设备突发大故障的自组织临界态风险度量及预警研究

【作者】 李斌；

【导师】 李波；

【作者基本信息】 电子科技大学 ， 模式识别与智能系统（专业学位）， 2012， 硕士

【摘要】 论文基于的研究项目是来源于2007年国家自然科学基金面上项目“面向动态不确定性的计划产能预测与风险决策研究”(批准号：70701007),以及2010年国家自然科学基金面上项目“设备突发大故障的自组织临界态辨识与风险度量研究”(批准号：51075060)。前者通过对产能规划过程中的不确定因素的研究进行计划产能的预测,从而进行过计划与欠计划的风险评估与控制；后者主要研究突发大故障设备的自组织临界态的判断、辨识和风险度量。本文对前一个自然基金在风险研究成果的基础上进一步深入研究,以解决临界突发大故障设备的风险度量问题。基于半导体制造企业的设备自组织临界态辨识,对已识别的设备自组织临界态风险因子进行分析,从而对临界突发大故障设备进行风险度量。本文将设备突发大故障特征研究的基础上,提出设备突发大故障符合自组织临界性机制的假设,并从定性分析和定量研究两个角度验证了这一假设。将金融领域中的投资风险度量理论CVaR风险度量模型引入到设备突发大故障的风险度量领域,并验证了其可靠性和有效性。在研究设备突发大故障的风险识别时,发现了突发事件链(Emergency Chain)分析法和故障树分析法(FTA)在动力学机制方面的内在关联,基于此提出了基于以上两种方法优势互补的EC-FTA风险因子识别和重要度排序的方法。并运用EC-FTA分析法对某半导体生产制造企业进行了风险因子识别与重要度排序的实证研究。对基于正态分布的CVaR风险度量算法进行改进,得到符合分形分布的CVaR风险度量算法,并对设备突发大故障实例进行风险度量建模和实证研究,验证了此改进算法的可靠性和优越性。针对性提出了设备突发大故障风险过程控制和决策模型,同时引入BP神经网络模型,进行了设备突发大故障的风险预警研究。更多还原

【Abstract】 The research of this thesis originates from two programs, one is the National Natural Science Foundation program (70701007) in2007--Project relate to the Dynamic Uncertain Planning Capacity Prediction and the Research of Risk Decision. The other is the National Natural Science Foundation Item (51075060) in2010--Self-organized critical state identification and research of risk measurement about equipment’s sudden large failure. The former aims to evaluate and control the risk based on excessive and insufficient plan through the anticipation of capacity planning, the planning, which is from the research of the uncertain factors during the capacity planning process. The latter is mainly to research the judgment, identification, and risk measurement of self-organized critical state about sudden faulted equipment. This thesis aims to carry out an in-depth research based on the risk measurement achievement of the former item, thus to solve the risk measurement problem of the critical big-sudden-faulted equipment. And paper is also based on the equipment’s self-organized critical state identification of the semiconductor manufacturing enterprise, to analysis the identified self-organized critical state risk factor of equipment, and thus to measure the risk of critical big-sudden-faulted equipment.The purpose of this thesis is to control the damages that caused by equipment’s sudden large failure in the random dynamic uncertain production environment. In order to control the disadvantage consequence during the productive process within customer’s acceptable range, the thesis needs to make risk measurement to self-organized critical state identification result of equipment’s sudden failure. So the paper based on the characteristic research of equipment’s sudden large failure, proposes the assumption of self-organized criticality mechanism that correspond with equipment’s sudden large failure, and has proved the assumption from the two aspects of qualitative analysis and quantitative research, and the thesis also introduces the investment risk measurement theory (CVaR risk measurement model) of the financial field into the risk measurement field of equipment’s sudden large failure, which is proved to be reliable and valid. Thesis discovered the internal relationship between Emergency Chain analysis method and FTA about the geodynamic mechanism when researched the risk identification of equipment’s sudden large failure, based on this and the two methods above, the paper proposed an advantage complementary method of EC-FTA risk factor identification and sequencing according the importance degree, and did the empirical research of this advantage complementary method by using the EC-FAT analysis method to the semiconductor manufacturing enterprise. It obtains the algorithms that are coincident with CVaR risk measurement of fractal distribution by improving the CVaR risk measurement algorithms of normal distribution, and makes risk metric modeling and empirical research of the equipment’s big-sudden-faulted case, and this improved algorithms also proved to be reliable and ascendant. This thesis proposed risk process controlling of equipment’s sudden large failure and decision model pertinently, and also did the research of risk warning on equipment’s sudden large failure with the introduction of BP neural network model at the same time.更多还原