Numerical investigation and theoretical description of flame propagation of unsaturated hydrofluorocarbon refrigerants

为了优化使用下一代制冷剂和开发未来环境友好的替代品，迫切需要从根本上了解它们的燃烧特性。层流燃烧速度(LBV)是衡量制冷剂火灾安全性的最重要指标之一，它在可燃物质的安全评估中起着核心作用。对于缓慢燃烧的火焰，如氢氟碳制冷剂，其火焰传播速度可能低于10厘米/S，由于浮力和辐射的影响，准确测定LBV是非常具有挑战性的。只有在微重力条件下才能消除前者的影响，而微重力条件并不广泛。此外，对于HFC制冷剂的安全评估，迫切需要可靠的实验方法和后处理技术，以便在正常重力条件下提供准确的层流火焰速度数据。然而，数值模拟可以极大地支持这种稳健方法的开发、验证和不确定性量化。该项目旨在开发准确和可靠的方法，以更好地了解、测量和预测现有和未来HFC制冷剂的燃烧特性和安全相关参数。HFC火焰的高保真详细数值模拟有助于分析，这些火焰考虑了浮力和辐射效应。第一个目标是与实验子项目2联合评估LBV。模拟和分析将有助于减少和量化通过测量评估火焰速度的潜在不确定性。此外，还将开发不同大小和保真度的简化动力学机制，以支持数值模拟和渐近分析。第二个目标是开发描述Markstein数对LBV的影响的模型，重点是火焰中的传输现象如何受到差异扩散的影响。现有的基于伴随的框架将扩展到高阶灵敏度，以解释Markstein数对热化学和输运机制的影响。第三个目标是开发基于机器学习(ML)和结构基团分析相结合的降阶模型，以预测现有和未来HFC制冷剂的LBV和Markstein数。训练数据将由来自子项目2的实验数据组成，并补充来自模拟的数据，并将在项目期间根据实验设计(DOE)方法不断改进。所得到的ML-模型将用于HFC制冷剂的基于模型的设计和安全评估。

For the optimal use of next generation refrigerants and the development of future environmentally friendly alternatives, a fundamental understanding of their combustion properties is urgently needed. One of the most important metrics used to classify fire safety of refrigerants is the laminar burning velocity (LBV), which plays a central role in the safety assessment of flammable substances. For slow-burning flames, such as hydrofluorocarbon (HFC) refrigerants, which may propagate at flame speeds of less than 10 cm/s, it is very challenging to accurately determine LBVs due to the effects of buoyancy and radiation. The elimination of the former effect can only be obtained under microgravity conditions, which are not widely accessible . Further, for the safety assessment of HFC refrigerants, there is a strong need for robust experimental methods and post-processing techniques that provide accurate laminar flame speed data under normal gravity conditions. However, numerical simulations can greatly support the development, verification, and uncertainty quantification of such robust methods. This project aims at developing accurate and robust methods to better understand, measure, and predict the combustion properties and safety-related parameters of existing and future HFC refrigerants. The analysis is facilitated by high-fidelity detailed numerical simulations of HFC flames that account for buoyancy and radiation effects. The first objective is the joint evaluation of LBV with the experimental subproject 2. Simulation and analysis will contribute to reduce and quantify potential uncertainties in the evaluation of flame speeds from measurements. Further, reduced kinetic mechanisms of different size and fidelity will be developed to support numerical simulations and asymptotic analysis. The second objective is the development of models describing the effects of Markstein numbers on LBV with a focus on how transport phenomena in the flame are affected by differential diffusion. An existing adjoint-based framework will be extended to higher-order sensitivities to explain the impact of Markstein numbers on thermo-chemistry and transport mechanisms. The third objective is the development of reduced-order models based on machine-learning (ML) combined with structural group analysis to predict LBV and Markstein numbers for existing and future HFC refrigerants. The training data will consist of experimental data from subproject 2 supplemented with data from simulations and will be continuously refined during the project based on the Design-of-Experiment (DoE) approach. The resulting ML-model will be useful for the model-based design and safety assessment of HFC refrigerants.