Knowledge-based interpretation of sensor patterns

In this paper we define knowledge-based interpretation of sensor patterns as the Qualitative Interpretation ( QI) problem. In our definition, QI corresponds to the generation of useful symbolic abstractions from numeric plant sensor data. The lack of robust QI capability represents a key bottleneck in the widespread application of knowledge-based system (KBS) technology within the chemical process industries (CPI). We characterize QI as an adaptive pattern recognition task which can be addressed in a supervised manner. Consequently, the potential utility of the back-propagation network (BPN) is examined. Representational adequacy is identified as the controlling issue. A process example illustrates the use of BPN for QI.