Sensor Fault Detection via Multiscale Analysis and Dynamic PCA

In this paper, 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 for different frequency ranges, (3) formation of a matrix of lagged details from a window of the original sensor data at different frequencies, (4) application of PCA decomposition to the matrix of details, and (5) diagnosis of the existence of faults via tests on the PCA statistics such as T2 and Q. The proposed strategy is able to isolate the effect of noise and process changes from the effects of physical changes in the sensor itself. For comparison we applied all of the tests and diagnoses to the original signal itself as well as to the wavelet details. To demonstrate the circumstances under which the above strategy might be used, a simulated noisy signal from a CSTR and a temperature signal from an operating pilot distillation column were analyzed. Faults were introduced into the thermocouple, and the diagnosis 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.