A Data-Driven Scheme for Quantitative Analysis of Texture

Texture is the orientation distribution of crystallites in polycrystalline materials. Given the discrete orientations, Schaeben suggested to adopt statistics for quantitative analysis of texture from discrete orientations, and he also conceived a clustering algorithm to facilitate the applications of statistical methods (H. Schaeben, J Appl Crystal 26:112–121, 1993). This data-driven scheme becomes more urgent and more necessary for the oncoming fourth paradigm: data-intensive scientific discovery, which follows after experimental science, theoretical science, and computational science paradigm. This research adopts a density-based clustering algorithm, DBSCAN, to process the orientation data from an austenitic stainless steel 316 L sample fabricated by selective laser melting. It is validated that the algorithm can robustly identify the orientation cluster (or texture component or preferred orientation). The statistical methods can successfully quantify the features of the identified orientation cluster with quantified uncertainty (statistical significance), which is often lacked in the general method of orientation distribution function. It is believed that this data-driven scheme can be applied to the many aspects of texture analysis.