On the choice of prior for induction motor parameters estimation using MCMC methods

Processing of the stator current of three-phase induction machines is a widely used technique for obtaining health state information. Most of the spectral components of the current depend on the slip, a parameter related to the load. A Bayesian approach associated with a Monte Carlo Markov chain algorithm is proposed to analyze the stator current of the healthy machine during steady-state operation, i.e., to estimate the slip and the noise variance. This approach allows us to take into account a priori information on the signal and to eliminate all the unknown and uninteresting stator current parameters. Several parameter prior assignments are discussed and results on real data are given.