【作者】 董玉亮；

【导师】 杨昆； 顾煜炯；

【作者基本信息】 华北电力大学（北京） ， 热能工程， 2005， 博士

【摘要】 随着电力工业的迅速发展,长期以来发电设备采用的计划维修模式逐渐暴露出维修不足和维修过剩等问题,正在向先进的状态维修模式转变。发电设备状态维修的成功实施离不开其支持系统,但目前人们对支持系统的研究主要集中在监测与诊断系统和维修管理系统,因而研究利用电站各种信息将监测诊断系统与维修管理系统连接起来的运行与维修决策支持系统具有重要意义和实用价值。 以可靠性为中心的维修(RCM)不仅可根据设备的故障模式确定针对性的维修策略,而且充分考虑设备的性能、维修经济性和维修策略之间的关系,越来越成为复杂系统维修中应用的首选维修分析方法。但是传统RCM在设备的重要度分析、状态评价、维修优化等方面缺少定量化的工具,因而降低了维修分析的效率和精确性。本文提出了一种改进的RCM分析方法,确定了发电设备运行与维修决策过程,并着重研究了决策过程中的设备重要度分析、状态评价及预测、维修决策及优化等关键技术。 针对现有设备重要度分析受主观因素影响过多而不准确的问题,分析了影响发电设备重要度的关键因素,提出了基于蒙特卡罗模拟的设备重要度分析方法,并根据重要度判据将设备分类,建立了各类设备维修方式决策规则。针对复杂设备提出了故障模式及影响分析(FMEA)与灰色理论相结合的定量故障风险分析模型,提高了风险分析的精度,为状态评价的特征参数提取提供依据。 提出了多状态特征参数变权模糊综合状态评价模型,充分利用设备在线和离线监测诊断数据、运行实时数据、可靠性分析数据、设备寿命评价数据、运行与维修历史数据等信息,使状态评价结果更加贴近设备实际运行状态。变权综合理论的引入,大大提高了状态综合评价的精度。提出了设备运行状态综合预测模型,神经网络和灰色理论的组合应用,提高了状态预测的准确性。对于具有较多状态特征参数的系统,提出了基于主成分分析(PCA)的神经网络状态预测模型,通过降低神经网络的输入维数,提高了预测的速度和精度。 利用维修历史数据和状态评价及预测结果,建立了各类设备的维修任务决策及优化模型,并给出了求解方法。对具有很强关联性的设备组(即复杂可修复系统),提出了短期维修决策模型,该模型充分考虑了设备的技术状态、可靠性、维修费用和设备大小修计划等因素,实现了整个系统的全局维修优化。 以Powerbuilder 9.0为基本开发工具,综合利用数据库、多线程、数据接口等技术设计并开发了发电设备运行与维修决策支持系统。作为一通用平台,该系统集状态监测、故障诊断和维修决策为一体,通过选择对象和模型实现了各类设备的运行与维修决策。更多还原

【Abstract】 With the rapid development of electric power, the problem of maintenance deficiency and surplus emerges in the schedule maintenance mode, which is adopted for power plant equipment in the past. So, it is turning into advanced condition base maintenance mode. The successful implementing of condition based maintenance for power plant equipment needs its support systems. Now, the researches of the support systems are focused on monitoring and diagnosis system (MDS) and computerized maintenance management system (CMMS). It is of great significance to research on operation and maintenance decision support system, which connects MDS with CMMS using useful information in power plant.In reliability-centered maintenance (RCM), maintenance strategies are made not only through analyzing failure mode but also the relationship of performance, economic and maintenance strategy. RCM becomes a more and more popular maintenance analysis method for complex system. But, in traditional RCM, the efficiency and accuracy of maintenance analysis are degraded for the absence of quantificational tools for equipment importance analysis, condition evaluation and maintenance optimization etc. Based on such a consideration, a streamlined RCM analysis method is put forward and the decision process of operation and maintenance for equipment in power plant is determined. Then some key technologies of equipment important analysis, condition evaluation and prediction, maintenance decision and optimization are studied.Aiming at the imprecision of existing equipment important analysis methods resulted from using too many subjective factors, the pivotal factors effecting equipment importance are analyzed, and an importance analyzing method based on Monte Carlo simulation is provided. Then, equipment is classified according to importance criterion, and the rules of maintenance mode decision are established. For complex equipment, a quantificational failure criticality analysis model based on failure mode and effect analysis (FMEA) and grey theory is put forward. The precision of criticality analysis is improved, and can be a support for characteristic parameters extraction in condition evaluation.A multi-parameters synthetic condition evaluation model based on variable weight and fuzzy theory is used. In the model, the on-line and off-line monitoring and diagnosis data, operation real-time data, reliability analysis data, life assessment data and the history data of operation and maintenance are fully used, the evaluation result is closer to the real condition. A synthetic condition prediction model is presented, using neural network and grey theory together make it possible to predict accurately. For the system with many characteristic parameters, a neural network condition prediction model based on principle component analysis (PCA) is studied. Prediction speed and precision are improved by decreasing the dimensions of neural network input.For maintenance decision and optimization, some maintenance task decision and optimization models using maintenance history data and results of condition evaluation and prediction are introduced, and their solutions provided. A short-term maintenance decision method is given for complex repairable system. By fully considering technique更多还原