Data-driven modelling of thermally coupled fluids and hydraulic efficiency optimisation in pipe flows

该项目与英国原子能机构库勒姆科学中心合作，目的是最大限度地提高聚变反应堆托卡马克设施所使用的冷却管道系统的热能输出。托卡马克是一台以环形形式构思的实验机器，旨在利用聚变能量，即轻元素氘和氚碰撞产生的能量。通过施加尖端超导线圈产生的高强度定制磁场，确保这些氢同位素始终以极端温度等离子体的形式被限制在内部真空容器中，从而促进了聚变反应。在托卡马克内部，一个称为包层的结构组件屏蔽了真空容器，使其免受聚变反应产生的高能中子的影响，从而保护托卡马克的其余组件免受热降解。此外，包层还负责a)通过模块部件将热功率输出传输到主动冷却的管道系统，以及b)为聚变反应堆的自我维持而滋生氚。行动总的动机是为碳中性能源做出贡献。具体地说，1.聚变反应堆中面向等离子体的结构部件，如包层组件的设计和热管理，对于防止磁约束聚变中昂贵和不可修复的损害至关重要。聚变发电厂的净发电量可以通过最小化冷却系统中的热水力损失(即粘性、湍流、传导)来增加，这是此类能源在经济上取得成功的关键方面。3.新的表面摩擦减阻技术，如鲨鱼皮、肋骨或各向异性多孔壁面介质，以其降低压降(即能量损失)从而最大化水力效率而闻名，但它们对传热的影响仍未量化，特别是在聚变反应堆环境中。AIM鉴于上述三个因素，本博士研究项目的目的是设计和实现一个数据驱动的计算模型，用于分析、评估和优化受限聚变环境下冷却管道系统中表面减阻技术的热水力效率。目的上述目标将通过以下目标得以实现：1.设计和实现一个用于分析减阻技术热工水力效率的高保真计算模型。重点将放在获得可靠结果所需的必要组件的评估上，包括最佳湍流模型、探索不可压缩到可压缩的流动状态、共轭换热通量条件。2.对照现有的实验、半分析和数值结果，对计算模型进行验证和基准比较。通过开发新的深度学习范式，开发快速且计算高效的代理降阶数据驱动计算模型。目标是最大限度地提高计算速度，同时保持高保真计算模型的大部分精度。4.现有的(可能是新的)用于减少壁面摩擦和增加换热的拓扑结构将被量化，并使用新的替代模型进行比较。在使用新的表面摩擦技术时，洞察潜在的热液机制，使热传递表面的新创新成为可能。作为最终目标，如果时间允许，将进一步利用深度学习范例来驱动自适应模拟，以便为偏滤器或包层冷却通道内的冷却剂管道找到最佳的表面类型几何形状。

THE CONTEXTThis project is in collaboration with UKAEA Culham Science Centre, with the aim of maximizing heat energy output from the cooling pipe systems employed by the fusion reactor Tokamak facility. The Tokamak is an experimental machine conceived in the form of a toroid and designed to harness the energy of fusion, that is, the energy resulting from the collision of light elements deuterium and tritium. The fusion reaction is facilitated by ensuring that these hydrogen isotopes are always confined in the form of an extreme temperature plasma, contained within an inner vacuum vessel, via the application of high-strength tailor-made magnetic fields generated through cutting-edge superconducting coils.Within the Tokamak, a structural component, named the blanket, shields the vacuum vessel from the high-energy neutrons produced during the fusion reaction and thus protects the rest of the components of the tokamak from thermal degradation. In addition, the blanket is responsible for a) the transfer of thermal power output across its module components into an actively cooled pipe system and b) the breeding of tritium for the self-sustainability of the fusion reactor. THE MOTIVATIONThe general motivation is contributing towards carbon neutral energy. Specifically, 1. The design and thermal management of plasma facing structural components in a fusion reactor, such as blanket modules, is critical in order to prevent costly and irreparable damage in magnetically confined fusion.2. Net power output of a fusion power plant can be increased through the minimisation of thermo-hydraulic losses (i.e. viscous, turbulence, conduction) in the cooling systems, a key aspect for the economic success of this type of energy. 3. Novel skin friction reducing technologies, such as shark skin, riblets or anisotropic porous wall media, are known for their effectiveness to reduce pressure drops (i.e. energy losses), thus maximising hydraulic efficiency, but their impact on heat transfer remains unquantified, particularly in the case of a fusion reactor environment.THE AIMGiven the above three factors, the aim of this PhD research project is the design and implementation of a data-driven computational model for the analysis, assessment and optimisation of the thermo-hydraulic efficiency of skin friction reducing technologies in cooling pipe systems in the context of confined fusion. THE OBJECTIVESThe above aim will be crystallised through the following objectives:1. Design and implementation of a high-fidelity computational model for the analysis of the thermo-hydraulic efficiency of skin friction reducing technologies. Emphasis will be placed in the appraisal of the necessary components required to obtain credible results, include an optimal turbulence model, explore the incompressible to compressible flow regime, conjugate heat transfer flux conditions. 2. Validation and benchmarking of the computational model versus available experimental, semi-analytical and numerical results.3. Development of a fast and computationally efficient surrogate reduced order data-driven computational model, through the exploitation of a new deep learning paradigm. The goal will be to maximise computation speed whilst preserving much of the accuracy of the high-fidelity computational model. 4. Existing (and possibly novel) topologies for reducing skin friction and increasing heat transfer will be quantified and compared using the new surrogate model.5. Gain an insight into the underlying thermo-hydraulic mechanisms when using new skin friction technologies, enabling the possibility for new innovations in heat transfer surfaces.6. As a final objective, if time allows, the use of the deep learning paradigm will be further exploited to drive adaptive simulations in order to find the best surface type geometry for a coolant pipe within a divertor or blanket cooling channel.