Original scientific paper

https://doi.org/10.21278/TOF.473048022

Fault Diagnosis Based on Optimized Wavelet Packet Transform and Time Domain Convolution Network

Dengxue Cao orcid.org/0009-0000-7602-3215 ; School of electrical and electronic engineering, Shanghai Institute of Technology, Shanghai, China
Yu Gu ; School of electrical and electronic engineering, Shanghai Institute of Technology, Shanghai, China
Wei Lin ; School of electrical and electronic engineering, Shanghai Institute of Technology, Shanghai, China

Abstract

In the past, the method of combining signal processing and neural network was widely used in the fault diagnosis of mechanical rolling bearings to achieve the purpose of fault signal detection and identification. However, traditional methods cannot fully extract fault features. In this paper, taking mechanical rolling bearings as an example, a new rolling bearing fault diagnosis model is proposed by combining the improved wavelet packet transform with time domain convolution network. The improved Northern Goshawk algorithm is used to optimize the wavelet packet transform and analyse the fault signal; then, the time domain convolution network is used to establish a fault diagnosis model. When compared with the wavelet packet decomposition model optimized by the improved Northern Goshawk algorithm, the accuracy of the improved wavelet packet transform model in detecting inner ring faults and rolling element faults is found to be increased by 8.3% and 7.4%, respectively.

Keywords

Wavelet packet transform; Fault diagnosis; Time domain convolution network; Rolling bearings

Hrčak ID:

305466

URI

https://hrcak.srce.hr/305466

Publication date:

5.9.2023.

Visits: 578 *