Auxiliary signals for improving on-line fault detection

The paper is concerned with the use of auxiliary (test) signals for improving the detection and diagnosis of faults in linear (or linearized) systems that may be subject to random disturbances. The detection method is based on Wald´s sequential probability ratio test (SPRT) using input-output rather than state-space process models to avoid Kalman Filtering. This `steady state´ approach is very simple to implement but does not allow exact estimation of the occurrence time of a fault. It is well known that the SPRT yields, on the average, the fastest detection time and the paper investigates the influence of test signals in accelerating detection of a fault as well as their influence on the frequency of missed alarms and false alarms. This approach parallels the use of test input signals to enhance the accuracy of parameter estimation