Design of H_/H∞ fault detection observer for closed-loop nonlinear system with disturbance

Abstract

Purpose This paper proposes a novel method of fault detection, which is based on H_/H∞ Runge–Kutta observer and an adaptive threshold for a class of closed-loop non-linear systems. The purpose of this paper is to improve the rapidity and accuracy of fault detection. Design/methodology/approach First, the authors design the H_/H∞ Runge–Kutta fault detection observer, which is used as a residual generator to decouple the residual from the input. The H_ performance index metric in the specified frequency domain is used to describe how sensitive the residual to the fault. The H∞ norm is used to describe the residual robustness to the external disturbance of the systems. The residual generator is designed to achieve the best tradeoff between robustness against unknown disturbances but sensitivity to faults, thus realizing the accurate detection of the fault by suppressing the influence of noise and disturbance on the residual. Next, the design of the H_/H∞ fault detection observer is transformed into a convex optimization problem and solved by linear matrix inequality. Then, a new adaptive threshold is designed to improve the accuracy of fault detection. Findings The effectiveness and correctness of the method are tested in simulation experiments. Originality/value This paper presents a novel approach to improve the accuracy and rapidity of fault detection for closed-loop non-linear system with disturbances and noise.

Fault detection and identification for rolling mill main drive system based on integrated observer under iterative learning strategy

“…To address the limitations of this method, Han et al 17 proposed a fault detection strategy based on the H − / L ∞ observer, which makes the residuals sensitive to faults and robust to disturbances at the same time. Chen et al 18 combine the performance of H − / L ∞ and design a Langer-Kutta observer, which can simultaneously balance the robustness of the observer to disturbance signals and the sensitivity to fault signals, and construct an adaptive threshold generator to improve the fault detection accuracy. Morato et al 19 designed a multi-linear variable parameter observer to solve the system fault diagnosis problem, solving the observer gain by linear matrix minimisation and satisfying the H 2 / H ∞ mixing performance requirement to reduce the influence of external disturbances on fault estimation.…”