A numerical model for translational and rotational momentum transfer of soft deformable micro particles in dilute two-phase flows

稀两相流中软可变形微粒的平移和旋转动量传递数值模型

• 批准号: 510050592
• 负责人: Professor Dr.-Ing. Paul Steinmann
• 金额: --
• 依托单位: Lehrstuhl für Technische Mechanik (LTM)
• 依托单位国家: 德国
• 项目类别: Research Grants
• 资助国家: 德国
• 项目状态: 未结题
• 来源: https://gepris.dfg.de/gepris/projekt/510050592?language=en
• 关键词: numerical; model; translational; rotational; momentum

The main objective of the project is to extend the numerical model developed in the previous project (rigid micro particles in dilute dispersed two-phase flow using the Lagrange-Euler approach) to a numerical model for translational and rotational momentum transfer of soft deformable micro particles. The first part is devoted to the development of a direct numerical simulation (DNS) tool capable of locally computing the flow-induced tractions and the resulting force and torque on a finite-sized soft deformable micro particle along with the particle deformation. Furthermore, the coupling between a CFD solver for the global-scale flow and Lagrangian particle tracking based on the novel soft deformable point-particle approach is established. This combination is achieved by an online approach based on the DNS tool, as well as by an offline approach based on a machine learned (ML) surrogate model. In the ML framework, a large DNS database of flow-induced forces and torques is built for different particle shapes, orientations, and flow conditions, allowing the trained ML model to be applied in computationally challenging flow cases with hundreds of thousands of soft deformable micro particles. Finally, the derived model is applied to a technologically relevant example of dilute multiphase flow with distributed lenticular vesicles.

该项目的主要目标是将以前项目中开发的数值模型（使用拉格朗日-欧拉方法的稀分散两相流中的刚性微粒）扩展到软变形微粒的平移和旋转动量传递的数值模型。第一部分是致力于发展的直接数值模拟（DNS）工具，能够本地计算的流动引起的牵引力和所产生的力和扭矩的有限尺寸的软变形微粒沿着颗粒变形。在此基础上，建立了基于新的软变形点粒子方法的拉格朗日粒子跟踪与全局尺度流场CFD求解器之间的耦合。这种组合是通过基于DNS工具的在线方法以及基于机器学习（ML）代理模型的离线方法实现的。在ML框架中，针对不同的颗粒形状、方向和流动条件构建了一个大型的流致力和扭矩DNS数据库，从而使经过训练的ML模型能够应用于具有数十万个软变形微粒的具有计算挑战性的流动情况。最后，推导出的模型被应用到一个技术相关的例子，稀多相流与分布的透镜状囊泡。