Fault estimation for a class of nonlinear dynamical systems

Model based fault estimation for a class of nonlinear dynamical systems is investigated. The state of the system is assumed unavailable, and a nonlinear observer is used to estimate the state. In the observer, a neurofuzzy network is used as the approximator to estimate faults. The network is trained online and the convergence of the proposed learning algorithm is established. Abrupt faults and incipient faults are analyzed in the paper and they can be estimated accurately using a neurofuzzy network with the proposed learning algorithm.