A Computational Workflow to Study Particle Transport in Porous Media: Coupling CFD and Deep Learning

In this work, we studied the coupling of CFD simulation with machine learning models, by using a large set of computational result as the training dataset of a simple fully-connected neural network. The focus of the CFD investigation is the flow and colloid transport in porous media models, both simple and complex, with the end result of obtaining a computationally inexpensive data-driven surrogate model able to replace the CFD simulation, while keeping the same accuracy. While considerable success was obtained in the case of simpler geometries, more sophisticated deep learning models are needed to treat cases characterized by non-trivial fluid dynamic structures.