【作者】 谭树彬；

【导师】 徐心和；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 故障诊断是近年来发展起来的一门综合性的边缘学科，它涉及到的理论基础和技术手段十分广泛。随着科学技术和经济建设的发展，大型设备与控制系统的状态监测和故障诊断已成为保证系统运行稳定性、可靠性和安全性，提高产品质量和生产效率的关键技术和重要手段，并日益引起国内外学者的广泛重视。 轧机厚度自动控制系统(AGC)是现代板带轧机实现高精度轧制必不可少的关键环节。该系统是一复杂的综合控制系统，包括机械、液压、电气等方面的综合信息。因此，除了机械设备中所有可能发生的故障在该系统中都可能发生之外，厚控系统还具有液压系统特有的失效形式。故该系统的故障率较高且故障原因复杂，是维护轧机的重点和难点，也是造成故障停产和产品质量下降的主要原因。 本文就是基于此背景下所进行的研究工作，以某轧机厚控系统为对象，以故障诊断技术的实际应用为目的，主要开展了以下研究： 首先，筛选出了用于故障诊断的26种特征信号，并建立起信号的采集、处理和存储的状态监测系统； 然后，首次提出了包括从设备硬件到控制系统、直到最终产品检验三个层次的故障诊断策略； 其次，提出了基于轧机解折模型的故障诊断方案，建立了轧机厚控系统的数学模型，设计出了适合于该轧机厚控系统的故障检测观测器，实现了几类典型故障的检测与诊断； 提出了基于规则和案例的故障诊断方案，归纳出用于故障诊断具有代表性的18条规则和10个案例，并给出了具体的实现方法和步骤； 接着，基于神经网络的故障诊断思想，对该系统进行了功能上的分解，设计出了用于故障诊断的BP网络结构和参数，并就具体的应用实例进一步说明了基于神经网络诊断方法的可行性； 推导出了变刚度控制的控制模型，用仿真的方法对该模型进行了验证，并综合三个方面的效果，提出了变刚度的合理取值范围。然后，结合生产实践，在对轧机厚控系统机理分析的基础上，实现了对其控制系统几类典型故障的检测与诊断；更多还原

【Abstract】 Condition Monitoring and Fault Diagnosis (CMFD), involved many theories and technical methods, is a comprehensive edge subject, which has been developed in recent years. Along with the development of sciences, technologies and economic, condition monitoring and fault diagnosis of large-scale equipments and their control systems have become an essential technology and important method to guarantee equipments or system to work stably, reliably and safely, to improve the product quality and the production efficiency. Now, CMFD is attached importance widely by many domestic and foreign scholars day by day.Automatic gauge control (AGC) system is the key control system of strip rolling mills to realize high accuracy of strip exit thickness. AGC system is a complex integrated control system, including mechanical, hydraulic and electrical comprehensive information. As a result, not only faults in mechanical devices but also those in hydraulic system could happen possibly in the AGC system. Therefore, failure rate of AGC system always is high, and its fault reasons are always complex, which is the key and difficult point to maintain strip rolling mills, and is also the main reason causing strip rolling mills to break down and product quality to become poor.Based on the study background described above, this dissertation focuses on the key technologies and methods of CMFD for the AGC system of a strip rolling mill. Taken the practical application as the goal, main work below has been studied:First, we picked out 26 kinds of characteristic signals used in the fault diagnosis, and established a condition monitoring system of the strip rolling mill for signals gathering, processing and memory.Then, the fault diagnosis idea including three levels, namely equipment hardware, control system and finished product, was proposed for the first time.Next, the fault diagnosis idea based on system model was brought forward, and the mathematical model of AGC system in the strip rolling mill was established. Two suitable fault detection observers for this AGC system were designed, and several kinds of typical fault detection and diagnosis were realized.The fault diagnosis idea based on rule and case was proposed. On the same time, 18 fault更多还原