Estimation of the Reynolds number in a Poiseuille flow using artificial neural networks

Abstract

In this work the estimation of Reynolds number in a 2-dimensional Poiseuille flow is explored employing artificial neural networks (ANNs). The velocity fields of the fluids were generated evaluating the Hage-Poiseuille equation for different Reynolds (Re) from 20 to 2000. The velocity profile obtained for each case is used as input data for the ANNs, which is then trained to predict the Re. The results show an accuracy of at least of 99.5% in all prediction cases. This analysis is the first step towards the construction of a Machine Learning algorithm capable of computing physical parameters in more general scenarios.