硕士学位毕业论文 硕士研究生: 孙翔 指导教师:沈斌 教授 答辩时间:2014.06 摘要 数控机床是制造企业生产线上的关键设备,为了避免由于发生故障时机床长时间停机给企业造成巨大的经济损失,需要运用故障诊断技术及时判断机床故障状态并进行排除,因此对数控机床展开相关故障诊断技术的研究具有重大的现实意义。但是,目前数控机床的故障诊断过分依赖维修服务人员的经验,尚缺乏一种集信息采集和诊断推理于一体的具有专家水平的智能化诊断系统。 因此本文系统研究了人工智能、案例推理、粗糙集的理论和方法,并针对数控机床的故障特点对上述理论在数控机床故障诊断领域中的应用做了详细的分析,在此基础之上,设计开发了一套智能化的数控机床故障诊断专家系统,具有一定的实用性。 论文所做的主要工作有: 1. 研究分析了数控机床故障常用的诊断方法以及专家系统的原理和粗糙集理论的基本概念,提出本系统故障诊断方法以专家系统为基础、案例推理技术为核心的设计思路,并通过案例的粗集表示和属性约简,阐明了本系统以报警号码作为故障诊断主要依据的合理性。 2. 对案例推理的关键技术案例检索的原理进行了研究,建立了故障案例的两级检索模型,通过采用最近邻相似性度量方法与贝叶斯排序算法相结合的案例检索方法,向用户提供一个综合决策支持,显著提升了案例检索的效率和性能。同时针对传统的最近相邻算法存在的一些不足进行了改进,提出了基于报警号码的相似度算法模型,提高了案例匹配的精度;设计了二级案例检索的影响值算法,并设计实验进行了验证。 3. 研究分析了专家系统的一般性结构与工作原理,结合数控机床故障诊断的特点,对基于粗糙集和 CBR 的数控机床故障诊断专家系统进行了详细的方案设计,给出了案例库以及各基本功能模块的可行方案。 4. 基于 MVC 设计模式并采用 Visual Studio 2010 和 SQL Server 2005 为软件平台开发了该智能化的数控机床故障诊断系统,通过诊断实例,对推理算法和原型系统的运行情况进行了验证,证明其在实际诊断中效果良好,并对下一步的研究方向进行了展望。 关键词: 案例推理,粗糙集,数控机床,故障诊断,专家系统 ABSTRACT CNC machine tool is the key equipment in production line of manufacturing enterprise, in order to avoid the huge economic losses caused by the failure of CNC, it is necessary to timely troubleshoot the CNC fault with fault diagnosis technique. Therefore, it has a great practical significance to carry out the research of CNC machine tool fault diagnosis technology. However, CNC troubleshooting is too dependent on the knowledge and experience of the maintenance engineer at present, it still lacks an intelligent diagnosis system with expert level of understanding, which has the integrated function of information collection and troubleshooting. For this reason, this paper studied the theory and method of artificial intelligence, case-based reasoning and rough set. According to the feature of CNC fault, analyzed the application of the above mentioned theory in CNC machine tool fault diagnosis field in detail. On this basis, designed and developed a set of intelligent CNC machine tool fault diagnosis expert system, with some practicality. The main contents are summarized as follows: 1. The basic concept of rough set, the most common methods of CNC fault diagnosis and basic principle of expert system were researched and analyzed in this paper. Put forward the design idea of diagnosis method of this system, namely based on expert system and case-based reasoning. Used rough set theory in case representation and attribute reduction to illustrate the reasonableness of taking alarm as the main basis in the process of troubleshooting. 2. The key technology of case-based reasoning was studied, established the model of two-stage case retrieval, by using Nearest-Neighbor Approach combined with Bayesian sorting algorithm to provide customers with a comprehensive decision support, it significantly improved the efficiency and performance of case retrieval. According to the shortages of traditional Nearest-Neighbor algorithm, improved it and put forward the model of similarity algorithm based on alarm. Designed the affect value algorithm and verified it with experiment. 3. Studied the general structure and working principle of expert system. Carried out a detailed scheme design of CNC machine tool fault diagnosis expert system based on rough set and case-based reasoning, gave the solution of knowledge base and the basic function modules. 4. Designed and developed database of the expert system and each function module making use of Visual Studio 2010 and SQL Server 2005 as software platform and MVC design model. Through a diagnosis example, the reasoning algorithm and the prototype system running status was verified and the next research direction was prospected. Key Words: case-based reasoning, rough set, CNC machine tool, fault diagnosis, expert system
硕士学位毕业论文 硕士研究生:陈敏 指导教师:沈斌 教授 答辩时间:2012.06 摘要 电动助力转向系统近年来得到了迅速发展,随着被越来越广泛地应用在轿车上,人们对 EPS 系统的安全性和可靠性也提出了更高的要求。电动助力转向由电机来提供助力,作为一项新技术,和液压动力转向相比, EPS 存在不同的故障模式,而作为 EPS 最重要的部件,传感器的故障研究,及其故障诊断技术就显得十分重要。本论文对 EPS 中最主要的扭矩传感器和车速传感器的各种故障进行研究和分析,建立了一套基于 LM 神经网络的传感器故障诊断系统。 本论文工作主要包括以下几个方面的内容: 首先,基于扭矩传感器和车速传感器的结构和工作原理的深入分析,应用故障树分析,建立了传感器的故障树模型,得出了故障原因;应用失效模式与影响分析得到了传感器故障的相关重要度,找出传感器故障诊断的最优诊断模式。 其次,通过对人工神经网络和传感器故障原因和特点的研究,在 Matlab 平台上设计了基于 LM 神经网络的 EPS 传感器故障诊断系统,且具有很强的兼容性。 最后,通过台架实验得到相关故障模式数据作为样本数据,对设计的传感器故障诊断系统进行试验论证。结果表明 LM 神经网络收敛快,其精度和准确度达到了正确诊断的要求,可以用于 EPS 传感器的实时故障诊断。 本文设计的 EPS 传感器故障诊断系统,切实有效,对其他传感器的故障诊断具有指导意义,有很大的实用价值。 关键词 :电动助力转向,故障诊断,故障树,失效模式与影响分析,神经网络, Matlab ABSTRACT The electric power steering system has been developing rapidly in recent years. As more and more widely applied in the automobile industry, people put forward higher requirements to the safety and reliability of the EPS system. As a new technology, the EPS is provided the impetus by an electric motor. Compared with the hydraulic power steering, the EPS existences of different failure modes. Because the EPS is the most important component in the auto, so that the sensors' failure research and their fault diagnosis technologies in EPS ,are particularly important. This thesis researches with the faults of the torque sensor and speed sensor , which are the most important components in the EPS, and has established a sensor fault diagnosis system , which is based on the LM neural network. The principal contents studied in this thesis are as follows: First, based on in-depth analysis of the structure and working principles of the torque sensor and speed sensor, the thesis constructs the sensor fault tree models and identifies the causes of the malfunction, which are with the application of fault tree analysis. And then through the failure mode and effects analysis determines the several importances of the sensor failure, which can identify the optimal diagnostic mode of the sensor fault diagnosis. Second, with the research on the the artificial neural network and the causes and characteristics of the sensor failure, the thesis designs a EPS sensor fault diagnosis system on the Matlab platform, which is based on the LM neural network and has a strong compatibility. Finally, demonstrate the sensor fault diagnosis system with the sample data of failure modes from the bench test. The results show that the LM neural network convergence fast, and its precision and accuracy can meet the requirements of the correct diagnosis. So that it can be used for real-time fault diagnosis of EPS sensors. The design of the EPS sensor fault diagnosis system is effectively .It has a guiding significance for the other sensor fault diagnosis and great practical value. Keywords : Electric power steering, fault diagnosis, fault tree, failure mode and effects analysis, neural networks, Matlab