Pitting Life Optimization for Helical Gears Based on Interfacial Lubrication and Machine Learning

Abstract Helical gears play a crucial role in mechanical transmission due to their superior load-bearing capacity and stability. However, achieving high-performance lubrication and extending pitting life remain significant challenges. This study develops a mixed elastohydrodynamic lubrication (EHL) model and introduces a novel pitting life optimization framework with machine learning. The methodology integrates a contact line percentage method for meshing parameters, a discrete convolution-fast Fourier transform (DC-FFT) for surface deformation, and Zaretsky's theory for predicting pitting life. Data-driven approaches, including back propagation neural network (BPNN), random forest (RF), gradient boosting decision tree (GBDT), and non-dominated sorting genetic algorithm-II (NSGA-II) are employed for prediction and multi-parameter optimization. Comprehensive investigations reveal that structural parameters exhibit varying degrees of influence, and coordinated adjustment significantly enhances fatigue reliability while restricting volume, providing guidance for practical design.