An artificial neural network constitutive model to predict high temperature flow behaviour in 18Ni(250) maraging steel

Hot compression experiments and microstructure observation investigations are utilized to analyze the hot deformation behavior and flow characterization of 18Ni(250) maraging steel. Arrhenius model, Strain-Compensated Arrhenius-type (SCA) model, Johnson-Cook (JC) model, Zerilli-Armstrong (ZA) model and ANN model were developed for forecasting the flow characteristics, the prediction of each constitutive model was quantitatively assessed using statistical parameters. To determine the ideal deformation settings, two-dimensional and three-dimensional hot deformation activation energy maps were created, and the affect of deformation parameters on the development of the microstructure was demonstrated. The result shows that the ANN model's coefficient of determination ( R ² ) is 99.679 % and average relative error ( ARE ) is 2.43 %, indicating that it has a greater prediction accuracy than other constitutive models. The dynamic recovery and flow localization in various deformation areas are examined in conjunction with the activation energy maps, and the optimal hot processing window were achieved. Embed the ANN constitutive model into the expert system to develop the “Forging Forming Force Prediction Module”. This module can calculate the stress and forging forming force in real time according to different deformation conditions, providing a theoretical basis for the selection of forging equipment and the verification of the effectiveness of the process scheme, and improving the intelligent manufacturing level of maraging steel forgings.