Sensor Fault Detection via Multiscale Analysis and Nonparametric Statistical Inference

Sensor validation is a topic of widespread importance. A new approach to sensor validation in real time is described that is based on (1) representation of the sensor signal by wavelets, (2) decomposition of the signal into different frequency ranges, (3) calculation of useful features of the signal at different frequencies, and (4) diagnosis of faulty operation via nonparametric statistical tests. The proposed strategy is able to isolate the effect of noise and process changes from the effects of physical changes in the sensor itself. To clarify the circumstances under which the above strategy could be used, a noisy signal from a simulated thermocouple in a dynamic continuous nonlinear unsteady state stirred tank reactor (CSTR) was analyzed. Faults were introduced into the thermocouple, and the diagnosis was carried out. The results of the diagnosis indicated that the proposed strategy had low type I (false alarm) and type II (failure to detect faults) errors and was distinctly better than a standard test for changes in a nonstationary signal of unknown characteristics.