【作者】 王英；

【导师】 方淑芬； 王文彬；

【作者基本信息】 哈尔滨工业大学 ， 管理科学与工程， 2007， 博士

【摘要】 现代生产设备技术含量高、结构复杂、系统特性强,其故障不但表现为很强的随机性,而且故障损失严重。传统的维修理念和方法受到严峻挑战,而状态维修作为一种更科学、更先进、更有效的维修理念和维修方式日益受到学术界和企业界的密切关注,并成为当前维修领域研究的热点问题。状态维修系统结构框架的构建有助于指导状态维修的研究和发展,完善和拓展状态维修的研究内容。状态维修决策作为状态维修系统的重要组成部分,是状态维修工作不可忽视的重要内容。对于状态维修系统结构的设计以及状态维修决策过程的深入研究具有非常重要的意义和实用价值。本文在梳理和分析状态维修相关研究文献的基础上,界定了状态维修的内涵,分析了状态维修的特点,并根据状态维修系统应实现的一系列功能和实际需要,构建了状态维修系统结构框架。根据状态维修系统结构框架,给出了状态维修的工作内容,提出了状态维修工作过程的三个阶段和决策过程的关键环节。此外,根据我国企业设备维修管理的现状,提出了状态维修实施的技术支持和管理保障基础,指出了企业开展状态维修的基本原则,并详细阐述了企业实施状态维修的工作流程。本文对状态维修决策过程进行了系统深入的研究和分析。首先,对缺陷状态进行早期识别。在分析设备运行特点的基础上,针对传统方法存在的缺陷和不足,分两种情况阐述了识别缺陷发生时刻的建模方法,重点探讨了测量信号为单值的情况,针对测量信号为多维向量的情况,提出了利用主成分分析方法进行特征量提取,提取出的主成分作为建模的输入数据,并建立了多维主成分与被监测对象状态之间的关系。同时,应用EM算法对模型中的未知参数进行估计,为了提高算法的收敛速度和估计的准确性,对EM算法进行了相应的改进。此外,对缺陷发生时刻的识别方法和过程进行了计算机仿真研究,验证了该方法的有效性和准确性。其次,对缺陷状态的劣化程度进行预知。在分析缺陷状态的可预测性和当前状态预知建模方法不足的基础上,阐述了基于随机滤波理论的残余寿命预知建模方法和过程,并在原有建模方法的基础上进行了重要改进。通过假设被监测设备在正常运行阶段和缺陷运行阶段的状态监测信号服从尺度参数存在一定关系、形状参数不同的两参数分布,而将一阶段的状态预知模型扩展到两阶段的状态预知模型。同时,为提高估计的有效性,减少估计误差,提出了在故障数据较少的情况下,利用极大似然方法进行模型参数估计的方法和过程。此外,对残余寿命预知建模方法和过程进行了计算机仿真研究,验证了该建模方法和参数估计方法的有效性和准确性。最后,在识别缺陷发生时刻、预知缺陷状态劣化程度的基础上进行决策优化,建立了维修行为决策优化模型,并给出模型的求解方法,同时运用计算机仿真方法,建立了两阶段的状态监测间隔期决策优化仿真模型,并给出主要的仿真步骤和过程。论文通过案例研究对状态维修决策过程相关模型的可行性和有效性进行了检验和分析。案例结果表明,本文所提出的状态维修决策过程和相关决策模型对于状态维修决策实践有较好的指导意义,使维修决策更具有动态性和科学性。更多还原

【Abstract】 Modern production equipment is high technology, complex in structure and has strong system performance, and its failure not only displays the strong randomness, but also the loss is very serious. The traditional maintenance ideas and the methods have received the stern challenge, and the condition-based maintenance (CBM) that is more scientific, advanced and effective maintenance idea and way receives the academic circles and enterprise’s close attention day by day, and becomes the current hot research topic in maintenance field.Establishing the structural frame of CBM system is helpful to construct the research and development of CBM and to perfect and extend research content of CBM. The decision-making as an important component of CBM system is noticeable important content in CBM work. The structure design of CBM system and thorough research of CBM decision-making process have very important significance and the practical value.The article defines the connotation of CBM, analyzes the characteristics of CBM based on the combing and analysis of research literature of CBM, and has constructed the basic structure frame of CBM system according to a series of functions that CBM system should implement and the need in reality. According to the basic structural frame of CBM system, work content of CBM is given, and three stages of CBM work process and key taches of decision-making process are proposed. According to the present situation of equipment maintenance management of enterprises in our country, The foundations of technology support and management guarantee in implementing CBM are proposed, and the basic principle of implementing CBM is pointed out, and the work process of implementing CBM is elaborated in detail.The article studies and analyzes the decision-making process of CBM systematically and thoroughly.First, the early identification of the defect state is carried on. Based on analysis of the equipment operating characteristics and in view of the deficiencies of the traditional methods, Modeling method of identifying the initiation time of defect is elaborated from two situations, and the situation in which the measured signal is single value is emphasized. Aiming to the situation in which the measured signal is a multi-dimensional vector, using principal component analysis method to draw the characteristic quantity that is data input of modeling is proposed, and the relationship between the multi-dimensional principal components and the state of monitored equipment is established. At the same time, EM algorithm is applied to estimate unknown parameters in the model. In order to enhance the convergence rate of EM algorithm and the accuracy of estimate, the corresponding improvement to the EM algorithm is made. In addition, the computer simulation research for the method and process of identifying the staring time of defect is conducted, and the validity and the accuracy of the method are testified.Next, deterioration degree prediction of defect state is carried on. Based on analysis of the foreseeability of defect state and of the deficiency of the current state prediction modeling methods, the residual life prediction modeling method based on the stochastic filtering theory is elaborated, and bigger improvement is done. One–stage prediction model is extended to two-stage prediction model by supposing the condition monitoring signal in the normal and defective operating stages of the monitored equipment be subject to the distribution of two parameters where the scale parameters have some relationship and the shape parameters are different. At the same time, in order to improve estimate validity and reduce estimate error, the method and the process of model parameter estimate using the maximum likelihood method in the situation of less failure data are proposed. In addition, the computer simulation research for the method and process of the residual life prediction is conducted, and the validity and the accuracy of the method are testified.Finally, based on the identification of defect starting time and prediction of defect state deterioration degree, the decision optimization is made. The maintenance action decision optimization model is established and the solution method is given. Meanwhile two-stage condition monitoring interval optimization simulation model is built using simulation method, and the main simulation step and process are given. Case study testifies and analyzes the feasibility and validity of models related to CBM decision-making process. The result shows that the decision-making process and models can direct the practice of CBM decision very well and make the decision more dynamic and scientific.更多还原