【作者】 李刚；

【导师】 宋福根；

【作者基本信息】 东华大学 ， 管理科学与工程， 2008， 博士

【摘要】 随着工业化与信息化的融合,制造企业管理信息系统日益向智能化的方向发展,制造企业中知识管理的地位与作用日益显著。对于制造企业而言,产品的设计、制造、服务等全生命周期过程中的知识管理至关重要,对于企业的持续发展起着关键的作用。产品知识管理的重要性已经引起很多学者的关注,但国内目前绝大部分研究还停留在概念与理论框架方面,鲜有对产品知识管理系统应用理论方面的研究,极大地制约了知识管理在制造企业的应用实效。本文以制造企业的产品知识管理为主线,分别从设计知识、制造知识、服务知识等角度,对产品知识管理中的若干关键问题进行了深入研究。本文工作的主要内容分为如下几部分:1.产品设计知识建模重点研究了产品设计知识的建模过程以及产品设计知识的半自动化建模问题。综合了现有主要的本体论建模过程的方法,提出了一种改进的最小关联综合法本体建模过程方法,探索更为实用的本体建模过程。本文对于不同性质的产品设计知识采取不同的半自动化建模方法:对于半结构化、非结构化产品设计知识——Web文本,提出了本体学习技术,研究了本体学习中领域本体的创建与领域本体的分类问题,提出了领域本体的分类算法,并对算法的有效性进行了验证;为使用原有的公共语词对异构Web本体进行知识表示,本文提出了一个异构Web本体模式识别的算法;为解决本体语义的更新以及补充问题,本文提出了一种获取新语义关系的算法,及时有效更新领域本体模型。对于结构化产品设计本体的半自动化构建,本文提出了通过UML——本体映射方法,将这些信息系统开发中的UML半自动地转化为本体,以方便产品设计知识本体构建的思路;并研究了UML向OWL映射的语义连接机理,解决了UML向本体的半自动化建模问题,从而可以提高本体的构建效率,提高本体的复用和更新效率,并有助于普及本体的应用、扩充本体的研究队伍,加强企业信息系统的集成。2.产品设计知识检索讨论了目前常用的本体知识表示语言,提出了产品概念的知识表示方法;产品开发过程设计知识的表示方法;并研究了基于Prot(?)g(?)的产品设计知识库存储方法。提出了基于依存关系的语义标注方法,将半结构化、非结构化Web文本与领域本体知识库进行语义关联,并进行了实例验证。构建了基于Jena的设计知识检索引擎,本引擎采用了最新的SPARQL语言;而且应用于产品知识管理系统,具有一定的先进性。经过实例验证,此检索引擎具有直观性与简明性,实现了产品设计知识中的智能检索和推理。为实现知识检索过程中用户偏好的管理,提出了用户偏好库特征词条权重的扩展算法。实验表明,本算法的查全率与查准率与相关算法相比具有一定的优点。3.产品工艺决策知识管理研究当前国内工艺知识管理的研究多局限于具体的特征零件与机械产品,制约了产品制造知识管理理论与系统的应用推广。为克服工艺知识管理的局限性和对应用环境的依赖性,本文提出了具有普适性的通用工艺决策知识管理模型:构建统一的零件特征信息模型;利用抽象的更具一般性的“属性”或“对象”设计工艺规范知识库;并构建了普适性的通用工艺知识推理模型,有效地保证工艺知识管理系统的通用性和移植性。4.产品故障诊断知识管理研究在已报道的故障诊断专家系统中,大多数都是针对某一具体设备的专用专家系统,通用性不强。为了开发高效且适用性广的故障诊断知识推理模型,本文提出了基于神经网络的故障诊断知识推理模型;为克服全连接神经网络结构复杂、训练速度慢等缺点,提出非全连接神经网络(PCBP)的构建理论与相关算法,并通过实验进行了验证。5.产品知识管理系统的建设企业的知识管理系统的建设是产品知识管理得以顺利实施与应用的有效手段,综合本文设计知识建模、设计知识检索、工艺决策知识管理、故障诊断知识管理理论的研究成果,建立了产品知识管理系统软件原型,原型系统验证表明,本文关于产品知识管理中关键问题的研究成果是有效的。更多还原

【Abstract】 With the integration of industrialization and informationisation,the management information system in manufacturing enterprises is evolving to be more intelligent, and Knowledge Management System is playing more important role.The knowledge management in product life cycle of design,manufacture and service is vital to manufacturing enterprises,especially for the sustainable development.The importance of product knowledge management has attracted attentions from many scholars,however most of whom are focusing on the concept and theoretical framework.In China only few have researched on the application of product knowledge management system,and this is one of the primary factors that restrict the effectiveness of the knowledge management in manufacturing enterprises.The dissertation focuses on the principles,methods and applications of key issues of Products Knowledge Management in manufacturing enterprises,and it includes the following sections:1.Product design knowledge modeling technologyThe dissertation focuses both on the product design process modeling as well as the semi-automated modeling problems of product design knowledge.A combination of existing major ontology modeling method is proposed-minimal associated integrated approach.The approach is proposed for exploring a more practical ontology modeling process.For different kinds of product design knowledge,different semi-automated modeling methods are proposed:For the semi-structured and unstructured product design knowledge,for example, Web text,ontology learning technology method is employed.The dissertation researches the creation of domain ontology and the classification of domain ontology. A classification algorithm of domain ontology is proposed,and its effectiveness is verified.The original ontology is used for the knowledge representation of heterogeneous Web text,and a pattern recognition algorithm is proposed.For the update and supplement of the domain ontology,a new semantic relationship algorithm is proposed.For the structured product design knowledge modeling,an UML-OWL mapping methods is proposed,these UML models developed in information systems can be transformed into ontology model for facilitating the model of structured product design knowledge.And the principle of the mapping methods is studied.The modeling method can increase the efficiency of developing ontology and reusing ontology,contribute to the universal application of the ontology implementation, expand the research team of ontology,and strengthen the integaration of enterprise information systems.2.Product design knowledge retrievalAfter discussing the common knowledge representation languages,the dissertation proposes a method of the knowledge representation method of the product concept,and the knowledge representation method of the product design knowledge among the product development process.The method of the storing method of prot(?)g(?)-based research and product design knowledge base is discussed.The semantic tagging method based on semantic dependency relations is proposed,and the dissertation also semanticly joins with the semi-structured and unstructured Web text and the domain ontology.And the effectiveness of the semantic dependency is validated.A product design knowledge retrieval engine based on Jena is developed.The engine uses the latest SPARQL language and is implemented in the product knowledge management.The advanced of the engine has been validated,and this retrieval engine is simple but intuitive.It realizes the intelligent retrieval and reasoning of the product design knowledge.For the realization of user preferences during the process of the product design knowledge retrieval,an extended algorithm of the feature lemma’s weight in the users’ preference database is proposed.Experiments indicate that the algorithm recall and precision compared with the other algorithm has certain advantages.3.The research of product process knowledge managementAt present,the most domestic research of knowledge management of process is limited to the specific characteristics of parts and machinery products,which is a limitation of the application and promotion of the process knowledge management system.While the dissertation presents a universal product process decision-making knowledge management model,using a unified Parts characteristics information model;the process knowledge base building method with the abstract and general“properties”or“classes”;and a universal process knowledge reasoning model.These methodologies effectively guarantee the general use of process knowledge management system.4.The research of product fault diagnosis knowledge managementAt present the most domestic research of fault diagnosis are focused maily on device-specific expert systems.The versatility of these methodologies is not strong.In order to develop efficient and common applicability of the fault diagnosis knowledge management model,the dissertation proposes a fault diagnosis knowledge reasoning model using artificial neural network technology.In order to overcome the shortcomings of complex structure and slow training of neural network, a partially connected neural network(PCBP)is constructed.Some related algorithms are discussed and the usability of PCBP was validated by experiments.5.Product knowledge management constructionEnterprise knowledge management construction is vital to the implementation and application of product knowledge management system.By combining the research results of the product design knowledge model,the product design knowledge retrieval,product process decision-making knowledge management,and product fault diagnosis knowledge management that proposed in the dissertation,a software prototype is developed.The application of the software prototype has satisfied achievements.The validation proves that the product prototype knowledge management system can promote the enterprise knowledge management level.The reseach results of the key issues of product knowledge management in this dissertation are validated.更多还原