Prediction of Double‐Pass Hot Deformation Behavior in 22Cr‐25Ni Austenitic Heat‐Resistant Steel Based on Multiple Machine Learning Models

Double‐pass hot compression tests of 22Cr‐25Ni austenitic heat‐resistant steel were conducted at the deformation temperature of 950–1150 °C, strain rate of 0.01–1 s −1 , and interpass time of 1–200 s. On this basis, four machine learning models including artificial neural network (ANN), random forest (RF), XGBoost, and LightGBM were established to predict the flow stress. XGBoost model achieved the optimal performance, with the coefficient of determination ( R 2 ) of 0.9988, root mean square error (RMSE) of 2.9514 MPa, mean absolute error (MAE) of 0.8711 MPa, and average absolute relative error (AARE) of only 0.0097, which can accurately characterize the evolution law of double‐pass flow stress under different process parameters. Feature importance revealed that deformation temperature (64%) and strain rate (29%) are the dominant influencing parameters, while strain (6.8%) and interpass time (0.2%) contribute less. Based on the data above, the static recrystallization (SRX) fraction during interpass times was further calculated, indicating that increasing deformation temperature, raising strain rate, and prolonging interpass time all promote the SRX process. This study provides data and theoretical support for the optimization of hot working processes of 22Cr‐25Ni steel and also offers a reference for the intelligent characterization of hot deformation behavior of other austenitic heat‐resistant steels.