Robust Fuzzy Fault Detection for Non-Linear Stochastic Dynamic Systems

ABSTRACT

One of the difficults for fault detection techniques for non-linear stochastic systems via model-based methods is the design of residual generation. In this paper, a new fault detection (FD) approche for non-linear stochastic systems is proposed. The non-linear system is represented by a discrite Takagi-Sugeno (TS) fuzzy model. The use of (TS) theory allows to represent non-linear systems as a set of linear systems, which represent the local system behavior around different operating points. The global system behaviour is described by a fuzzy fusion of all systems. The FD system for each local sub system is designed by solving the corresponding Discrite Algebra Recati Equation (DARE). Optimization algorithm based on minimizing the residual covariance matrix is used to obtain a robust FD for global system behavior. The observer gain matrices are solved using a set of Linear matrix Inequalities (LMIs).