节点文献
面向故障过程的多设设可靠性分析与维修决策
Reliability Analysis and Maintenance Decision-making Method of Multi-machine for Failure Process
【作者】 刘繁茂；
【作者基本信息】 华中科技大学 ， 工专工程， 2011， 博士
【摘要】 车间制造系统的稳定性是保证产品质量和快速响应市场的关键性工程问题之一。其中,设备的稳定性是制造系统稳定性的基础,而准确的制造设备服役可靠性分析和合理的维修计划是保持设备稳定运行的关键因素。本文从设备故障数据出发,建立了多设备可靠性分析和维修决策的统一框架,在综述了目前单设备与多设备可靠性分析、维修决策模型及其相关领域的理论方法、模型和研究现状的基础上,提出了若干针对多设备可靠性分析和维修决策的新方法,具体包括:（1）分析了设备故障数据的特点,指出其具有重复测量数据的特性。综合考虑了同型设备之间的异质性和重复测量之间的自相关性,建立了针对多设备可靠性分析的广义线性混合效应模型,克服了传统单设备可靠性模型的不足。详细论述了模型参数估计计算理论和方法,拟合优度检验的新方法。模型之优点在于不仅可以得到单个设备的可靠性变动趋势,而且还可以得到总体的可靠性变动趋势。（2）分析了维修效应和劣化趋势对设备可靠性的冲击影响,提出了综合比例强度和非线性混合效应的多设备可靠性分析集成模型。详细讨论了模型参数估计计算理论和方法。与传统模型相比,该模型融合了设备相关工程信息,不仅可以对维修效果进行定量评估,而且还可以得到单设备和总体的相关可靠性指标之封闭解。（3）分析了选择性维修方法在解决生产线多设备维修决策问题的优势。首次将ARI1和ARI∞两种算术回退故障强度模型和选择性维修方法相结合,给出了设备在小修、预防维修、大修等不同维修方式下的故障强度的演化规则,以此为基础,建立了基于约束规划的多设备选择性维修决策模型,并给出了结合贪婪规则和遗传算法的求解算法和相关工程实例。（4）分析了引入役龄回退因子和故障率因子的复合故障率计算方法,首次将复合故障率模型和多设备维修选择优化方法相结合。从与故障强度不同的视角,定义了设备在小修、预防维修、大修等不同维修方式下的故障率演化规则。以此为基础,建立了基于约束规划的多设备选择性维修决策模型,并给出了结合结构启发式规则和禁忌搜索算法的求解算法和相关工程实例。（5）在上述研究基础上,开发了汽车发动机车间生产线维修决策分析支持的原型系统。该系统基于工业以太网络运行,采用流行的软件架构,具有良好的扩展性和跨平台性。详细介绍了该系统的工业以太网络体系,相关功能模块及运行实例。最后,对全文进行总结,指出进一步的研究方向。
【Abstract】 The stabilization of enterprise manufacturing system is one of the key problems to restrict the quality of production and the rapid response to market. The stabilization of equipment is the basis of manufacturing system’s stabilization, and its two key factors are the reliable analysis of operational reliability of equipment and the suitable maintenance plan of equipment. Based on the failure data of machine, a unified framework for reliability analysis and maintenance-decision of multi-machine is proposed firstly by this dissertation. Firstly, this thesis provides a systematic literature review on the theories, methods and research situations for reliability analysis and maintenance decision model about single machine and manufacturing system and its related fields. Then, several new methods addressing how to make the reliability analysis and maintenance-decision of multi-machine are implemented as follows:（1）The characteristic of machine failure data is illuminated, which is similar to the characteristic of repeated measurement data and recurrent event data. GLMM model used widely in medical and biological field are proposed. Unlike the classical single machine reliability model, this model has ability to handle the heterogeneity of same type equipment and the self-correlation of repeated measurement. Furthermore, the theory and computational method of parameter estimation and a new method of goodness-of-fit are discussed in detail. In general, this model can provide the reliability changing trend for both the population and each individual machine.（2）Similar to the above linear model, reliability analysis model for multiple machine is built based on proportion intensity model and nonlinear mixed-effects model, which takes account of the influence to the reliability of machine caused by the maintenance effect and deterioration level. Furthermore, the theory and computational method of parameter estimation are discussed in detail. Specially, related engineering information of equipment is considered by this model; Not only quantitative assessment of maintenance effect but also analytical solution of related reliability index for both the population and each individual machine is provided by this model.（3）It is brought forward that selective maintenance decision model has the advantage of solving the production line maintenance decision-making problem in engineering. ARI1 and ARI∞model, two kinds of arithmetic reduction of intensity model, are combined with selective maintenance method for the first time. The influence on machine failure intensity by different maintenance actions （PM, minimal repair, overhaul） is described. Furthermore, a maintenance decision-making model for multi-machine system is built based on constraint programming and a resolution algorithm is provided by combining the greedy heuristic rules with genetic algorithm. Finally, a case study is provided.（4）The computational method of mixed failure rate base on age reduction factor and hazard rate increase factor is discussed. The mixed failure rate model is combined with selective maintenance optimizing method for the first time.Then, the influence on machine failure rate by different maintenance actions （PM, minimal repair, overhaul） is provided. Furthermore, a maintenance decision-making model for mutil-machine system is built based on constraint programming and a resolution algorithm is provided by combining the heuristic rules with tabu search. Finally, a case study is provided.（5）A maintenance decision-making analysis system of machining line in car engine plant is designed and developed based on the research founding above. Industrial Ethernet network is the hardware base of this system, and popular architecture is adopted as the software architecture of this system. The popular architecture makes it has good extension and across platform. Its Ethernet network, software architecture, function modules and some running case are presented in detail.Finally, a conclusion is drawn and the trend on maintenance decision-making model of manufacturing system is anticipated.