DEM 2024

Geostatistical deconvolution to map punctual sources from radiation measurements

Authors: Yvon Desnoyers, GEOVARIANCES and Chaly Mahe, CEA Marcoule

Tuesday 28 May – 13:50 – 15:30 – T04: Initial Radiological State Characterization – Stats / Sampling / Optimisation

Abstract

The purpose of geostatistics deconvolution is to improve the spatial processing of nondestructive radiological data (dose rate measurement, in situ gamma spectrometry…) using an advanced algorithm. The objective is to estimate better the actual activity content of the measured object (tank, pit, concrete slab, wall…). This deconvolution method can be used from Very Low-Level Waste characterization to High-Activity cases.

Due to the fact input data are not punctual (nor only the average value on a supposedly homogeneous area), the spatial response of the detector can advantageously be considered during the variographic step. This weighting function is derived from the probe characteristics and the measuring conditions (distance to the surface, possible collimation…). Thus, the map of the radiological sources can be directly obtained from the radiation map by deconvolution, benefiting from the spatial correlation between partially overlapping measurements. This is a smart resolution of a challenging inverse problem. This paper first illustrates promising results on simple 2D cases with one or two punctual sources with different measuring conditions.

This improved knowledge of the radiological state of the characterized surface then significantly impacts the destructive sampling campaign, which is necessary for the complete radiological characterization (including hard-to-detect radionuclides).