Automated Grain Yield Behavior Classification

Abstract

A method for classifying grain stress evolution behaviors using unsupervised learning techniques is presented. The method is applied to analyze grain stress histories measured in situ using high-energy x-ray diffraction microscopy from the aluminum–lithium alloy Al-Li 2099 at the elastic–plastic transition (yield). The unsupervised learning process automatically classified the grain stress histories into four groups: major softening, no work-hardening or -softening, moderate work-hardening, and major work-hardening. The orientation and spatial dependence of these four groups are discussed. In addition, the generality of the classification process to other samples is explored.