For complex and nonlinear chemical industry processes, the performance of fault detection is very poor when traditional principal component analysis (PCA) is used. Thus the concept of using kernel principal component analysis (KPCA) to conduct fault detection is proposed, which will make the complex nonlinear problem in input space convert into linear problem in feature space. This method is applied in Tennessee Eastman (TE) chemical industry process. The simulation result shows that KPCA method is obviously better than common PCA method in fault detection for complex nonlinear chemical process.%针对传统主元分析法(PCA)应用于复杂非线性的化工过程故障检测时存在性能差的问题,提出利用核主元分析法(KPCA)来进行故障检测的思想.从而将输入空间中复杂的非线性问题转化为特征空间中的线性问题.将上述方法应用于Tennessee Eastman(TE)化工过程模型,仿真结果表明,KPCA方法在复杂非线性化工过程敝障检测方面的应用明显优于普通的PCA方法.
展开▼
