Perception based hybrid intelligent systems in petroleum applications

We describe the methods of processing of perception based information in hybrid intelligent systems. Several innovative techniques like a multi-set based algebra of qualitative perception-based uncertainties and perception-based data mining form the technological framework of the approach. In the paper, we discuss the algebra of strict monotonic operations and inference procedures based on perception-based evaluations of uncertainty of facts and rules. They are characterized by multi-set-based representation of evaluations of uncertainty and by multi-valued inference of conclusions in expert system rules. The proposed method is implemented in the CAPNET Expert System Shell. We also discuss the method of evaluation of perception-based patterns in time series data bases. The approach is illustrated by examples of diagnostics of excessive water production in petroleum wells combining both methods.