Input Design for Active Fault Detection: Reconciling System Control Objectives.

Abstract

Active fault detection (AFD) is the newest frontier in the field of fault detection and has drawn increasing amounts of research attention. AFD technology can enhance fault detection performance by injecting a predesigned auxiliary input signal for a specific fault. In most existing studies, system control objectives are not fully considered in the auxiliary input design of AFD. This article investigates a new reconciliatory input design problem for both achieving control objectives and improving fault detection performance. An exemplary algorithm for the reconciliatory input design is proposed, by using a trajectory optimization approach. The proposed algorithm consists of three parts: 1) residual generation; 2) trajectory optimization; and 3) input design. A state observer is designed to obtain residual signals used as fault indicators. Considering the optimization index composed of the fault indicators, a trajectory optimization technique is carried out to find an optimal system trajectory which can improve the fault detection ability to the greatest extent. The control input is designed to track this optimal trajectory while complying with system physical constraints. In order to demonstrate the effectiveness of the proposed methodology, simulation cases on an underwater manipulator are conducted.