LS-DYNA 并行计算和粒子法
课程目标(Objectives):

并行计算:网格的不断细分,以及 新的更为复杂的算法在LS-DYNA中的不断应用,带来了对计算能力的极大考验。LS-DYNA的并行计算成为了解决这一问题的最佳工 具。它可以有效地减少运算时间,从而在合理时间内求得工程问题的解。

本课程涵盖LS-DYNA并行计算的一般介绍和网格分解。它将会帮助用户来熟悉有限元模型的拆分过程并对这一过程产生直观的理 解和想象。 我们还将演示各种特殊的拆分方式。进而,用户可以更深刻地了解MPP的内部操作并应用在日常工程问题中,来达到 减少运行时间的目的。通常MPP的模型排错是非常困难的。这里,课程将会提供一些基本方法和用于检测模型的环境设置。

课程中我们也会讨论MPP下一代的技术-HYBRID。 它将应用在超大模型和超大系统中。

粒子法:粒子法是LSTC最近开发的一种新的模拟气囊 的方法,能够很好的模拟气囊的展开过程,适用于离位分析(OOP)。计算时间,参数设置以及可靠性都比ALE优越。本课程包括讲座和适用案例分析, 旨在介绍LS-DYNA中的粒子气囊法(CPM)的 基本理论及背景;本课程还将提供一些简单的例子帮助参加者理解粒子气囊法的优缺点。

MPP:Finer mesh and more sophisticated algorithm are continuous implemented into LS-DYNA to get better simulation prediction. In order to obtain result within reasonable time, LS-DYNA/MPP becomes the most important tool.

This class will explain the general background and the decomposition method used inside LS-DYNA/MPP. It will help users to "visualize" how the model is divided among processors to gain the speedup. Furthermore, it will also demonstrate how special division will help during run time. Users could follow the idea and buildup more knowledge to practice in daily model to shorten the turnaround time. In general model debugging is very difficult under distributed environment. Here, the class will also provide general debug procedures and environment setup for easier debugging. We will also introduce the next generation technology so called HYBRID code to target very large model and systems.

CPM: The corpuscular particle method (CPM) is newly developed for airbag deployment simulation in LS-DYNA. In this method, the gas is modeled as a set of individual particles. The method could model the out-of-position (OOP) occupant interaction; it is simple, numerically robust, easier and faster than ALE. This course describes the corpuscular particle method (CPM) in LS-DYNA. It is compiled as a one-day training class, covering both background theory and practical usage of the method. The course is accompanied by a set of simple test models that help bringing insight into possibilities and limitations of the method.

课时:

一天

Duration:

1 days

授课语言:

中文

Language:

Chinese

主要内容(Main Contents):


MPP概览:


  • 历史沿革及简介  Introduction
  • 问题区域划分  Domain decomposition
  • MPP优化指标  Scalability
  • 系统优化设置  System tuning and model debugging
  • 高性能计算在研方向(Hybrid等)  HYBRID - the next generation technology

  • 培训课程专家(Instructor):


    王季先/Jason Wang 博士毕业于加州伯克利大学,1997 年加入LSTC,负责任意拉格朗日欧拉方法(ALE), 流固耦合(SFI), 并行计算(MPP)及其它离散单元方法如Discrete Element, SPH, CPM的开发和研究.