| description abstract | Abstract. The interaction between flow complexity and heat transfer processes within spiral wound heat exchangers (SWHEs) remains incompletely understood due to the disparate length and time scales involved. Consequently, traditional computational fluid dynamics (CFD) methods have been extensively employed to elucidate the underlying transport mechanisms and optimize heat exchanger design. However, microscale modeling of transport processes in SWHEs is computationally expensive, while macroscale modeling using homogenization cells reduces computational costs at the expense of accuracy. To address these limitations, an integrated deep neural network (DNN) and CFD method for multiscale modeling was developed and implemented to investigate transport phenomena in SWHEs. In this approach, DNN techniques are utilized to correlate behaviors between microscale and macroscale levels. The DNN model is trained on a dataset generated from numerous microscale simulations and subsequently coupled with the macroscopic governing equations. Experimental validation was conducted to evaluate the capability of the integrated DNN-CFD approach in accurately capturing steady-state flow behavior and thermal transport phenomena within the SWHE. The results demonstrate that the integrated method achieves a significant reduction in computational time, requiring only 3.36% of the time needed for full-scale CFD simulations while utilizing the same computational resources. Furthermore, when compared with experimental data, the relative error in the heat transfer coefficient is less than 20%, indicating that the proposed DNN-CFD integrated method offers considerable practical utility for the numerical study of large-scale SWHEs. | |
