A confidence-based approach to the self-validation, fusion and reconstruction of quasi-redundant sensor data

Often is the case in industrial applications that multiple sensors are used to measure similar quantities. These sensors may not be truly redundant in that they are not placed to measure exactly the same parameter. However, these parameters may be very well correlated. In this paper, we address three aspects of dealing with data from such quasi-redundant sensors. Specifically, we (1) employ fuzzy logic rules for self-validation and self-confidence; (2) exploit the near linear relationship between sensors for reconstructing missing or low-confidence data; and (3) fuse this data to determine a single measure and a qualitative value for its reliability. The methodology presented is illustrated on experimental temperature data from a cupola iron-melting furnace.