【作者】 张卫华；

【导师】 吴重光；

【作者基本信息】 北京化工大学 ， 控制理论与控制工程， 2010， 博士

【摘要】 石油化工生产过程由于传感器漂移、设备失效、工艺波动或操作错误等原因,导致生产运行中经常出现异常工况状态,轻则影响产品质量、生产调度计划,重则会出现生产事故,造成人员伤亡和巨大的经济损失。如何及时识别出存在的异常工况,找出导致异常工况产生的原因,预测该异常工况发展下去可能产生的后果,提出相应的措施,进行有效的异常工况管理,避免产生严重后果,确保石化生产装置的长期稳定运行是当前国际上研究的热点。本文在国家863课题“定性定量复杂故障诊断技术研究”(2009AA04Z133)的支持下,在前人研究成果的基础上,研究基于符号有向图SDG (Signed Directed Graph)方法,结合专家规则、模糊逻辑、主元分析等定性定量故障诊断方法,探索建立异常工况管理系统的实现方法和评测手段。异常工况管理系统是基于计算机系统的自动监测、推理、预测和指导为一体的集成系统,它有四个层次的目标：v保持生产运行正常；v当生产发生工艺波动时,能够在异常工况管理系统的指导下使工艺操作恢复到正常状态；v当生产发生较大偏离时,能够在异常工况管理系统的指导下使工艺状态处在一个安全的状态；v当生产发生紧急状况或事故时,最小化事件的严重性影响。为了达到上述目标,本文开发了数据处理、状态识别、原因诊断、可能后果和处理措施功能模块,结合数据采集模块和专家知识管理模块,做到：v在异常工况发生时,能实时、快速地对问题进行本源定位、后果评估、方案提出以及方案的再论证；v具有自学习功能,能够使操作经验得到继承。对异常工况的处理过程也可以通过计算机重放,使管理层对操作处理的审查有所依据。操作工也可以利用计算机进行历史事故的学习,有利于经验在员工中的移植。v历年积累的安全信息,经过数据挖掘,可以为管理层提供事故类型、故障源分布、事故概率等等统计信息,为安全管理工作提供决策支持。异常工况管理系统的核心指标是推理引擎的推理速度的快捷和推理结果的准确。推理速度的快捷取决于推理引擎算法合理性,推理结果的准确性依赖于所建立的模型的准确性。为了验证模型的准确性,本文对模型的校验和验证方法和手段进行了初步探索。本文的主要工作和取得的成果如下：1、进行了基于SDG的多种故障诊断方法融合的技术研究；2、基于Visual Studio,开发了异常工况管理系统软件平台；3、建立了定性定量检验和验证(Verification & Validation,V&V)试验平台；4、建立了用于常减压装置的异常工况管理系统；5、完成了常减压装置异常工况管理系统在定性定量V&V平台上的试验验证；6、完成了常减压装置异常工况管理系统在工业现场的应用测试。更多还原

【Abstract】 The petrochemical production process often emerge abnormal situation state due to sensor drift, equipment failure, operator error, process fluctuations or other reasons. It would cause product quality decline, production scheduling plan delayed, a serious accident occurs, even casualties and huge economic losses happens.How to timely identify the existence of abnormal situations, find the reasons leading to abnormal situations, predict the possible consequences, propose appropriate measures to assist operator to cope timely and correctly, avoiding serious effects appear, ensure that petrochemical production plant the long-term stable operation is the hot spots of current international study.Based on the results of predecessors research, supported by the national 863 project 2009AA04Z133 "Qualitative and Quantitative Study of Complex Fault Diagnosis", the research study mainly on signed directed graph (SDG, Signed Directed Graph) technology, combined with expert rules, fuzzy logic, principal component analysis etc. qualitative and quantitative fault diagnosis methods, explored the establishment of Abnormal Situation Management System implementation methods and evaluation tools.Abnormal Situation Management System is based on automatic monitoring, reasoning, forecasting and guidance as one integrated computer system, which has four levels of objectives:v keep production running normal;v when a volatility process occurs, can be resumed to normal state under the guidance of management system;v when a large deviation occurs, can enable process state in a safe state under the guidance of management system;v when an emergency situation or incident occurs, can minimize the seriousness of the impact the incident under the guidance of management systems.In order to achieve the above objectives, this paper had developed a state identification, the reasons for the diagnosis, possible consequences, and treatment measures etc. functional modules, combined with data acquisition module and expert knowledge management module, to:v when an abnormal situation occurs, in real time, quickly locate origin of the problem, assess the consequences, present the program and program argument;v having self-learning function, make the experience of how to cope with the abnormal situation to be inherited.v security information accumulated over the years, through data mining, can provide management with incident type, fault source distribution, incident probability, and so on the statistical information for security management to provide decision support.The core of Abnormal Situation Management System is the fast inference speed and accurate inference results of the reasoning inference engine. The fast inference speed depends on the rational algorithm of the inference engine, and the inference accuracy of the results depends on the accuracy of the model established. In order to verify the accuracy of the model, this paper maked a preliminary exploration of model validation and verification methods.The main work and achievements of this paper are as follows:1. Carried out a variety of SDG-based fault diagnosis method of fusion technology;2. Developed abnormal situation management system software platform which is based on Visual Studio;3. Established a qualitative and quantitative V&V (Verification & Validation) test platform;4. Established an Abnormal Situation Management System for CDU unit;5. Completed V&V of CDU Unit Abnormal Situation Management System in the qualitative and quantitative platform;6. Completed CDU Unit Abnormal Situation Management System application testing in the industrial field.更多还原