Control of convergence in a computational fluid dynamics simulation using ANFIS

Abstract

Adaptive-network-based fuzzy inference system (ANFIS) was used to control convergence of a computational fluid dynamics (CFD) algorithm. Normalized residuals were used to select the relaxation factors on each iteration of the computation. The ratio between the residual of the current iteration and of the previous iteration was used as a control input. ANFIS was designed to keep the tuning index at or near one throughout the fluid dynamics simulation. The finite volume CFD algorithm SIMPLER was used as the platform for this study. Four different benchmark cases were examined.