Adaptive lattice filter based approach to fault detection in the absence of a system model

This paper proposes a novel approach to the incipient and abrupt fault detection for a class of unknown linear dynamic systems. The proposed method generates a set of residual sequences from the system input and measured output using an adaptive bi-channel lattice filter. The fault detection is carried out by performing statistical tests on the residuals. The distinctive advantage of this method is that a model of the unknown system is not required a priori for the purpose of fault detection. An application of this approach in a practical control system has been implemented, and satisfactory results have been obtained.