| description abstract | In this paper, an effective strategy is proposed to realize the smooth visualization of large-scale finite element models on a desktop computer. Based on multicore parallel and graphics processing unit (GPU) computing techniques, the large-scale data of a finite element model and the corresponding graphics data can be handled and rendered effectively. The proposed strategies mainly consist of four parts. First, a parallel surface extraction technology based on the dual connections of elements and nodes is developed to reduce the graphics data. Second, the OpenGL vertex buffer object (VBO) technology is used to improve the rendering efficiency after surface extraction. Third, the element-hiding and cut-surface functions are implemented to facilitate the observation of the interior of the meshes. Finally, the stream/filter architecture, which has the advantages of efficient computation and communication, is introduced to meet the needs of large-scale data processing and various visualization methods. These strategies are developed on the general visualization system SiPESC.Post. Using these strategies, SiPESC.Post implements high-performance display and real-time operation for large-scale finite element models, especially for models containing millions or tens of millions of elements. To demonstrate the superiority and feasibility of the presented strategies, large-scale numerical examples are presented, and the strategies are compared with several commercial finite element software systems and open-source visual postprocessing packages in terms of visualization efficiency. | |
