Combination of non destructive assay and machine learning techniques to detect replaced pins in spent fuel assemblies

This Master's thesis deeply explores the integration of advanced non-destructive assay (NDA) techniques combined with the capabilities of machine learning to bolster the detection of replaced pins within spent fuel assemblies. Such endeavors are fundamental for the safeguarding of nuclear materials, ensuring the utmost integrity and safety throughout their lifecycle.Key to this research are innovative technologies like the ForkBall detector, Self-Indication Neutron Resonance Densitometry (SINRD), and Partial Defect Tester (PDET). Each of these tools offers unique insights and advantages in pinpointing discrepancies within the fuel assemblies. A cornerstone of this study is the utilization of a comprehensive database of Monte Carlo simulations, meticulously curated to represent a wide spectrum of fuel assembly configurations and real-world scenarios. The primary ambition of this investigation is to harness the power of machine learning to accurately identify replaced pins. Essential to this task is determining the best practices for detector placement, calibration, and measurement duration, ensuring that the results are both reliable and actionable. A significant portion of this research is devoted to refining and evaluating the developed machine learning algorithms. Misclassified observations, rather than being seen as mere errors, are perceived as essential feedback, driving further refinement. By understanding the nuances of these inaccuracies, the research seeks to improve the robustness and reliability of the algorithms. In essence, this study aims to significantly elevate the efficiency and accuracy of nuclear safeguards inspections, making a pivotal contribution to global nuclear security efforts. It underscores the importance of seamlessly merging traditional nuclear assay methodologies with the potential of contemporary computational techniques, creating a robust framework to tackle today's nuclear challenges.