Prediction method of root stress distribution in herringbone gears

WANG Cong; LENG Sheng; JIANG Zenghua; CHEN Fuxing; LU Fengxia

doi:10.16579/j.issn.1001.9669.2026.04.003

Abstract：[Objective]To achieve rapid prediction of the root stress distribution in herringbone gears under variable operating conditions and address the high computational cost of the finite element method,a prediction model for root stress distribution based on a BP neural network was constructed.[Methods]Firstly,finite element simulations were conducted on a herringbone gear pair under 20 sets of operating conditions with varying rotational speeds and load torques to obtain stress distribution data in the tooth root region.Secondly,tooth root stress tests were performed,and the simulation results were compared with the experimental data to verify the validity of the finite element model.Then,a stress subregional prediction method was adopted,in which the tooth root region was divided into three subregions,and BP neural network models were independently constructed and subsequently integrated to form a global prediction model.Finally,a set of operating conditions outside the training samples was selected,and both the finite element simulation and the proposed model were employed to obtain the root stress distribution and its maximum value,thereby validating the predictive performance.[Results]The results indicate that the root stress distribution predicted by the model is highly consistent with that obtained from finite element simulations.The prediction error for the maximum tooth root bending stress is 4.6%,and the mean squared error of the prediction is 3.017 MPa,demonstrating the effectiveness of the proposed model.This study provides a reference for the rapid evaluation of tooth root stress in herringbone gears under variable operating conditions.