Machine learning in nuclear safeguards

Current nuclear safeguards analysis techniques either separately analyse signatures measured by different instruments, or combine them in a simplified way. These techniques also rely heavily on operator declared information, which, from a safeguards point of view, is not a desirable property. Multivariate analysis (MVA) techniques, or machine learning algorithms, have been used in several other fields characterized by complex systems or large amount of data. The use of MVA is however still limited in nuclear safeguards research and non-existing in nuclear safeguards implementations. For this reason we have over the past years, been exploring both classification and regression techniques, aiming to predict fuel parameters such as initial enrichment, burnup and cooling time, as well as quantify plutonium content and identify partial defects.

Selected publications

S. Grape, E. Branger, Zs. Elter, L. Pöder Balkeståhl, Determination of spent nuclear fuel parameters using modelled signatures from non-destructive assay and Random Forest regression, Nuclear Inst. And Methods in Physics Research, A 969 (2020)

M. Åberg Lindell, P. Andersson, S. Grape, A. Håkansson, M. Thulin, Estimating irradiated nuclear fuel characteristics by nonlinear multivariate regression of simulated gamma-ray emissions. Nuclear Instruments and Methods in Physics Research, A, Volume 897, pp 85-91 (2018)

M. Åberg Lindell, P. Andersson, S. Grape, C. Hellesen, A. Håkansson, M. Thulin, Discrimination of irradiated MOX fuel from UOX fuel by multivariate statistical analysis of simulated activities of gamma-emitting isotopes, Nuclear Instruments and Methods in Physics Research, A, 885, pp 67–78 (2018)

L. Caldeira Balkeståhl, Z. Elter, S. Grape, C. Hellesen, Nuclear safeguards verification of modelled partial defect PWR fuel using multivariate analysis. IAEA Symposium on International Safeguards: Building Future Safeguards Capabilities, Nov 5-8, Vienna (2018).

Z. Elter, L. Caldeira Balkeståhl, S. Grape, C. Hellesen. Nuclear safeguards verification of modelled BWR fuel using a multivariate analysis approach. IAEA Symposium on International Safeguards: Building Future Safeguards Capabilities, Nov 5-8, Vienna (2018).

C. Hellesen, S. Grape, P. Jansson, S. Jacobsson Svärd, M. Åberg Lindell, P. Andersson, Nuclear Spent Fuel Parameter Determination using Multivariate Analysis of Fission Product Gamma Spectra. Annals of Nuclear Energy, Volume 110, pp 886-895 (2017)