A genetic algorithm for fault identification in electrical drives: a comparison with neuro-fuzzy computation

Industrial applications require suitable monitoring systems able to identify any decrement in the production efficiency involving economical losses. The information coming from a general purpose monitoring system can be usefully exploited to implement a sensorless instrument monitoring an AC motor drive and a diagnostic tool providing useful risk coefficients. The method is based on a complex digital processing of the line signals acquired by means of a virtual instrument. In this paper a genetic algorithm, implemented in a Mathcad environment, performs the evaluation of the risk indexes from the processed line signals. The combination of genetic algorithms and neural network is also investigated as a promising possibility for the development of a reliable diagnostic tool. The risk coefficients derived from this approach are evaluated, discussed and compared to other indexes - in particular fuzzy indexes - introduced by the authors in previous papers.