Accession Number : ADA115979

Title :   A System Failure Detection Method: Failure Projection Method.

Descriptive Note : Master's thesis,

Corporate Author : MASSACHUSETTS INST OF TECH CAMBRIDGE LAB FOR INFORMATION AND DECISION SYSTEMS

Personal Author(s) : Lou,Xi-Cheng

PDF Url : ADA115979

Report Date : Jun 1982

Pagination or Media Count : 94

Abstract : This thesis proposes a system failure detection method -- Failure Projection Method (FPM) which provides a geometric pictures of the problem of failure detection in the presence of model uncertainties and noise. The concept of FPM is thoroughly studied in this thesis. In particular, two groups of formulations have been developed. One gives distinct geometrical interpretation while the other is based on assuming that one has available a priori information on the system state. Within two groups three formulations which are based on slightly different criteria and have decreasing complexity of calculation are developed. The simplest require only a singular value decomposition. Also two numerical examples are given which show the relationship among these formulations and thus provide a deep understanding of their nature. An algebric approach is developed for the generation of a complete set of minimal length parity checks. The FPM is extended to including measurement and process noise. Again a formulation is developed which only involves a singular value decomposition. The FPM is demonstrated on a model of three machine power system to indicate how it can be used as a design tool in assessing system redundancy and in determining parity checks.

Descriptors :   *Error detection codes, *Failure, *Parity, *Mathematical prediction, *Systems analysis, Noise, Decision making, Residuals, Geometry, Vector analysis, Problem solving, Theses, Orthogonality

Subject Categories : Theoretical Mathematics

Distribution Statement : APPROVED FOR PUBLIC RELEASE