气固两相流是一个非线性、非平衡系统，具有复杂的非均匀介尺度结构，EMMS范式是解决这类介尺度问题的重要手段，实现虚拟过程工程（Virtual process engineering，简称VPE）是对这类介尺度问题进行模拟的工程应用目标。然而，EMMS范式的成功实施、VPE的最终实现离不开对多尺度模拟方法的使用，其中比较典型的有双流体模型（Two-fluid model，简称TFM）和离散颗粒法（Discrete particle method，简称DPM）。为TFM模拟和DPM模拟选用准确度较高的封闭模型，比如固相应力模型（仅应用于TFM模拟）和气固相间曳力封闭模型（同时应用于TFM模拟和DPM模拟），提高TFM模拟和DPM模拟的准确性，无论对辅助工业实践还是对探究气固两相流的流动规律，都有重要意义。本论文为提高模拟准确性，首先修改针对光滑颗粒的程序以使其可以模拟更为真实的粗糙颗粒气固两相流，然后使用这套程序在Geldart D颗粒浓相鼓泡床的背景下以尺度相对最小的DNS数据为参照标准，对尺度相对较大的DPM模拟、TFM模拟所使用的封闭模型，尤其是对会给模拟结果带来很大影响的气固相间曳力系数关联式展开了一系列的评估与探讨。本论文的具体内容如下：论文第一章总结了粗糙颗粒气固两相流数值模拟的研究进展，从表达形式与来源方式两个角度概述了不同气固相间曳力系数关联式，介绍了模型对比在评估固相应力模型有效性、开发新封闭模型等方面所起的作用。第二章在实现粗糙颗粒TFM模拟程序后，利用模型对比的方式系统详细地对比了粗糙颗粒鼓泡床模拟结果和光滑颗粒鼓泡床模拟结果，说明了实现粗糙颗粒气固两相流数值模拟的必要性以及所改得的粗糙颗粒TFM模拟程序的有效性。第三章以已有工作中的DNS数据和实验数据为标准，详细对比了相同工况下使用十四种不同气固相间曳力系数关联式的DPM模拟结果和TFM模拟结果。研究发现这十四种气固相间曳力系数关联式均无法准确复现DNS的模拟结果，同时也发现TFM模拟所使用固相应力模型在某些情况下会导致模拟结果失真。第四章以已有工作中的DNS数据和实验数据为标准，采用模型对比的方式系统考察了考虑状态变量涨落（颗粒温度或固相体积分数涨落）的气固相间曳力系数关联式对提升DPM模拟和TFM模拟结果准确性的作用，发现这些模型只能给计算出的曳力带来很小的提升，并不足以消除DPM模拟结果、TFM模拟结果和DNS数据之间的差距。最后，总结了本研究所获得的主要成果，给出了获取、使用更加准确的气固相间曳力系数关联式和固相应力模型以及通过模拟多粒径分布颗粒体系、计算流场中热量传递、模拟工业中某实际Geldart D颗粒等操作来使当前工作更具针对性和指导意义的意见和建议。;Gas-solid two phase flows are non-linear and non-equilibrium systems with complicated non-uniform mesoscale structures, EMMS paradigm is an important way to solve the above-mentioned problem while realization of virtual process engineering (VPE) is the ultimate engineering goal. Multiscale simulation methods, such as two fluid model (TFM) and discrete particle method (DPM), are essential to both implementation of EMMS paradigm and realization of VPE. Thus, selecting appropriate closure models for TFM and DPM simulations and improving accuracy of the above two simulation methods play a critical role in supporting engineering practice and exploring the nature of gas-solid two-phase flow.In order to improve the simulation accuracy, an in-house code with the ability to simulate gas-solid two phase flow of rough spheres was realized by modifying the code which was designed for smooth spheres. The code was then used to assess the closure models that should be used in DPM and TFM simulations which were at a relatively larger scale with applying the DNS data which was at the relatively smallest scale as the benchmark, especially the closure models of interphase drag force which would exert important effects on TFM and DPM simulation results. The contents of the dissertation are as follows:Firstly, a literature review of the difference between the gas-solid two-phase flow of rough spheres and of smooth spheres was done; an introduction of interphase drag coefficient correlations from the aspect of different classes of relations and different origins was then delivered; additionally, the application of model comparison into different works such as assessing and constructing clouse models would be described.In Chapter 2, a comprehensive comparison between simulation results of gas solid two phase flow of rough spheres and of smooth spheres were carried out after extending the in-house code to simulate the hydrodynamics of rough spheres. This part of work set forth the necessity of simulating the gas-solid two-phase flow of rough spheres, while showed the effectiveness of the in-house code.In Chapter 3, extensive TFM and DPM simulations were carried out to study the hydrodynamics of gas-solid flows of monodisperse, rough spheres in a dense fluidized bed using fourteen drag coefficient correlations available in literature, and the simulation results were compared to the direct numerical simulation (DNS) and experimental data from an existing publication. It was found that the fourteen different drag coefficient correlations could not accurately reproduce the results of DNS and the used particle stress model for TFM simulation may result in abnormal fluidization phenomena under certain conditions.In Chapter 4, extensive TFM and DPM simulations were performed to assess how much improvement could be achieved when the drag coefficient correlations that considered the effect of granular temperature and solid concentration fluctuation were used, using the experimental and DNS data from existing publications as the benchmark. It is found that all currently available drag correlations that included the fluctuations of state variables could only bring a minor improvement and they were insufficient to fill in the gap between DPM/TFM and DNS results. At last, the main conclusions were drawn, while advices of acquiring and applying more accurate interphase drag correlations and particle stress models, making the work more instructive by simulating polydisperse system of particles or calculating heat transfer or simulating Geldart D particles used in industry were given.