TEMPERATURE FIELD RECONSTRUCTION of SURFACES HEATED THROUGH RADIATIVE HEAT TRANSFER USING CONVOLUTIONAL NEURAL NETWORKS

Diverse applications require us to know the value of a given physical quantity in real time. For many applications, we can rely on probes to fulfill this task. However, this approach can be challenging when we need to know these quantities in many locations over a given domain due to the spatial limitation involved. In this context, using Convolutional Neural Networks (CNN) can offer a good option to deal with such a problem. A well-Trained physics-informed CNN can reconstruct the distribution of a given physical quantity over a domain using only a few sensors, allowing us to reconstruct the desired field distribution even in a limited space or complex geometries where a large array of sensors remains impractical. In this work, we will present the initial approach to developing a real-Time tool for monitoring the thermal behavior of nuclear reactor pressure vessels. Based on an experimental setup, the team developed a computational model using the Multiphysics Object-Oriented Simulation Environment (MOOSE) framework, where the Ray Tracing and Heat conduction modules were used to evaluate the temperature distribution over a convex metal surface heated through radiative heat transfer. This metal surface represents a section of a heated nuclear reactor vessel wall. The team verified our computational model against the available experimental data. Part of the data generated by the MOOSE model was used to train the Convolutional Neural Network to reconstruct the vessel wall's outer surface temperature. The CNN generalization was then verified against the experimental and computational data. The predicted temperature distribution over the vessel's outer wall presented an R-square metric of 0.9991 when compared against the MOOSE model.