Position error prediction for interference fit assembly of die insert rings in high-precision motor cores

The insert ring assembly, a core component of motor core progressive dies, experiences nonlinear deformation during interference fit assembly, which can lead to die accuracy failure. This paper proposes a neural network model based on a sparrow search algorithm (SSA)-optimized back propagation neural network (BPNN) to predict hole center position errors induced by insert ring interference-fit assembly in high-precision motor core progressive dies. First, a finite element simulation model for insert ring interference fit assembly was developed. A multi-software collaborative simulation platform was independently constructed to generate numerous geometric simulation models of multi-axis hole interference fit with varying structures, collecting corresponding position error data. Second, through insert ring interference fit experiments, a high-precision coordinate measuring machine (CMM) was employed to measure the center coordinates of arrayed circular holes on insert rings post-assembly. The radial displacement of the arrayed circular holes was calculated to validate the simulation accuracy. Finally, an SSA-BP model for position error prediction was built, using insert ring structural parameters as inputs and arrayed circular hole position errors as outputs. Results demonstrate that when predicting position error of insert rings with different structures using the SSA-BP neural network model, the prediction metrics achieved MAE = 6.0224e-05, MSE = 7.3169e-09, MAPE = 1.0082%, and R 2 = 0.997. Compared with conventional BPNN, the SSA-BP model shows 84%, 97%, 84%, and 9% improvements in MAE, MSE, MAPE, and R 2 respectively, significantly enhancing both prediction accuracy and computational efficiency. Therefore, this study provides technical guidance for die manufacturers in insert ring position error prediction, compensation, and optimization.