Classification method for faults diagnosis in reluctance motors using Hidden Markov Models

The Switched Reluctance Machine (SRM) is ideal for safety critical applications due to its superior fault-tolerance characteristics. The switched reluctance drive is known to be fault tolerant, but it is not fault free. Fault diagnosis of SRM in the critical applications is often a difficult and daunting task. Thus, finding efficient and reliable fault diagnostics methods especially for SR machines is extremely important. This paper focuses on the development, and application of modern statistical classifier method, namely Hidden Markov Model (HMM) associated with a smoothed ambiguity plane Time-Frequency Representation (RTF) for the diagnosis based classification of electrical faults in this particular machine. The RTF-HMM Technique is composed of two steps: the Feature Extraction step based on the smoothed ambiguity plane designed for maximizing the separability between classes using Fisher's discriminant ratio and Hidden Markov Model algorithm applied for the classification step. The algorithm of each step is well developed. Classifier development and training data is carried out by the HMM using a set of fault scenarios, between healthy, single and combined faults, in terms of torque at different load level in order to deduce the fault severity. Parameter training of Hidden Markov Models generally need huge a mounts of historical data. Experimental results proves that the use of RTF-HMM based approaches is a suitable strategy for the automatic classification of new sample independent from de type of fault signal.