Data validation of intelligent sensor using predictive filters and fuzzy logic

A new approach using polynomial predictive filters and fuzzy logic for the online validation of sensor measurements is proposed. Polynomial predictive filters are applied to measured data records during the fault free learning stage. Predictions derived from the predictive filters are compared with the actual measurements to generate an error sequence. Fuzzy rules are then derived from the error sequence together with the physical constraints of a sensor to classify the quality of measurements. Faulty measurements can then be picked up by the fuzzy detection rules to ensure the correctness of measurements. Experimental results for detecting the quality of measurements from a temperature sensor are presented.