A Study on Intelligent CAD Systems based on Hybridization of Artificial Intelligence Approaches and Optimization Techniques

通过对几个设计问题的混合型CAD系统的开发，研究了人工智能（AI）技术与数学优化技术相结合的智能CAD系统的设计方法。所开发的系统及其特点概括如下：1.电厂布置设计系统，采用约束导向搜索方法确定布置构件间的空间关系，采用优化技术确定布置构件的实际尺寸.用于自动生成布局约束的专家系统，布局约束是上述布局设计系统中的设计条件.支持飞机初步设计过程的集成CAD系统，采用面向对象的建模方法表示设计对象，采用优化过程细化设计.平面连杆机构形状优化设计系统，采用面向对象的建模方法表示连杆机构，采用符号处理技术和代数编程语言自动生成优化数学模型.基于遗传算法和局部极小化算法的混合优化排样方法，通过以上研究，得出了人工智能技术与优化过程混合的准则.人工智能技术对于设计问题的概念和组合方面是有效的，而优化技术对于它们的具体化和数值方面是有效的。两种技术应根据各自设计问题的特点进行混合.人工智能技术中的建模技术是自动生成优化计算数学模型和交互式支持设计过程的有效手段。

Design methodology for intelligent CAD systems by combining artificial intellignece(AI)techniques and mathematical optimization techniques are studied through the development of the hybrid-type CAD systems for several design problems. The systems developed and their characteristics are summarized as follows :1. Layout design system for power plants, in which a constraint-directed search procedure is used for fixing the spatial relations among layout components and an optimization technique is used for determining the actual sizes of layout components.2. Expert System for automatically generating layout constraints which are design conditions in the above layout design system.3. Integrated CAD system for supporting preliminary design process of aircraft, in which an object-oriented modeling method is used for representing design object and an optimization procedure is used for refinement of a design.4. Automated optimal shape design system of planar link mechanisms, in which an object-oriented modeling method is used for representing link mechanisms, and symbol processing technique and an algebraic programming language are used for automatically generating the mathematical model for optimization.5. Hybrid approach for optimal nesting problems using a genetic algorithm and a local minimization algorithm.Through the above studies, the following criteria for hybridization of AI techniques and optimization procedure are concluded.1. AI techniques are effective for the conceptual and combinatorial aspects of design problems, and optimization techniques are effective for the embodiment and numerical aspects of them. Both of techniques should be hybridized corresponding to the characteristics of respective design problems.2. The modeling techniques in AI techniques are effective for automatically generating the mathematical models for optimization calculation and for interactively supporting the design process.