Collaborative Research: Simulation of Multibody Dynamics. Leveraging New Numerical Methods and Multiprocessor Capabilities

合作研究：多体动力学模拟。

• 批准号: 0700191
• 负责人: Dan Negrut
• 金额: --
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-09-01 至 2010-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0700191&HistoricalAwards=false
• 关键词: Collaborative; Research; Simulation; Multibody; Dynamics.

This project capitalizes on new numerical integration methods and recent breakthroughs in multi-processor technologies (software and midrange hardware) to provide the theoretical and computational foundation that will enable a paradigm shift in mechanical system simulation. Recent numerical integration methods developed for mechanical system simulation of large industrial problems (e.g. vehicles, aircraft, spacecraft, gears, chains/track, contact/impact, etc.) have demonstrated the ability to reduce simulation times by a factor of two to three. The proposed work draws on these results to (a) investigate and extend the use of Hilber-Hughes-Taylor (HHT)-type integrators in mechanical system simulation, and (b) investigate and extend new specialized partitioned additive Runge-Kutta methods that have excellent stability properties, adjustable numerical damping, and a wide range of convergence orders. This effort will also lead to accurate integration formulas with adjustable numerical damping for the solution of first order differential equations. These formulas will enable the simulation of mechatronic systems (mixed multibody dynamics and controls problems) in a unified framework. A second direction of research will investigate mechanical system simulation techniques that leverage recent advances in software and midrange hardware support for parallel computation. Multibody dynamics specific load balancing and inter-process communication management that account for the topology of the mechanical system, along with an asynchronous Jacobian evaluation strategy, will enable scalable equation formulation and numerical solution on multiprocessor platforms.Together, the two thrusts of this research (new numerical integration techniques and the associated parallel computation support) will yield one to two orders of magnitude reduction in simulation times for large industrial-type mechanical systems. In economic terms, this work benefits design engineers relying on simulation-based engineering by enhancing their productivity. In educational terms, this work provides material and examples for hands-on classes in Computational Dynamics and Numerical Analysis, as well as the foundation for a longer term initiative aimed at introducing under-represented groups to the field of Computational Science through a series of seminars and workshops offered to High School students and teachers.

该项目利用新的数值积分方法和多处理器技术（软件和中端硬件）的最新突破，提供理论和计算基础，从而实现机械系统仿真的范式转变。 最近为大型工业问题（例如车辆、飞机、航天器、齿轮、链条/轨道、接触/冲击等）的机械系统仿真开发的数值积分方法已经证明能够将仿真时间减少两到三倍。 拟议的工作利用这些结果来（a）研究和扩展希尔伯-休斯-泰勒（HHT）型积分器在机械系统仿真中的使用，以及（b）研究和扩展新的专用分区加性龙格-库塔方法，该方法具有优异的稳定性能、可调的数值阻尼以及广泛的收敛阶数。 这项工作还将带来具有可调数值阻尼的精确积分公式，用于求解一阶微分方程。 这些公式将使机电系统（混合多体动力学和控制问题）在统一框架中进行仿真成为可能。 第二个研究方向将研究机械系统仿真技术，该技术利用软件和中端硬件支持的最新进展来支持并行计算。 考虑机械系统拓扑的多体动力学特定负载平衡和进程间通信管理，以及异步雅可比评估策略，将在多处理器平台上实现可扩展的方程公式和数值求解。这项研究的两个主旨（新的数值积分技术和相关的并行计算支持）将使大型工业型机械系统的仿真时间减少一到两个数量级。 从经济角度来看，这项工作使依赖基于仿真的工程的设计工程师提高了生产力。 在教育方面，这项工作为计算动力学和数值分析的实践课程提供了材料和示例，并为旨在通过一系列研讨会将代表性不足的群体引入计算科学领域的长期计划奠定了基础以及为高中生和教师提供的讲习班。