Measurement validation via expected uncertainty

Validated sensor data is critical to the operation of a modem process plant as erroneous data can degrade process performance. Consequently measuremen t validation is a necessary part of process management. Self-validating sen sors reference observed sensor performance against a local model of ideal s ensor behaviour. Discrepancies between the observed and desired sensor resp onse are used to resolve faulty sensor operation and indicate the resultant measurement quality. However, non-ideal aspects of the validation problem lack treatment in many existing self-validation methods and require the val idated-measurement model to have a probabilistic foundation. This paper pre scribes a formal basis for integrating non-ideal diagnostic information int o a stochastic model of sensor operation. The analysis is presented in the context of an existing self-validation framework with a mass-flow simulatio n illustrating the concept. (C) 2001 Elsevier Science Ltd. All rights reser ved.