【作者】 孙红岩；

【导师】 龚宪生；

【作者基本信息】 重庆大学 ， 机械设计及理论， 2007， 博士

【摘要】 大型旋转机械是广泛应用于各行各业的重点关键设备,如:汽轮机发电机组、水轮机发电机组、涡轮发动机、压缩机、通风机、水泵等,建立大型旋转机械智能诊断系统是企业为保证设备安全运转的迫切需要。由于企业中大型旋转机械分布的分散性及大型旋转机械智能诊断的复杂性,分布式智能、具有自适应功能的多算法“动态融合”智能诊断是大型旋转机械智能诊断的理论研究和工程应用中急需解决的问题。多Agent系统(Multi-Agent System,简称MAS)作为分布式人工智能的一个分支,将问题域分解为多个自治或者半自治的Agent,Agent与所处环境、人以及个体之间进行自主交互、协商与合作表现出“集体智能”,以解决大规模问题的求解。多Agent系统在机器人智能判别、互联网搜索引擎等领域得到了应用。本论文在国家科技攻关计划项目《设备故障网络化智能诊断系统》(编号:2001BA201A0610)和有关横向课题的支持下,根据多Agent系统的基本理论和大型旋转机械特点,研究了大型旋转机械智能诊断的知识体系、多Agent系统总体设计、个体Agent、多Agent系统模型与策略,建立了适用性强的新型的大型旋转机械智能诊断多Agent系统,在有关工程项目中得到了初步验证。本文主要研究内容有:(1)大型旋转机械智能诊断知识体系的研究。根据对多家大型企业的调研成果、ISO 18436国际标准及大型旋转机械智能诊断的特点,首次将教育学中知识体系的概念及研究方法引入到大型旋转机械智能诊攭的知识表示中,深入研究了知识体系系统的结构、内容及系统实现方法,以Authorware软件为平台,建立了新型的、系统的、开放式、模块化的大型旋转机械智能诊断知识体系及软件系统。两个现场实例初步验证系统作为人工诊断知识库的作用。(2)大型旋转机械智能诊断多Agent系统的总体设计。在大型旋转机械智能诊断知识体系的基础上,研究了大型旋转机械智能诊断任务分解方法,建立了多Agent系统的体系结构及个体Agent模型。在多Agent系统体系结构的设计中,研究了多Agent系统的组成、各部分功能、系统工作机制及多Agent系统与知识体系系统之间的关系,实现了由管理Agent总体协调的多个功能Agent协商决定的多种方法“动态融合”完成诊断任务的大型旋转机械智能诊断多Agent系统。在个体Agent模型的设计中,首次提出了在现有BDI模型的基础上引入基于兴趣指标和信心指标的心理状态的心智模型,建立了由心智层和行为层组成的两层个体Agent结构,分析了心智层和行为层的结构及功能。(3)大型旋转机械智能诊断个体Agent的研究。在现有BDI模型基础上,提出了个体Agent心智中思维状态具体结构及行为流程。研究了个体Agent心智中心理状态的影响因素,建立了基于二维模糊隶属度的信心指标算法及基于信心指标和信号特点的CBR兴趣指标算法。使个体Agent在解决诊断任务的过程中模拟人类专家的思维方式,根据具体任务自主地选择合适的诊断行为,完成诊断任务。(4)大型旋转机械智能诊断多Agent系统模型与策略的研究。研究了多Agent系统的协商策略、诊断流程、通讯方式、通讯协议、管理Agent的功能结构及有关算法。提出了基于诊断准确率和诊断时间的效用评估算法、基于效用评估的协商策略、分区黑板和消息通讯方式相结合的通讯方式及共享本体论的通讯协议。使多Agent系统能模拟人类专家集体诊断方式,实现多个Agent之间的交互、协商及协作解决诊断问题。(5)大型旋转机械智能诊断多Agent系统的工程应用研究。在某大型钢铁企业现有的网络基础上,设计了该企业大型旋转机械智能诊断多Agent系统的网络体系结构。并针对该企业的两台汽轮鼓风机组,从系统的结构设计、系统的诊断策略及软件开发等方面,研制了大型旋转机械智能诊断多Agent系统的汽轮鼓风机组诊断实例系统。初步验证了大型旋转机械智能诊断多Agent系统研究的正确性和可行性。更多还原

【Abstract】 Large rotating machinery are extensively used in key equipments such as steam turbine generator set, hydraulic generator set, turbojet engine, compressor, fan, water pump and so on. In order to ensure the equipments working safely, it is of urgent demand to build large rotating machinery intelligent diagnosis system.With the large rotating machinery in industries development oriented to be distributing, open, complicate and intensive, multi-algorithm‘dynamic fusion’intelligent diagnosis with distributing intelligence and self-adjusting function is the urgent problem to be solved in theoretical research and engineering application.Multi-Agent System, abbreviated to MAS, as a branch of distributed artificial intelligence, will divide the problem domain into multiple autonomous or half-autonomous Agents. These Agents interact, negotiate and collaborate with environment around, human and individuals, which will perform the collectivity intelligence for solving the complicted problems. So Multi-Agent System is well applied in areas such as robot intelligent decision-making, web search engine and so on.The work in this thesis was supported by National Science & Technology Development Program‘Plant fault internet-based intelligent diagnosis system’and relative horizontal projects. Based on the basic theory of Multi-Agent System and the characteristics of large rotating machinery, the knowledge system of large rotating machinery, Multi-Agent System general design, single Agent, Multi-Agent System model and policy were studied, large rotating machinery intelligence diagnosis Multi-Agent System with strong applicability were built and then was examined primarily in some relative engineering projects.Main research objectives in this thesis are listed as follows:(1) Study on knowledge system of large rotating machinery intelligent diagnosis. Based on investigations of many relative large companies, ISO 18436 International Standard and characteristics of large rotating machinery intelligent diagnosis, the concept and research methods of knowledge system in Education were introduced into large rotating machinery intelligent diagnosis system for the first time. The structure, contents and system implementation method were systemically studied. Using Authorware software as the platform, a new style, systemic, open and modular knowledge system and software of large rotating machinery intelligent diagnosis were built in this thesis.(2) General design of large rotating machinery intelligent diagnosis Multi-Agent System. Based on the knowledge system of large rotating machinery intelligent diagnosis, task decomposition method of large rotating machinery intelligent diagnosis system was studied. The structure of Multi-Agent System and single Agent model were built at the same time. In the process of designing the structure of Multi-Agent system, the components, the function of each part and system working mechanism of Multi-Agent System were studied. Large rotating machinery intelligent diagnosis Multi-Agent System was built, which completed the diagnosis task by‘dynamic fusion’which is the result of coordination multiple function Agents negotiation with principal Agent. During designing the single Agent model, based on the traditional BDI model, introducing the mental model which is based on interest index and confidence index to the thinker model was first proposed. The two layer single Agent structure composed of thinker and behavior was built and the function and the structure was analysed.(3) Study on large rotating machinery intelligent diagnosis single Agent. Based on the BDI model, the thinking state specific structure and behavior flow of single Agent were proposed. The affecting factors of mentation in single Agent mental were studied. The confidence index algorithm based on two dimensional fuzzy membership grade and the interest index algorithm based on CBR were built, which made the single Agent simulate the human expert in the process of solving diagnosis problem and autonomously select the appropriate diagnosis action and complete the diagnosis by the specific task.(4) Study on the Multi-Agent System model and policy of large rotating machinery intelligent diagnosis. The specific structure and relative althorithm of the negotiation policy, diagnosis flow, communication protocol, the tingker of principal Agent in Multi-Agent System were studied. The communication mode which integrates the partitioned blackboard and information transporting mode, and the communication protocol of shared ontology were proposed, which make the Multi-Agent System simulate the collectivity diagnosis mode of human expert so that interaction, negotiation and collaboration to solve the diagnosis problems were realized.(5) Study on application of large rotating machinery intelligent diagnosis Multi-Agent System. Based on the current network of one large steel industry, for the two stream turbine fan set, from the views of the structure design, diagnosis policy of the system and software developing and so on, the on-line diagnosis system of stream turbine fan set which based on large rotating machinery intelligent diagnosis multi-Agent system was built. Finally, the feasibility and validity of the large rotating machinery Multi-Agent System were primarily examined.更多还原