Please use this identifier to cite or link to this item:
Title: Stray magnetic field based health monitoring of electrical machines
Authors: Lui, Zheng
Issue Date: 2018
Publisher: Newcastle University
Abstract: Electrical machines are widely used in industrial and transportation applications which are essential to industrial processes. However, the lack of reliability and unpredictable life cycles of these machines still present opportunities and challenges for condition monitoring research. The breakdown of an electrical machine leads to expensive repairs and high losses due to downtime. The motivation of this research is to improve the reliability of electrical machines and to classify different kinds of failures via non-intrusive methods for condition-based maintenance and early warning of failure. Major potential failure types in electrical machines are winding and mechanical failures, which are caused by dynamic load state, component ageing and harsh working environments. To monitor and characterise these abnormal situations in the early stages, this thesis proposes stray magnetic field-based condition monitoring allowing fault diagnosis with the help of finite element models and advanced signal processing technology. By investigating the interaction between stray flux variations and machine failure, different kinds of faults can be classified and distinguished via numerical and experimental studies. A non-intrusive stray flux monitoring system has been developed and can provide both static and transient stray flux information and imaging. The designed monitoring system is based on a giant magnetoresistance (GMR) sensor used to capture low stray flux fields outside the electrical machine’s frame. Compared with other monitoring systems, its small size, low cost, non-inventive and ease of setting up make the designed system more attractive for in many long-term monitoring applications. Additionally, integration with the wireless sensor network (WSN) means that the latter’s unique characteristics makes the proposed system suitable for electrical machine monitoring in industrial applications replacing existing expensive wired systems. The proposed system can achieve real-time data collection and on-line monitoring with the help of spectrogram and independent component analysis. Three cases studies are conducted to validate the proposed system with different failures and loading states, using load fatigue, winding short-circuit failure and mechanical testing. In these case studies, electrical and mechanical failures and dynamic loads are investigated, collecting stray flux information with different kinds and sizes of electrical machines using both simulation and experimental approaches. Stray flux information is collected for different situations of winding failure, unbalanced load and bearing failures. Comprehensive transient feature extraction using spectrogram is implemented with respect to multiple failures and load variations. Spectrograms of stray flux can provide time-frequency information allowing the discrimination of different failures and load states. Different faults can be distinguished through independent component analysis of stray flux data. Compared with traditional and current detection strategies, stray flux-based monitoring can not only provide failure indicator and better resolution but also gives location information. Additionally, by applying different feature extraction methods, different failure types can be separated based on stray flux information, which is likely to be difficult to achieve using traditional monitoring approaches. However, stray flux monitoring systems suffer from issue of noise and instability, and more case studies and investigations are needed for further refinement.
Description: PhD Thesis
Appears in Collections:School of Engineering

Files in This Item:
File Description SizeFormat 
Lui, Z, 2018.pdfThesis5.4 MBAdobe PDFView/Open
dspacelicence.pdfLicence43.82 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.