AI-Enabled Optimization of Radioactive Waste Management at the Nigeria Research Reactor-1

Abstract

Efficient management of radioactive waste generated in research reactors remains an important operational and regulatory challenge, particularly in facilities with limited storage capacity and heterogeneous waste streams. This study presents an AI-enabled decision-support framework for optimizing radioactive waste management at the Nigeria Research Reactor-1 (NIRR-1), a 30 kW miniature neutron source reactor. The framework integrates deterministic radioactive decay modeling, machine learning–based waste classification, activity forecasting, and constrained multi-objective optimization. Waste inventory data were analyzed using a Random Forest classifier and regression-based activity prediction models, followed by Pareto-based optimization of storage scheduling under regulatory clearance constraints. The classification model achieved an accuracy of 93.7%, while activity forecasting produced a coefficient of determination R² = 0.993 with normalized prediction errors below 5%. Optimization results indicate that systematic decay-informed scheduling can reduce projected storage congestion from 82% to 61%, while the integrated AI–optimization framework further reduces storage utilization to approximately 50.8%. The results demonstrate that combining decay physics with data-driven optimization can significantly enhance operational efficiency without compromising regulatory compliance. The proposed framework provides a practical computational tool for improving radioactive waste management at research reactor facilities.