Remaining Useful Life Prediction of Gearbox Using Discrete Wavelet Transform and BP Neural Network

In the realm of gearbox research, predicting the remaining service life of gear shafts is a crucial task. This paper investigates the essential techniques for gear shaft life prediction. Firstly, the time synchronous averaging algorithm and discrete wavelet transform are implemented to preprocess and extract features from the vibration signals of the gear shafts. Secondly, to more accurately reflect the changing patterns of remaining service life of gear shafts, the vector autoregression (VAR) model, which considers the mutual effects of different independent variables, is employed to obtain the residual of the entire dataset. Finally, the Back-Propagation (BP) neural network was employed to forecast the remaining operative lifespan of the gear shaft, precluding possible safety hazards and financial losses. The relative error between the predictive output and actual values of the model was within 6. The research results indicate that this technique enables precise prognostication of the service life of gear shafts, offering sound support for the maintenance and preservation of mechanical equipment.