基于混合碰撞模型与机器学习的带电粒子输运高效数值方法
An efficient numerical method for charged particle transport based on hybrid collision model and machine learning
AI总结：
本文针对惯性约束聚变中带电粒子输运的计算瓶颈，提出混合碰撞模型与机器学习神经网络模型，并集成到RDMG和LARED-S程序中，实现高效准确的蒙特卡罗模拟。
摘要：
带电粒子输运是惯性约束聚变等离子体燃烧过程中的重要能量输运方式。一方面，热斑内部的带电粒子具有强非平衡效应，因此需要求解玻尔兹曼输运方程以准确模拟带电粒子的能量输运过程。另一方面，带电粒子输运具有高碰撞频率和复杂阻止本领的特点，传统蒙特卡罗算法的计算量在现有计算条件下难以承受。针对大库仑势碰撞截面带来的计算瓶颈，我们开发了一种混合碰撞模型，在保持碰撞过程二阶精度的同时大幅降低了计算成本。为解决复杂阻止本领模型带来的计算瓶颈，我们开发了一种基于机器学习的神经网络模型，实现了不同阻止本领的形式统一和高效计算。基于该计算方法，我们开发了RDMG程序和LARED-S程序的带电粒子输运蒙特卡罗功能模块，并将其应用于惯性约束聚变临界靶性能研究，展示了良好的计算效率和精度。
英文摘要：
Charged particle transport is an important energy transport mode in the combustion process of inertial confinement fusion plasma. On the one hand, charged particles inside the hot spot have a strong non-equilibrium effect, so it is necessary to solve the Boltzmann transport equation to simulate the energy transport process of charged particles accurately. On the other hand, charged particle transport has the characteristics of high collision frequency and complex blocking power, so the calculation amount of the traditional Monte Carlo algorithm is difficult to bear under the existing calculation conditions. Aiming at the computational bottleneck caused by the large Coulomb potential collision cross-section, we developed a hybrid collision model which greatly reduced the computational cost while maintaining the second-order accuracy of the collision process. In order to solve the computational bottleneck caused by the complex blocking power model, we developed a neural network model based on machine learning to achieve formal unity and efficient calculation of different blocking power. Based on the calculation method, we developed the charged particle transport MC function modules of the RDMG program and LARED-S program and applied them to the study of critical target performance of inertial confinement fusion, which showed good computational efficiency and accuracy.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
