Computational Mechanics Workbench

9616764苏斯曼，杰拉尔德J.这项研究的目标是为海量数字数据集的智能分析和解释创建体系结构和算法基础。其目的是提供降低特定工程问题分析复杂性的技术，同时将这些特定问题作为重点来制定推进人工智能前沿的一般原则。智能数据解释程序必须利用视觉识别和符号抽象能力。跟踪大量相互作用、变形的物体对人类视觉系统来说可能太繁重了。赋予我们的计算机画图的能力，让他们自己看到并使用他们所看到的来指导他们未来的行动，这既是必要的，也是令人兴奋的。将图像推理与符号计算相结合的人工智能程序已被用于解决动力系统分析、控制器设计、机构运动学等应用领域的问题。这一范例有望成为未来工程分析和设计环境中不可或缺的一部分。将为流体动力学家提供一个计算环境，以便有效地从模拟或实验湍流数据中提取结构和分离相互作用模式，并对结构相互作用进行控制实验。其目标是：(1)将CFD数据的后处理时间缩短一到两个数量级；(2)提供数据的符号描述以供总结和比较；(3)开发符号和几何技术来编码连贯的结构、它们的演化和相互作用的规则。***

9616764 Sussman, Gerald J. The goal of this research is to create the architectural and algorithmic foundations for intelligent analysis and interpretation of massive numerical datasets. The aim is to provide the technology to reduce the complexity in the analysis of specific engineering problems, and at the same time use these specific problems as a focus to develop general principles that advance the frontiers of Artificial Intelligence. Smart data interpretation programs must exploit visual recognition and symbolic abstraction capabilities. Tracking a multiplicity of interacting, deforming objects might well be too taxing for the human visual system. It is both necessary and exciting to endow our computers with the capability to draw pictures for themselves to see and use what they saw to guide their own future actions. AI programs that integrate 'reasoning with images" with symbolic computations have been implemented to solve problems in several applications areas, such as dynamical system analysis, controller design, mechanism kinematics. This paradigm can be expected to be integral part of future engineering analysis and design environments. A computational environment will be provided for fluid dynamicists to efficiently extract structures and isolate interaction patterns form simulated or experimental turbulence data, and to perform control experiments on structure interactions. The objectives are : (1) to cut post-processing time of CFD data by one to two order of magnitude, (2) to provide symbolic descriptions of data for summarization and comparison, and (3) to develop symbolic and geometric techniques to encode coherent structures, their evolution, and interactions rules. ***