Microstructure-induced fatigue scatter of additively manufactured inconel 718: Insight from multilevel simulations and dislocation-based strain gradient crystal plasticity

Abstract

The fatigue performance of additively manufactured (AM) Inconel 718 is intrinsically governed by its grain morphology, necessitating a predictive understanding of the underlying plasticity-dominated mechanisms. To address this challenge, this study applies an integrated multilevel computational framework that explicitly bridges the process–structure–property–performance chain by coupling finite-element and cellular-automata (FE–CA) simulations of grain growth during laser powder bed fusion (LPBF), a deep neural network (DNN) for efficient material parameter calibration, and a strain-gradient crystal plasticity finite element (CPFE) model for fatigue life prediction. This unified framework enables, for the first time, a rigorous like-for-like comparison of three characteristic AM microstructures—equiaxed, columnar, and mixed grains—under a consistent computational and experimental calibration protocol, and thereby reveals new micromechanical insights into potential fatigue damage initiation from the plasticity perspective. Our simulations indicate that fatigue resistance is predominantly controlled by grain morphology and further modulated by morphology-induced anisotropy. Among them, equiaxed grains exhibit superior fatigue resistance to columnar and mixed grain morphologies, which is attributed to the activation of multiple slip systems and the resulting homogeneous deformation. In contrast, the strong texture in columnar grains gives rise to a pronounced “channeling effect”, leading to highly localized slip and a mismatch between regions of elevated plastic strain and actual damage accumulation. In terms of loading direction, the fatigue resistance under loading along the building direction (BD) is higher than that under loading along the transverse direction (TD). Crack initiation is predominantly predicted at high-angle grain boundaries and triple junctions, with the specific patterns highly sensitive to both grain morphology and loading direction. A key finding is the identification of a critical fatigue indicator parameter (FIP) threshold, beyond which fatigue life scatter intensifies significantly. While the CPFE model provides accurate predictions at intermediate strain amplitudes, its efficacy diminishes at higher strains due to the activation of alternative failure mechanisms. Overall, by integrating established computational methods, this work provides microstructure-sensitive insights and a practical framework for fatigue life prediction of AM materials, offering a potential pathway for AM process and microstructure optimization to achieve superior fatigue performance.