Fault detection for continuous-time systems, the PMF identification approach

In this paper, we present an application of Poisson Moment Functionals (P.M.F) identification method in fault detection for continuous-time systems. The application concern an electro-mechanical system which can be used in chemical industry as a mixer, A continuous-time model is derived in the form of differential equation. The fault detection procedure uses the parameter estimation approach via the implementation of the P.M.F identification method, The PMF method is implemented far determining the nominal model and the corresponding bias of the ordinary least-squares method by the use of a set of training data, Based upon the bias analysis of the observed model we have detected the faulty system behaviour. The advantages of the PMF identification approach is that it does not need any direct transformation of the continuous-time model and permits to handle the derivative terms occurring in the continuous-time model. easily and directly, Copyright (C) 1998 IFAC.