Learn Generalization Feature via Convolutional Neural Network: A Fault Diagnosis Scheme Toward Unseen Operating Conditions

Yuantao Yang,Jiancheng Yin,Yushu Chen,Yuqing Li,Huailiang Zheng,Minqiang Xu

doi:10.1109/access.2020.2994310

Yuantao Yang, Jiancheng Yin + Show 4 more

Open Access

https://doi.org/10.1109/access.2020.2994310

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2020
Citations: 25	License type: CC BY 4.0

Affiliation: Harbin Institute of Technology

Abstract

In recent years, Convolutional neural networks (CNNs) have achieved start-of-art performance in the fault diagnosis field. If there is no available information on the unseen operating conditions, the model trained on the seen operating condition cannot perform well. One of the feasible strategies is to enhance the generalization ability of the network on various seen operating conditions. We introduce the center loss to the traditional CNN and build an end-to-end fault diagnosis framework (called CNN-C). By minimizing the intra-class variations, center loss cluster the learned features across various seen operating conditions. With the joint supervision of the center loss and the softmax loss, the learned features of the same class could minimize the domain difference across various seen operating conditions while the features of different classes are separable. The generalization ability of network is improved on unseen operating conditions. Compared with the shallow methods and traditional CNN, the proposed method is promising to deal with the fault diagnosis tasks of the bearing and gearbox.

Highlights

Fault diagnosis of rotating machinery is curial to reduce maintenance costs and improve the safety and reliability of the systems [1]–[3]
We focus on the scenario that the fault diagnosis tasks on the unseen operating conditions
The main contributions of the proposed method can be summarized as follows: (1) We propose Convolutional neural networks (CNNs)-C network with the center loss to enhance the property of clustering within-class and the property of distinction among the classes

Summary

INTRODUCTION

Fault diagnosis of rotating machinery is curial to reduce maintenance costs and improve the safety and reliability of the systems [1]–[3]. Based on the plenty of the vibration signals are collected by the sensors, signal processing skills and deep learning algorithms are utilized to extract the fault characteristics of different health conditions These features are extracted from the seen operating conditions, which means that both the training samples and testing samples are followed the same distribution. It is necessary to discuss the performance of these well-trained models on the unseen operating conditions In this situation, there is an immediate need is to further improve the generalization ability of the fault diagnosis framework. Different from [14], we improve the generalization ability of the network by eliminating the distribution difference across the various operating conditions, instead of utilizing tricks and strategies during the training progress To this end, we introduce a regularization term called center loss into the network objective function, which intently restricts the clustering property of the learned features.

CNN WITH THE CENTER LOSS

NETWORK ARCHITECTURE

PROPERTIES OF THE PROPOSED METHOD

EXPERIMENTAL VERIFICATION

EXPERIMENTAL SETUP

Findings

CONCLUSION