Total Measurement Uncertainty Estimation For Tomographic Gamma Scanner

ABSTRACT

Radioactive waste contained in drums can be highly heterogeneous in matrix distribution and may also exhibit a non-uniform and unrelated distribution of radionuclides. Under such circumstances, accurate quantitative results can be obtained by performing non-destructive assay of the waste using a Tomographic Gamma Scanner (TGS). The TGS combines high-resolution gamma spectrometry and low spatial resolution 3-dimensional transmission and emission imaging techniques to accomplish assay goals. The transmission image is a voxel-by-voxel distribution of linear attenuation coefficients throughout the drum volume, and the emission image is a voxel-by-voxel distribution of the source activity. The TGS technique is well suited for low to moderate density waste matrices, say 1.0 g.cm ^-3 or below for 55 U.S. gal. drums, although it can be extended to higher densities by using alternative approaches to the analyses. These include the uniform layer and the bulk density analyses. In the uniform layer approach, all the voxels in a given drum layer (or segment) are populated with the same average value of linear attenuation coefficient. In the bulk density approach all of the voxels in all the drum layers are populated with the same value of linear attenuation coefficient.

The TGS technique can tolerate a higher degree of matrix heterogeneity and a greater non-uniformity of the source distribution than other, non-imaging, g-ray techniques. However, the method is not immune to radial biases and source distribution dependent errors. For source positions that are at the boundary between two layers, the assay results may be biased depending on the location of the emission image. In general, these biases tend to get worse with increasing matrix densities. An extensive measurement campaign was undertaken by CANBERRA Industries to study the above mentioned biases and quantify the Total Measurement Uncertainty (TMU) for the TGS technique. Point sources of ¹³⁷Cs and ⁶⁰Co were located inside drum matrices whose densities ranged from 0 g.cm ^-3 to 2.9 g. cm ^-3 and assayed using the TGS. To study the radial bias, single point sources of ¹³⁷Cs and ⁶⁰Co were co-located at different radial positions and assayed. To study the source distribution dependent error, three point sources of ¹³⁷Cs and three of ⁶⁰Co were distributed randomly inside a given drum matrix and assayed using the TGS. This is based on the postulate that a typical waste drum is likely to contain at least 3 equivalent point sources randomly distributed in the drum volume rather than a single localized source. For each matrix drum, fifteen such random source distributions were generated and assayed. This was repeated for six matrix drums with densities ranging from 0 g.cm ^-3 to 2.9 g.cm ^-3. To study the effect of an emission source at a layer boundary, a point source of ¹³⁷Cs and ⁶⁰Co were co-located inside a drum matrix at a given radial position and at the boundary between 2 segments and assayed. The measurements were repeated for a few radial positions and for drums with different matrix densities.

The present paper describes the challenges posed by difficult to assay waste along with the experiments and analyses used to construct the uncertainty contributions.

INTRODUCTION

The Tomographic Gamma Scanner (TGS) is a High Resolution Gamma Spectrometry (HRGS) based instrument that is being increasingly employed to perform non-destructive assay of radioactive waste. The TGS methodology combines low spatial resolution imaging techniques with HRGS and in certain situations can yield quantitative results that are more accurate when compared to non-imaging methods ^[1-3]. In a TGS assay, the waste drum is scanned with three degrees of freedom, (i.e. rotation, translation and elevation). The waste drum is typically divided into 16 vertical segments, and at every segment two scans are performed. In the transmission scan, a highly collimated gamma ray source is used to interrogate the waste matrix. In the emission scan, the transmission source is not exposed, and the detector views the gamma ray emissions from radionuclides within the item. The TGS scan sequence generates a series of data grabs or views which over determine the contents of the voxels. Algebraic reconstruction in real space is used to extract a best fit solution consistent with the data and as free from spurious features as possible. The transmission data is used to determine the linear attenuation coefficient map (transmission image). The emission data is used to solve for the radionuclide distribution on a voxel-by-voxel basis, which is then corrected for photon attenuation using the transmission map.

As in any NDA method, the Total Measurement Uncertainty (TMU) budget must be quantified for the TGS for a proper interpretation of the assay results. In this paper, we estimate the TMU contributions due to radial biases, source non-uniformity, and errors due to source being located at the boundary between 2 segments for 208 liter drums. The TMU estimates are obtained for several drums matrices in the density range between 0.4 g.cm ^-3 to 1.0 g.cm ^-3. The results for high densities will be presented in a future report.