Approach for Stabilized Peak/Valley Stress Modeling of Non-Isothermal Fatigue of a DS Ni-Base Superalloy

Abstract

Abstract Turbine blades derived from directionally solidified (DS) Ni-base superalloys are increasingly employed in the first and second stages of gas turbine engines, where thermal and mechanical cycling facilitate cyclic plasticity and creep. The elongated grains, which are aligned with the primary stress axis of the component, provide (1) greater creep ductility, and (2) lower minimum creep rate in solidification direction compared to other directions. Tracking the evolution of deformation in these structures necessitates a constitutive model having the functionality to capture rate, temperature, history, and orientation dependence. Historically, models rooted in microstructurally based viscoplasticity simulate the response of long-crystal, dual-phase, Ni-base superalloys with extraordinary fidelity; however, a macroscopic approach having reduced order is leveraged to simulate low-cycle fatigue (LCF), creep, and creep-fatigue responses with equally high accuracy. This study applies uncoupled creep and plasticity models to predict the thermomechanical fatigue (TMF) of a generic DS Ni-base, and an anisotropic yield theory accounts for transversely isotropic strength. The microstructure of the subject material contains γ-matrix (FCC Ni) and γ′-particles (cuboidal Ni3Al). Because of the fully analytic determination of material constants from mechanical test data, the model can be readily tuned for materials in either peak- or base-loaded units. Application of the model via a parametric study reveals trends in the stabilized hysteresis response of under isothermal fatigue, creep fatigue, idealized thermomechanical fatigue, and conditions representative of in-service components. Though frequently considered in design and maintenance of turbine materials, non-isothermal fatigue has yet to be accurately predicted for a generalized set of loading conditions. The formulations presented in this study address this knowledge gap using extensions of traditional Ramberg-Osgood and Masing models.