基于数据驱动的仪表故障检测

发布时间：2018-10-13 09:07

【摘要】：随着半导体技术、制造工艺、通讯技术和网络技术的飞速发展,现代工业生产自动化水平日益提高,控制系统复杂程度随之加大。生产过程中的各仪表测量值是否能准确反应生产过程的状态,对于工业生产过程的安全性,控制系统的可靠性,保证工业产品质量有着至关重要的作用。工业仪表作为控制系统的关键部件之一,由于其材料工艺,制造生产技术以及工作环境等因素,工业仪表在整个控制系统中较容易发生故障。快速、准确检测出仪表故障并采取相关的正确策略,是保证控制系统稳定运行、消除生产安全隐患的关键,具有十分重要的意义。本文针对生产工业过程中多仪表的具体情况,采用基于数据驱动的故障检测方法,实现了对生产工业过程中多仪表的故障检测,并将所研究方法应用于Tennessee Eastman仿真平台,模拟不同的仪表故障信号,对检测结果进行对比分析。本文的主要工作如下： 1)针对工业生产数据并不符合高斯分布这一特殊的数据分布情况,并结合生产过程中多仪表的状况,引入了基于独立元分析的故障检测方法。通过对历史正常数据进行分离提取源信息,建立故障检测模型,并通过核密度估计的方法确定控制限,实现了多仪表故障检测。通过与基于主元分析的故障检测方法进行比较分析,验证了基于独立元分析的故障检测方法更适于工业生产过程中的仪表故障检测。 2)针对工业生产数据中存在的高斯源信息和非高斯源信息,以主元分析法和独立元分析法为主要理论依据,通过对工业生产过程数据不同信息的提取和分离,分别采用相应的分析方法,建立不同的故障检测模型。仿真结果表明,相较于采用单一的基于独立元分析的故障检测方法,将独立元分析法和主元分析方法结合使用具有更好的检测效果。 3)在基于贡献度的独立元子空间理论方法基础上,对子空间的故障检测模型进行完善。在仪表微小故障难以检测的问题上加以应用,根据不同的实际需求提供相应的故障集成检测策略。仿真结果基于贡献度的独立元子空间方法和完善的故障检测模型提高了仪表微小故障的检测效果,不同的集成策略更加具有灵活性和应用性。本文通过引入不同的故障检测理论方法对工业生产过程多仪表情况下的故障进行检测,对检测结果进行分析探讨。可以对工业过程多仪表的故障检测提供新思路。
[Abstract]:With the rapid development of semiconductor technology, manufacturing technology, communication technology and network technology, the automation level of modern industrial production has been improved day by day, and the complexity of control system has increased. It is very important for the safety of the industrial production process, the reliability of the control system and the quality of the industrial products whether the measured values of each instrument in the production process can accurately reflect the state of the production process. As one of the key components of the control system, industrial instruments are prone to malfunction in the whole control system due to the factors such as material technology, manufacturing technology and working environment. It is of great significance to detect the instrument failure quickly and accurately and adopt the correct strategy to ensure the stable operation of the control system and eliminate the hidden trouble of production safety. In this paper, according to the specific situation of multi-instrument in the process of production industry, the fault detection method based on data drive is adopted to realize the fault detection of multi-instrument in the process of production industry, and the research method is applied to the simulation platform of Tennessee Eastman. The fault signals of different instruments are simulated and the test results are compared and analyzed. The main work of this paper is as follows: 1) aiming at the fact that the industrial production data do not accord with the special data distribution of Gao Si, and considering the situation of multiple instruments in the production process, a fault detection method based on independent element analysis is introduced. Through separating and extracting the source information from the historical normal data, the fault detection model is established, and the control limit is determined by the method of kernel density estimation, which realizes the fault detection of multiple instruments. Compared with the fault detection method based on principal component analysis, It is verified that the fault detection method based on independent element analysis is more suitable for instrument fault detection in industrial production process. 2) aiming at Gao Si source information and non-Gao Si source information in industrial production data, Based on principal component analysis (PCA) and independent component analysis (ICA), different fault detection models were established by extracting and separating different information from industrial process data. The simulation results show that, compared with the single fault detection method based on independent element analysis, The combination of independent component analysis and principal component analysis has better detection effect. 3) the fault detection model of subspace is improved on the basis of independent subspace theory and method based on contribution degree. It is applied to the problem that it is difficult to detect the small fault of the instrument, and the corresponding integrated detection strategy is provided according to the actual demand. The simulation results show that the independent subspace method based on the contribution degree and the perfect fault detection model can improve the detection effect of the instrument micro-fault, and the different integration strategies are more flexible and applicable. In this paper, different fault detection theories and methods are introduced to detect the faults under the condition of multiple instruments in industrial production process, and the detection results are analyzed and discussed. It can provide a new idea for the fault detection of multiple instruments in industrial process.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TH165.3

【参考文献】