Preliminary Results in the Use of Dempster Shafer Theory for a Radiological Waste Characterization Expert System

This report describes an investigation into the applicability of Dempster Shafer theory for representing the uncertainty inherent in the characterization of containerized radiological waste. This research has been conducted by scientists at the Mississippi State University Diagnostic Instrumentation and Analysis Laboratory in collaboration with scientists at the Idaho National Engineering Laboratory. The goal is to develop an expert system that will assist in the determination of the proper disposition of radiological waste containers. Because the waste is typically assayed and examined nondestructively (i. e., without opening the containers), it is impossible to characterize the contents with certainty. The expert system will examine information including assay data and documentation for each waste container to determine if the waste meets the criteria for shipment to a permanent storage facility. Dempster Shafer theory provides a mechanism for representing the uncertainty in a domain without specifying a complete model. 1.0 Introduction A large number of containers of radiological waste from the Rocky Flats Plant at Golden, Colorado, are stored at the Radioactive Waste Management Complex (RWMC) in Idaho. The Idaho National Engineering Laboratory (INEL) is responsible for determining which of these containers “are in compliance with storage, shipping, and disposal requirements” and thus may be shipped to the Waste Isolation Pilot Plant (WIPP) in New Mexico (Matthews et al. 1993). The Stored Waste Examination Pilot Plant (SWEPP) is one nondestructive examination/assay system that has been developed to characterize this waste. Other assay systems are under development. Each of these systems typically perform the characterization on unopened containers. This introduces a large number of complicating factors that can cause the assumptions of the assay systems to be violated or that can attenuate the signals that the assay systems use for calculations. INEL is developing a Waste Assay Measurement Integration System (WAMIS) “to improve confidence in and lower the uncertainty of waste characterization data” (Mousseau