The methodology of auxiliary signal design for robust fault detection based on a multi-model (MM) formulation of normal and faulty systems is used to study the problem of incipient fault detection. The fault is modelled as a drift in a system parameter, and an auxiliary signal is to be designed to enhance the detection of variations in this parameter. It is shown that it is possible to consider the model of the system with a drifted parameter as a second model and use the MM framework for designing the auxiliary signal by considering the limiting case as the parameter variation goes to zero. The result can be applied very effectively to many early detection problems where small parameter variations should be detected. Two different approaches for computing the test signal are given and compared on several computational examples.

No abstract

No abstract

Failure detection is an active area of Navy research due to its many important applications. Recently, an approach for multi-model identification and fault detection in the presence of bounded energy noise over finite time intervals has been introduced. This family of algorithms was originally designed to work on linear systems that can be modeled analytically. In this paper, efforts made toward extending this algorithm for fault detection to nonlinear systems along with efforts in testing this family of algorithms on real systems are discussed.

No abstract