Data Preprocessing Techniques in Convolutional Neural Network Based on Fault Diagnosis Towards Rotating Machinery

Yong Zhu,Shengnan Tang,Shouqi Yuan

doi:10.1109/access.2020.3012182

Yong Zhu, Shengnan Tang + Show 1 more

Open Access

https://doi.org/10.1109/access.2020.3012182

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2020
Citations: 79	License type: CC BY 4.0

Affiliation: Jiangsu University, Zhejiang University

Abstract

Rotating machinery plays a critical role in many significant fields. However, the unpredictable machinery faults may lead to the severe damage and losses. Hence, it is of great value to explore the precise approaches for fault diagnosis. With the development of the intelligent fault diagnosis methods based on deep learning, convolutional neural network (CNN) has aroused the attention of researchers in machinery fault diagnosis. In the light of the reduction of difficulty in feature learning and the improvement of final diagnosis accuracy, data preprocessing is necessary and crucial in CNN-based fault diagnosis methods. This review focuses on CNN-based fault diagnosis approaches in rotating machinery. Firstly, data preprocessing methods are overviewed. Then, we emphatically analyze and discuss several main techniques applied in CNN-based intelligent diagnosis, principally including the fast Fourier transform, wavelet transform, data augmentation, S-transform, and cyclic spectral analysis. Finally, the potential challenges and research objects are prospected on data preprocessing in intelligent fault diagnosis of rotary machinery.

Highlights

Rotating machinery is widely used in many fields, including aerospace, navigation, wind turbines and so on
APPLICATIONS OF DATA PREPROCESSING APPROACHES IN convolutional neural network (CNN)-BASED FAULT DIAGNOSIS With the increase of mechanical failure data and coupling complexity, intelligent fault diagnosis on the basis of deep neural network (DNN) has drawn the interest of researchers owing to the potent capability of extraction and admirable learning ability [61]–[63]
By means of discrete wavelet transform (DWT), raw data were transformed into time-frequency matrixes, which were taken as input of the following CNN model for fault diagnosis of planetary gearbox [82]

Summary

INTRODUCTION

Rotating machinery is widely used in many fields, including aerospace, navigation, wind turbines and so on. APPLICATIONS OF DATA PREPROCESSING APPROACHES IN CNN-BASED FAULT DIAGNOSIS With the increase of mechanical failure data and coupling complexity, intelligent fault diagnosis on the basis of DNN has drawn the interest of researchers owing to the potent capability of extraction and admirable learning ability [61]–[63]. From the perspective of the requirements for input data, not all DNN-based methods could directly process raw vibration data for the final defect classification and prediction Such methods are not viewed as end-to-end approaches, data preprocessing technique plays a crucial role in the intelligent fault diagnosis [64]–[66]. By means of DWT, raw data were transformed into time-frequency matrixes, which were taken as input of the following CNN model for fault diagnosis of planetary gearbox [82]. The employment of time-frequency matrixes made the obtained feature representation more comprehensive, in addition, the learning of nonlinear relationships was achieved as well

S-TRANSFORM

Findings

CONCLUSION AND PERSPECTIVES