Studies On A Distributed Visualization System Oriented To Engineering And Scientific Computation

In engineering and scientific computation, it is common for scientists to develop visual representations of ideas and data and share with each other. The types of such data may be various, the size may be huge and the store locations may be remote. Visualizing this type of data can not resort to stand-alone mode visualization system, while needing to research distributed visualization software tools. Coupled with increasing compute power and graphics capabilities, recent advancements in the computer network have offered a technical support for the development of distributed and collaborative scientific visualization tools.This paper presents a software framework and kernel algorithms of a distributed and collaborative visualization prototype system which is developed for large scale engineering and scientific data. The design aspect of the software framework focuses on three problems: how to manage the collaborative control, how to reestablish a collaborative session and how to construct a collaborative framework of a share visualization process. In the aspect of the kernel algorithms, a parallel cell projection volume rendering algorithm for large scale unstructured scalar data and a 3D vector field flow topology extraction algorithm for vector data are developed.In the first place, three collaborative visualization models are discussed in detail. We then present a framework for distributed and collaborative visualization. This framework contains two parts: One is a distributed visualization subsystem, and the other is a collaborative component named CESC_Remote_DI. By using CESC_Remote_DI we implement collaborative visualization operations between researchers. Collaboration control management is implemented by an adopting master-slave management method. By using compact, intelligent controlling information and recording each operation in a text file, we implement the reestablishment remote collaboration sessions. This software framework provides many advantages besides its simpleness. The bandwidth is small because only script commands are transferred between sites. The system response time is nearly the same as that of the stand-alone mode. Therefore, all remote scientists appear to be seeing the same high resolution, dynamic, and 3D scenes simultaneously. CESC_Remote_DI is normally used along with a desktop video tool if the network bandwidth permits, or along with a normal phone conference if the network bandwidth does not permit. These remote collaboration sessions can be recorded and posted onto the Web for other scientists to playback and modify.The cell projection volume rendering method is a classic algorithm for visualizing large-scale unstructured grid data. However, it is a challenging task to run this method on graphics hardware due to the memory and bandwidth limitations. For overcoming these limitations and providing flexibility in handling various types of cells and meshes, we develop a high accuracy Parallel Software Scanned Cell Projection (PSSCP) algorithm. By constructing an A-Buffer data structure of the image space coordinate of outside faces of the volume grid, the PSSCP can handle non-convex meshes. By using the static round-robin scheme, the PSSCP can obtain a good load balance to achieve better scalability. By using an asynchronous communication strategy on the phase of scan conversion and image composition, the PSSCP obtains a good parallel efficiency.To avoid the limitation of traditional vector field algorithms, such as a cluttered display or the inconvenience to identify controlling parameter, a 3D vector field flow topology extraction algorithm is developed based on critical point analysis for CFD (Computational Fluid Dynamics) datasets. The algorithm can automatically identify the critical points in a vector field. Curves integrated from a small distance around critical points can clearly show the fluid flow features. In many CFD computations, no-slip boundary conditions are imposed on the velocity field. On these boundaries, the velocity is zero. The algorithm analyzes this special case by examining the skin friction field and shows the corresponding fluid flow feature round the wall boundary.