Heat Exchanger Design, Analysis and Evaluation by Data-Driven Computational Fluid Dynamics

Abstract Heat exchanger design, analysis and evaluation are traditionally based on experimental data, empirical relationships and theoretical equations. Recently, due to the recent development of data-driven computational fluid dynamics, we are able to conduct “smart” analysis based on a large amount of data. In the present research, we developed a fast CFD data generation method. Based on the large amount of data, we developed two data-driven models: 1) a machine learning model that can predict the heat transfer behavior, and 2) a tree-based model that can capture the feature importance. This methodology can potentially be applied to future high-efficiency heat exchanger designs.