面向产品设计的知识工程若干关键问题研究
发布时间：2018-11-02 10:10
【摘要】：知识是人类智能的基础,所以知识也是人工智能问题的求解的基础。人工智能领域研究的主要问题就是如何从现实世界人类智慧的积累中获取知识、如何合理地把已获得的知识表示成计算机能够识别和处理的形式以便于存储、以及如何基于这些知识进行推理来解决实际的问题,即知识获取、知识表示和基于知识推理。上世纪90年代以来,人工智能、计算机技术和CAD技术不断发展和结合,形成了基于知识工程(KBE)的研究,并且很快KBE已经成为一个研究热点。本论文围绕着知识工程的一些关键问题进行了系统的分析,并以此为基础,对知识建模、知识表达、知识推理和知识繁衍进行了深入的和较为全面的研究,主要包括：(1)基于QFD的知识建模方法通过把顾客的需求有效地转化为技术措施,并将顾客需求的价值排序反应到技术措施的选择当中,从而能够确定出满足顾客需求的设计方案。以QFD为主线,采用情景分析法、KJ法、AHP法等工具,构建基于QFD的产品概念设计和产品方案设计的知识模型。(2)面向对象和框架方法相结合的知识表示方法现在产品设计的复杂性,不仅需要数值计算,而且需要用到非参数化知识,如启发式经验等。单一知识表示法难以全面、有效地表达各种设计所需求的知识。本文基于面向对象思想和框架知识表示法的优势,提出了基于面向对象设计思想,借用框架表示法中槽结构的产品设计知识表示方法。(3)基于案例和基于规则集成的知识推理方法本文提出了一种智能化的产品设计知识推理机制,帮助设计人员解决技术难度大和设计时间长的问题。首先根据基于QFD产品设计知识建模方法确定的产品技术参数；然后采用基于案例推理的人工智能技术,根据设计专家经验知识进行推理,搜索产品设计案例库中与QFD得出的产品技术参数最相似的产品设计方案；最后采用基于规则推理的方法,根据规则库中领域专家和设计工程师设定的产品设计修改规则,提出修改建议,制定产品设计方案。(4)基于混沌开关的遗传算法的知识繁衍方法因为传统遗传算法的局限性,同时复杂产品的优化设计是一个离散性、多变量、非线性、多目标混合型规划问题,所以很难求解产品设计优化问题。本文把遗传算法和混沌理论相结合起来,提出了利用混沌搜索有效基因的方法：利用混沌的遍历性特点,动态地搜索个体的基因座空间,并使搜索到的基因座上的等位基因产生变异,从而产生一个新的个体。现代产品设计包括对知识的继承、集成、创新和管理,它是一个知识驱动的创造性过程。本论文以变压器智能设计为背景,设计了变压器的主体和构件的对象模型,以及以槽表示的数据和属性模型；设计了产品和零件主模型的案例表示、索引机制、相似度模型,以及案例的搜索算法；设计了基于规则的推理网络、规则的表示方法,以及基于规则的变压器设计案例修改算法；最后,提出了基于混沌开关的遗传算法求解模型,并基于变压器的优化设计模型进行了实例验证。
[Abstract]:Knowledge is the foundation of human intelligence, so knowledge is also the foundation of solving artificial intelligence problem. The main problem in the field of artificial intelligence is how to acquire knowledge from the accumulation of human intelligence in the real world, how to properly represent the acquired knowledge as the form of computer can recognize and process in order to facilitate storage, And how to solve practical problems based on these knowledge, namely knowledge acquisition, knowledge representation and knowledge-based reasoning. Since the 1990s, the development and integration of artificial intelligence, computer technology and CAD technology have formed the research of KBE based on knowledge engineering (KBE), and KBE has become a hot spot soon. This paper systematically analyzes some key problems about knowledge engineering, and makes a thorough and more comprehensive study on knowledge modeling, knowledge representation, knowledge reasoning and knowledge reproduction. (1) The knowledge modeling method based on QFD effectively transforms the customer's needs into technical measures and sorts the value ranking of customer requirements into the selection of technical measures, so as to determine the design scheme to meet the requirements of customers. Taking QFD as the main line, the knowledge model of product conceptual design and product design based on QFD is constructed by using tools such as scene analysis method, AHP method and AHP method. (2) The knowledge representation method combined with object-oriented and framework methods now has the complexity of product design, not only numerical calculation but also non-parametric knowledge, such as heuristic experience, etc. The single knowledge representation is difficult to fully and effectively express the knowledge of various design requirements. Based on the advantage of object-oriented idea and framework knowledge representation, this paper proposes a method for expressing product design knowledge based on the object-oriented design idea and the groove structure in the frame representation. (3) Based on case-based and rule-based knowledge reasoning method, an intelligent reasoning mechanism of product design knowledge is put forward to help designers solve the problems of large technical difficulty and long design time. firstly, according to the product technical parameters determined by the QFD product design knowledge modeling method, then adopting the case-based reasoning based artificial intelligence technology, reasoning according to the experience knowledge of the design experts, The product design scheme is the most similar to the product technical parameters obtained from QFD in the product design case library, and finally, based on the rule-based reasoning method, according to the product design modification rules set by the domain expert and the design engineer in the rule base, the modification suggestions are put forward and the product design plan is formulated. (4) The knowledge reproduction method of genetic algorithm based on Hash switch is the limitation of traditional genetic algorithm, and the optimization design of complex products is a problem of discrete, multi-variable, non-linear and multi-objective hybrid programming, so it is difficult to solve the problem of product design optimization. In this paper, the genetic algorithm and the genetic algorithm are combined, and the method for searching the effective gene by using the genetic algorithm is put forward: using the ergodic characteristics of the genetic algorithm, the locus space of an individual is dynamically searched, and the allele of the searched locus is mutated, thereby creating a new individual. Modern product design includes the inheritance, integration, innovation and management of knowledge, which is a creative process of knowledge-driven. Based on the transformer intelligent design, this paper designs the object model of the main body and the component of the transformer, and the data and attribute model represented by the slot, and designs the case representation, index mechanism, similarity model of the product and part master model, as well as the search algorithm of case. This paper designs a rule-based reasoning network, a rule-based representation method, and a rule-based transformer design case modification algorithm. Finally, a genetic algorithm solution model based on the rule-based switch is put forward, and an example verification is carried out based on the optimized design model of the transformer.
【学位授予单位】：东北大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TB472;TP18
，
本文编号：2305725
[Abstract]:Knowledge is the foundation of human intelligence, so knowledge is also the foundation of solving artificial intelligence problem. The main problem in the field of artificial intelligence is how to acquire knowledge from the accumulation of human intelligence in the real world, how to properly represent the acquired knowledge as the form of computer can recognize and process in order to facilitate storage, And how to solve practical problems based on these knowledge, namely knowledge acquisition, knowledge representation and knowledge-based reasoning. Since the 1990s, the development and integration of artificial intelligence, computer technology and CAD technology have formed the research of KBE based on knowledge engineering (KBE), and KBE has become a hot spot soon. This paper systematically analyzes some key problems about knowledge engineering, and makes a thorough and more comprehensive study on knowledge modeling, knowledge representation, knowledge reasoning and knowledge reproduction. (1) The knowledge modeling method based on QFD effectively transforms the customer's needs into technical measures and sorts the value ranking of customer requirements into the selection of technical measures, so as to determine the design scheme to meet the requirements of customers. Taking QFD as the main line, the knowledge model of product conceptual design and product design based on QFD is constructed by using tools such as scene analysis method, AHP method and AHP method. (2) The knowledge representation method combined with object-oriented and framework methods now has the complexity of product design, not only numerical calculation but also non-parametric knowledge, such as heuristic experience, etc. The single knowledge representation is difficult to fully and effectively express the knowledge of various design requirements. Based on the advantage of object-oriented idea and framework knowledge representation, this paper proposes a method for expressing product design knowledge based on the object-oriented design idea and the groove structure in the frame representation. (3) Based on case-based and rule-based knowledge reasoning method, an intelligent reasoning mechanism of product design knowledge is put forward to help designers solve the problems of large technical difficulty and long design time. firstly, according to the product technical parameters determined by the QFD product design knowledge modeling method, then adopting the case-based reasoning based artificial intelligence technology, reasoning according to the experience knowledge of the design experts, The product design scheme is the most similar to the product technical parameters obtained from QFD in the product design case library, and finally, based on the rule-based reasoning method, according to the product design modification rules set by the domain expert and the design engineer in the rule base, the modification suggestions are put forward and the product design plan is formulated. (4) The knowledge reproduction method of genetic algorithm based on Hash switch is the limitation of traditional genetic algorithm, and the optimization design of complex products is a problem of discrete, multi-variable, non-linear and multi-objective hybrid programming, so it is difficult to solve the problem of product design optimization. In this paper, the genetic algorithm and the genetic algorithm are combined, and the method for searching the effective gene by using the genetic algorithm is put forward: using the ergodic characteristics of the genetic algorithm, the locus space of an individual is dynamically searched, and the allele of the searched locus is mutated, thereby creating a new individual. Modern product design includes the inheritance, integration, innovation and management of knowledge, which is a creative process of knowledge-driven. Based on the transformer intelligent design, this paper designs the object model of the main body and the component of the transformer, and the data and attribute model represented by the slot, and designs the case representation, index mechanism, similarity model of the product and part master model, as well as the search algorithm of case. This paper designs a rule-based reasoning network, a rule-based representation method, and a rule-based transformer design case modification algorithm. Finally, a genetic algorithm solution model based on the rule-based switch is put forward, and an example verification is carried out based on the optimized design model of the transformer.
【学位授予单位】：东北大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TB472;TP18
，
本文编号：2305725
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2305725.html
