Selective Visualization of Martensite in Bainitic Steel Using Backscattered Electron Images and Phase Fraction Evaluation Using Machine Learning
Hiroshi Imoto, Kaoru Sato, Kenji Ogata
pp. 1-8
Abstract
Multi-phase steels are often used to realize a combination of high strength and toughness and/or ductility. To optimize their mechanical properties, it is vital to accurately evaluate the grain size, hard phase size and distribution, and dislocation density. In this paper, we studied a new method for evaluating the morphology and phase fraction of the hard phase, i.e., the martensite-austenite constituent (M-A), which is an important component that governs the mechanical properties of high strength steels. Using a scanning electron microscope, martensite can be selectively visualized with a bright contrast by collecting high-angle backscattered electrons. This method identifies only martensite in isolation from other phases, whereas both martensite and austenite are highlighted with the conventional two-step etching method. In addition, machine learning image analysis allows accurate extraction of martensite even in the presence of inhomogeneous backscattered electron image contrast in the matrix. This method provides an accurate and simple evaluation of the morphology of martensite in multi-phase steels over a large area.