Distributed/Parallel Analysis and Optimization of Structural Systems Using Intelligent Autonomous Element

首先，对分布式和协作式问题求解的研究现状进行了综述。研究发现，分布式协作问题求解是一种由解决局部问题的智能主体通过契约、冲突、调整和委托等交互来解决整个问题的方法，适用于并行处理和优化设计问题。其次，研究了用于结构分析和优化的同步和异步算法。结果表明，当单元数变大时，采用异步算法是可行的。然后，将该算法在大规模并行计算机上实现，解决了几个结构问题。利用该方法可以对变几何桁架进行几何和结构分析，并发现建模的简便性和计算的加速比是该方法的主要因素。特别是，可以构造一种新的分布式优化方法，该方法可以在大规模并行计算机上成功实现。此外，分析与优化相结合是该方法的重要特点。我们称这种方法为自主智能单元法，因为这种方法是基于元素的。并且利用包含结构系统中的所有元件的每种行为的知识库，所述结构系统中的所有元件包括主动元件中的致动器。

First, the researches on distributed and collaborated problem solving were reviewed. It is found that the distributed and collaborated problem solving is an approach where intelligent agents which solve the local problems perform the interaction including contracts, conflicts, adjustments, and delegation in order to solve the entire problem, and this approach is suitable for parallel processing and optimum design problem. Next, a synchronous and asyncronous algorithms for analyzing and optimizing structures are investigated. It is found that the asynchronous algorithm is usuful when the number of elements becomes large. Then, the algorithm is implemented into a massively parallel computers, and several structural problems were solved. The geometric and structural analyzes of variable geometry trusses can be performed using this approach, and it is found that the easiness in modeling and the computational speedup are the main important factors in this approach. Especially, a new distributed optimization method can be constructed, and this method can be implemented in a massively parallel computer successively. Further the integrationof analysis and optimization is the important feature in this method. We call this method the autonomous intelligent element method since this method is a element-based. and utilizes a knowledgebase which contains every behavior of all the elements in structural systems including actuators in active menbers.