Direct Numerical Simulation and Advanced Modelling of Turbulent Flame Kernels for High-Efficiency Low-Emission Spark Ignition Engine

1导言公路运输导致碳排放和气候危机。向电动汽车的过渡将需要中间解决方案，特别是重型货车。开发高效率、低排放的新型发动机成本高、耗时长，因此需要快速、准确的计算模型。在实践中，快速工业软件牺牲了许多详细的小规模物理，而这些细节必须被工程模型所取代。开发这样的模型是具有挑战性的，并且依赖于高保真的模拟，其中包括基于基本原理的基本物理；这在工业环境中是不可能完成的。2概述Ricardo是以发动机效率方面的专业知识为基础的，它仍然是提高效率和减少汽油、柴油、新兴生物燃料和天然气发动机排放的发动机技术的先驱。目前应用最先进的火花点火发动机模型经常依赖于湍流火焰速度的经验闭合。这受到与发动机运行参数范围内的有效性和适用性以及模型中固有的调谐系数的影响相关的大量不确定性的影响。目前，现有的实验数据(大多数集中在平面火焰上)不能提供足够的信息来减少这种不确定性。本项目将使用典型发动机条件下的火核的高保真模拟，以提高我们的基本理解，增强准确性和稳健性，并减少与湍流火焰速度预测相关的不确定性。3研究问题本项目的目的是提高湍流火焰模型的预测能力，用于工程应用。该项目由Ricardo UK赞助，其软件将为模型开发和评估提供试验台。目标：*使用PeleLM创建一个由湍流火焰核和火焰壁相互作用组成的高细节和高精度的直接数值模拟数据库。*利用该数据库来加深对湍流燃烧行为以及内燃机设计如何影响火焰行为从而影响发动机性能的基本了解。*使用该数据库，比较Vectis使用的湍流火焰模型的性能和准确性，并在必要时开发和实施修改。4项目计划将建立火焰核心高保真模拟数据库，涵盖与SI发动机相关的一系列条件；Ricardo将提供代表天然气和汽油发动机的流动和湍流场条件，基于可用于DNS初始化的调谐RANS模拟。通过对湍流火焰响应的详细分析和对湍流动能和能量耗散率输运方程中各项的评估，将加深对基本知识的理解。这样的分析是一个独特的优势，只有通过分析域名系统才可能实现，并将使开发包含积分长度标度、热释放和曲率等影响的湍流火焰模型的重要术语得以识别。由此产生的湍流火焰模型可以在Vectis CFD套件中实施，并在真实发动机模拟中进行评估，作为学生在Ricardo UK的时间的一部分。域名系统的计算将使用PeleLM进行，这是由劳伦斯伯克利国家实验室计算科学和工程中心开发的。规模越来越大的模拟将分阶段进行，从一个高规格的工作站开始，通过纽卡斯尔的高性能计算设施“火箭”，直到国家超级计算机“阿彻”(UKCTRF的计算时间应用程序)。

1 IntroductionRoad transportation contributes to carbon emissions and the climate emergency. The transition to electrified vehicles will require intermediate solutions, especially for heavy goods vehicles. Developing new engines with high efficiency and low emissions is expensive and time intensive, hence fast and accurate computational modelling is necessary. In practise, fast industrial software sacrifices much of the detailed small-scale physics, which has to be replaced by engineering models. Developing such models is challenging and relies upon high-fidelity simulations that include fundamental physics based on first principles; this cannot be done in an industrial setting.2 OverviewRicardo was founded on expertise in engine efficiency, and remains a pioneer in engine technology for improving efficiency and reducing emissions from gasoline, diesel, emerging bio-fuel and gas engines. Current state-of-the-art applied spark-ignition engine modelling frequently relies upon empirical closures for turbulent flame speed. This suffers from substantial uncertainties associated with validity and applicability across the range of the engine operation parameters and the effect of the tuning coefficients inherent in the model. At the moment, available experimental data (with the majority focusing on planar flames) does not provide sufficient information to reduce this uncertainty. This project will use high-fidelity simulation of flame kernels under typical engine conditions to improve our fundamental understanding, enhance accuracy and robustness, and to reduce uncertainty associated with turbulent flame speed prediction.3 Research QuestionThe aim of this project is to improve predictive capabilities of turbulent flame models for engineering applications. The project is sponsored by Ricardo UK, whose software will provide the test bed for model development and evaluation.Objectives:* Create a high detail and high accuracy direct numerical simulation database using PeleLM consisting of turbulent flame kernels and flame wall interactions.* Exploit the database to develop fundamental understanding of the behaviour of turbulent combustion and how the design of an IC engine can affect the flame behaviour, and therefore engine performance.* Using the database, compare the performance and accuracy of the turbulent-flame models used by Vectis, and develop and implement modifications as necessary.4 Project PlanA database of flame kernel high-fidelity simulations will be constructed that span a range of conditions relevant to SI engines; Ricardo will supply flow and turbulent field conditions representative for natural gas and gasoline engines, based on tuned RANS simulations that can be used for DNS initialisation. Fundamental understanding will be developed through detailed analysis of turbulent flame response and assessment of individual terms in the transport equations for turbulent kinetic energy and energy dissipation rate. Such analysis is a distinct advantage only possible through analysis of DNS, and will enable identification of important terms for developing turbulent flame models that incorporate effects such as integral length scale, heat release and curvature.The resulting turbulent flame models can then be implemented in the VECTIS CFD suite and evaluated in real engine simulations, as a part of the student's time at Ricardo UK. The DNS calculations will be carried out using PeleLM, which was developed at the Center for Computational Sciences and Engineering at the Lawrence Berkeley National Laboratory. Simulations increasing in size will be carried out in stages, starting from a high-specification workstation, through Newcastle's HPC facility "Rocket", up to the national supercomputer "Archer" (compute time application to UKCTRF).