Nondestructive assay in complex, self-attenuating radioactive materials by gamma spectroscopy: A mathematical model and empirical determination of error.

摘要

Over the past several years, portable High Purity Germanium (HPGe) gamma spectroscopy systems have been used with nondestructive assay techniques to characterize waste items at Los Alamos National Laboratory (LANL) and other Department of Energy (DOE) sites. Accurate quantification of the radionuclide contamination required that an analytical model be developed. The resulting model, based on a point kernel volumetric approach, is used to translate the raw spectral information into total activity for each detected radionuclide. Using uncollimated, in situ counting techniques, a relatively high-efficiency n-type HPGe detector, a portable laptop computer, and a multi-channel analyzer, the nondestructive assay system has superior detection limits with state of-the-art accuracy. Not only can the system be used to successfully count drums, but also large items such as 90 cubic foot boxes, gloveboxes and heavy machinery. Additionally, contaminated materials such as floors, walls, soil and water have been characterized with the same model. Because of the versatility of the model and the very low detection limits attainable, tremendous cost savings have been realized from low-level/TRU waste segregation activities, and free-release/low-level determinations.; The model has been used to determine the activity in a number of contaminated and spiked items and matrices with both known and unknown quantities. In an extensive study to determine the empirical accuracy of the model, a number of measurements were made on sources of known activity. The results show that measurement errors of 20 to 50 percent are achieved, depending on item size, geometry and radionuclide contamination. For larger items, results were typically 30 to 50 percent from the known value. Measurements of smaller and homogeneously contaminated items showed that the measurement errors were the same order of magnitude as the uncertainty of the source (10 to 20 percent). Furthermore, although individual error sources are biased in either the positive or negative, no bias error is apparent in the data collected to date.; This paper will present the model theory and algorithms, detail the methodology for acquiring and analyzing data, and present theoretical and measured errors of the analyses.