Online fault detection and isolation of nonlinear systems based on neurofuzzy networks

Online fault detection and isolation (FDI) of dynamic systems provides early warnings to operators for scheduling suitable maintenance, leading to improved reliability and safety of systems. Fault detection and isolation based on neurofuzzy networks is discussed in this chapter. The proposed scheme involves two steps. In the first step, a neurofuzzy network, derived from B-spline neural networks (BSNN), is trained to model the nonlinear plant, from which residuals are generated for fault diagnosis. In the second step, another neurofuzzy network is trained online to model the residuals. Qualitative description of the faults is then extracted and encoded for fault detection and isolation from fuzzy rules obtained from the second neurofuzzy network. The performance of the proposed scheme is illustrated by simulation examples involving fault diagnose of a DC motor and a two-tank control system. © 2007