Estimated Uncertainty in Segmented Gamma Scanner Assay Results due to the Variation in

General purpose gamma scanners are often used to assay unknown drums that differ from those used to create the default calibration. This introduces a potential source of bias into the matrix correction when the correction is based on the estimation of the mean density of the drum contents from a weigh scale measurement. In this paper we evaluate the magnitude of this bias that may be introduced by performing assay measurements with a system whose matrix correction algorithm was calibrated with a set of standard drums but applied to a population of drums whose tare weight may be different. The matrix correction factors are perturbed in such cases because the unknown difference in tare weight gets reflected as a bias in the derived matrix density. This would be the only impact if the difference in tare weight was due solely to the weight of the lid or base, say. But in reality the reason for the difference may be because the steel wall of the drum is of a different thickness. Thus, there is an opposing interplay at work which tends to compensate. The purpose of this work is to evaluate and bound the magnitude of the resulting assay uncertainty introduced by tare weight variation. We compare the results obtained using simple analytical models and the 3-D ray tracing with ISOCS software to illustrate and quantify the problem. The numerical results allow a contribution to the Total Measurement Uncertainty (TMU) to be propagated into the final assay result. INTRODUCTION The non destructive assay of drummed gamma-ray emitting radioactive waste is commonly performed using a spectroscopic drum scanner. Correction factors are needed to account for the attenuation of the characteristic gamma-ray lines by the contents of the drums. One way to estimate these corrections is to estimate the mean density of the contents and use it in a suitable mathematical algorithm. At each energy of interest the algorithm estimates the matrix correction factor from the mean density assuming a particular material composition. Generally the matrix is taken to be homogeneous and uniform with activity uniformly distributed throughout. A common practice is to determine the parameters used in the algorithm from calibration measurements. These are typically performed using a set of standard drums and matrices chosen to simulate the waste forms. In application however it is possible to encounter a variety of drum types at a given facility. Where the differences are significant and can be recognized allowances can be made and an item specific calibration can be used. But, sometimes, especially in automated waste treatment plants, the drum type may not always be known to the assay system. In this case the assay will be based on the default settings. If the drum does not share the same characteristics as the calibration drum set, and in this study we are concerned principally with tare weight, then a bias can result. The purpose of this work is to evaluate and bound the magnitude of the assay uncertainty introduced by tare weight variation. We use a combination of simple analytical models and 3-D ray-tracing to illustrate and quantify the problem. The numerical results allow a contribution to the Total Measurement Uncertainty (TMU) to be propagated into the final assay result. WM2009 Conference, March 1-5, 2009, Phoenix, AZ PARKER FORMULA APPROACH The matrix attenuation correction factor for each segment of a Segmented Gamma Scanner (SGS) [1] can be estimated using the Parker equivalent slab model [2]. In this case the efficiency correction factor for each layer is determined using the following formula: