Life Prediction Feasibility in Tmf via Stress/strain Data from a Viscoplasticity-based Numerical Model

Contemporary computing packages handle a wide variety of stress analysis types, but are yet to provide an optimal way to handle certain load cases and geometries. Blades in gas turbine systems, for instance, undergo repetitive thermal and mechanical load cycles of varied shape and phasing. Complexly-shaped airfoils create non-uniform stress paths that exacerbate the problem of FEA software attempting to determine the correct states of stress and strain at any point during the load history. This research chronicles the update and integration of Miller’s original viscoplasticity model with ANSYS finite element analysis software. Elevated temperature strain-controlled LCF and strain-controlled TMF loadings were applied to single-element, uniaxial simulation runs and the results were then compared to data from duplicate experimental testing. Initial findings indicate that the model maintains significant accuracy through several cycles, but longer tests produce varying error in hysteretic response. A review of the modernized implementation of Miller’s viscoplasticity model is presented with a focus on modifications that may be used to improve future results. INTRODUCTION Efficient gas turbine operation without the need for overly conservative service intervals is of paramount importance to the energy and aerospace industries. Thermomechanical Fatigue (TMF) -capable models are a core essential in creating accurate numerical simulations that ultimately can be used as a lifeprediction tool for turbine components [1]. It has been theorized that contemporary computing packages can be used to augment viscoplasticity models that display a wide range of applicability. While it is not expected that such a constitutive model could properly predict times for fracture initiation or failure, it is reasoned that predictions of approximate stress/strain states in hardening, softening, or stable regions during the lifetime of a part are quite useful. The model selected for review in this study is the 1976 Miller viscoplasticity model, which has been demonstrated to be accurate in a variety of monotonic, cyclic, high-temperature, and creep loadings [2]. The commercial computing package ANSYS was utilized to supply loadings that simulate elevated temperature low cycle fatigue (LCF) as well as TMF to the model. While Miller’s model does not explicitly support nonisothermal cases, ANSYS can supply the model updated temperature-dependent parameters when it passes the boundary conditions with each successive simulation step [3]. Although TMF loadings can incorporate many additional submechanisms and interactions not present in LCF, [4-6] it is reasoned that simulations of the current level of sophistication can already meet an intermediary goal of providing accurate initial stress/strain responses and stress histories through the first 100 cycles of a load history. In the present study, simulation data is gathered from both Miller's unaltered model and the ANSYS augmented model under elevated LCF and TMF conditions. These results are then compared with a mixture of historical and new experimental data from matching load conditions. It is shown that the ANSYS-adapted Miller model maintains a notable degree of accuracy for simulated fully-reversed cyclic loadings in steel with significant plasticity at elevated temperatures. However, examination of the hysteresis loops and stress histories beyond the region where initial work-hardening occurs reveals increasing error versus the experimental LCF cases. For TMF load simulations, both in-phase (IP) and out-ofphase (OP) TMF cases with similar fully-reversed strain ranges initially match the stress and strain responses of some similar experimental data well [7]. Even so, successive cycling leads to error that increases versus the experimental data earlier in the load history than in the LCF case. A mismatch or misformulation of the parameters that govern the isotropic and Proceedings of ASME Turbo Expo 2012 GT2012 June 11-15, 2012, Copenhagen, Denmark 1 Copyright © 2012 by ASME GT2012-69138 Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/06/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use 2 Copyright © 2012 by ASME kinematic hardening behavior may constitute the driving mechanism behind the progressive error in both the TMF and LCF cases. The handling of the non-isothermal loads externally specifically seems to inadvertently induce an artificial kinematic hardening effect not observable in the LCF cases, as the yield surfaces can be observed to be translating with each successive cycle and increasing the peak stress errors asymmetrically.