On-line sensor validation of single sensors using artificial neural networks

We examine how the signal from a single sensor from a dynamic process might be analyzed to ascertain whether the sensor signal is valid or not. It is quite possible for the signal to indicate a change has occurred in the process when the sensor itself is what has changed. Two possible approaches to sensor validation discussed here are (1) use of second and higher order statistics rather than the mean of a signal, and (2) modeling the sensor signal via artificial neural networks.