Application of scheduled H∞ polynomial filters to fault detection

This paper presents an approach to robust fault detection for linear stochastic systems with time-varying parameters. This approach combines the advantage of a H-infinity polynomial estimator and gain scheduling techniques. This will allow us, using the estimation error equation, to generate a robust residual which is sensitive to faults and insensitive to perturbations due to noises (white or coloured) and parameters variation. The residual design procedure and the approach to gain scheduling H-infinity filters are developed. To illustrate the efficiency of the method, a comparative simulation on a laboratory process is presented.