Fault detection via on-line parameter estimation for a strong variable parameters PM brushless machine

With reference to a strong variable parameters AC brushless machine, the paper deals with an online method of fault detection for closed loop controlled PMSM. The proposed technique exploits the monitoring of the motor d,q axes inductances via a Recursive Least Square Algorithm (RLS). The motor parameters, which have been pre-mapped with reference to the d,q axes current components, are compared to the results of the online identification procedure and their discrepancy is evaluated in real time. A fault indicator index is derived and its behavior conditioned in order to properly recognize any anomalous condition. The proposed method is tested via numerical simulations on the basis of a real PM brushless machine whose parameter have been evaluated in different operating conditions by means of several off-line experimental measurements.