Data-driven Bayesian model-based prediction of fatigue crack nucleation in Ni-based superalloys

Abstract This paper develops a Bayesian inference-based probabilistic crack nucleation model for the Ni-based superalloy René 88DT under fatigue loading. A data-driven, machine learning approach is developed, identifying underlying mechanisms driving crack nucleation. An experimental set of fatigue-loaded microstructures is characterized near crack nucleation sites using scanning electron microscopy and electron backscatter diffraction images for correlating the grain morphology and crystallography to the location of crack nucleation sites. A concurrent multiscale model, embedding experimental polycrystalline microstructural representative volume elements (RVEs) in a homogenized material, is developed for fatigue simulations. The RVE domain is modeled by a crystal plasticity finite element model. An anisotropic continuum plasticity model, obtained by homogenization of the crystal plasticity model, is used for the exterior domain. A Bayesian classification method is introduced to optimally select informative state variable predictors of crack nucleation. From this principal set of state variables, a simple scalar crack nucleation indicator is formulated.