Just another WordPress site - Ruhr-Universität Bochum

Abstract

In this study, a machine learning-based technique is presented to automatically identify dislocation structures from experimental Laue microdiffraction patterns. We present a framework that may be used for direct comparisons between experimental Laue microdiffraction patterns and virtual diffraction patterns based on discrete dislocation dynamics simulations. The mapping of patterns to dislocation structures from a catalog of structures requires an indexing scheme. We present such a scheme based on clustering techniques that can be browsed and compared to experimentally observed patterns in an automated fashion. The effectiveness of the model along with domain expertise from the underlying physics will ultimately provide a better insight into dislocation structure evolution and, therefore, material behavior.