The IEEE International Symposium of Diagnostics for Electrical Mahcines Power Electronics and Drives (SDEMPED´99), Gijon (Spain). 01 September 1999
The on-line diagnosis of an electrical motor is the first step fowards the application of an effective predictive maintenance plan on such electrical machines. This will save costs caused by unexpcted failures causing unavailabilities, it will improve the quality of service and it will delay the need to make new investments to replace equipment. A fast detection of features of possible failures will permit taking corresponding actions in order to mitigate or eliminate its impact. In this paper a method based on the use of neural networks allows the construction of models that characterize the normal behaviour of particular variables of an electrical motor uner several working conditions. The models obtained can be used for on-line detection of deviations in the normal behaviour predicted. This can be done by making a comparison between real and predicted values of variables of the different models. The models formulated use non-intrusive measurements. They are based on the characterization on different electrical and mechanical effects.
Keywords: Diagnosis, predictive maintenance, neural networks, modelling, electrical motor
Publication date: September 1999.
M.A. Sanz-Bobi, M.A. Donaire, Diagnosis of electrical motors using artificial neural networks, The IEEE International Symposium of Diagnostics for Electrical Mahcines Power Electronics and Drives (SDEMPED´99), Gijon (Spain). 01-03 September 1999.