A novel reconstruction approach for sensor faults using an adaptive unknown-input observer with dual information channels

In this study, a fault diagnosis approach is proposed to reconstruct sensor faults, which integrate uncertainty decoupling and adaptive techniques. Specifically, an unknown input estimator is addressed to dissociate undetermined partial process uncertainties, and an adaptive estimation method is presented to achieve a reconstruction for sensor faults, where both proportional and integral information channels of the residuals are utilized. The estimator gains are secured by solving strict linear matrix inequalities to ensure the dynamics of the estimation error to be robustly stable. Both linear dynamic systems and nonlinear dynamic systems with Lipschitz constraints are examined, and the resulting fault reconstructed algorithms are developed. The applicability of the proposed adaptive estimation approaches is demonstrated via simulation studies by using a civil aircraft system and a robotic manipulator system. The advantages of the proposed estimation technique are illustrated by comparison studies.