GPU-based Parallel Solution Algorithm For The Laser Radiation Symmetry Analysis On Capsule In Indirectly Driven

In the indirectly driven approach to inertial confinement fusion, controlling the radiationdrive symmetry of capsule is a crucial task for achieving controlled nuclear fusion ignition.Using numerical methods to simulate capulse implosion physical processes can evaluate thequality of the radiation symmetry, and then optimize target design. This method can improvethe experiment efficiency. However, the complexity of the simulation model’s structure andintensive numerical computation is a serious hamper for simulation analysis. In this thesis, thesystematic analysis and research on how to accurately and efficiently calculate the capsuleradiation driven symmetry.Calculation of capsule surface flux distribution is the premise of the symmetry analysisand the beam-hohlraum-capsule’s shape parameter optimization. Presently, it mainly uses theview factor to solve this problem. The analytic view factor method is fast, but have lowapplicability. The discrete view factor is applicable to arbitrarily shaped ones. Based on theequivalent model of two-region uniformly, discreting the three-dimensional simulation modeland analysing the capsule receiving radiation energy calculation process. Against self-occlusion, using Quad-tree structure hierarchical to construct hierarchical bounding box.Itmakes quickly to exclude the obstucted cells.Discretization of grid cells has led to sharp increase of the numerical computation, andthe time is greatly extended. Parallel technology is an important method to achieve highperformance computation. The CUDA architecture provides a way to use CPU/GPUcollaborative heterogeneous computing model, which is a cost-effective means of parallelcomputing.An efficient parallel algorithm based on CUDA with occlusion culling is proposed.Based on the parallel analysis of3D discrete view factor algorithm, the calculation processesof view factor and occlusion detection are mapped to GPU threads, and algorithm with CUDAis implemented.The accuracy and efficiency of the algorithm is verified by experiments on differentsimulation models. The validity is proved by normalization of view factor and comparison of analytical solutions. Compared with result on CPU, while maintaining the precision,accelerated algorithm obtain a satisfactory result by CUDA. Finally, the frame and functionalmodules of the software for ICF simulation analysis are introduced. Then, a typical projectcase of optimization for the ICF device is given out.