机器学习在沸腾传热过程中的应用进展

Boiling heat transfer, widely applied to energy and power equipment, such as boiler water-cooled walls and nuclear reactor evaporators, is critical to their efficient operation. Accurate prediction of boiling heat transfer is therefore essential to prevent such equipment from local overheating, dry-out, over-temperature, and other abnormalities effectively. While traditional prediction models have been constrained by the challenge of coupling complex multi-physical fields, machine learning offers new ideas for addressing this challenge through data-driven modeling and intelligent analysis. First of all, the application of artificial intelligence (AI) techniques to boiling heat transfer prediction is reviewed, with recent years of work using machine learning algorithms for heat transfer coefficient prediction and bubble dynamics parameter extraction summarized. The results indicate that, although notable advantages of Al in boiling heat transfer have been demonstrated, challenges remain, including strong dependence on data, obvious "black box" nature of models, and substantial computational resource requirements. Furthermore, future research directions for machine learning in boiling heat transfer are identified, including but not limited to rapid and low-cost multiphase flow simulation development of generalized models across different scenarios, and real-time dynamic control of heat transfer systems. Finally, it is promisingly expected that the integration of physical laws with data-driven modeling, together with the building of datasets and the provision of open-source algorithms, will be pivotal in deepening the application of AI to boiling processes, further empowering the development of efficient energy and power systems. This study can provide a reference for the application of machine learning methods to intelligent prediction, optimal design, and safety control of boiling heat transfer processes.