Integrated data-driven local characteristic general framework for heat exchanger modeling for supercritical pressure fluids

Abstract

Supercritical pressure (SCP) fluid heat exchangers are increasingly utilized in industrial applications owing to their high compactness and superior heat transfer efficiency. However, existing modeling methods cannot achieve both effective and accurate predictions of heat exchanger performance owing to the significant variations in the thermal properties of SCP fluids. In this study, we propose an integrated data-driven local characteristic general framework for heat exchanger modeling (IDLC-HEM) for SCP fluids that aims to achieve effective dimensionality-reduction modeling and performance assessment for a wide range of heat exchangers. Here, to improve the applicability of the IDLC-HEM to various heat exchanger configurations, we develop a heat transfer surrogate model based on the light gradient boosting machine algorithm. This model, built on a SCP CO₂ convective heat transfer database encompassing 428 working conditions, is integrated into the iterative procedure. The performance of the IDLC-HEM in one- and two-dimensional heat exchanger modeling is evaluated using printed-circuit and microtube heat exchangers, respectively. The results indicate that the deviations between the calculated results of the IDLC-HEM and the ground truth remain below 10%, while the deviation derived by logarithmic mean temperature difference method range from 81% to 473%. Furthermore, a comprehensive performance analysis is conducted for heat exchangers under varying operating parameters. IDLC-HEM offers a practical framework for artificial intelligence-assisted heat exchanger design. As the database expands and surrogate model develops, it can be further promoted and applied in complex scenarios such as phase-change heat changers.