Multi-processing based computer-aided design systems for real design problems

本课题研究基于多处理的计算机辅助设计系统来解决实际设计问题。随着计算机支持在产品开发中发挥越来越重要的作用，计算机技术的发展方向也从高速的单CPU系统转向多CPU的多处理系统。为了满足日益增长的设计自动化等方面的要求，新型的设计技术将与这一方向相配合。在这种情况下，我们开发了一种设计对象和设计过程的分布式建模方法、一个基于agent的分布式设计系统以及几种多cpu实现的优化方法。研究结果总结如下：1。由于实际设计问题包含多学科、多粒度的信息，因此任何一种设计对象和设计过程的分布式建模方法都是必不可少的。组态是系统组件和几何实体之间的一组声明性条件，是建模方法的有效媒介。它具有促进计算机支持的并行工程的能力。利用基于agent的知识工程技术，可以建立计算机支持的并行工程环境。计算机系统是通过工作站集群上的套接字通信实现的。针对实际设计问题具有组合条件多、多模态、非线性、仿真计算量大等不适定特性，将遗传算法与多cpu实现相结合，成功地应用于能源装置配置设计问题、汽车动力总成优化设计。研究了复杂系统设计问题的建模方法和基于分解的优化方法。这些结果表明了多加工技术在大规模设计应用中的可行性。明确了基于多处理的计算机辅助设计系统的研究方向。少

This project studied multi-processing based computer-aided design systems for real design problems. While the computer support is getting to play more important and essential roles in the product development, the direction of computer technologies shifts from high-speed single CPU systems to multi-processing systems with plural CPUs. The new type of design technology will be necessary to cooperate with such direction for meeting the increasing demands on design automation and so forth. Under this circumstance, we developed a distributed modeling method of design object and design process, an agent-based distributed design system, and several optimization methods which are implemented with multiple CPUs. The research results are summarized as follows :1.Since real design problems include multi-disciplinary and multi-grained information, any distributed modeling method of design object and design process is indispensable. Configuration, which is a set of declarative conditions between sy … More stem components and geometric entities, is effective as a media of the modeling method. It has an ability for facilitating computer-supported concurrent engineering.2.A computer-supported concurrent engineering environment can be established with agent-based knowledge engineering technique. The computer system is implemented via socket communication over a workstation cluster.3.Since the real design problems have ill-posed characteristics such as a large number of combinatorial conditions, multi-modality, nonlinearity, computational expense in simulation calculations, etc., the combination of genetic algorithms and multi-CPU implementation is successfully applied to the energy plant configuration design problem, the vehicle power train optimization design. Also, a modeling procedure and a decomposition-based optimization method for complicated system design problems are studied, respectively.These results show the feasibility of multi-processing techniques in large-scale design applications. Moreover, they clarify the research direction of multi-processing based computer-aided design systems. Less