A methodology to reduce the computational effort in 3D-CFD simulations of plate-fin heat exchangers

Abstract

The analysis of a plate-fin heat exchanger performance requires the evaluation of key parameters such as heat transfer and pressure drop. In this regard, Computational Fluid Dynamics (CFD) can be proficiently adopted, at the design stage, to predict the performance of plate-fin heat exchangers. However, these last are often characterized by a complex geometry, such as in the case of plate exchangers with turbulators, leading to a huge computational effort, which often exceeds the available resources. In this study, a numerical methodology for the simulation of plate heat exchangers is proposed, to bypass the limits imposed by the computational cost. The methodology relies on the simulation of a minimal portion of the exchanger (two plates, one per fluid) characterized by periodic boundary conditions (that mimic the presence of several layers). The total heat exchanged is obtained simply multiplying the calculated heat transfer by the number of plate couples composing the device. Moreover, the two plates allow to calibrate porous media which are adopted to rebuild (in a simplified version) the two fluid circuits of the whole exchanger and obtain the overall pressure drop across the device for both the hot and cold fluids.The proposed approach is validated against experimental data of an oil cooler for automotive application, that is a plate-fin heat exchanger characterized by the presence of turbulators. The numerical outcomes are compared to the experiments in terms of pressure drop and heat transfer for a wide range of volumetric flow rates. Particular attention is devoted to the mesh sensitivity and the adopted computational grid minimizes the number of cells (and, thus, the computational cost), without compromising the accuracy. Moreover, the Reynolds-Stress-Transport turbulence model is accurately selected among the most diffused ones, in order to properly match the test bench data.The proposed methodology allows to reduce of nearly one order of magnitude the total number of cells required for the simulation of the heat exchanger performance. The heat transfer is predicted with high accuracy, i.e. error is always lower than 4%. As for the pressure loss, the deviation compared to the experiments increases up to nearly 15% (for one of the simulated conditions) but it is considered still acceptable.