An Automatic Sensor Fault Detection and Correction Algorithm

Today’s turbofan engines are electronically controlled using sensor measurements and pilot commands. The behavior of the system depends on the quality of each input. However acquisition chains are subject to failures and even if redundancy exists one have to detect which measurements are reliable and which are poor. Once detected the false data may be temporarily replaced by approximation or redundancy. Then further investigation proceeds to detect and fix the faulty component. This article gives clues to detect a faulty acquisition chains among a set of measurements and proposes a correction scheme. The fault diagnostic is realized with two steps: the detection of a problem, then its localization. The replacement uses a physical but rough model and a mathematic algorithm to correct the physical estimation according to the observation of validated measurements. Each measurement may be subject to different kind of faults: sensor drifts, harness intermittencies, regulation trends, etc. The goal is to detect which measurements are reliable and which are not.