High Performance Computational Methods for Structural Mechanics

结构力学的高性能计算方法

• 批准号: 9208980
• 负责人: Charbel Farhat
• 金额: $ 4.4万
• 依托单位: University of Colorado at Boulder
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1992
• 资助国家: 美国
• 起止时间: 1992-06-15 至 1994-11-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9208980&HistoricalAwards=false
• 关键词: Performance; Computational; Methods; Structural; Mechanics

Growing interest in sophisticated structures such as next generation composite aircraft and large space structures, places increasing demands on the optimality of the structural components used. These demands are met by improving both the accuracy and efficiency of the computational tools used to develop and analyze these structures. However, the computational demands of these systems are so great, that engineers must frequently be satisfied with rather crude and simplified models in order to speed up the design cycle time. The true potential for execution improvement lies in improved solution algorithms and parallel computing. This interdisciplinary research program will focus on the design of efficient parallel computational methods for the finite element analysis of such complex structures. The research includes three components: computational mechanics, applied mathematics and applied computer sciences. The first component involves the development of discrete mathematical models based on finite elements for representing aerospace structures as assemblies of beams, plates and shells. The second component addresses the design of domain decomposition methods for static, transient, and vibration analyses using those models. The last research component aims to tuning these methods to massively parallel supercomputers.

人们对下一代复合材料飞机和大型空间结构等复杂结构的兴趣日益浓厚，对所用结构部件的优化性提出了越来越高的要求。 通过提高用于开发和分析这些结构的计算工具的准确性和效率来满足这些要求。 然而，这些系统的计算需求非常大，工程师必须经常满足于相当粗糙和简化的模型，以加快设计周期。 执行改进的真正潜力在于改进的解决算法和并行计算。 该跨学科研究计划将重点关注设计用于此类复杂结构的有限元分析的高效并行计算方法。 该研究包括三个组成部分：计算力学、应用数学和应用计算机科学。 第一个部分涉及开发基于有限元的离散数学模型，用于将航空航天结构表示为梁、板和壳的组件。 第二部分涉及使用这些模型进行静态、瞬态和振动分析的域分解方法的设计。 最后一个研究部分旨在将这些方法调整为大规模并行超级计算机。