Geostatistical Sampling Optimization of Contaminated Facilities
Résumé
Geostatistics applied to radiological evaluation of nuclear premises provides methods to estimate radiological activities, together with their associated uncertainty. These tools enable to set up a sophisticated sampling methodology combining radiation map and destructive samples. The radiological assessment is divided into two steps: first, a systematic (exhaustive) control of the surface activity is performed. Then, to assess the true contamination, concrete samples are collected and analyses are performed at several locations within the premises. These two types of measurement are first dealt separately, then cokriging is applied to estimate the contamination over the premise, taking both measurements into account. This work provides a methodological study of geostatistical and computational approaches to target suitable areas for additional radiological measures. In order to compare the proposed augmented sampling designs, several optimization criteria can be taken into account. Their diversity ensures the coverage of a wide range of practical problematics. Two algorithms (greedy algorithm and simulated annealing) are developed to optimize the chosen criterion value as a function of the location of the additional points. The sampling scenarios obtained with the different algorithms are compared in terms of optimization performance and computational efficiency.