Deep learning-based prediction of flow and heat transfer in pyrolytic hydrocarbon fuels at supercritical pressure
Shuai Xu, Yu Feng, Feng Chen, Cheng‐Peng Li, Xingguo Wei, Jiang Qin
Abstract
The prediction of flow and heat transfer characteristics of supercritical hydrocarbon fuels is crucial for the research of regenerative cooling systems in hypersonic vehicles. This paper proposes a deep learning model for predicting the flow and heat transfer characteristics of supercritical hydrocarbon fuels in regenerative cooling channels. Through testing and comparative analysis, the results indicate that the model achieves high accuracy in predicting the flow and heat transfer characteristics during the pyrolysis of hydrocarbon fuels. Analysis of the model's predictions for fuel temperature, velocity, heat transfer coefficient, conversion, and the mass fraction of cracking products reveals linear correlation coefficients exceeding 0.99 when compared to computational fluid dynamics (CFD) calculation results. Additionally, the model performs well on datasets beyond the training set. In the new test dataset, the mean relative errors (MREs) for temperature, velocity, heat transfer coefficient, fuel conversion, and methane mass fraction are 0.45%, 7.4%, 1.0%, 7.3%, and 8.4%, respectively. These results attest to the model's robust generalization capabilities, indicating its utility for predicting fluid heat transfer under diverse operating conditions. This study enables the rapid prediction of the heat transfer capabilities of supercritical hydrocarbon fuels with pyrolysis reactions, which is crucial for the design of thermal management systems in aircraft.