【作者】 刘道华；

【导师】 原思聪；

【作者基本信息】 西安建筑科技大学 ， 机械电子工程， 2006， 硕士

【摘要】 由于机械设计具有自身设计的复杂性、设计目标的多解性,同时建造设计型专家系统具有自身的难点,因此当前在构建机械设计专家系统开发工具方面发展得相当缓慢,但为了提高机械设计的质量和效率,构建机械设计专家系统仍成为机械设计制造业研究的热点问题。随着计算机、小波理论、神经网络及遗传演化等技术的快速发展,以这种多学科理论支持的专家系统开发工具也迅速发展起来,并且构建机械设计专家系统开发工具也成为机械工业未来发展的必然趋势。 本研究课题分析了机械设计专家系统开发工具的国内外研究现状和优缺点,探讨了专家系统开发工具的基本性能指标。在此基础上,对机械设计专家系统开发工具的关键技术进行了较为深入的研究:提出了用知识的数据预处理技术构建知识库的方法;提出了以知识的关系化表示方法为代表的多种知识模型及表达方式;提出了以基于自适应神经网络的模糊推理方法、基于带区间值的产生式规则及其模糊推理机制为代表的多种推理方法及控制策略;提出了以神经网络、粗糙集理论、在线拾取及模糊分类等自动知识获取的方法;此外,还给出了几种解释机制,并研究了开发工具的整体界面设计问题。 构建机械设计专家系统开发工具时,在提高系统存储的空间复杂度方面采用了稀疏矩阵的存取方式;在提高系统的运算时间复杂度方面采用了并行处理技术;在提高系统的鲁棒性方面采用了具有良好的信息保持性及适应性的神经网络技术。 上述关键技术的研究,对建造实用的机械设计专家系统具有重要的理论意义,对提高机械设计的质量和效率具有潜在的经济效益和广阔的应用前景。更多还原

【Abstract】 Because mechanical design has its self-complexity and the multi-answer, and building the design of expert system also possesses its difficulties, the developing tool building the expert system of mechanical design is developed very slowly. But to improve the quality and efficiency of mechanical design, building the expert system of mechanical design is still becoming the crucial topic in the mechanical design and manufacture. Along with the rapid development of some technology such as computer, little wave theory, nervous network and hereditary evolution, the developing tool of expert system based on these multi-subject theory is also developing rapidly, and building this developing tool also becomes the inevitable tendency in the future mechanical industry.This paper has analyzed the domestic and international research status and the strongpoints and shortcomings of the developing tool in mechanical design, and has discussed the basic performance parameter of it. On this foundation, the essential technology about the developing tool is deeply studied, including of carrying out the data pretreatment of various method on the knowledge database, and proposing to take the expression method of knowledge relations as representative’s many kinds of knowledge modes and the expression ways, and proposing to take the auto-adapted nerve network’s method and the belt sector value production pattern as representative’s many kinds of inference method and the control strategy, and giving the automatic gained method of knowledge by the nerve network and the rough sets theory and on-line collection and the fuzzy classification and so on, furthermore, several explanation mechanism is provided,and the whole interface design problem of the developing tool is studied.When building the developing tool, the system adopted the access way of sparse matrix from to raise the systematical capacity and parallel handling technology to raise the operational time complex degree and the nervous network technology that has good information maintenance and adaptability to raise the systematic robustness.The study of this key technology has important theoretical significance to build the practical expert system of mechanical design. It also has latent economic benefits and the wide application prospect for enhancing mechanical design quality and efficiency.更多还原