Abstract:In this paper, the true stress-strain data of Haynes 282 alloy were obtained by conducting isothermal compression tests on a thermal-mechanical simulator.Haynes 282 alloy shows typical dynamic recrystallization characteristic during the deformation process at elevated temperature.Moreover, the flow stress was quite sensitive to the thermodynamic parameters and represents complicated highly-nonlinear relationship with the thermodynamic parameters. In order to accurately describe and predict the true stress-strain relationship of Haynes 282 alloy, a back-propagation neural network was constructed by employing hot deformation parameters as the inputs, and employing flow stress as the output. The evaluation results of the constructed neural network show that the constructed neural network in this research can accurately characterize the hot flow behavior of Haynes 282 alloy. Accurate simulation of the hot flow behavior of Haynes 282 alloy is achieved by implanting the neural network into a finite element software in the form of material subroutine and constructing the finite element model of isothermal compression test.