Year
1988
Abstract
An Artificial Intelligence-Expert System called APES (Analysis of Proliferation by Expert System) has been developed and tested to permit a non proliferation expert to evaluate the capability and capacity of a specified LWR reactor and PUREX reprocessing system for producing and separating plutonium even when system information may be limited and uncertain. APES employs an expert system coded in LISP and based upon an HP-RL (Hewlett Packard-Representational Language) Expert System Shell. The user I/O interface communicates with a blackboard and the knowledge base which contains the quantitative models required to describe the reactor, selected fission product production and radioactive decay processes, Purex reprocessing and ancillary knowledge. The expert shell supports about 150 forward and backward chaining rules used for inferencing, processing and producing active model values, static knowledge structures, methods, inheritance plans, frames and slots, rule domains, numerical algorithms, inferencing plans and agenda control. The shell also provides textual English explanations and default data to the user when requested. APES checks for consistency of both the data supplied by the user and data inferred and developed by the expert system. Sample data from APES compares well with ORIGEN2 and SEPHIS/MOD4 code outputs for U-235 enrichments from about 1 to 10 percent and burnup of 1500 to 30,000 MWD/MTHM.