Neural-network-based adaptive fault estimation for a class of interconnected nonlinear system with triangular forms

Abstract

In this paper, a novel fault estimation methodology is proposed for a class of interconnected nonlinear continues-time systems with triangular forms. In the distributed fault estimation architecture, a fault detector is utilized to generate a residual between the subsystem and its detector or observer. Moreover, a threshold for distributed fault detection and estimation in each subsystem is designed. Due to the universal approximation capability of the radial basis function neural networks, it is used to estimate the unknown fault dynamics. The time-to-failure is determined by solving the adaptive law from the current time instant to a failure threshold. Finally, the proposed methods are verified in the simulation.