is the deviatoric back stress, ε̇c0, σ c 0, andn are creep parameters, θp and θc are variables representing the effects of cyclic hardening on ε̇p and ε̇c, respectively, and yeq and σeq are defined as yeq = √ 3 2 ‖s−a‖ σeq = √ 3 2 ‖s‖ (3.3) It is further assumed that the back stress can be decomposed into M parts (Chaboche et al., 1979; Chaboche, 2008)2, and that cyclic hardening equally affects the drag and back stresses (Ohno et al., 1998, 2017a)3. In addition, it is assumed that the evolution of each part of the back stress is represented by the Ohno-Wang model (Ohno and Wang, 1993). We thus use the following equations for a a=(1+θp)ã ã= M∑ i=1 h(i)b(i) ḃ (i) = 2 3 ε̇ p− ζ(i)(ζ(i)b(i)eq ) k(i) 〈 ε̇ p : b(i) b (i) eq 〉 b (i) (3.4) where ã is the deviatoric back stress free of the effect of cyclic hardening, b(i) is the i-th non- -dimensional back stress related to ã, h(i) is the i-th incipient kinematic hardening modulus, ζ(i) and k(i) are parameters of the back stress evolution, and b (i) eq is defined as b(i)eq = √ 3 2 ‖b(i)‖ (3.5) Austenitic stainless steels exhibitamarkeddependenceof cyclic hardeningonthe strain range (e.g., Chaboche et al., 1979; Ohno, 1982; Kang et al., 2003). Hence, we assume the following equation for θp in Eqs. (3.2)1 and (3.4)1 θp =φ(∆εp)κ (3.6) where φ(∆εp) is the material function representing the dependence of cyclic hardening on the viscoplastic strain range∆εp, and κ is the cyclic hardening parameter, which evolves as κ̇=L(κ0−κ)ṗ−Rκ ω (3.7) Here,L and κ0 are strain hardening parameters,R andω are thermal recovery parameters, and ṗ denotes the accumulating rate of viscoplastic strain ṗ= √ 2 3 ‖ε̇p‖ (3.8) 1Eqs. (3.1)-(3.3) based on the decomposition of the inelastic strain-rate into viscoplastic and creep components were assumed for solders in the absence of cyclic hardening (Kobayashi et al., 2003). 2Themultiple back stresses can be transformed to the multiple surfaces proposed byMróz (1967), as shown byOhno andWang (1991). 3Trampczynski (1988) experimentally showed the effect of cyclic hardening on the back stress using the technique of successive unloading. 502 N. Ohno et al. Equation (3.7) does not represent the cyclic softening following cyclic hardening that was ob- served in the creep-fatigue tests (Fig. 3). However, this limitation is allowed for the purpose of simulating the cyclic hardening and stress relaxation behavior discussed in Section 2.2. The creep strain-rate ε̇c may be affected by cyclic hardening because grain boundary sliding can be accommodatedwith dislocation viscoplasticity as demonstrated byCrossman andAshby (1975). This effect is represented by θc in Eq. (3.2)2. We assume θc = cθp (3.9) where c is a parameter representing the effect of cyclic hardening on ε̇c. The constitutivemodel described in this Section needs∆εp to be evaluated during computa- tion. We can use the resetting scheme of a viscoplastic strain surface to correctly evaluate ∆εp (Ohno et al., 2017b). This plastic-strain-range (PSR) surface has the same expression as the memory surface of Chaboche et al. (1979), and follows the same evolution rule as that of Ohno (1982). In the resetting scheme, however, the PSR surface is reset to a point and re-evolves every cycle under cyclic loading. The resetting thus provides a definite value for the evolution parameter η of the PSR surface irrespective of the amount of cyclic hardening, pre-straining, and ratcheting. In this study, η is set to 0.4, as verified by Ohno et al. (2017b). 4. Determination of material parameters Table 4 gives thematerial parameters used in this study, whichwere determined using the follo- wing procedure. In the table,E and ν denoteYoung’smodulus andPoisson’s ratio, respectively. Table 4.Material parameters with stress inMPa, strain in mm/mm, and time in s Elastic E =1.44 ·105, ν =0.30 Viscoplastic ε̇ p 0 =10 −3, σ p 0 =7.53 ·10 1,m=20.0 Creep ε̇c0 =10 −3, σc0 =2.72 ·10 2, n=10.9 Kinematic hardening h(1) =1.63 ·105, h(2) =3.81 ·104, h(3) =9.27 ·103, h(4) =1.59 ·103, h(5) =7.24 ·102 ζ(1) =6.67 ·103, ζ(2) =2.00 ·103, ζ(3) =6.67 ·102, ζ(4) =2.50 ·102, ζ(5) =1.25 ·102 k(i) =3.0, (i=1,2, . . . ,5) Cyclic κ0 =0.726, L=13.4, R=0.411, ω=13.6, c=0.32 hardening λ=4.00 ·102,∆ε p 0 =6.61 ·10 −3 PSR surface evolution η=0.40 1. The initial tensile curve at ε̇ = 10−3 s−1 was fitted, as shown in Fig. 4. This fitting was made using in-house developed Excel software assuming that cyclic hardening and creep strain-ratewere negligible under initial tensile loading. The initial tensile curvewas on-line fitted by numerically integrating the constitutive equations in the Excel software. Among the parameters, ε̇ p 0 was selected to be ε̇ p 0 = 10 −3 s−1, and k(i), responsible for ratcheting, was set to 3.0 (Ohno et al., 2016a). Theviscoplasticity exponentmhadalmost no influence on ε̇p at ε̇=10−3 s−1 because we selected ε̇ p 0 =10 −3 s−1 in Eq. (3.2)1. Thus, σ p 0, M, h (i), and ζ(i) were determined. 2. Variations in σ+peak with N in the fatigue tests at ε̇ = 10 −3 s−1 with ∆ε = 0.01, 0.007 and 0.004 were used to determine L, κ0, and φ(∆ε p). It is shown that the constitutive Effect of cyclic hardening on stress relaxation in SUS316HTP... 503 Fig. 4. Tensile stress-strain relationship at ε̇=10−3 s−1 model gives the following relationships to σ+peak in the absence of thermal recovery of cyclic hardening (Appendix A)4 σ+peak −σ ini +peak σsat+peak −σ ini +peak =1− exp(−Lp) σsat+peak −σ ini +peak σini+peak =φ(∆εp)κ0 (4.1) whereσini+peak andσ sat +peak indicate the initial and saturated values ofσ+peak . Equations (4.1) were used to determineL,κ0, andφ(∆ε p) (Figs. 5a and 5b). The following formofφ(∆εp) was found appropriate in the present study φ(∆εp)= 1− exp(−λ∆εp) 1− exp(−λ∆ε p 0) (4.2) whereλ is a fitting parameter, and∆ε p 0 is selected to be equal to the saturated viscoplastic strain range in the fatigue test at ε̇=10−3 s−1 and∆ε=0.01. Fig. 5. (a) Change in σ+peak with the accumulated viscoplastic strain p and (b) dependence of saturated σ+peak on the viscoplastic strain range∆ε p in the fatigue tests at ε̇=10−3 s−1 3. The thermal recovery parameters R and ω in Eq. (3.7) were determined to represent the effect of strain hold time th on σ+peak atN ≈ 20 in the creep-fatigue tests at ε̇=10 −3 s−1 with th =10min and 60min at∆ε=0.01 (Appendix B). 4Goodall et al. (1981) first showed Eq. (4.1)1 for fitting the tensile peak stress data of 316 stainless steel at 600◦C. 504 N. Ohno et al. 4. The Norton-stage data under strain holding at N =1 and 20 in the creep-fatigue test at ε̇=10−3 s−1 with th =60min at∆ε=0.01 were fitted, as shown by the solid and dashed lines in Fig. 2a. The fitting at N = 1 was used to determine σc0 and n in Eq. (3.2)2 by selecting ε̇c0 = 10 −3 s−1 with negligible cyclic hardening, θc ≃ 0, at N = 1. The fitting at N =20was then used to estimate c in Eq. (3.9) to reproduce the small increase in σ10E−7 depicted in Fig. 3 (Appendix C) c≃ σN=2010E−7/σ ini 10E−7−1 σN=20+peak/σ ini +peak −1 (4.3) where σN=2010E−7 and σ N=20 +peak denote the values of σ10E−7 and σ+peak atN =20, respectively. 5. The saturated hysteresis loops in the fatigue tests at ε̇ = 10−3 s−1 and 10−4 s−1 at ∆ε=0.01 were fitted to determine the viscoplasticity exponentm (Fig. 6). Fig. 6. Saturated stress-strain hysteresis loops in the fatigue tests at ε̇=10−3 s−1 and 10−4 s−1 at∆ε=0.01 5. Comparison of simulated and experimental results The creep-fatigue tests were simulated using the constitutive model described in Section 3 with the material parameters given in Table 4. The constitutive model was implemented in Abaqus using a user subroutine UMAT by extending the UMAT program developed by Ohno et al. (2016b, 2017b). From here on, t∗ denotes the time elapsed after the onset of strain holding, and σrelax indicates the stress attained at the end of stress relaxation under strain holding. It is restated that the cyclic softening following cyclic hardening is disregarded in the constitutive model. This limitation is allowed in simulating the transient and Norton stages affected diffe- rently by cyclic hardening. Accordingly, this Section compares the simulated and experimental results with emphasis on the stress relaxation behavior under strain holding at cycles where cyclic softening was not significant. The tensile peak stress variations and stress relaxation curves observed in the creep-fatigue tests at ε̇ = 10−3 s−1 with th = 60min at ∆ε = 0.007 and 0.01 were simulated with good accuracy, as shown in Figs. 7 and 8. The variations in σrelax with N in the two tests were also simulatedwell, thoughσrelax was slightly inaccurate in the case of∆ε=0.007 (Fig. 7).The stress relaxation becamemore significant as cyclic hardening developed in both the experimental and simulated results. Ignoring ε̇c in the constitutive model did not affect the transient stage under strain holding, but resulted in considerably under-predicting the stress relaxation, as shown in Figs. 9a and 9b in the case of ∆ε = 0.01 with th = 60min. Hence, accurate simulation of the Effect of cyclic hardening on stress relaxation in SUS316HTP... 505 stress relaxation shown inFigs. 7 and 8was owing to the dominance of ε̇p and ε̇c in the transient andNorton stages, respectively. Therefore, the addition of ε̇c to ε̇p and theNorton type of creep equation expressed asEq. (3.2)2 for ε̇ c enabled accurate simulation of the stress relaxation under strain holding. Fig. 7. Variations in σ+peak and σrelax withN and∆ε under creep-fatigue loading at ε̇=10 −3 s−1 with th =60min Fig. 8. Stress relaxation under creep-fatigue loading at ε̇=10−3 s−1 with th =60min at: (a)∆ε=0.01 and (b) ∆ε=0.007 Fig. 9. Effect of the creep strain-rate on (a) variations in σ+peak and σrelax withN and (b) stress relaxation atN =20 under creep-fatigue loading at ε̇=10−3 s−1 with th =60min at∆ε=0.01 506 N. Ohno et al. The effect of cyclic hardening on ε̇c was taken into account through θc in Eq. (3.2)2 in the constitutive model, and θc was assumed to be proportional to θp, θc = cθp, in Eq. (3.9). Figure 10 demonstrates the effect of c on the stress relaxation in the simulation of the creep- -fatigue test at ε̇ = 10−3 s−1 with th = 60min at ∆ε = 0.01. As shown in Fig. 10b, the stress relaxation at N = 20 was slightly over-predicted if c = 0, whereas it was noticeably under- -predicted if c = 1. If c = 0 cyclic hardening had no effect on ε̇c through θc, and if c = 1 cyclic hardening had the same effect on ε̇c and ε̇p. Selecting c = 0.32 (i.e., θc ≈ θp/3) was found to be appropriate for simulating the stress relaxation. It was thus shown that ε̇c was much less affected by cyclic hardening than ε̇p, leading to suggestion that inelastic deformation mechanisms different from viscoplasticity started to operate shortly after the onset of strain holding, as discussed in Section 2. Fig. 10. Effect of the cyclic hardening parameter c on (a) variations in σ+peak and σrelax withN and (b) stress relaxation atN =20 under creep-fatigue loading at ε̇=10−3 s−1 with th =60min at ∆ε=0.01 Figure 11 illustrates the effects of th on σ+peak and σrelax measured in the creep-fatigue tests at ε̇=10−3 s−1 with th =10min and60min at∆ε=0.01. InFig. 11a,σ+peak for th =0 is shown for reference. In the two testswith th =10and60min, the effect of th onσ+peak appeared slightly after the near-saturation of cyclic hardening, whereas the effect on σrelax became rather large with increasing N before the near-saturation of cyclic hardening. These experimental features were well reproduced by the constitutive model. The creep-fatigue tests at ∆ε = 0.007 with th =10 and 60min were also simulated accurately, though not shown here to save the space. Fig. 11. Effect of the strain hold time th on (a) variations in σ+peak and σrelax with N and (b) stress relaxation atN =1 and 20 under creep-fatigue loading at ε̇=10−3 s−1 at∆ε=0.01 Effect of cyclic hardening on stress relaxation in SUS316HTP... 507 The slight effect of th on σ+peak described above was successfully simulated because of the thermal recovery of cyclic hardening represented by the second term on the right-hand side of Eq. (3.7). Here, it is noted that the thermal recovery exponent ω in Eq. (3.7) is large (Table 4); as a result, the thermal recovery of cyclic hardening occurred non-linearly to yield the slight effect of th on σ+peak despite the factor of six difference in th in the two tests with th =10 and 60min. However, the comparatively large effect of th on σrelax was well simulated owing to ε̇ c expressed as Eq. (3.2)2, as depicted inFig. 11b. Because σrelax denotes the stress attained at the end of stress relaxation, it is seen fromFig. 11b that the difference in σrelax in the two tests was caused by the stress relaxation during 10¬ t∗ ¬ 60min in the test with th =60min; the stress relaxation from t∗ = 10min to t∗ = 60min was about 10MPa and 25MPa at N = 1 and 20, respectively. The stress relaxation during 10¬ t∗ ¬ 60min was in the Norton stage. Therefore, the difference in σrelax in the two tests was well simulated because of the Norton type of creep equation expressed as Eq. (3.2)2 for ε̇ c. The stress relaxation under strain holding became larger with the development of cyclic hardening or with the increase in the strain hold time, as shown in this Section. Goodall et al. (1981) observed this feature in creep-fatigue tests of 316 stainless steel at 600◦C, andNouailhas (1989) simulated the tests using a unifiedmodel of cyclic viscoplasticity. However, Goodall et al. (1981) andNouailhas (1989) did not notice the transient andNorton stages in stress relaxation, which were studied in this work; moreover, Nouailhas (1989) did not show stress relaxation curves under strain holding. 6. Concluding remarks In this work, the cyclic hardening and stress relaxation behavior of SUS316HTP was experi- mentally and numerically studied under cyclic loading with tensile strain holding at 700◦C. Creep-fatigue tests were performed to show that the slow stress-relaxation stage satisfying Nor- ton’s law followed the transient stress-relaxation stage under strain holding. The Norton stage was much less affected by cyclic hardening than the transient stage. Since the transient stage was rate-controlled by viscoplasticity in the presence of the increase in dislocation density in grains to cause cyclic hardening, it was suggested that inelastic deformation mechanisms, such as diffusion creep and grain boundary sliding, operated in the Norton stage. A cyclic viscoplastic-creepmodel was developed based on the experimental results described above. In this model, the inelastic strain-rate ε̇in was decomposed into viscoplastic and creep strain-rates, which were dominant in the transient and Norton stages in stress relaxation, re- spectively. The viscoplastic strain-rate ε̇p was expressed by incorporating the noticeable effect of cyclic hardening on the drag and back stresses, while the creep strain-rate ε̇c was ruled by Norton’s law andwas assumed to beweakly affected by cyclic hardening. Thematerial parame- ters in the constitutivemodel were determined to verify the decomposition of ε̇in into ε̇p and ε̇c, which were affected differently by cyclic hardening. Finally, the cyclic viscoplastic-creep model was used to simulate the creep-fatigue tests per- formed in the present study. The constitutive model successfully simulated the stress relaxation behavior in the presence of cyclic hardening, and the stress relaxation in the simulation became more significant as cyclic hardening developed, as observed in the creep-fatigue tests. This was owing to the dominance of ε̇p and ε̇c in the transient and Norton stages, respectively, resulting in the transient stage being much more affected by cyclic hardening than the Norton stage. The stress-relaxation curves were also accurately simulated, and the effect of th on the stress relaxation was attributed to ε̇c in the Norton stage. 508 N. Ohno et al. Appendix A. Change in tensile peak stress Let us consider rapid, uniaxial cyclic loading with th =0 to ignore ε̇ c. On the tension side, Eqs. (3.2)1, (3.4)1 and (3.6) give the following equation in the viscoplastic region, where ε̇ p ≃ ε̇ σ≃ (1+φκ) [ σ p 0 ( ε̇ ε̇ p 0 )1/m + α̃ ] (A.1) where α̃ indicates theuniaxial component of the cyclic-hardening-free back stress.Because cyclic hardening is negligibly small under the initial loading to the first tensile peak, Eq. (A.1) allows the initial tensile peak stress σini+peak to be expressed as σini+peak ≃σ p 0 ( ε̇ ε̇ p 0 )1/m + α̃ini+peak (A.2) where α̃ini+peak denotes the initial peak value of α̃. Here, let us assume that the tensile peak value α̃+peak does not change from α̃ ini +peak with the increasing N because α̃ is regarded as the back stress in the absence of cyclic hardening. Eqs. (A.1) and (A.2) thus provide σ+peak ≃ (1+φκ)σ ini +peak (A.3) When the thermal recovery of cyclic hardening is negligible under rapid cyclic loading, Eq. (3.7) is integrated to give κ=κ0[1− exp(−Lp)] (A.4) Hence, Eq. (A.3) leads to Eqs. (4.1). Appendix B. Determination of thermal recovery parameters Let us consider rapid, uniaxial cyclic loading with th 6= 0 to determine the thermal recovery parameters R and ω in Eq. (3.7). Let us suppose that κ decreases from κ+peak to κrelax under tensile strain holding, and that κ increases from κrelax to κ+peak under rapid cyclic loading in one cycle. Here, we assume that the second and first terms on the right-hand side in Eq. (3.7) are active under tensile strain holding and rapid cyclic loading, respectively, to provide κrelax = [κ 1−ω +peak +R(ω−1)th] 1/(1−ω) κ+peak =κrelax +(κ0−κrelax)[1− exp(−Lp ∗)] (B.1) where p∗ denotes the change in p due to rapid cyclic loading in one cycle p∗ =2 ( ∆ε− σrelax + |σ−peak| E ) (B.2) To determine R and ω using Eqs. (B.1), the tensile peak stresses at N = 30 in the creep- -fatigue tests at ε̇ = 10−3 s−1 with th = 10min and 60min at ∆ε = 0.01 are used to evaluate κ+peak10 and κ+peak60 using Eq. (A.3) as κ+peak10 = 1 φ (σ+peak10 σini+peak −1 ) κ+peak60 = 1 φ (σ+peak60 σini+peak −1 ) (B.3) where the subscripts 10 and 60 indicate th =10min and 60min. Then, κrelax10 and κrelax60 are calculated using Eq. (B.1)2. Here, it is noted that φ, κ0, and L are determined in Step 2 in Section 4. Finally,R andω are evaluated by numerically solving the following equations derived from Eq. (B.1)1 κ1−ω relax10−κ 1−ω +peak10 =R(ω−1)th10 th10 =600s κ1−ω relax60−κ 1−ω +peak60 =R(ω−1)th60 th60 =3600s (B.4) Effect of cyclic hardening on stress relaxation in SUS316HTP... 509 Appendix C. Cyclic hardening parameter for creep strain-rate To evaluate c in Eq. (3.9), we consider the changes in σ+peak and σ10E−7 with N in the creep- -fatigue test at ε̇ = 10−3 s−1 with th = 60min at ∆ε = 0.01. For σ+peak in this creep-fatigue test, Eq. (A.3) is valid, though κ is affected by th 6= 0. For σ10E−7, Eqs. (3.2)2, (3.6) and (3.9) provide σ10E−7 =(1+ cφκ)σ ini 10E−7 σ ini 10E−7 =σ c 0 ( ε̇c ε̇c0 )1/n (C.1) where ε̇c =10−7 s−1. The change in κ is considered small under strain holding when ω is large in Eq. (3.7). Thus, using Eqs. (A.3) and (C.1)1, c is estimated as c≃ σ10E−7/σ ini 10E−7−1 σ+peak/σ ini +peak −1 (C.2) Acknowledgment We thankMelissa Gibbons, PhD, fromEdanzGroup (www.edanzediting.com/ac) for editing a draft of this manuscript. References 1. Ainsworth R.A., 2006, R5 procedures for assessing structural integrity of components under creep and creep-fatigue conditions, International Materials Reviews, 51, 107-126 2. Chaboche J.L., 2008, A review of some plasticity and viscoplasticity constitutive theories, Inter- national Journal of Plasticity, 24, 1642-1693 3. Chaboche J.L., Dang Van K., Cordier G., 1979, Modelization of the strain memory effect on the cyclic hardening of 316 stainless steel, Proceedings of the 5th International Conference on Structural Mechanics in Reactor Technology, L, L11/3 4. Crossman F.W., Ashby M.F., 1975, The non-uniform flow of polycrystals by grain-boundary sliding accommodated by power-law creep,Acta Metallurgica, 23, 425-440 5. Goodall I.W., Hales R., Walters D.J., 1981, On constitutive relations and failure criteria of an austenitic steel under cyclic loading at elevated temperature, Creep in Structures, Springer- -Verlag, 103-127 6. Hales R., 1980, A quantitative metallographic assessment of structural degradation of type 316 stainless steel during creep-fatigue,Fatigue of Engineering Materials and Structures, 3, 339-356 7. Hales R., 1983, A method of creep damage summation based on accumulated strain for the assessment of creep-fatigue endurance, Fatigue of Engineering Materials and Structures, 6, 121-135 8. InoueT., IgariT.,OkazakiM., SakaneM.,TokimasaK., 1989,Fatigue-creep life prediction of 2.25Cr-1Mo steel by inelastic analysis,Nuclear Engineering and Design, 114, 311-321 9. Kang G., Ohno N., Nebu A., 2003, Constitutive modeling of strain range dependent cyclic hardening, International Journal of Plasticity, 19, 1801-1819 10. Kobayashi M., Mukai M., Takahashi H., Ohno N., Kawakami T., Ishikawa T., 2003, Implicit integration and consistent tangent modulus of a time-dependent non-unified constitutive model, International Journal for Numerical Methods in Engineering, 58, 1523-1543 11. Krausz A.S., Krausz K., 1996, Unified Constitutive Laws of Plastic Deformation, Academic Press, San Diego 510 N. Ohno et al. 12. Miller A., 1976, An inelastic constitutive model for monotonic, cyclic, and creep deformation: Part 1 – equations development and analytical procedures, Journal of Engineering Materials and Technology, 98, 97-105 13. Mróz Z., 1967, On the description of anisotropic workhardening, Journal of the Mechanics and Physics of Solids, 15, 163-175 14. Nam S.W., 2002, Assessment of damage and life prediction of austenitic stainless steel under high temperature creep-fatigue interaction condition,Materials Science and Engineering A, 322, 64-72 15. NouailhasD., 1989,Unifiedmodelling of cyclic viscoplasticity:Application to austenitic stainless steels, International Journal of Plasticity, 5, 501-520 16. OhnoN., 1982,Aconstitutivemodel of cyclic plasticitywith anonhardening strain region,Journal of Applied Mechanics, 49, 721-727 17. Ohno N., Abdel-Karim M., Kobayashi M., Igari T., 1998, Ratchetting characteristics of 316FRsteel at high temperature,Part I: strain-controlled ratchetting experiments and simulations, International Journal of Plasticity, 14, 355-372 18. Ohno N., Mizushima S., Okumura D., Tanie H., 2016a, Warpage variation analysis of Si/solder/Cu layered plates subjected to cyclic thermal loading, Advanced Methods of Continu- um Mechanics for Materials and Structures, Advanced StructuredMaterials, 60, 185-204 19. Ohno N., Tsuda M., Sugiyama H., Okumura D., 2016b, Elastic-viscoplastic implicit integra- tion algorithmapplicable to both plane stress and three-dimensional stress states,Finite Elements in Analysis and Design, 109, 54-64 20. Ohno N., Wang J.D., 1991, Transformation of a nonlinear kinematic hardening rule to a mul- tisurface form under isothermal and nonisothermal conditions, International Journal of Plasticity, 7, 879-891 21. Ohno N., Wang J.D., 1993, Kinematic hardening rules with critical state of dynamic recovery, Part I: formulation and basic features for ratchetting behavior, International Journal of Plasticity, 9, 375-390 22. OhnoN.,YamamotoR.,OkumuraD., 2017a,Thermo-mechanical cyclic hardeningbehavior of 304 stainless steel at large temperature ranges:Experiments and simulations, International Journal of Mechanical Sciences (article in press) 23. Ohno N., Yamamoto R., Sasaki T., Okumura D., 2017b, Resetting scheme for plastic strain surface in constitutive modeling of cyclic plasticity, ZAMM – Zeitschrift für Angewandte Mathe- matik und Mechanik (article in press, DOI: 10.1002/zamm.201700298) 24. Priest R.H., Ellison E.G., 1981, A combined deformation map-ductility exhaustion approach to creep-fatigue analysis,Materials Science and Engineering, 49, 7-17 25. Takahashi Y., 1998, Evaluation of creep-fatigue life predictionmethods for low-carbon nitrogen- added 316 stainless steel, Journal of Engineering Materials and Technology, 120, 119-125 26. Takahashi Y., Shibamoto H., Inoue K., 2008, Study on creep-fatigue life predictionmethods for low-carbon nitrogen-controlled 316 stainless steel (316FR), Nuclear Engineering and Design, 238, 322-335 27. TrampczynskiW., 1988,The experimental verificationof the evolutionof kinematic and isotropic hardening in cyclic plasticity, Journal of the Mechanics and Physics of Solids, 36, 417-441 28. YanX.L., ZhangX.C., Tu S.T.,Mannan S.L., XuanF.Z., LinY.C., 2015,Reviewof creep- -fatigue endurance and life prediction of 316 stainless steels, International Journal of Pressure Vessels and Piping, 126-127, 17-28 Manuscript received December 28, 2017; accepted for print March 1, 2018