Microstructural damage sensitivity prediction using spatial statistics

Abstract

The vast compositional space of metallic materials provides ample opportunity to design stronger, more ductile and cheaper alloys. However, the substantial complexity of deformation micro-mechanisms makes simulation-based prediction of microstructural performance exceedingly difficult. In absence of predictive tools, tedious experiments have to be conducted to screen properties. Here, we develop a purely empirical model to forecast microstructural performance in advance, bypassing these challenges. This is achieved by combining in situ deformation experiments with a novel methodology that utilizes n-point statistics and principle component analysis to extract key microstructural features. We demonstrate this approach by predicting crack nucleation in a complex dual-phase steel, achieving substantial predictive ability (84.8% of microstructures predicted to crack, actually crack), a substantial improvement upon the alternate simulation-based approaches. This significant accuracy illustrates the utility of this alternate approach and opens the door to a wide range of alloy design tools.