Advanced microstructural characterization in high-strength steels via machine learning-enhanced high-speed nanoindentation and EBSD mapping

Abstract

This research investigates the nanoscopic features of Advanced High-Strength Steels (AHSS) through a bottom-up approach employing high-speed nanoindentation mapping (HSNM) to elucidate structure-property relationships. The influence of grain boundaries on nanomechanical properties was documented, highlighting the challenge of SEM-EBSD analysis in differentiating phases with identical crystal structures (BCC, FCC, etc.). Integrating SEM-EBSD with HSNM in the same region of interest is essential for detailed insights into phase/microstructure distribution and accurate grain boundary identification. A modular four-step analysis protocol, designed and validated on ferritic-bainitic TRIP steels (TBF), leverages machine learning-enhanced HSNM for significant advancements in AHSS design. The initial phase involves the application of the expectation-maximization algorithm for probability distribution fitting of HSNM data, deriving primary mechanical phase statistics. This exclusively facilitates the correlation of elastic modulus and hardness for each phase/microstructure using nanoindentation data. Further refinement of phase/microstructure to mechanical property correlations was achieved through a supervised machine learning approach, ensuring precise association between EBSD and nanoindentation data. This includes detailed image analysis and clustering of nanoindentation data, enhancing the precision in phase recognition. This methodology addresses the critical challenges in developing 3rd Generation AHSS, aiming to fill the gap in accurately identifying and quantifying phases such as martensite, austenite, bainite, and ferrite, thereby reducing classification and measurement uncertainties. The approach contributes to the fundamental understanding of AHSS microstructures and provides a scalable framework for the comprehensive characterization of structural materials.