Uma nova formulação fuzzy/bayesiana para detecção de ponto de mudança em séries temporais: aplicações na detecção de falhas incipientes.

Abstract

This work presents a novel approach for incipient fault detection in dynamical systems which is based on a Fuzzy/Bayesian technique for change point detection in time series. The proposed change point detection technique is based on a two step formulation. In the first step, a result from the fuzzy set theory is applied to the time series in order to cluster and transform the initial data, with arbitrary distribution, into new data that can be approximated with a beta distribution. The second step consists in using a Markov chain Monte Carlo (the Metropolis-Hastings algorithm) to the change point detection in the transformed time series. In this work, the incipient faults are characterized by small changes in the time series. As special features of the proposed methodology for incipient fault detection, it does not require any mathematical model of dynamic system, any threshold specification. To illustrate the efficiency of the proposed methodology, some results in the problem of detection of incipient faults in an industrial actuator of the FDI benchmark in the European RTN DAMADICS and in the stator winding of induction motors are presented.