Turbulent Heat Transfer Characteristics of Supercritical Carbon Dioxide for a Vertically Upward Flow in a Pipe Using Computational Fluid Dynamics and Artificial Neural Network

Abstract

Modelling of turbulence heat transfer for supercritical fluids using Computational Fluid Dynamics (CFD) software is always challenging due to the drastic property variations near critical point. Use of Artificial Neural Networks (ANN) along with numerical methods have shown promising results in predicting heat transfer coefficients of heat exchangers. In this study, accuracy of four different turbulent models available in the commercial CFD software - Ansys Fluent is investigated against the available experimental results. The k-e Re Normalization Group (RNG) model with enhanced wall treatment is found to be the best-suited turbulence model. Further, K-e RNG Turbulence Model is used in CFD for parametric analysis to generate the data for ANN studies. A total of 1,34,698 data samples were generated and fed into the ANN program to develop an equation that can predict the heat transfer coefficient. It was found that, for the considered range of values the absolute average relative deviation is 3.49%.