Fault detection for non-Gaussian stochastic systems via augmented Lyapunov functional approach

In this paper, a new fault detection (FD) scheme is studied for non-Gaussian stochastic dynamic systems using output probability density functions (PDFs). Different from the classical FD problems, the measured information is the PDFs of system output rather than its value, where the B-spline expansion technique is applied so that the considered FD problem is transformed into a nonlinear FD problem. In this context, feasible FD method is presented by combining linear matrix inequality (LMI) technique with augmented Lyapunov functional, which involves a tuning parameter and a slack variable. Furthermore, in order to improve the detection sensitivity performance, an optimal algorithm is applied to minimize the threshold by tuning the parameter. Simulation for a model in the paper-making process is given to demonstrate the efficiency of the proposed approach.