The Role of System Behavior in Diagnostic Performance

Abstract

The diagnostic performance of system built-in-test has historically suffered from deficiencies such as high false alarm rates, high undetected failure rates and high fault isolation ambiguity. In general these deficiencies impose a burden on maintenance resources and can affect mission readiness and effectiveness. Part of the problem has to do with the blurred distinction between physical faults and the test failures used to detect those faults. A greater part of the problem has to do with the test limits used to establish pass/fail criteria. If the limits do not reflect system behavior that is to be expected, given its current no fault (or fault) status, then a test fail result can often be a false alarm, and a test pass result can often constitute an undetected fault. A model based approach to prediction of system behavior can do much to alleviate the problem.