Development of Artificial Neural Network-Based Model for Prediction of Temperature Field in Host Rock of a Geological Disposal Facility for Radioactive Waste

Calculation of temperature field in a deep geological repository (DGR) after emplacement of a large number of heat emitting vitrified radioactive canisters is important and requires large computational time and hence in this study an effort has been made towards development of artificial neural network (ANN) based model that can predict the temperature quickly. The datasets required to train the ANN model were generated using an in-house developed GUI tool for simulating heat diffusion process. Various numerical studies were conducted with different configurations of the ANN model and different datasets of size 50, 100, 150, 200, to optimize the number of input data required to train the model. The results in the form of temperature values predicted by the trained ANN model have been compared with those for the same problem calculated using analytical and finite difference based methods. The trained ANN model can predict temperature values with less than 0.001% error.