基于CAD邻居列表和包围盒的蒙特卡罗粒子输运几何跟踪加速方法研究

摘要

Geometry navigation plays the most fundamental role in Monte Carlo particle transport simulation. It’s mainly responsible for locating a particle inside which geometry volume it is and computing the distance to the volume boundary along the certain particle traj ectory during each particle history. Geometry navigation directly affects the run-time performance of the Monte Carlo particle transport simulation, especially for complicated fusion reactor models. Thus, two CAD-based geometry acceleration algorithms,the neighbor search and the bounding box,are presented for improving geometry navigation performance. The algorithms have been implemented in the Super Monte Carlo Simulation Program for Nuclear and Radiation Process (SuperMC). The fusion reactors of FDS-Ⅱ and ITER benchmark models have been tested to highlight the efficiency gains that can be achieved by using the acceleration algorithms. Testing results showed that efficiency of Monte Carlo simulation can be considerably enhanced by 50% to 60% with the acceleration algorithms.%几何跟踪主要进行蒙特卡罗粒子输运计算中粒子位置和径迹长度的计算，它是蒙特卡罗粒子输运计算的关键技术之一。由于聚变堆几何结构极其复杂，使得几何跟踪在整个蒙特卡罗粒子输运计算中占据30％~80％的计算时间，因此几何跟踪方法的效率是决定聚变堆蒙特卡罗粒子输运计算效率的重要因素之一。本文提出了基于CAD的邻居列表和包围盒加速方法，并基于 FDS 团队自主研发的超级蒙特卡罗核计算仿真软件系统 SuperMC进行实现。利用聚变堆 FDS-Ⅱ和 ITER模型对本文方法进行了数值验证，测试结果表明本文方法不影响计算结果，并能使程序计算效率提高50％~60％，证明了本文方法的正确性和有效性。