Multi-objective process parameter optimization for GH2132 superalloy milling based on IPSO-BP neural network

Tool wear significantly affects cutting performance and workpiece quality, while also leading to increased machining process complexity and diminishing tool longevity. Consequently, identifying the appropriate process parameters to guarantee high efficiency and quality in material processing has emerged as a significant topic of study. This research enhanced the Back Propagation Neural Network (BPNN) method with an improved particle swarm optimization (IPSO-BP) algorithm. A multivariate model was subsequently created to investigate the correlation between material removal rate and tool wear rate, which was validated with both experimental data and network-predicted values. This study employed an orthogonal experimental design to evaluate the IPSO-BP algorithm against both the traditional BP model and factory empirical baseline in terms of solving accuracy and practical machining performance. The findings showed that the developed multivariate model attains exceptional prediction accuracy. Furthermore, this study took the removal rate of GH2132 superalloy and the tool wear rate as optimization objectives, formulated a corresponding fitness function, solved it using the IPSO-BP algorithm, and ultimately identified the optimal process parameters: cutting speed v  = 30.23 m/min, feed rate f  = 0.038 mm/rev, and depth of cut a _p  = 0.35 mm. Under these optimized parameters, the validation error of tool wear rate was 4.21%, and the machined surface showed improved topographical uniformity compared with that obtained using the factory empirical parameters, confirming that the proposed method can enhance machining efficiency and surface integrity.