Numerical investigation of water forced convection inside a copper metal foam tube: Genetic algorithm (GA) based fuzzy inference system (GAFIS) contribution with CFD modeling

Abstract

Computational fluid dynamics (CFD) is an accurate method, but it is not always simple. In this approach, the partial differential governing equations are discretized and solved for many divisions in the domain of the fluid. The solution to such equations takes so much time and needs a computer with high hardware capacity. Although many studies have focused on the accuracy improvement of the CFD method, a few investigations have considered CFD optimization, especially in prediction time. The machine learning (ML) of the CFD data using artificial intelligence (AI) algorithms could reduce CFD calculations. For the first time, this study aims to analyze the efficiency of an AI method, namely the genetic algorithm-based fuzzy inference system (GAFIS), in contribution to CFD modeling. A turbulent flow of water through a heated metal foam tube is considered. The wall heat flux and water inlet velocity are constant, while the water inlet temperature changes from 295 ok to 335 ok. To precisely predict the pressure, the intelligence status of the GAFIS must be evaluated. So, the effects of GAFIS parameters, including the number of populations, number of clusters, and percentage of mutation, on the prediction error are investigated. For validation, the GAFIS performance is verified by the fuzzy adaptive network (ANFIS).The results revealed that the best intelligence condition (regression number of 0.996) is related to the number of populations of 30, the number of clusters of 10, and the percentage of mutation of 0.6. The intelligent GAFIS predicted the same pressure as the CFD. The ANFIS and the GAFIS showed the same predictions as the CFD data. The GAFIS learning and prediction times were 223 and 345 s, respectively, while those times were 3 and 10 s for the ANFIS. The AI algorithms reduced the computational time comparing to the CFD calculation time (∼ 5,409 s).