Model based diagnostics and prognostics of three-phase induction motor for vapor compressor applications

In this work, a novel scheme for detection and prediction of multiple simultaneous faults in a three-phase induction motor in the context of vapor compression applications is presented. Induction motors used in vapor compression systems operate under variable speed conditions with variable frequencies. Such dynamic operating conditions may cause an occurrence of multiple, simultaneous faults including insulation degradation and a rotor bar breakage. These faults, when left undetected, lead to the failure of the motor and the entire vapor compression system. Hence, a condition monitoring of induction motors is essential. Conventional fault detection methods have various drawbacks including high implementation costs, requirement of extensive testing and offline training, and are difficult to implement for small machines. In this study, a model-based fault detection approach is used where fault detection and prediction employs an online estimation of system states. The faults under consideration are incipient electrical faults: insulation degradation and broken rotor bars. A nonlinear observer with neural network online approximator is employed to discover the system parameter degradation thus learns the unknown fault dynamics. Another online approximator is used to facilitate fault isolation, or root-cause analysis, and a time to failure (TTF) prediction before the occurrence of a failure.