Rotating Machinery Fault Diagnosis Using Long-short-term Memory Recurrent Neural Network

Rui Yang, Mengjie Huang, Qidong Lu, Maiying Zhong

Research output: Contribution to journalArticlepeer-review

71 Citations (Scopus)

Abstract

With the fast development of science and industrial technologies, the fault diagnosis and identification has become a crucial technique for most industrial applications. To ensure the system safety and reliability, many conventional model based fault diagnosis methods have been proposed. However, with the increase in the complexity and uncertainty of engineering system, it is not feasible to establish accurate mathematical models most of the time. Rotating machinery, due to the complexity in its mechanical structure and transmission mechanics, is within this category. Thus, data-driven method is required for fault diagnosis in rotating machinery. In this paper, an intelligent fault diagnosis scheme based on long-short-term memory (LSTM) recurrent neural network (RNN) is proposed. With the available data measurement signals from multiple sensors in the system, both spatial and temporal dependencies can be utilized to detect the fault and classify the corresponding fault types. A hardware experimental study on wind turbine drivetrain diagnostics simulator (WTDDS) is conducted to illustrate the effectiveness of the proposed scheme.

Original languageEnglish
Pages (from-to)228-232
Number of pages5
Journal10th IFAC Symposium on Fault Detection, Supervision and Safety for Technical Processes SAFEPROCESS 2018: Warsaw, Poland, 29-31 August 2018
Volume51
Issue number24
DOIs
Publication statusPublished - 2018
Externally publishedYes

Keywords

  • AI
  • FDI methods
  • Mechanical
  • electro-mechanical applications

Cite this