Evaluation of machine learning methodologies using simple physics based conceptual models for flow in porous media

Machine learning (ML) techniques have drawn much attention in the engineering community due to recent advances computational and an enabling environment. However, often they are treated as black-box tools, which should be examined for their robustness range of validity/applicability. This research presents evaluation application flow/transport porous media, where exact solutions (obtained from physics-based models) used train ML algorithms establish when how these fail predict first order flow-physics. Exact so not introduce artifacts numerical solutions. To test, validate, physics flow media using algorithms, one needs a reliable set data that may readily available and/or might suitable form (i.e. incomplete/missing reporting, metadata, or other relevant peripheral information). overcome this, we generated structured datasets simple representative building blocks such Buckley-Leverett, convection-dispersion equations, viscous fingering. Then, outcomes those equations fed into examine predictive strength key features, breakthrough time, saturation component profiles. In this research, show physics-informed algorithm can capture physical behavior effects various parameters (even shocks sharp gradients present) on features flow. Furthermore approach utilized solve inverse problems estimate parameters. study, focused capturing dominant selected processes constituting outcome algorithms. We expect per help solving more complex by providing correlative proxies.