基于多分类支持向量机的工业故障分类

发布时间：2019-06-28 17:43

【摘要】：本文中主要研究了复杂工业过程中的故障分类问题。当今,工业过程变得更加大规模化、复杂化、高耦合化。任何一个异常情况都可能被传播和放大,从而对整个工业生产过程造成不必要的财产的损失和人员伤亡。因此,复杂工业过程故障分类问题具有很强的实际意义。到目前为止,基于数据的故障检测和诊断的方法得到了很好的发展。如一些多元统计的方法被提了出来,这包括主成分分析(Principal Component Analysis,PCA)、独立成分分析(Independent Component Analysis,ICA)和支持向量机(Support Vector Machine,SVM)等。在高维数据分类时,太多的变量将会导致更高的计算量。且数据中所含的噪声也会降低分类的准确率。因此,数据降维就显得很重要了。目前,已经出现了许多数据降维的方法。例如在本文中使用的主成分分析、核主成分分析(Kernel Principal Component Analysis,KPCA)、独立成分分析和偏最小二乘(Partial Least Squares,PLS)的降维方法。本论文首先使用支持向量机和主成分分析支持向量机(Principal Component Analysis based Support Vector Machine,PCA-SVM)进行故障分类。由于主成分分析降维损失了分类准确率,进而使用核主成分分析支持向量机(Kernel Principal Component Analysis based Support Vector Machine,KPCA-SVM)进行故障分类以提高分类准确率。在核主成分分析降维过程中应用了核函数,引入了未知参数,使得计算过程变得复杂化。为了避免这一问题,使用独立成分分析支持向量机(Independent component analysis based support vector machine,ICA-SVM)进行故障分类。在研究中发现,基于主成分分析和核主成分分析的支持向量机对复合干扰引起的故障分类性能不佳。随后,使用偏最小二乘法支持向量机(Partial Least Squares based support vector machine,PLS-SVM)对此类故障进行分类。传统的偏最小二乘法类别编码方法不能很好的体现类别间的相关性,因此对于类别编码方法进行了改进,提出了基于改进偏最小二乘法支持向量机方法的故障分类,获得了较好的分类效果。
[Abstract]:In this paper, the problem of fault classification in complex industrial process is studied. Nowadays, the industrial process has become more large-scale, complex and highly coupled. Any abnormal situation may be spread and magnified, resulting in unnecessary loss of property and casualties throughout the industrial production process. Therefore, the problem of fault classification in complex industrial processes is of great practical significance. So far, data-based fault detection and diagnosis methods have been well developed. For example, some multivariate statistical methods have been proposed, including principal component analysis (Principal Component Analysis,PCA), independent component analysis (Independent Component Analysis,ICA) and support vector machine (Support Vector Machine,SVM). In the classification of high-dimensional data, too many variables will lead to higher computational complexity. And the noise contained in the data will also reduce the accuracy of classification. Therefore, data dimension reduction is very important. At present, there have been many methods of data dimension reduction. For example, the dimensionality reduction methods of principal component analysis, kernel principal component analysis (Kernel Principal Component Analysis,KPCA), independent component analysis and partial least squares (Partial Least Squares,PLS) are used in this paper. In this paper, support vector machine (SVM) and principal component analysis support vector machine (Principal Component Analysis based Support Vector Machine,PCA-SVM) are used for fault classification. Because the dimension reduction of principal component analysis loses the classification accuracy, the kernel principal component analysis support vector machine (Kernel Principal Component Analysis based Support Vector Machine,KPCA-SVM) is used for fault classification to improve the classification accuracy. The kernel function is applied to the dimension reduction process of kernel principal component analysis, and the unknown parameters are introduced, which complicates the calculation process. In order to avoid this problem, independent component analysis support vector machine (Independent component analysis based support vector machine,ICA-SVM) is used for fault classification. It is found that the support vector machine based on principal component analysis and kernel principal component analysis has poor performance in fault classification caused by compound interference. Then, the partial least square support vector machine (Partial Least Squares based support vector machine,PLS-SVM) is used to classify this kind of faults. The traditional partial least square class coding method can not reflect the correlation between categories well, so the category coding method is improved, and the fault classification based on improved partial least square support vector machine method is proposed, and good classification effect is obtained.
【学位授予单位】：渤海大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP18

【相似文献】