节点文献

基于直觉启发和改进遗传算法的形状概念设计

Shape Conceptual Design based on Intuition Heuristic and Improved Genetic Algorithm

【作者】 赵婷婷

【导师】 魏小鹏;

【作者基本信息】 大连理工大学 , 机械设计及理论, 2006, 博士

【摘要】 针对目前计算机辅助概念设计(CACD)领域的研究现状,采用思维模拟的方法,对计算机辅助概念设计系统进行了研究。本文结合国家自然科学基金项目“基于演化的概念设计生物建模理论与方法研究”(No.60174037)和“基于知识进化的人机协同方案创新设计理论与方法研究”(No.50275013)的理论研究,针对概念设计中几个热点问题进行了深入地研究,主要研究成果如下: 提出了思维的基本单元变异联想的定义。对思维的本质特征以及基本思维形式进行了研究,并且提出了以激励事件为分段节点、以变异联想为基本单元的思维过程的分段连续函数表示方法,同时提出了用变异联想表示其它思维形式的计算模型。 提出了直觉启发的计算模型。针对直觉思维的产生过程,采用了Hopfield神经网络以及交叉变异等实现方法,对直觉、经验、联想和可视激励之间的相互关系进行了定量的描述,建立了模拟直觉启发的认知模型以及计算模型。最后给出了利用直觉启发模型生成新分形图的应用实例,计算结果表明此算法能够实现创新。 提出了采用直觉启发模型进行创新设计的一种新方法。根据思维的突变产生直觉的观点,给出了实现创新设计的多种基本运算规则,并且对Hopfield神经网络实现联想记忆的算法进行了改进,从而给出了利用直觉启发模型进行创新设计的计算过程。最后以桌子的自动造型设计为例对此算法进行了验证,计算结果表明此算法能够产生创新。 提出了产品的设计元素的统一的基因表达方法,本文称之为0-1分段基因表达方法,这种表示方法既表示了产品的功能需求特征,又便于计算,并且将其应用于概念设计和创新设计过程中。 提出了两种改进的遗传算法,一种本文称之为分段遗传算法,这种方法采用多参数级联编码方法,遗传算子采用分段交叉算子和分段互异算子,解决了产品结构概念设计的多目标优化模型的计算问题;另一种改进的遗传算法本文称之为最优蔓延遗传算法,这种方法的特点是:(1) 最优解是一个群体。(2) 优化的目标函数表达的是一类目标。(3)遗传运算的目的是使群体中的最优个体逐渐扩大。(4) 遗传终止条件是当群体中所有的个体都是最优个体时,则结束循环。 提出了概念设计过程中选择最优结构的定量求解方法。由于概念设计中每个功能都对应着多种实现结构,哪一种结构组合更符合顾客需求,无法靠人工选择,需要选择合适的算法进行计算。给出了产品功能与结构、结构与结构之间的相互关系,定义了产品的相关矩阵,建立了产品结构概念设计的多目标优化数学模型。对产品的功能、行为、

【Abstract】 According to the developing state in the field of conceptual design, the systems of compute Aided conceptual design are studied based on mathematical algorithms of genetic algorithm, neural networks etc.The background of this research is analyzed firstly on the conceptual design, creative design, evolutionary design, shape design etc. Both advantages and disadvantages of existing methods of conceptual design are discussed and the existent problems are pointed out as well. Finally, the research methods and contents of this dissertation are introduced.The segment-continue function processes of thinking are proposed. The basic forms ofthinking are studied. Because intuition plays an important role in creativity, the process of generating intuition is simulated. The relationship between intuition, experience, association and stimuli is described in a quantitative way. The experience is achieved by Hebb’s law. On the bases of definition of mutation association, the process of intuition is modeled by Hopfield neural networks, crossover and mutation operators. The cognitive and computational models for simulating intuition are established. Finally, an example for rendering fractal graphs is given to show the efficiency of the methods presented here.A new method about intelligent design is proposed by modeling the process to generate intuition. In this method, the concepts of gene expression about the design elements were defined. The improved algorithm of Hopfield neural networks for associative memory was used. So the cognitive and computational models for simulating intuition were established on the bases of crossover and mutation operators. In addition, by the models and the basic computational rules for creative design, the computational process of creative design was given. Finally, an example for creative design of table is illustrated to verify the feasibility and validity of the methods presented here.The optimization algorithms for selecting optimization structure are proposed. The problem for selecting optimization structure is the key problem of the researchers in the fields of conceptual design. According to the relationships between structure and function, the mathematical models of multi-object optimization structure design are provided. Some concepts about gene expression and correlation matrix are defined. An improved genetic algorithm called segmenting genetic algorithm is proposed. An example is given to show that this method is able to realize the automatic design of selecting optimization structure. In addition, this segmenting genetic algorithm can provide good on-line and off-line performances.The approaches selecting creative solutions are proposed in this paper. In the creative design based on Genetic algorithm, the problems for selecting creative solutions have not been solved. Firstly, an improved genetic algorithm is provided. This paper called it spreading genetic algorithm. Furthermore, good design results are found out by iterative operations. Then the results are visualized for selection of man or fitness. Finally, an example is given to show that this method is able to realize creative design.Then Cell genetic algorithm and it’s application in the creative design are studied. Cell genetic algorithm is improved and the new cell genetic operators about crossover operator, mutation operator, substitution operator, compress operator, combination compress operator, compress extending operator etc. Two or many genetic operator are used in the computation process.On the basis of theoretic research above, software platform of Integrated design system for conceptual design and creative design is developed. The successful implementation in enterprise testifies the feasibility of the theories and methods presented in this dissertation.The works are concluded and recommendations for future research are also included.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络