Development of mechanistic-artificial intelligence model for simulation of numerical data of water flow in porous materials

Abstract

Fluid dynamics of water flow through porous metallic media is significant for cooling and heating applications. The prediction of the velocity of fluid flowing inside the porous media could provide useful data for pressure drop calculations. This prediction is usually performed by a mathematical modeling approach like computational fluid dynamics (CFD). The computational method is a powerful means of precision calculation, although it could take more time and expenditure for more complex geometries or more complex fluid flow regimes. Artificial intelligence (AI) algorithms could learn and map CFD data under several conditions. AI algorithms could continue the prediction of physical data, saving computing time and performance. The present study focuses on CFD modeling of water flow inside a pipe filled with copper porous media. For the first time, the fuzzy bee algorithm (BAFIS) maps the generated CFD data to inlet velocities of 0.5, 0.7, 1.1, and 1.3 m/s. The BAFIS intelligence assessment for accurate convective flow prediction in porous media is not available in the literature. Additionally, the reliability of this method for missing CFD data prediction has not been considered yet. The results showed that the maximum intelligence (regression=0.97) is for a bee number of 140. Using the most intelligent BAFIS, the outlet velocity can be predicted by the artificial intelligence method for further nodes and inlet velocities without CFD modeling. BAFIS could precisely predict the outlet velocity for missing data with an inlet velocity of 0.91 m/s based on previously mapped data. A comparison was made between BAFIS and the fuzzy neural network (ANFIS) for CFD data predictions. The mean-square error and the root-mean-square error of ANFIS were slightly more than BAFIS (i.e., 0.1% and 3%, respectively).