Machine Learning–Assisted Processing Map and Hot Deformation Behavior of 00Cr18Mn21Mo2Ni4N Austenitic Stainless Steel
TL;DR
This integrated flow‐stress modeling, processing‐map construction, and EBSD verification provides guidance for optimizing the hot deformation of this high‐nitrogen austenitic stainless steel within the investigated thermomechanical range.
Abstract
To investigate the hot deformation behavior and hot‐working window of 00Cr18Mn21Mo2Ni4N austenitic stainless steel, isothermal compression tests were performed at 900–1200 °C and strain rates of 0.01–10 s −1 . Arrhenius‐type, physics‐based, and GA‐BPNN constitutive descriptions were evaluated to obtain a reliable flow‐stress field for processing‐map construction within the dynamic material model framework. After validation against the measured flow curves, the GA‐BPNN model provided the closest agreement with the experimental stresses, with a mean absolute relative error of 4.24% and a correlation coefficient (R) of 0.9952. The model was therefore used as a data‐driven interpolation tool to support processing‐map analysis, rather than as a newly proposed machine‐learning architecture. The resulting stable and unstable processing domains were further examined by electron backscatter diffraction (EBSD), which linked the predicted hot‐working window to dynamic recrystallization, dynamic recovery, and microstructural heterogeneity. This integrated flow‐stress modeling, processing‐map construction, and EBSD verification provides guidance for optimizing the hot deformation of this high‐nitrogen austenitic stainless steel within the investigated thermomechanical range.