Printed circuit plate heat exchanger (PCHE) has the characteristics of high efficiency and compact heat transfer, but the analysis method of PCHE flow and heat transfer based on finite volume method (FVM) is inefficient and difficult to carry out large-scale optimization calculation.

This study aims to improve the analysis and optimization efficiency of three-dimensional models by constructing a fusion field reduction model for PCHE low and heat transfer.

Firstly, the trapezoidal PCHE was taken as the research object, supercritical CO₂ as the working fluid for the hot channel, and water as the working fluid for the cold channel. Then, the proper orthogonal decomposition (POD), truncated singular value decomposition (tSVD), and gaussian process regression (GPR) were combined to construct a reduced-order model by Latin hypercube sampling and a small amount of sample data generated by Fluent software. Thereafter, the reduced-order model was applied to the prediction of flow and heat transfer characteristics and overall performance of PCHE under new working conditions.

The prediction results show that the fusion field reduction model efficiently decomposes the high rank matrix and accurately predict the multi-physics field distribution of trapezoidal PCHE. For the temperature field and velocity field of PCHE, the root mean square error of the new working condition in the sample space is 1.5×10^-3, and the root mean square error of the new working condition outside the sample space is 6.2×10^-3, and the calculation efficiency is improved by 270 times. However, the prediction effect of the fusion field reduction model on the overall performance of PCHE is poor.

The reduced-order model of the fusion field has certain reference significance for the future optimization analysis work.