Data-Driven Damage Model Based on Nondestructive Evaluation

Abstract

A computational damage model, which is driven by material, mechanical behavior, and nondestructive evaluation (NDE) data, is presented in this study. To collect material and mechanical behavior damage data, an aerospace grade precipitate-hardened aluminum alloy was mechanically loaded under monotonic conditions inside a scanning electron microscope, while acoustic and optical methods were used to track the damage accumulation process. In addition, to obtain experimental information about damage accumulation at the laboratory scale, a set of cyclic loading experiments was completed using three-point bending specimens made out of the same aluminum alloy and by employing the same nondestructive methods. The ensemble of recorded data for both cases was then used in a postprocessing scheme based on outlier analysis to form damage progression curves, which were subsequently used as custom damage laws in finite element (FE) simulations. Specifically, a plasticity model coupled with stiffness degradation triggered by the experimentally defined damage curves was used in custom subroutines. The results highlight the effect of the data-driven damage model on the simulated mechanical response of the geometries considered and provide an information workflow that is capable of coupling experiments with simulations that can be used for remaining useful life (RUL) estimations.