Fault monitoring and residual life estimation of electric motors

Techniques for predictive maintenance of plant components primarily focus on fault monitoring and diagnostics of rotating machinery. Consequently, the analysis systems used by plant personnel perform the function of vibration monitoring and trending of rotating machinery. The important next step in advancing the technology and for providing decision-making information to plant personnel is the system prognosis phase. The estimation of residual life of plant components (both rotating machinery and stationary systems) falls into this category. Another important issue to be addressed is that, vibration monitoring by itself may not always provide machinery degradation information. Furthermore, an attempt towards a root cause analysis of component degradation with time is seldom addressed. The purpose of the research reported in this thesis is to develop and demonstrate a methodology for the estimation of residual life of plant components. The research focus was on the degradation monitoring and characterization of small horsepower induction motors. Both experimental and analysis techniques were developed to establish the solution to this important problem. Experimental data were acquired from a motor testing laboratory. The motors were operated continuously for a long period of time under externally imposed anomalies. The data were used to trend degradation of motors and to establish predictive models for residual life estimation. Dynamic regression models and neural network models were established to characterize the trend information and to determine the useful remaining life of small horsepower induction motors for a specified alarm level. A PC-based analysis system was developed using the MATLAB platform. Both laboratory and industrial predictive maintenance data were used to demonstrate the applicability of this new approach. A model, describing the dynamics of induction motors, was developed to understand the characteristics of anomalies and the test data. This research has shown that just the monitoring of vibration of certain rotating machinery is not sufficient to trend the degradation of these machinery. Electrical parameters and internal changes in motor characteristics must also be monitored as part of a predictive maintenance program. The technology developed in this research may easily be extended to large horsepower motors used in nuclear and fossil power plants.