[A-HW25] Near Surface Investigation and Modeling for Groundwater Resources Assessment and Conservation

[AHW25-10] Artificial neural networks for the transport of multi-member radionuclide decay chain: A computationally efficient alternative to numerical methods

*Uyen Thi Thu Nguyen¹, Jui-Sheng Chen¹ (1.National Central University )

Keywords:Radionuclide Transport, Artificial Neural Networks (ANNs), Groundwater Contamination

The transport of radionuclides in subsurface environments is governed by complex processes such as advection, dispersion, sorption, and radioactive decay. Traditional numerical methods, including the finite difference method (FDM) and finite element method (FEM), are commonly used to solve the advection-dispersion equation (ADE) for modeling contaminant transport. However, these approaches often require extensive computational resources, especially for large-scale or multi-species systems. In this study, an Artificial Neural Network (ANN) model is developed to predict the transport and decay of radionuclides in groundwater systems. The ANN is trained on a dataset generated from FDM solutions of the ADE, incorporating both parent and daughter species to ensure realistic representation of radioactive decay. Monte Carlo sampling is used to optimize dataset selection, reducing computational cost while maintaining prediction accuracy. The ANN model demonstrates high precision in replicating FDM results, achieving low mean absolute error (MAE) and root mean squared error (RMSE) values across various Peclet numbers (Pe). Additionally, the ANN significantly reduces computational time compared to conventional numerical methods, making it a viable alternative for rapid and accurate simulations of radionuclide transport. These findings highlight the potential of AI-driven models in enhancing environmental risk assessment and decision-making for radioactive waste management.