EAGER: Accurate and Efficient Surrogate Modeling Applied to Computational Mechanics

EAGER：准确高效的代理建模应用于计算力学

• 批准号: 1132364
• 负责人: Spandan Maiti
• 金额: $ 10.61万
• 依托单位: University of Pittsburgh
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2010
• 资助国家: 美国
• 起止时间: 2010-08-01 至 2013-01-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1132364&HistoricalAwards=false
• 关键词: EAGER; Accurate; Efficient; Surrogate; Modeling

The objective of this EArly-concept Grant for Exploratory Research (EAGER) project is to apply High Dimensional Model Representation (HDMR) technique to construct highly accurate and efficient surrogate computational mechanics models. Advanced computational mechanics simulations typically deal with very large scale models that are extremely time consuming. HDMR technique uses systemic sampling procedure relating outputs of the original physical model to its inputs to produce highly accurate and computationally cheap equivalent models. These low order surrogate models are expected to dramatically reduce the cost of very large scale simulations while maintaining the accuracy of the original models. The proposed methodology is extendable to other fields of engineering besides computational mechanics.This research has the potential to transform current state of modeling philosophy and simulation capability in computational mechanics, and enable very large scale simulations at a reasonable cost. Simulations hitherto intractable due to very high computational burden can be performed efficiently with the proposed methodology. Additionally, it will provide new avenues in the development of accurate yet efficient simulation methods. Educational objectives will focus on (1) training graduate as well as undergraduate students in the area of surrogate modeling, and (2) developing new course materials for the existing courses on computational mechanics. The research being at the confluence of diverse areas of science and engineering such as mechanics, statistics and computational methods, will provide rich educational and research experiences for undergraduate and graduate students in state-of-the-art computational modeling and simulation.

这个早期概念探索性研究资助（EAGER）项目的目标是应用高维模型表示（HDMR）技术来构建高精度和高效率的替代计算力学模型。高级计算力学模拟通常处理非常耗时的超大规模模型。HDMR技术使用系统采样过程将原始物理模型的输出与其输入相关联，以产生高度准确且计算成本低的等效模型。这些低阶代理模型有望大大降低超大规模模拟的成本，同时保持原始模型的准确性。所提出的方法可扩展到计算力学以外的其他工程领域，这一研究有可能改变当前计算力学建模理念和仿真能力的现状，并以合理的成本实现超大规模的仿真。迄今为止棘手的模拟由于非常高的计算负担，可以有效地进行所提出的方法。此外，它将提供新的途径，在准确而有效的模拟方法的发展。教育目标将集中在（1）培训研究生以及本科生在代理建模领域，（2）开发新的课程材料，为现有的课程计算力学。该研究是在科学和工程的不同领域，如力学，统计和计算方法的融合，将提供丰富的教育和研究经验，为本科生和研究生在国家的最先进的计算建模和模拟。