Induction machine condition monitoring with higher order spectra. I. Fundamentals and fixed frequency operation

This paper details a novel method of detecting and unambiguously diagnosing various induction machine fault conditions from measurement of only the machine electromagnetic vibration. The detection method introduced involves the induction machine being considered as a simple system thereby precluding the need for a priori data or complex modelling techniques. The induction machine fault conditions discussed are common failure mechanisms, and represent faults of both an electrical and mechanical nature. A novel signal processing technique for condition monitoring called higher order spectra is introduced and is shown to be the optimal detection mechanism for fault diagnosis using the machine vibration. These techniques are then verified in a rigorous experimental investigation and the detection mechanism is shown to be largely independent of the induction machine load condition. In addition, it is shown that this technique is entirely extensible to other fault types without loss of generality. Part I of the paper concentrates predominantly on the fundamental aspects of this novel detection method, and it is shown that the developed techniques improve on conventional methods. It also details a rigorous experimental investigation and highlights pertinent results.