Fault Detection in Multiply-Redundant Measurement Systems via Sequential Testing

The paper presents the theory and application of a sequential test procedure for fault detection and isolation (FDI). The test procedure is suited for development of intelligent instrumentation in strategic processes like aircraft and nuclear plants where redundant measurements are usually available for individual critical variables. The test procedure consists of (1) a generic redundancy management procedure which is essentially independent of the fault detection strategy and measurement noise statistics, and (2) a modified version of sequential probability ratio test (SPRT) algorithm for fault detection and isolation, which functions within the framework of the aforesaid redundancy management procedure. The sequential test procedure is suitable for real-time applications using commercially available microcomputers and its efficacy has been verified by on-line fault detection in an operating nuclear reactor.