Stochastic Fault Diagnosis using a Generalized Polynomial Chaos Model and Maximum Likelihood

A novel approach has been developed to diagnose intermittent stochastic faults by combining a generalized polynomial chaos (gPC) method with maximum likelihood estimation. The gPC is used to propagate stochastic changes in an input variable to a measured output variable from which the fault is to be inferred. The fault detection and diagnosis (FDD) problem is formulated as an inverse problem of identifying the unknown input from a maximum likelihood based fitting of the predicted and measured output variables. Simulation studies compare the proposed method with a Particle Filter (PF) to estimate the value of an unknown feed mass fraction of a chemical process. The proposed method is shown to be significantly more computational efficient and less sensitive to user defined tuning parameters than PF.