Speaker
Description
Authors (affiliation): Hugues Paradis (CEA, France), Anthony Der Mesrobian-Kabakian (CEA, France), Antoine Cagniant (CEA, France), Olivier Delaune (CEA, France), Charles Philippe Mano (CEA, France), Luc Patryl (CEA, France)
Underground or underwater nuclear tests can eventually lead to radioxenon releases in the atmosphere. In this scope, the verification regime of the Comprehensive nuclear Test Ban Treaty installed dedicated noble gas systems for the detection of low-level radioxenon activity in the air.
Recent development of these noble gas systems included new detector technologies. They exhibits very low background count rates. In this case, radioxenon signal as low as a few counts per 12h are expected. Therefore, for such low count measurements, classical Currie law estimation for measurement detection thresholds and detection limits are not precise enough. In this context, the CEA/DAM implemented several algorithms (matrix inversion, iterative process), and keeps the effort to improve the data analysis with innovative tools, such as spectral unmixing.
Due to the lack of statistics, it is not convenient to test and compare these algorithms on measured low-level radioxenon spectra, therefore a Monte Carlo simulated database of spectra was generated, for several detection configuration (high resolution beta/gamma spectra, low resolution gamma/ high resolution beta spectra, lowresolution beta/gamma spectra). To optimize the analysis of this database, the spectral unmixing algorithm was ported on GPU, leading to a drastic decrease in computation time and allowing for the processing of large simulated datasets in reasonable delay.
This work will be presented, and algorithms key performance indicators for each detector configuration will be compared.
