颗粒流体两相流的直接数值模拟与非均匀结构分析
文献类型：学位论文
| 作者 | 周国峰 |
| 学位类别 | 博士 |
| 答辩日期 | 2014-06 |
| 授予单位 | 中国科学院研究生院 |
| 导师 | 葛蔚 |
| 关键词 | 气固两相流 介尺度结构 直接数值模拟 曳力关联式 团聚物 |
| 其他题名 | Direct Numerical Simulation of Particle-Fluid Flows and the Analysis of Heterogeneous Structures |
| 学位专业 | 化学工程 |
| 中文摘要 | 颗粒流体两相流是典型的复杂系统，呈现复杂的时空多尺度结构，特别是团聚物和气泡等介尺度结构，对理论和实验研究都构成极大的挑战。计算流体力学（computational fluid dynamics, CFD）方法以其独特的优势，正日益成为研究颗粒流体两相流的重要手段。CFD研究颗粒流体两相流的方法主要分为三大类，分别是：流体、颗粒两相都基于连续介质假设的双流体模型（two-fluid model, TFM）；流体相沿用基于连续介质假设的Navier-Stokes方程，但颗粒相采用拉格朗日方法逐个跟踪的离散颗粒模型（discrete particle model, DPM），也叫颗粒轨道模型；以及不借助任何相间作用经验关联式，直接求解Navier-Stokes方程的直接数值模拟（direct numerical simulation, DNS）。实际应用中，由于传统的TFM和DPM的计算网格远大于介尺度结构的特征尺寸，并且尚未考虑网格内介尺度结构的影响，因此普遍存在较大的误差，严重限制了其应用。而关于细网格模拟及基于其结果的亚格子模型能否有效降低误差还存在争论。 多尺度结构特征在气固两相流中表现得更为明显，其介尺度结构的研究是目前颗粒流体两相流的热点和难点能量最小多尺度（Energy-Minimization Multi-Scale, EMMS）模型把复杂的气固两相系统分解为稀相、密相和相间相，再由控制机制的协调提出了相应的稳定性条件，最终得到了考虑介尺度结构影响的EMMS曳力。将该曳力与TFM或DPM耦合可有效地提高它们的预测精度。 但是，无论是对细网格模拟有效性的考察还是对EMMS模型内在机理的探索与求证，都需要自下而上地从颗粒尺度开始研究介尺度结构的形成与演化。在此，DNS将发挥不可替代的作用。本论文基于具有优越并行性和扩展性的格子Boltzmann方法（lattice Boltzmann method, LBM）和浸入移动边界方法（immersed moving boundary, IMB），并借助具有多尺度体系结构的新型超级计算机系统，有效突破了DNS计算量巨大的瓶颈，使得介尺度结构DNS研究成为可能。 论文第一章介绍了颗粒流体两相流的三类数值模拟方法，重点阐述了DNS及其在介尺度结构方面的研究进展。论文第二章详细地论述了本论文采用的DNS算法，实现了针对颗粒的时驱硬球模型（time-driven hard sphere model，TDHS）和离散单元模型（discrete element method, DEM），并改进了原始的IMB方法。改进后的IMB方法不仅提高了计算的精度，还减少了计算流固耦合参数的复杂度。考虑到气固两相流DNS巨大的计算量，本章还在多尺度计算系统中实现了上述DNS算法基于通用图形处理器（general propose graphics processing unit, GPGPU）的并行计算。论文第三章用三个典型算例集中论证了第二章提出的DNS算法的有效性和合理性。 论文第四章在静态近似假设下，考虑了三种简单非均匀结构对曳力的影响，验证了当非均匀结构存在时，只关联固相体积分率和颗粒雷诺数的传统曳力模型存在不足，有必要显式地考虑非均匀结构的影响。论文第五章通过对气固两相流的动态大规模DNS，细致地考察了模拟区域大小、外力场强度、颗粒碰撞恢复系数和气体黏度等对介尺度非均匀结构形成的影响，分析了静态近似假设研究气固两相流动力学行为的适用性。最后，论文讨论了气固两相流中介尺度非均匀结构的形成过程，并重点分析了团聚物的聚并与破碎机制。 论文最后总结了所取得的主要成果，展望了基于DNS的气固两相流研究的前景。 |
| 英文摘要 | The particle-fluid flows featuring prevalent spatio-temporal multi-scale structures, especially mesoscale structures, such as clusters and bubbles, are typical complex systems, which bring great challenges to either experimental or theoretical study. Computational fluid dynamics (CFD) is becoming a promising tool in studying particle-fluid flows for its exclusive advantages. The CFD methods for particle-fluid flows can be divided into three categories roughly, the two-fluid model (TFM) where both the fluid and the solid phases are assumed to interpenetrate continuously; the discrete particle model (DPM) where the fluid phase is treated the same as in TFM, but each particle is tracked with a Lagrangian method; and the direct numerical simulation (DNS) which directly solves the Navier-Stokes equations with the no-slip boundary condition satisfied at the surfaces of all particles so that any empirical models for interphase interaction is no longer needed. In practice, since the grid size of TFM or DPM is usually much larger than the characteristic size of meso-scale structures, and the effect of sub-grid structures is not accounted for yet, the predictability of TFM or DMP is not good enough in general, which limits its applicability. Meanwhile, because of the inherent defect of the assumption of uniformity within the grid in attached to TFM or DPM, the sub-grid models filtered from fine-grid simulation are still in debate. Gas-solid flows are especially characterized by complex multi-scale structures, and the study of the mesoscale structures in gas-solid flows by far is the hot and difficult for particle-fluid flows. The Energy Minimization Multi-Scale (EMMS) model accounts for the effects of meso-scale structures by decomposing the complex particle-fluid system into dilute phase, dense phase and inter-phase between them. Conservation equations are established for each phases, and these equations are closed with stability conditions based on the compromising of different dominate mechanisms, which finally results in the EMMS drag. Applying the EMMS drag can significantly improve the accuracy of TFM and DPM. However, either the validity of the fine-grid simulation or the exploration of the EMMS model requires the study of the formation and evolution of meso-scale structures from the particle scale based in a bottom-to-up method, where DNS will play an exclusive role. Because of the excellent parallelism and scalability inherent of the coupled lattice Boltzmann method (LBM)-immersed moving boundary (IMB), it is the nature choice of for this study. Given the huge computation cost of DNS, a well-established multi-scale supercomputing system is employed, which makes the DNS of meso-scale structures feasible. Chapter 1 introduces the three categories of CFD methods for particle-fluid flows simulation, with the emphasis on DNS and its application in studying meso-scale structures. Chapter 2 elaborates the DNS algorithm employed in this thesis. LBM is adopted to solve the fluid phase considering its inherent parallelism. The time-driven hard sphere model (TDHS) and the discrete element method (DEM) are implemented for the solving discrete particle phase. The immersed moving boundary (IMB) is employed for the fluid-particle interaction. The study improves the original IMB, which not only improves the accuracy, but also reduces its complexity in determining the interphase interaction parameter. Considering the huge computational cost of DNS for particle-fluid flows, parallel implementation is carried out for this DNS algorithm based on general propose graphics processing units (GPGPU) of the multi-scale supercomputing system. Chapter 3 is devoted to demonstrate the validity and reasonability of the proposed DNS algorithm with three typical cases. With the assumption of static state approximating dynamic system of gas-solid flows, Chapter 4 analyzes three simple fixed heterogeneous structures in detail to understand their effect on t |
| 语种 | 中文 |
| 公开日期 | 2015-07-08 |
| 源URL | [http://ir.ipe.ac.cn/handle/122111/15524]  |
| 专题 | 过程工程研究所_研究所（批量导入） |
| 推荐引用方式 GB/T 7714 | 周国峰. 颗粒流体两相流的直接数值模拟与非均匀结构分析[D]. 中国科学院研究生院. 2014. |
入库方式： OAI收割
来源：过程工程研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。