Data-Driven Fault Detection, Isolation and Identification of Rotating Machinery: with Applications to Pumps and Gearboxes

  • Author / Creator
    Zhao, Xiaomin
  • Fault diagnosis plays an important role in the reliable operation of rotating machinery. Data-driven approaches for fault diagnosis rely purely on historical data. Depending on how a diagnosis decision is made, this thesis divides data-driven fault diagnosis approaches into two groups: signal-based approaches and machine-learning-based approaches. Signal-based approaches make diagnosis decisions directly using signal processing techniques. Machine-learning-based approaches resort to machine learning techniques for decision making. There are three main tasks in fault diagnosis: fault detection (detect the presence of a fault), fault isolation (isolate the location / type of the fault), and fault identification (identify the severity of the fault). This PhD research studies signal-based approaches for fault identification and machine-learning-based approaches for fault detection, isolation and identification. In signal-based approaches for fault identification, generating an indicator that monotonically changes with fault progression is a challenging issue. This thesis proposes two methods to generate such indicators. The first method uses multivariate signal processing techniques to integrate information from two sensors. The second method uses fuzzy preference based rough set and principal component analysis to integrate information from any number of sensors. In machine-learning-based approaches, feature selection is an important step because it improves the diagnosis results. For fault detection and isolation, classification is often used as the machine learning algorithm. In this thesis, a feature selection method based on neighbourhood rough sets is proposed for classification. Coming to fault identification, classification is not suitable because classification does not utilize the ordinal information within the different fault severity levels. Therefore, this thesis proposes to use another machine learning algorithm, ordinal ranking, for fault identification. A feature selection method based on correlation coefficient is proposed for ordinal ranking as well. Moreover, an integrated method which is capable of conducting fault detection, isolation and identification is proposed by combining classification and ordinal ranking. The proposed methods are applied to fault diagnosis of impellers in slurry pumps and fault diagnosis of gears in planetary gearboxes. Experimental results demonstrate the effectiveness of the proposed methods.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
    • Department of Mechanical Engineering
  • Supervisor / co-supervisor and their department(s)
    • Zuo, Ming J (Mechanical Engineering, University of Alberta)
  • Examining committee members and their departments
    • Ma, Yongsheng (Mechanical Engineering, University of Alberta)
    • Pedrycz, Witold (Electrical and Computer Engineering, University of Alberta)
    • Lipsett, Mike (Mechanical Engineering, University of Alberta)
    • Lange, Carlos (Mechanical Engineering, University of Alberta)
    • Jiang, Jin (Electrical and Computer Engineering, University of Western Ontario)