Using Deep Learning Based Approaches for Bearing Fault Diagnosis with AE Sensors

Miao He, David He, and Eric Bechhoefer
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_16_039.pdf336.28 KBAugust 15, 2016 - 11:06am

In the age of Internet of Things and Industrial 4.0, the prognostic and health management (PHM) systems are used to collect massive real-time data from mechanical equipment. Mechanical big data has the characteristics of large-volume, diversity and high-velocity. Effectively mining features from such data and accurately identifying the machinery health conditions with new advanced methods become new issues in PHM. A major problem of using the existing PHM methods for machinery fault diagnosis with big data is that the features are manually extracted relying on much prior knowledge about signal processing techniques and diagnostic expertise, limiting their capability in fault diagnosis with big data.

This paper presents a deep learning based approach for bearing fault diagnosis using acoustic emission (AE) sensors with big data. Different from widely used shallow neural network architecture with only one hidden layer, the branch of machine learning methods with multiple hidden layers are regarded as deep learning method. The presented approach pre-processes AE signals using short time Fourier transform (STFT) and extract features. Based on the simply processed AE features, an optimized deep learning structure, large memory storage retrieval neural network (LAMSTAR) is used to perform bearing fault diagnosis. The unique structure of LAMSTAR enables it to establish more efficient and sparse distributed feature maps than traditional neural networks. By leveraging the labelled information via supervised learning, the trained network is endowed with discriminative ability to classify bearing faults. The AE signals acquired from a bearing test rig are used to validate the presented method. The test results show the accurate classification performance on various fault types under different working conditions, namely input shaft rotating speeds. It also proves to be effective on diagnosing bearing faults with relatively low rotating speeds.

Publication Year: 
2016
Publication Volume: 
7
Publication Control Number: 
039
Page Count: 
10
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed