Expert System for Determining Clastic Depositional Environments
A. W. Shultz, J. H. Fang, M. R. Burston, H. C. Chen, M. Beasley
We developed a prototype of a knowledge-based expert system for interpreting clastic sedimentary depositional systems. The present structure is modeled in part on established expert systems such as PROSPECTOR and LITHO, yet addresses a different goal and uses an original and flexible rule base.
Interpretation can be made on two levels: (1) probable depositional environments can be determined given a set of observed features such as lithology, texture, fossil content, and sedimentary structures; and (2) a depositional system can be determined on the basis of vertical or lateral disposition of different assemblages of features. The first level is a rule-based AI application program that applies abductive reasoning to assign certainty factors to probable sedimentary environments, whereas the second level adds a degree of sophistication, accuracy, and context sensitivity typically provided by facies modeling. Certainty factors are derived by combining the user's certainty that a feature has been observed (observational certainty), and the certainty that observation of the feature is related to a depositional environment (relational certainty).
The system is coded in PROLOG on an IBM personal computer. PROLOG was chosen for its amenability to task-oriented programming, sequential inference logic, and semantic substitution (aliases). The program is both antecedent- and consequent-driven using a rule base of approximately 1,000 relationships between observable features and associated environments. When initial observations are entered, either interactively or by batch entry, the program is antecedent driven. The system operates in the consequent-driven mode when directed to evaluate the working hypotheses.
AAPG Search and Discovery Article #91043©1986 AAPG Annual Convention, Atlanta, Georgia, June 15-18, 1986.