Fault Diagnosis of Gearbox Based on Deep Residual Shrinkage Network in Noise Environment

Accurate fault diagnosis of gearbox is helpful to ensure the safe and stable operation of mechanical system, reduce the economic losses caused by faults and reduce the occurrence of catastrophic accidents. However, in practical tasks, the samples will contain some noise and irrelevant information, which may affect the effect of feature learning. In this paper, a fault diagnosis method of gearbox based on residual contraction neural network is designed, which can notice unimportant features through attention mechanism and set them to zero through soft threshold function, so as to strengthen the ability of depth neural network to extract useful features from noisy signals. The method is applied to the self built rotating machinery fault simulator, obtains excellent diagnosis results for gearbox faults under different noise environments.