Modeling of the hot-deformation behavior of Fe-Ni-Al maraging steel using constitutive equations and machine learning algorithms
Haiyan Wang,Xueyun Gao,Wenbo Fan,Lei Xing,Liangeng Hua
DOI: https://doi.org/10.1007/s10853-024-10179-5
IF: 4.5
2024-09-11
Journal of Materials Science
Abstract:The establishment of a constitutive model capable of accurately predicting the flow stresses for materials in hot-deformation processes can greatly aid in the design and optimization of hot-working procedures. In the present work, hot-compression experiments of low-cost Fe-Ni-Al maraging steel were performed in the temperatures ranging from 900 to 1150 °C and strain rates of 0.01 s −1 to 10 s −1 . The conventional constitutive models, i.e., modified-ZA model, optimized-ZA model and Arrhenius model, were established using the experimental data. Meanwhile, multiple ML models, namely RF, GBR, XGB, SVR, DT, KRR and DNN, were developed to map the flow stresses to temperature, strain and strain rate. A thorough comparison demonstrates that the DNN model surpasses the traditional models and other ML approaches in predicting the experimental values. In addition, the accurate prediction of stresses from the DNN model for dense temperature intervals facilitates the construction of more accurate processing maps.
materials science, multidisciplinary