Mixed residual life prediction method of planetary gear reducer

Given the significant importance of health monitoring and life prediction research for planetary gear reducers, a hybrid life prediction method is proposed. The firefly algorithm (FA) is introduced to optimize the variational mode decomposition (VMD) process, thereby enhancing noise reduction of the vibration signal from the reducer shell. On this basis, multi-dimensional degradation features are extracted, and health indicators are constructed. Subsequently, a hybrid prognostic model for remaining useful life (RUL) is constructed, fusing the nonlinear Wiener process and the long short-term memory network (LSTM). Testing demonstrates that the proposed approach achieves MAE and RMSE values of 3.750 and 4.583. In comparison, the LSTM model yields values of 11.375 and 12.937, respectively, demonstrating that the proposed method significantly reduces prediction errors and achieves higher accuracy and reliability.