Artificial neural network model development for prediction of nonlinear flow in porous media

Abstract

To explore non-linear flow behaviour in porous media, e.g., packed bed of particles, this study focuses on the prediction of pressure drop/gradient by establishing empirical relations for Forchheimer coefficients based on artificial neural network (ANN) development. To facilitate this process, a large set of experimental data for pressure drop of water flow in various porous media is collected: part from the present experiment and other from those available in the literature. The newly generated ANN model is developed by considering the key influencing factors such as particle density, diameter, and shape, and column porosity. The new model shows excellent performance on the prediction of flow behaviour over previous empirical formula, which is attributed to the ANN's advantages on considering the inter-factor connections and their influences on the results. This proposed approach can provide a new means for analysing the nonlinear flow regime through porous media based on limited dataset.