NEPTUNE:并行三维全电磁粒子模拟软件

为求解具有复杂几何的高功率微波电磁场问题,本文研制了一个三维全电磁粒子并行软件NEPTUNE。本文介绍了该并行软件的基本结构和采用的一些并行算法。目前,该软件已经成功模拟了多种高功率源器件,并可扩展到数千台处理器核上运行。     

基于Cart3D的全机数值模拟及并行计算

利用CFD商用软件Cart3D对亚声速飞行飞机的三维绕流流场进行了数值模拟以及并行计算,得到了飞机附近的流场,实现了此软件在高性能并行计算机上的并行;通过对比不同商用软件的计算结果,验证了用Cart3D软件进行数值模拟的有效性,为大规模科学工程计算提供了技术参照。     

基于集成众核的3D蒙特卡罗半导体器件模拟器

3D蒙特卡罗器件模拟计算量大,计算量随网格与粒子数增加而急剧增加。通过分析3D蒙卡模拟加速热点和进一步可并行性,研究有效电势方法的集成众核并行方案;研究粒子自由飞行、统计模拟信息、计算表面粗糙散射等热点并行方案,最终实现基于CPU/MIC的三级并行3D蒙特卡罗器件模拟软件。实验结果显示,三级并行比单级并行获得更好的性能;当提高模拟精度时,相比单级并行,三级并行蒙特卡罗模拟加速比增加。     

基于GPU粒子系统的烟花模拟

针对复杂3D烟花模拟方法存在计算量大,效率不高和逼真性不够等问题,提出了一种基于GPU粒子系统的烟花模拟方法.采用三维网格模拟烟花形状,通过扩展双面深度剥离算法在网格上快速均匀采样,提高顶点信息收集效率;提出并行迭代聚类算法,实现多层次的烟花爆炸,达到烟花的多重性效果;应用逆动力学原理控制烟花中粒子的运动,给出形状约束粒子的速度公式.实验结果表明,该算法对烟花模拟效果在真实性和速度上都有所提高.     

一种适于黑油模型并行的快速新解法--改进的GMRES算法

本文应用区域分解算法进行油藏模拟的并行计算研究,寻求可高效并行求解三维三相数值模拟问题的最优算法。在对流行的预处理共轭梯度算法及GMRES算法进行对比研究的基础上,提出了改进的GMRES算法,这种算法具有迭代参数不需优化、收敛快、可得到较精确解等优点。应用该解法对三维三相黑油模型软件进行并行化改造。通过模型及实际油藏计算,比软件原算法及GMRES算法的计算速度得到大幅度提高。并行效率较高,并行化后的模拟软件可以有效地解决大型整装构造油藏的数值模拟问题。     

大规模粒子模拟并行前处理系统的设计与实现

粒子模拟是目前化工、材料、生物等领域重要的研究手段之一.随着计算机软硬件的发展和大规模并行集群的出现,可模拟的粒子规模越来越大,模拟对象也越来越复杂.前处理足粒子模拟初始数据的生成环节,它负责将模拟对象转化为粒子系统,并按照模拟算例需求,将粒子数据输出为文件.前处理是连接模拟对象和模拟计算的纽带.是粒子模拟过程中关键的一步.本文提出的设计方案是首先使用BRLCAD建立模拟对象的三维模型,然后将二维模型转换为空间枚举,接着在空间枚举的规则块中填充粒子,同时通过使用元胞法检测粒子之间的冲突来保证粒子的合法性,最后根据粒子的类型和位置计算粒子的物性并将粒子数据输出到文件.本文根据该设计方案结合MPI并行计算技术,实现了大规模粒子模拟并行前处理系统,并进行了一系列的测试证明了该系统的实用性和可靠性.     

三维位错动力学并行算法与程序研制

为研究极端条件下金属材料的性质,在JASMIN 框架上研制了三维并行位错动力学程序PDD3D. 它集成了离散位错动力学模拟的物理方案和数值算法.通过设计实现高效的分布式数据结构、可扩展的快速多极子解法器以及基于影像区的拓扑操作通信方式,该程序具有较高的性能和较好的可扩展性.1024 个处理器上,对包含3 千万条位错线的物理模型的模拟结果显示,PDD3D 程序获得了81%的并行效率.     

三维有限差分计算中大规模网格生成及显示技术

针对爆炸与冲击问题并行仿真计算软件PMMIC-3D（Parallel Multi-Material in Cell 3D）的计算网格为正交六面体网格的特点,开发与PMMIC-3D接口统一的可对任意复杂三维实体模型进行大规模有限差分网格生成的三维前处理软件MESH-3D.MESH-3D采用CSG和STL模型两种建模方式进行复杂实体建模,并采用基于边的整体切片算法,借鉴计算机图形学中的扫描线填充算法完成三维有限差分网格划分.在绘制网格时,删除网格单元的公共面,大大缩短计算时间和减少存储空间,实现对网格的快速消隐显示.MESH-3D可实现百亿量级网格单元的生成和显示.三维前处理软件MESH-3D的开发有力地支持爆炸与冲击问题的仿真计算.     

基于ANSYS的平面电容传感器阵列三维仿真研究

通过电磁场有限元分析软件ANSYS,构建了电容层析成像(ECT)系统平面3×4电容传感器阵列三维模型.介绍了平面电容传感器阵列结构并分析了其测量原理.对电容进行了仿真计算.仿真与实验进行了对比,证明了仿真模拟的可行性.通过仿真计算电容值,研究了材料与电极间距和材料厚度对电容值的影响.计算了电极对长度和宽度方向的检测深度.通过ANSYS模拟,为损伤检测的特征提取和逆问题图像重构提供了仿真数据.     

计算与通信重叠和并行I/O在粒子模拟中的应用

三维电磁场粒子模拟是研究空间众多微观物理现象的一项先进数值模拟方法。虽然应用MPI和OpenMP混合编程技术实现了程序并行,但阻塞通信的通信同步和应用网络文件系统集中式数据I/O的数据传输降低了程序效率。介绍引入非阻塞通信法,最初计算需要通信部分,在其他计算继续时,进行非阻塞通信,最后接收全部数据,从而实现计算和通信重叠,减少通信等待时间;在分布式存储系统中,各节点同时把本节点数据输入输出到本地单独文件中,大幅度减少数据并行I/O时间,随着数据量和CPU数的增加,改善更加明显,从而提高程序性能。     

Electromagnetic particle-in-cell simulations on magnetic reconnection with adaptive mesh refinement

The electromagnetic particle-in-cell (EM-PIC) model using the adaptive mesh refinement (AMR) is reconsidered so that it is properly and efficiently applied to the current sheet evolution associated with magnetic reconnection. It is very important to adequately select the refinement criteria for cell splitting. It is demonstrated that fine cells have to be distributed not only in the region where the electron Debye length is small, but also in the region where the electron-scale structure is expected to be significant. While the AMR reduces the number of cells drastically, the total simulation cost is also controlled by the number of particles. In order to reduce the total number of particles in the entire system, the present code controls the local number of particles per cell by splitting or coalescing particles. It is shown that the particle splitting and coalescence are quite effective as well as the AMR to enhance the efficiency of the EM-PIC simulations. A new 3D code extended from the 2D code is also introduced. The code is checked against the tearing instability and the lower hybrid drift instability, and it is confirmed that the code has been successfully developed. It is also found that the 3D simulations can gain more efficiency by using the AMR than the 2D simulations.     

Three-dimensional electromagnetic particle-in-cell with Monte Carlo collision simulations on three MIMD parallel computers

A three-dimensional electromagnetic particle-in-cell code with Monte Carlo collision (PIC-MCC) is developed for MIMD parallel supercomputers. This code uses a standard relativistic leapfrog scheme incorporating Monte Carlo calculations to push plasma particles and to include collisional effects on particle orbits. A local finite-difference time-domain method is used to update the self-consistent electromagnetic fields. The code is implemented using the General Concurrent PIC (GCPIC) algorithm, which uses domain decomposition to divide the computation among the processors. Particles must be exchanged between processors as they move among subdomains. Message passing is implemented using the Express Cubix library and the PVM. We evaluate the performance of this code using a 512-processor Intel Touchstone Delta, a 512-processor Intel Paragon, and a 256-processor CRAY T3D. It is shown that a high parallel efficiency exceeding 95% has been achieved on all three machines for large problems. We have run PIC-MCC simulations using several hundred million particles with several million collisions per time step. For these large-scale simulations the particle push time achieved is in the range of 90–115 ns/particle/time step, and the collision calculation time in the range of a few hundred nanoseconds per collision.     

SPH 流固耦合模拟自适应采样固壁虚粒子边界处理方法

固壁虚粒子边界处理方法是流体模拟中一种主要边界处理方法,但其不能确保流 体粒子不穿透固体边界,并且计算量较大。为防止流体粒子穿透边界,在边界附近设置一个阻 尼区,阻尼区内的流体粒子被边界施加一个弹性力和一个和流体粒子运动速度方向相反的阻尼 力,使得边界附近流体粒子更加稳定。为减少计算量,提出两种边界粒子自适应采样法：一种 是依据边界周围粒子数目的不同,边界粒子自适应地采样质量不同的大小粒子;另一种是依据 边界周围粒子数目的不同,边界粒子自适应的采样不同层数的相同质量粒子。与传统的固体边 界粒子采样方法相比,该方法减少了边界粒子数目,加快了模拟速度,节省了计算机内存,基 于GPU 加速技术实现的三维流体模拟,能够进行实时交互。     

On the Dynamics of Charged Electromagnetic Particulate Jets

This work addresses the modeling and simulation of charged particulate jets in the presence of electromagnetic fields. The presentation is broken into two main parts: (1) the dynamics of charged streams of particles and their interaction with electromagnetic fields and (2) the coupled thermal fields that arise within the jet. An overall model is built by assembling submodels of the various coupled physical events to form a system that is solved iteratively. Specifically, an approach is developed whereby the dynamics of charged particles, accounting for their collisions, inter-particle near-fields, interaction with external electromagnetic fields and coupled thermal effects are all computed implicitly in an iterative, modular, manner. A staggered, temporally-adaptive scheme is developed to resolve the multiple fields involved and the drastic changes in the physical configuration of the stream, for example when impacting a solid wall or strong localized electromagnetic field. Qualitative analytical results are provided to describe the effects of the electromagnetic fields and quantitative numerical results are provided for complex cases.     

Particle Monte Carlo and lattice-Boltzmann methods for simulations of gas-particle flows

A numerical solution concept is presented for simulating the transport and deposition to surfaces of discrete, small (nano-)particles. The motion of single particles is calculated from the Langevin equation by Lagrangian integration under consideration of different forces such as drag force, van der Waals forces, electrical Coulomb forces and not negligible for small particles, under stochastic diffusion (Brownian diffusion). This so-called particle Monte Carlo method enables the computation of macroscopic filter properties as well the detailed resolution of the structure of the deposited particles. The flow force and the external forces depend on solutions of continuum equations, as the Navier-Stokes equations for viscous, incompressible flows or a Laplace equation of the electrical potential. Solutions of the flow and potential fields are computed here using lattice-Boltzmann methods. Essential advantage of these methods are the easy and efficient treatment of three-dimensional complex geometries, given by filter geometries or particle covered surfaces. A number of numerical improvements, as grid refinement or boundary fitting, were developed for lattice-Boltzmann methods in previous studies and applied to the present problem. The interaction between the deposited particle layer and the fluid field or the external forces is included by recomputing of these fields with changed boundaries. A number of simulation results show the influence of different effects on the particle motion and deposition.     

Application of a total variation diminishing scheme to electromagnetic hybrid particle-in-cell plasma simulation

A discretization procedure for a total variation diminishing (TVD) scheme is introduced to an electromagnetic hybrid particle-in-cell (PIC) plasma simulation code in order to improve the numerical stability and resolution when calculating the plasma flow field in which magnetic field discontinuities (for example, Rankine–Hugoniot jump conditions for shock waves) are generated. In the hybrid PIC code used in this study, ions are treated as particles and electrons are assumed to be an inertia-less (mass-less) fluid. In the numerical results of one-dimensional test simulations, the TVD scheme significantly prevents non-physical, numerical oscillations, which would ordinarily be produced in the solution when the convection term of the magnetic induction equation in the hybrid PIC code is discretized by central difference schemes at magnetic field discontinuities. Furthermore, a two-dimensional simulation of the global structure of a collision-less bow shock, which is suitable for practical use, makes it possible to clearly capture the bow shock by using the hybrid PIC code with the TVD scheme.     

A splitting integration scheme for the SPH simulation of concentrated particle suspensions

Simulating nearly contacting solid particles in suspension is a challenging task due to the diverging behavior of short-range lubrication forces, which pose a serious time-step limitation for explicit integration schemes. This general difficulty limits severely the total duration of simulations of concentrated suspensions. Inspired by the ideas developed in [S. Litvinov, M. Ellero, X.Y. Hu, N.A. Adams, J. Comput. Phys. 229 (2010) 5457–5464] for the simulation of highly dissipative fluids, we propose in this work a splitting integration scheme for the direct simulation of solid particles suspended in a Newtonian liquid. The scheme separates the contributions of different forces acting on the solid particles. In particular, intermediate- and long-range multi-body hydrodynamic forces, which are computed from the discretization of the Navier–Stokes equations using the smoothed particle hydrodynamics (SPH) method, are taken into account using an explicit integration; for short-range lubrication forces, velocities of pairwise interacting solid particles are updated implicitly by sweeping over all the neighboring pairs iteratively, until convergence in the solution is obtained. By using the splitting integration, simulations can be run stably and efficiently up to very large solid particle concentrations. Moreover, the proposed scheme is not limited to the SPH method presented here, but can be easily applied to other simulation techniques employed for particulate suspensions.     

Two particle-in-grid realisations on spacetrees

The present paper studies two particle management strategies for dynamically adaptive Cartesian grids at hands of a particle-in-cell code. One holds the particles within the grid cells, the other within the grid vertices. The fundamental challenge for the algorithmic strategies results from the fact that particles may run through the grid without velocity constraints. To facilitate this, we rely on multiscale grid representations. They allow us to lift and drop particles between different spatial resolutions. We call this cell-based strategy particle in tree (PIT). Our second approach assigns particles to vertices describing a dual grid (PIDT) and augments the lifts and drops with multiscale linked cells.Our experiments validate the two schemes at hands of an electrostatic particle-in-cell code by retrieving the dispersion relation of Langmuir waves in a thermal plasma. They reveal that different particle and grid characteristics favour different realisations. The possibility that particles can tunnel through an arbitrary number of grid cells implies that most data is exchanged between neighbouring ranks, while very few data is transferred non-locally. This constraints the scalability as the code potentially has to realise global communication. We show that the merger of an analysed tree grammar with PIDT allows us to predict particle movements among several levels and to skip parts of this global communication a priori. It is capable to outperform several established implementations based upon trees and/or space-filling curves.