A new technique for fault detection in multi-sensor probes

Proposes a novel approach to determine the self consistency of the observed values from a collection of sensors. While the original techniques were developed for deterministic chaotic systems, the generalizations presented apply to both deterministic and stochastic systems. The authors deal with the case where a particular process or system is observed by a large number of sensors, each measuring some distinct characteristic of the system. In brief, the approach distinguishes observations which correspond to physically realizable states from those which represent physically nonrealizable states; the latter implying error in the observation (i.e. sensor error).<
>