Optimization of Heat Treatment Process Parameters for 8Cr4Mo4V Bearing Ring Using FEA-NN- PSO Method

Abstract

This study combined finite element simulation and machine learning methods to optimize the heat treatment process parameters for 8Cr4Mo4V steel bearings. First, the stress evolution of quenching and tempering processes was numerically simulated. The stress during quenching is mainly influenced by thermal stress and phase transformation stress, which play dominant roles on the bearing surface before and after the martensitic phase transition, respectively. After quenching, the simulated retained austenite content was 18.7%, closing to the experimental value of 17.8%, verifying the accuracy of the simulation results. As the number of tempering cycles increased, the residual stresses generated by quenching were released. Based on the high-quality data obtained from finite element simulations, backpropagation neural network (BPNN) and generalized regression neural network (GRNN) were further applied to establish a heat treatment process-residual stress relationship model. By integrating the trained machine learning model with a particle swarm optimization algorithm (PSO) optimization algorithm, optimal heat treatment process parameters were successfully obtained. Validation simulations using the optimized parameters showed that the maximum radial residual tensile and compressive stresses in the bearing ring after heat treatment were reduced to 174 MPa and 201 MPa, respectively. This approach applicable to optimize heat treatment processes for other workpieces, offering broad prospects for engineering applications.

Graphical Abstract