Thermal signatures for pattern recognition approach applied to induction motor diagnosis

Electric drives condition monitoring is essential to optimize maintenance operations and to increase reliability levels. This paper presents a diagnosis method for electrical faults detection. Firstly some signatures representing induction motor thermal heating are developed. Indeed a motor provides normal losses (mechanical, electrical, magnetic, etc.) as well as additional losses due to some faults. Losses involve an operating temperature increase, which can be particularly damaging for insulation. Eventually this can bring partial or total destruction of this insulation and create a short circuit between turns. From a thermal modelling of induction motor, with a simplified model, the heating can be computed and used as faults signatures. Secondly in order to realize automatic diagnosis, theses signatures are associated with a pattern recognition approach. The aim is to detect faults appearing on the system and to define their severity level by reference to an initial data base. In order to prove reliability and efficiency, experimental results will be presented using an induction motor 5.5kW.