MULTIFREQUENCY ANALOG FAULT DIAGNOSIS FOR LINEARIZED CIRCUITS

LAWRENCE ANTHONY RAPISARDA, Purdue University

Abstract

This report addresses the problem of multifrequency fault diagnosis performed in the context of the component connection model (CCM). The discussion includes an introduction to the CCM as well as two approaches to the fault diagnosis problem, the transfer function matrix approach and the tableau approach. The first method generates the fault diagnosis equations by equating measurement data expressed in terms of a composite system transfer matrix whose structure is delineated in terms of the CCM matrices. The resulting equations are extremely nonlinear and lack the lucidity that is characteristic of problems posed in terms of the CCM. The equations of the tableau approach on the other hand result from a characterization of the ambiguity associated with the measurement data. Such a characterization allows the fault diagnosis equations to remain in tableau form, thereby retaining the clarity and numerical stability usually associated with the use of the CCM. Basic properties of tableau approach presented include the formulation of a diagnosability test and upper/lower bounds on the number of test frequencies necessary to conduct a diagnosis. In addition to the full diagnosability problem this report also investigates the properties of the tableau fault diagnosis equations under the assumption that the number of simultaneous faults in a system under test is limited. The approach employs multifrequency testing to extract a maximum of information from a given set of test points. The analysis specifically addresses the case in which the number of faulty components exceeds the number of measurement points. The development includes a detailed description of the solution algorithm and example problems. A discussion of the test frequency and test point selection problems is also included.

Degree

Ph.D.

Subject Area

Electrical engineering

COinS