기어 전달오차 최소화를 위한 베이지안 최적화 설계

This paper proposes an optimal design method for minimizing gear transmission error using Bayesian Optimization to reduce gear noise and vibration. Transmission error is significantly influenced by gear geometry. Conventional ISO 6336 and AGMA 2101 analyses cannot fully reflect the actual tooth shape or contact ratio, while finite element analysis provides high accuracy but entails substantial computational cost due to fine meshing and nonlinear contact conditions. In this study, a simplified finite element model considering contact ratio was developed to improve computational efficiency. A Gaussian process surrogate model was constructed to predict transmission error using a limited number of samples, and the expected improvement criterion was employed to efficiently search for new optimal points. The effectiveness of the proposed optimization framework was verified by comparing its results with those obtained from a genetic algorithm, demonstrating that the Bayesian Optimization approach can achieve high accuracy with reduced computational effort.