论文部分内容阅读
在基于蒙特卡罗粒子输运方法的反应堆模拟中,如裂变堆、聚变裂变混合堆等,达到可接受的统计误差需要大量的计算时间,这已成为蒙特卡罗方法的挑战问题之一,需通过并行计算技术解决。为解决现有方法中通信死锁的问题并保证负载均衡性,设计了基于双向遍历的临界计算并行算法。该方法基于超级蒙特卡罗核计算仿真软件系统SuperMC进行实现,以池式钠冷快堆BN600基准模型进行验证,并与MCNP进行对比。测试结果表明,串行和并行计算结果一致,且SuperMC并行效率高于MCNP。
In reactor modeling based on Monte Carlo particle transport methods, such as fission reactors, fusion fission hybrid reactors, achieving acceptable statistical errors requires a large amount of computation time, which has become one of the challenges of the Monte Carlo method, requiring Solved by parallel computing. In order to solve the problem of deadlock of communication in the existing methods and ensure the load balancing, a parallel algorithm of critical computing based on bidirectional traversal is designed. The method is based on the SuperMC Monte Carlo simulation software SuperMC, and is validated by a pool sodium sodium fast reactor BN600 benchmark model and compared with MCNP. The test results show that the serial and parallel computing results are consistent, and SuperMC parallel efficiency is higher than MCNP.