肺部支气管内气体流动和颗粒沉积的格子Boltzmann方法并行模拟研究

项目名称： 肺部支气管内气体流动和颗粒沉积的格子Boltzmann方法并行模拟研究

项目编号： No.11301549

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 数理科学和化学

项目作者： 刘红娟

作者单位： 中南大学

项目金额： 22万元

中文摘要： 流行病学研究的结果证实了呼吸系统疾病的症状多发生在肺部气管分岔部分，而颗粒物浓度增加与呼吸系统相关疾病的发病率及死亡率具有很强的相关性，因此研究真实人体肺部支气管内气体流动及颗粒物沉积的规律对毒理学、治疗学和人体健康评估等具有重要的指导意义。目前这方面的研究大多基于Weibel对称模型，数值方法主要采用传统的计算流体力学方法，难以得到真实流场的细致结构和瞬时特性。项目拟基于真实人体肺部图像 重建三维模型，提出自适应网格划分策略；在现有格子Boltzmann方法的基础上，重点研究弹性边界的处理方法、流固耦合模型及相应的并行算法设计；实现CUDA 环境下格子Boltzmann 方法对肺部支气管内气体流动的GPU 并行模拟，并利用格子气方法模拟颗粒运动。研究正常和病理状态下肺部支气管内流场以及壁面剪切应力的分布特性，探索颗粒物的沉积规律，推进格子Boltzmann 方法在生物医学临床领域的应用。

中文关键词： 肺部支气管；格子Boltzmann方法；气体流动；颗粒；GPU并行计算

英文摘要： The results of epidemiological studies have confirmed that symptoms of respiratory diseases always occur in the trachea bifurcation part. And the increase of particle pollution highly associated with the increase of occurrence and mortality of Pulmonary diseases.Thus the study of transport and deposition of particulate matter in the human lung bronchial airways has important guiding significance to Toxicology, Therapy and human health assessment. Research in this area is mostly based on the Weibel symmetric model, using the traditional computational fluid dynamics methods, which is difficult to get real flow field structure and transient behavior. Based on the real human lung three-dimensional image reconstruction model,the project will put forward the adaptive meshing strategy。On the basis of the existing lattice Boltzmann methods, we will mainly study the elastic boundary processing method ,fluid-structure coupling model and the corresponding parallel algorithm design;realize the lattice Boltzmann algorithm on GPU using CUDA to simulate the gas flow inside,where the lattice gas automata method is adopted to simulate the movement of the particle.We will analysis the characteristics of the flow field and wall shear stress distribution in the lung bronchus on the normal and pathological state, explore the deposit

英文关键词： hunman lung airways；lattice Boltzmann method；airflow；particle；GPU parallel computing