DenMap single crystal solidification structure feature extraction: Automation and application

Abstract

A fully automatic, standardised, and validated dendritic core detection algorithm, DenMap, is applied to analyse CMSX-4® and CMSX-10® single crystal microstructure. Through the utilisation of novel imaging techniques such as Gaussian deconvolution and minimisation-based thresholding, the newly enhanced version of DenMap demonstrates a significant improvement in detection accuracy and range of applications. The results indicate dendrites preferentially form channels, with array disorder increasing with height from the chill plate. A new twisting phenomenon has been observed that may have significant impact on the development of low and high angle grain boundaries and porosity formation within the microstructure. DenMap provides the foundation for 3D reconstruction methodologies and advanced pattern recognition techniques. This new standardised approach will facilitate the development of machine learning algorithms and eventually AI for single crystal solidification studies, particularly for the purpose of process optimisation.