Prediction of cyclic damage in metallic alloys with crystal plasticity modeling enhanced by machine learning

Abstract

A computational approach consisting of two frameworks is introduced to construct a fatigue model for the Inconel 718 superalloy. Due to the extreme nature of Inconel 718’s typical service environment, fatigue is predicted to be a primary mode of failure for the material. To this end, a crystal plasticity framework is employed to assess the typical columnar (as-built) and equiaxed (heat-treated) microstructures for fatigue life under repeated loading, using the criterion of stored energy to predict life cycle. A machine learning framework is then introduced to work as an extension to the model by learning the relationship between stored energy, maximum stress, and fatigue life. By comparison of the results of both frameworks, fatigue life prediction accuracy is validated, and both approaches are established as accurate predictors of fatigue life for this family of superalloys.