颗粒停留时间分布是气固流化床设计的一个关键参数。在许多非催化固相转化过程中，研究者期望将颗粒停留时间与颗粒完全转化时间相匹配以实现不同尺寸颗粒的同步转化。随着计算能力的提高，数值模拟已成为研究颗粒停留时间的重要手段。因此本文基于GPU（Graphics Processing Unit）大规模并行的粗粒化CFD-DEM（Computational Fluid Dynamics-Discrete Element Method）方法，对连续操作的内构件流化床和相应尺寸的自由床进行了长时间的颗粒停留时间分布离散模拟，主要研究成果如下：考虑到矿物颗粒的非规则性，将球形度这一参数引入曳力系数模型中，推导出了适合于多分散、非球形颗粒的曳力模型。考察了Ergun-Wen-Yu-Ganser和Di Felice-Ganser曳力模型的预测能力。模拟结果表明：相比Ergun-Wen-Yu-Ganser曳力模型，Di Felice-Ganser 模型预测的全床压降和颗粒停留时间和实验结果吻合得更好。模拟了多分散系统颗粒的停留时间，发现对于同一粒径分布，3种和5种颗粒的平均停留时间（Mean residence time，MRT）比单一尺寸颗粒更长，三粒径和五粒径系统最粗与最细颗粒MRT之比分别约为2和3。流化床放大是化学工程的一个巨大挑战。通过对不同尺寸（长度）流化床的离散模拟发现，当纵向挡板数量不变时，流化床越长，挡板间距越长，每个舱室内气相浓度随径向位置的波动越大。当舱室宽度不变时，流化床越长，挡板数量越多，此时每个舱室内气相浓度随径向位置变化不大。因此可以设置更多间距相同的纵向挡板使得舱室变窄，以达到抑制气泡变大，减小返混，提高气固接触的效果。此外，随着流化床长度的增加：颗粒RTD（Residence time distribution）峰值降低，分布变宽，长拖尾现象更加明显；MRT变长，挡板数量增多时，不同粒径颗粒MRT之间的差异变大。;Residence time distribution of particles is a critical parameter for proper design of gas-solid fluidized beds, especially in many non-catalytic solid conversion processes where it is highly desirable to match the residence time of a particle and its complete conversion time to achieve the synchronized conversion of particles of different sizes. With the improvement of computation capacity, numerical simulation has become an important means to study the residence time of particles in solid processing systems. Therefore, based on the large-scale GPU parallel coarse-grained CFD-DEM method, a long-time discrete simulation of the particle residence time distribution in the continuously operated fluidized bed with baffles and the corresponding free bed was carried out. The main research results and conclusions are as follows:Considering the irregularity of mineral particles, the drag model suitable for polydisperse and non-spherical particle was derived and coupled with the CFD-DEM model by introducing the particle sphericity into the drag coefficient on the basis of the drag model of Ergun, Wen & Yu, and Di Felice. The prediction ability of the Ergun-Wen-Yu-Ganser and Di Felice-Ganser drag model was investigated. The simulation results showed that compared with the Ergun-Wen-Yu-Ganser drag model, the predicted total bed pressure drop and particle residence time by Di Felice-Ganser drag model were in a better agreement with the experimental results. It is found that for the same particle size distribution, the mean residence time of three and five classes of particles is longer than that of one class of particle, and the ratio of MRT of the coarsest and the finest particles in three-particle size and five-particle size systems is about 2 and 3, respectively.Scale-up of fluidized bed is a great challenge in chemical engineering. Through discrete simulation of fluidized beds with different sizes (lengths), it is found that when the number of baffles remained unchanged, the longer the fluidized bed, the longer the baffle spacing, resulting in the greater the fluctuation of gas phase concentration in each chamber with the radial position. When the chamber width is constant, the longer the fluidized bed is, the more baffles there are, and the voidage in each chamber changes little with the radial position. Therefore, more vertical baffles with the same spacing can be set to narrow the chamber, so as to inhibit the growth of bubbles, reduce back-mixing, and improve the gas-solid contact effect. In addition, with the increase of the fluidized bed length, the peak of particle RTD decreases, the distribution becomes wider, the long tailing phenomenon is more obvious, the MRT becomes longer, and when the number of baffles increases, the difference between the MRT of particles with different sizes becomes larger.