The wideband signal taken directly from a sensor can be processed to provide two metrics for validating the measurement device. This can reduce reliance on hardware and functional redundancy, and minimise the complexities associated with redundancy based validation systems. This paper describes some of the results of an approach which is based on applying a number of signal processing algorithms on the signal representing the total response of the sensor to the process. The parameters derived from the algorithms provide the input vector to a feedforward backpropagation neural network which determines the value of two validation metrics namely device state and process state. The proposition is that validation should commence at the physical interface to the process right through the hierarchy of operations in plant control, maintenance, and management.
展开▼
