中文摘要：

高质量的网格生成是叶轮机械复杂流动准确预测的关键技术之一。目前在大部分结构网格生成过程中,网格拓扑及局部网格密度均基于经验或粗略的估计,此类网格一般需要根据计算结果进行手动调整。本文发展了基于多块贴体结构化网格的网格自适应技术以达到自动调整局部网格密度并在不严重增加计算量的前提下改进计算精度的目的。本文基于真实物理流动机理,提出了有效的网格加密判据,可以自动判断激波和涡等复杂流动。在此基础上本文发展了可以保证网格质量的贴体网格加密技术。通过一系列算例验证了本文发展的网格自适应技术的有效性和实用性.结果表明：本文方法可以自动捕捉流动细节并根据当地流动尺度自动调整局部网格尺度;同时相对于全局网格加密,本文方法可以达到改进精度和节省计算量之间的平衡。

英文摘要：

High quality mesh is among one of the key technologies required for accurate prediction of complex turbomachinery flow.In most of the mesh generation process,the mesh topology and density are pre-determined with experience or rough estimation and the resulted mesh needs refinement.In this work an efficient Adaptive Mesh Refinement strategy is proposed to automatically adapt the mesh density and improve the prediction without globally refine the mesh.A flow physics based refinement function is proposed which is able to capture the shock and vortex.A mesh refinement method which guarantees the mesh quality is also proposed.The efficiency of current method is verified with a series of test cases.According to the results,current method is shown to automatically capture local flow details.Also compared to the cost-extensive global refinement strategy,with current method the balance between improving accuracy and reducing cost can be achieved.