A decision system to detect failures in induction motors

Induction motors present numerous advantages due to their robustness and their power-weight ratio. Thus, they are widely used in the industry. The necessity to insure continuous and safety operation involves preventive maintenance programs. This paper presents the application of a pattern recognition approach to detect electrical faults in an induction motor. For this purpose, only electrical measurements are used. The origin and the causes of faults are presented, with their distribution on the machine components. Some time or frequency-dependent parameters, relevant for fault detection are described. They are used to build up a pattern vector. Then a decision system, based on this signature and the k-nearest neighbours is proposed.