论文部分内容阅读
复杂工业过程数据通常具有非高斯性和强非线性特征,为此提出了一种基于核独立成分分析和支持向量数据描述(KICA-SVDD)的非高斯非线性系统的故障检测方法.该方法首先运用核独立成分分析方法对数据进行特征提取,然后通过引入支持向量数据描述对独立主元成分进行建模,并计算相应的统计量及控制限,实现非高斯非线性系统下的故障检测.最后在Tennessee-Eastman(TE)过程上进行了仿真实验,结果表明所提出的方法降低了故障错分比例和漏检比例,验证了其可行性和有效性.
In order to solve this problem, a method based on KICA-SVDD for fault detection of non-Gaussian nonlinear systems is proposed in this paper.Firstly, the fault detection method of non-Gaussian nonlinear systems based on KICA- The method of kernel independent component analysis is used to extract the features of the data, then the independent principal components are modeled by introducing support vector data description, and the corresponding statistic and control limits are calculated to realize the fault detection in non-Gaussian nonlinear systems. The simulation experiments on the Tennessee-Eastman (TE) process show that the proposed method reduces the proportion of misclassification and the ratio of missed inspections, and verifies its feasibility and effectiveness.