ISS: Understanding the Gravity Effect on Flow Boiling Through High-Resolution Experiments and Machine Learning

ISS：通过高分辨率实验和机器学习了解重力对流动沸腾的影响

• 批准号: 2126437
• 负责人: Chen Li
• 金额: $ 40万
• 依托单位: University of South Carolina at Columbia
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-07-01 至 2025-06-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2126437&HistoricalAwards=false
• 关键词: ISS; Understanding; Gravity; Effect; Flow

Flow boiling plays an essential role in energy-water nexus in both terrestrial and space applications. These applications include thermoelectric power generation, thermal management of power electronics and microelectronics, water purification, and heating, cooling and air-conditioning systems. However, flow boiling is significantly affected by five major forces such as surface tension, inertia, shear, vapor evaporation momentum, and gravitational force. The significant changes of channel sizes and working conditions (such as flow rate, heat load, and temperature) result in various contributions of these five forces and hence drastic changes of flow boiling patterns and performance. In addition, it is extremely challenging to conduct experiments of flow boiling in a wide range of channel sizes and working conditions due to the prohibitive costs and efforts. In this project, a package of “Deep Models of Flow Boiling” will be developed to understand the effects of these major forces on flow boiling through the combined use of ground and microgravity experiments and the machine learning based techniques. The models are aimed to not only predicting flow boiling characteristics but also creating synthetic images of flow patterns. This project will pave the way for performing virtual flow boiling experiments under a wide range of working conditions. Furthermore, it would provide a powerful platform to study and design flow boiling-based water-energy systems in a significantly more comprehensive and economic way.The challenging objective of developing the deep models of flow boiling will be achieved by three major research tasks. First, high-resolution experiments and dataset will be constructed. In order to assure accurate and more continuous inputs for machine learning, a complete and accurate data pool of flow boiling will be built through high-resolution experiments under a wide range of working conditions in terrestrial conditions on a test setup that is identical with the test section of the NASA Flow Boiling and Condensation Experiment (FBCE) on the International Space Station (ISS). Experimental data on the FBCE in ISS will be also collected to provide a quality dataset in microgravity. Second, modeling of the force effect on physical variables will be achieved by machine learning. An end-to-end Multi-Target Hybrid Deep Regression (MTHDR) framework will be built to predict physical variables of flow boiling using the collected datasets from both ground and ISS experiments. Third, image synthesis will be performed for two-phase flow patterns. A generative adversarial network (GAN)-based model will be developed to create images of two-phase flow patterns so as to establish a framework to understand and even quantify the effects of major forces on extremely complex two-phase flow patterns.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

流动沸腾在地球和空间应用中的能量-水联系中起着重要作用。这些应用包括热电发电、电力电子和微电子的热管理、水净化、加热、冷却和空调系统。然而，流动沸腾受到表面张力、惯性、剪切、蒸汽蒸发动量和重力等五种主要力的显著影响。通道尺寸和工作条件（如流量、热负荷和温度）的显著变化导致这五种力的贡献不同，从而导致流动沸腾模式和性能的剧烈变化。此外，由于成本和努力过高，在大范围通道尺寸和工作条件下进行流动沸腾实验极具挑战性。在这个项目中，将开发一套“流动沸腾的深层模型”，通过结合使用地面和微重力实验以及基于机器学习的技术来理解这些主要力量对流动沸腾的影响。这些模型的目的不仅是预测流动沸腾特性，而且是创建流动模式的合成图像。该项目将为在广泛的工况下进行虚拟流沸腾实验铺平道路。此外，它将提供一个强大的平台，以更全面和更经济的方式研究和设计基于流动沸腾的水能系统。发展流动沸腾的深层模型这一具有挑战性的目标将通过三个主要的研究任务来实现。首先，构建高分辨率实验和数据集。为了保证机器学习的准确和更连续的输入，将在与国际空间站（ISS） NASA流动沸腾和冷凝实验（FBCE）测试段相同的测试装置上，通过在地面条件下广泛工作条件下的高分辨率实验，建立一个完整准确的流动沸腾数据池。还将收集国际空间站FBCE的实验数据，以提供高质量的微重力数据集。其次，力对物理变量的影响的建模将通过机器学习来实现。将建立端到端多目标混合深度回归（MTHDR）框架，利用地面和ISS实验收集的数据集预测流动沸腾的物理变量。第三，将对两相流型进行图像合成。将开发一个基于生成对抗网络（GAN）的模型来创建两相流模式的图像，以便建立一个框架来理解甚至量化主要力量对极其复杂的两相流模式的影响。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。