Knowledge acquisition and interpretation problem-solving methods for visual expertise: S study of petroleum-reservoir evaluation

The paper presents a practical study of expertise in sedimentary petrography and its incorporation into a knowledge-based computing system, PetroGrapher. The study identifies features of expert geologists' essentially visual knowledge that limit the usefulness of traditional logic-oriented symbolic methods for developing relevant knowledge-based systems. In particular, the experts' expressions of knowledge differ significantly from what is taught through the literature about their subject. Also, the means (ontology) of describing the explicit objects of geological data are not well-suited to describe how experts combine such objects and make inferences about them. The paper indicates how to deal with those two different levels of ontology, proposes “knowledge graphs” as an effective medium for linking them and explains their place in a general descriptive model of how experts solve problems in interpretation based on visual evidence. The model is being validated through the use of PetroGrapher, for management of data and knowledge about reservoir rocks, in a petroleum company.