The verification of spent fuel is important both for safeguards and for the safe, economic and ecological final disposal of spent nuclear fuel. The experimental observables associated to non-destructive assay measurements of spent fuel assemblies are often a complex function of the characteristics of the fuel, its irradiation history and other variables related to the used measurement setup and devices; nowadays one often assumes that some of the variables are known in order to interpret the data and draw conclusions.
A database of detector simulated responses corresponding to 8400 cases with different fuel characteristics and irradiation parameters is available at SCK•CEN. In this work, we propose the use of these simulated observables as input for data analysis algorithms aimed at characterizing the spent fuel and drawing safeguards conclusions.
We propose the application of artificial neural networks due to their ability to generalize non-linear relationships. This project builds on previous research at SCK•CEN in the field and aims at optimizing the performance of the neural network and defining a minimal training data set allowing to achieve the desired performance.