Use of neural networks to predict rear axle gear damage

Accurate rear axle damage prediction is very difficult because of the rotating speeds and the changing loads when the truck is running. In this paper, a new method, which consists of a data pretreatment (recursive processing) and artificial neural networks, is proposed to accurately predict rear axle damage. Simulated and the experimental results have shown the proposed method has relatively high prediction accuracy, and through comparison with traditional time series forecasting methods using the same parameters of vibration, it was found that the performance of artificial neural networks is better in forecasting accuracy. This study provides a new approach for predicting remaining gearing life.