【作者】 王珏；

【导师】 王印松；

【作者基本信息】 华北电力大学 ， 系统工程， 2020， 硕士

【摘要】 随着现代化水平的不断提高,控制系统日趋大型化和复杂化,这就对它的安全性和可靠性提出了更高的要求,因此控制系统的故障诊断技术受到了越来越广泛的关注。并且,由于现代传感技术的飞速发展及计算机水平的不断提高,大量的过程数据被记录存储下来,如何提取这些数据中的有效信息已经成为故障诊断领域的重要研究方向。在这种情况下,慢特征分析作为一种特征提取方法应运而生,它能够提取时序信号中变化最缓慢的成分,有效表示系统的固有属性。本文将慢特征分析法引入到故障诊断技术中,并以控制系统为研究对象进行了相关研究。首先,在对慢特征分析算法的思想和原理进行相关介绍后,研究了完整的基于慢特征分析的故障诊断模型:利用经过慢特征分析提取后的数据,构建能够反映系统运行状态的监控统计量,实现故障的检测;检测到故障后,通过贡献图法计算每个变量对统计量的贡献率实现故障的定位。其次,针对慢特征分析无法良好地处理非线性数据的问题,引入了高斯核函数对算法进行了非线性扩展,并将核慢特征分析法应用到控制系统的故障诊断中。提出了基于核慢特征分析的监控统计量法用于故障的检测;由于核方法无法找到高维特征空间到原始空间的逆映射函数,引入了核样本等效替换思想,通过揭示核矩阵与输入矩阵之间的关系,间接地确定输入变量与监控统计量间的对应关系,实现故障的定位。仿真结果表明,该方法与传统的慢特征分析算法相比,在处理非线性过程数据方面具有优越性,并且核样本等效替换思想在保证计算精度的同时,大大提高了诊断效率。最后,为实现控制系统执行器故障类型的精确辨识,以调节阀为研究对象,提出了基于特征指标信息融合的故障诊断方法。利用调节阀输入输出变量间的关系,提出了几种能够反映调节阀不同故障特点的特征指标,并利用D-S证据理论对各个指标进行信息融合,得到最终的诊断结果,最后通过实验验证了该方法的有效性与优越性。更多还原

【Abstract】 With the continuous improvement of modernization level,the control system is becoming more scale and more complex,which puts forward higher requirements for its safety and reliability,therefore,the fault diagnosis technology of the control system has received more and more attention.In addition,due to the rapid development of modern sensing technology and the continuous improvement of computer level,a large amount of process data is recorded and stored,how to extract valid information from these data has become an important research direction in the field of fault diagnosis.In this case,slow feature analysis appeared which acts as a feature extraction method,it can extract the slowest component of the time series signal and effectively represent the inherent properties of the system.This paper introduces the slow feature analysis method to the fault diagnosis technology,and related researches have been conducted with the control system as research object.Firstly,the paper introduces the idea and principle of slow feature analysis algorithm,and the complete fault diagnosis model based on slow feature analysis is studied: using the data extracted by slow feature analysis to construct monitoring statistics which can reflect the operating status of the system to achieve fault detection;if the fault occurs,the contribution graph method is used by calculating the contribution rate of each variable to statistics to realize the fault location.Secondly,In order to solve the problem that slow feature analysis cannot handle the nonlinear data well,the Gaussian kernel function is introduced to nonlinear expansion of the algorithm,and the kernel slow feature analysis method is applied to the fault diagnosis of the control system.The monitoring statistic method based on kernel slow feature analysis is proposed for fault detection,and because the kernel method can not find the inverse mapping function from high-dimensional feature space to original space,the idea of kernel sample equivalent replacement is introduced to realize the fault identification,which can find the correspondence between input variables and the monitoring statistics indirectly by revealing the relationship between the kernel matrix and the input variable matrix.The simulation results show that this method is superior to the traditional slow feature analysis in dealing with nonlinear process data,and the kernel sample equivalent replacement idea not only guarantees the calculation accuracy,but also greatly improves the diagnosis efficiency.Finally,In order to accurately identify the fault type of the actuator of control system,taking control valve as the research object,a fault diagnosis method based on information fusion of characteristic index is proposed.According to the relationship between the input and output variables of control valve,several characteristics indicators reflecting control valve’s different fault phenomena have been found;and then,the D-S evidence theory is used to integrate each index to get the final diagnosis result,and finally the effectiveness and superiority of the method were verified by experiments.更多还原