Fault detection and diagnosis of induction motors based on hidden Markov model
Résumé
Accurate fault detection and diagnosis in complex systems is necessary for economic and security reasons. In this paper, we present a novel approach for fault detection and diagnosis based on Hidden Markov Models. This approach uses pattern recognition combining motor current signature analysis and multiple features extracted from transformations made on current and voltage signals in order to build the representation space. If the representation space is well chosen, each operating mode can be represented as a class. A hidden Markov model is then designed for each class and used as classifier for the detection and diagnosis of faults. The proposed approach is tested on an induction motor of 5.5 Kw with bearing failures and broken rotor bars. Further, the effectiveness of this approach is compared with a neural-network-based approach. The experimental results prove the efficiency of the hidden Markov model-based approach in condition monitoring of electrical machines.