Fault diagnosis algorithm based on multi-channel neighbor feature convolutional network

Introduction The health status of bearings is an essential prerequisite to ensure the safe and stable operation of vehicles. However, the negative impact of covariate shifts among data channels on diagnostic accuracy is an issue that is often overlooked in data-driven algorithms based on multi-channel data. Therefore, extracting the most representative features from multi-channel data is key to achieving highprecision fault diagnosis. Methods To address this issue, this paper proposes a fault diagnosis algorithm based on a multi-channel neighbor feature convolutional network. First, to mitigate the covariate shift problem in the data, inverted mel-scale frequency cepstral coefficients are introduced to obtain domain-invariant features with high recognition accuracy. Furthermore, to fully leverage multichannel data and extract more discriminative features, we design a multi-channel adjacent feature convolutional module. This module employs sparse mapping to extract local neighboring features while preserving global constraint characteristics. Result and discussion Experiments are carried out on Xi’an Jiaotong University and Case Western Reserve University data. The results show that the proposed method can achieve high performance and high precision fault diagnosis.