Deep Learning based Diagnostics of Orbit Patterns in rotating machinery

Haedong Jeong, Sunhee Woo, Suhyun Kim, Seungtae Park, Heechang Kim, and Seungchul Lee
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_16_051.pdf802.42 KBSeptember 19, 2016 - 2:16am

Vibration-based orbit analysis has been employed as a powerful tool in diagnosing the operating state for rotating machinery in power plants. However, due to the difficulties of extracting mathematical features for data-driven approaches in the orbit analysis, it heavily depends on the expert knowledge or experience. In this paper, the deep learning algorithm in machine learning is used to develop autonomous orbit pattern recognition. In details, the convolutional neural network is implemented to build up weights between convolution kernels and pixels, and to construct the entire structure of the neural networks. Finally, the trained network enables us to classify the shapes of the orbit via orbit shape images and its result can estimate fault modes of the rotating machinery. The proposed framework is demonstrated with a rotating testbed.

Publication Year: 
2016
Publication Volume: 
7
Publication Control Number: 
051
Page Count: 
7
Submission Keywords: 
deep learning
Convolutional Neural Networks
rotating machinery
Orbit Analysis
Image Pattern Recognition
machine learning
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