High-performance computing cluster

高性能计算集群

• 批准号: 463921749
• 金额: --
• 依托单位: Technische Universität Berlin
• 依托单位国家: 德国
• 项目类别: Major Research Instrumentation
• 财政年份: 2021
• 资助国家: 德国
• 起止时间: 2020-12-31 至 2021-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/463921749?language=en
• 关键词: performance; computing; cluster

The focus of research at the chair of mechanical process engineering and materials processing is the modeling and simulation of particulate systems using methods such as the Discrete Element Method (DEM) and its coupling to numerical flow simulations (CFD). Regarding the latter, the solution of volume- and Reynolds-averaged Navier-Stokes equations (RANS) or the application of the Lattice Boltzmann method (LBM) as well as of Smoothed Particle Hydrodynamics (SPH) is pursued. In this context the Discrete Element Method is already a computationally intensive process that increases in numerical complexity depending on the number of particles, if e.g. non-spherical particle shape is taken into account or particles in the range down to 1 µm are considered. The flow simulations coupled to the DEM become particularly complex if large domains are considered volume-averaged with a corresponding number of cells, the description of the fluid phase is discrete (e.g. in the SPH) or the Direct Numerical Simulation (DNS) of particle-laden flows is intended (e.g. in the LBM or the SPH), whereby this increasingly is required to develop closure approaches. Regarding the listed aspects, a focus of the research at the chair of mechanical process engineering and materials processing is not only the application of the aforementioned methods to mechanical process engineering, but also in particular the method advancement. In this sense, the provision of local computing resources is of central importance in order to perform direct testing but also application with corresponding parallelization, which cannot be dispensed. With this in mind, an expansion of the existing computing infrastructure is applied for with this proposal, whereby the corresponding utilization can be satisfactory guaranteed (see section 3. - use in research). In the expansion, the data management, which is carried out by means of a RAID system, is particularly taken into account. The requested extension enables the chair of mechanical process engineering and materials processing to address future-oriented numerical questions with regard to particle-laden multiphase flows without resolution restrictions and thus to improve processes, to clarify the fundamentals of particle-laden flows and, in particular, to develop and extend new numerical methods to improve the description of the aforementioned systems.

机械过程工程和材料加工主席的研究重点是颗粒系统的建模和仿真，使用离散元法（DEM）及其与数值流动模拟（CFD）的耦合等方法。对于后者，研究了体积平均和雷诺平均Navier-Stokes方程（RANS）的解或晶格玻尔兹曼方法（LBM）以及光滑粒子流体力学（SPH）的应用。在这种情况下，离散元法已经是一个计算密集的过程，如果考虑非球形颗粒形状或考虑范围低至1 μ m的颗粒，则根据颗粒的数量，数值复杂性会增加。如果考虑具有相应数量的单元的体积平均的大域，流体相的描述是离散的（例如在SPH中），或者旨在对颗粒流进行直接数值模拟（DNS）（例如在LBM或SPH中），则耦合到DEM的流动模拟变得特别复杂，因此越来越需要开发封闭方法。对于上述方面，机械过程工程和材料加工主席的研究重点不仅是上述方法在机械过程工程中的应用，而且特别是方法的进步。从这个意义上说，本地计算资源的提供对于执行直接测试和具有相应并行化的应用程序至关重要，而并行化是不可分割的。考虑到这一点，本建议申请扩大现有的计算基础设施，从而可以令人满意地保证相应的利用（见第3节）。-用于研究)。在扩容过程中，通过RAID系统实现的数据管理尤为重要。所要求的扩展使机械工艺工程和材料加工主席能够在没有分辨率限制的情况下解决关于颗粒负载多相流的面向未来的数值问题，从而改进工艺，澄清颗粒负载流的基本原理，特别是开发和扩展新的数值方法来改进上述系统的描述。