(Improvement of quantitative data evaluation for the non-destructive characterization of radioactive containers and objects)

Projektleiter: Dr. Ch. Lierse v. Gostomski

Projektleitung: Dr. Thomas Bücherl

Wissenschaftler: Dr. Stefan Rummel, Marek Blaszczynski

Das diesem Statusbericht zugrundeliegende Vorhaben wird mit Mitteln des Bundesministeriums für Bildung und Forschung unter dem Förderkennzeichen 15S9411 gefördert.

 

Project description:

For the characterization of the radioactive waste packages (e.g. 200 L drums) preferably non-destructive methods are applied. For identification of the radioactive inventory segmented gamma scanning is commonly applied. Information on the matrix composition is achieved by a-priori information or transmission measurements, which can range from a single point transmission measurement over 1D- and 2D-radiography up to 2D and 3D-tomography, the latter giving the most detailed information but being time-consuming.

The conceptual formulation of a new R&D project funded by the German Federal Ministry of Education and Research (BMBF) is related to the improvement in quantification in the non-destructive characterization of radioactive waste packages applying segmented gamma scanning and transmission measurements. The boundary conditions for this project are (i) taking into account the typical instrumentation available at nuclear facilities, (ii) minimize the overall measurement time and (iii) combining data from emission and transmission measurements by bayesian methods.

At nuclear facilities, segmented gamma scanner are state-of-the-art and are mostly equipped with a vertically moveable transmission source. This hardware allows the registration of multi-rotational emission and transmission measurements, i.e. a 2D mapping of both, the isotope specific emission data and the (energy specific) transmission data, on the lateral surface of the waste package. Limiting the development on this data considers the first two boundary conditions, where the minimization of the overall measurement time depends on the spatial and time resolution the waste package is scanned and analyzed.

The basic idea is to backproject this data on a voxel model of the waste package, each voxel representing information on the isotopes and activities it contains, as well as on its matrix properties (e.g. material, density etc.). The filled voxel model finally must reproduce the measured data with sufficient accuracy. This approach has two main challenges: first, the backprojection will not give an unambiguous solution, as this will require data from 3D transmission or emission tomography, and second, the size of voxel model and the requirements on storage and data handling may become problematic. The latter results from a simple estimation: Assuming, a 200 L waste drum of 800 mm in height and 560 mm in diameter is characterized. As its wall thickness is about 1.5 mm, an edge length of each voxel of about 0.5 mm is required, i.e. the complete voxel model is set up by more than two billion individual voxels. Assuming that the information each voxel must store is about 100 byte, this gives an overall storage size of about 170 gigabyte. On the other hand, about 20% of the voxels represent areas outside the waste package contain no relevant information, and in other areas, the variation of information may be described by a much lower spatial resolution. Thus the application of special methods like adaptive refinement trees and adaptive mesh refinement in combination with special mathematical routines (e.g. sparse matrices) will tackle this challenge.

The first challenge, the ambiguity of the solutions, is countervailed applying a Bayesian approach and correlating the data by appropriate likelihood functions, taking into account all available a-priori information. The results are probability density functions for the parameters in each voxel. The required information is then determined by summing the functions for the relevant parameter, i.e. the activity of an isotope. The development is subdivided into three steps to overcome its complexity. First, an appropriate model (likelihood function) for the backprojection of the transmission data is developed referring to tomographic reconstruction techniques for its development. Image processing applied on the radiography give additional information on geometries, dimensions, attenuation properties etc. of objects the matrix is composed of, and is integrated in the evaluation procedure as well as a priori information on the parameter ranges etc. However, a Kriging surrogate model is applied to tackle the challenge of infinitive solutions. The second step is similar to the first one, but now working on the emission data. For simplification, the matrix is assumed to be known. The development will follow the ideas and the workflow of the first step. The third step will combine the results of the other steps, i.e. the known matrix of the second step is replaced by the result of the first.