The Research Of Three-Dimension Visualization Technologies In Nuclear Simulation And Implementation Of Prototype System

As an effective instrument used for analyzing, digging and processing tremendous dataset, Visualization in Science Computing is playing more and more important role in application domain of high-performance nuclear simulation computing. The dataset in this domain which is usually large-scale, high-dimension and complicated, bring forward higher challenge to the traditional technology and method in Visualization Modeling, Real Time Interaction and Parallel Rendering.Based on the research of traditional 3D visualization modeling technology, region picking algorithm and visualization software architecture, we have analyzed the requirements in nuclear simulation visualization, aiming at the research and realization of modeling and real time interaction technology, also at the research of flexible parallel rendering on high performance computing environment. The contributions and relevant work in the paper are as follows:Firstly, a compound modeling and description algorithm of complicated solid model is presented for high efficiency. We have analyzed some existed algorithms. To solve problems of large memory requirement and low rendering precision, the new algorithm with CSG description interface adopt optimize cell representation method for solid model record. It is indicated by experiment that the new algorithm has capabilities of exact model description, less memory requirement and higher rendering accuracy.Secondly, a solid-inside data picking method based on dynamic section is presented. We have analyzed some existed region picking algorithms, which have some problem as low efficiency. First we made the function of picking target point freely come true with dynamic section method. Then data picking as an advanced interact ional method is carried out depend on the CSG expression characteristic. Compared to the traditional ones, new method is excellent in target point orientation, interaction capability and picking efficiency.Thirdly, the architecture is presented to enhance the visualization efficiency in parallel mode. Based on the research of the architecture in some existed visualization systems, we present new architecture which can be used to configure resources freely. And a new screen region distribution strategy is presented to enhance the stability of static distribution algorithm.Finally, we have developed a prototype system YHDVS based on analyses and realization of algorithms above. YHDVS has been applied to the MCNP4C benchmark data successfully.