| Abstract Scope |
In many high-performance steels, the key physical properties are closely related to the fraction and distribution of different phases, including martensite, ferrite and austenite, as well as the microstructural constituent bainite. Although electron backscatter diffraction (EBSD) can discriminate austenite from other phases based on crystallographic differences, the separation of bainite, martensite and ferrite is much more challenging as all 3 can be reliably indexed using a body-centered cubic (BCC) structure.
Here we introduce a technique utilizing machine learning to classify EBSD datasets. Using a combination of diffraction pattern quality, grain-based, and plastic strain measurements coupled with a system training process, the BCC-based phases can be differentiated quickly and reproducibly. We will show how this approach can be used to separate the constituent phases of the microstructure in a range of advanced steels, including dual-phase and highly heterogeneous structural steels. |