Early Weak Fault Signal Enhancement and Recognition Method of Rudder Paddle Bearings Based on Parameter Adaptive Stochastic Resonance

Abstract

Aiming at the problem that the characteristic frequency amplitude of early bearing fault signals is weak and difficult to extract under the strong noise, a vibration signal enhancement method based on maximum overlapping wavelet packet transform (MODWPT) and adaptive stochastic resonance (SR) is proposed. Based on multiple single-component signals decomposed by MODWPT, a signal enhancement and reconstruction model of the ordinary variable scale SR system is constructed. At the same time, the whale optimization algorithm (WOA) algorithm is adopted to optimize the parameters of the SR system adaptively with the improved signal noise ratio (ISNR) index as the fitness function. Finally, a feature-enhanced version of the original signal is obtained and transformed into a time-frequency image dataset using a two-dimensional wavelet transform, which can be input into the Resnet network for fault identification and classification. Based on a set of early fault data from IMS and marine bearing data, the significant effect of the proposed method on early fault signal feature enhancement is verified. By comparing the fault characteristic frequency and its frequency doubling before and after the enhancement of the envelope spectrum, the component noise component is suppressed, and the fault frequency amplitude is highlighted under the action of SR potential function. This method can effectively improve the fault recognition accuracy, which has a certain application prospect.

Keywords

• Signal enhancement
• stochastic resonance
• maximum overlapping wavelet packet transform
• resnet