论文部分内容阅读
核主元分析(KPCA)在非线性系统的故障检测方面明显优于普通的PCA方法,但存在无法进行故障辨识以及在故障诊断过程常常出现核矩阵K计算困难等难题。针对上述问题,提出了一种基于特征样本核主元分析方法(FS-KPCA)非线性故障辨识方法。首先采用特征样本(FS)提取方法有效解决核矩阵K的计算量问题。然后利用计算核函数的偏导方法求取KPCA监控中每个原始变量对统计量T2和SPE的贡献率,利用每个变量对监控统计量贡献程度的不同,可以辨识出故障源。将上述方法应用到TE过程,仿真结果表明该方法不仅能够有效辨识故障,而且提高了故障检测和辨识速度。
KPCA is obviously superior to ordinary PCA in fault detection of nonlinear system, but it has some problems such as failure identification and calculation of nuclear matrix K often in fault diagnosis. In view of the above problems, this paper proposes a nonlinear fault identification method based on the principal component analysis (FS-KPCA) of feature samples. Firstly, the method of feature extraction (FS) is used to solve the problem of calculating the kernel matrix K effectively. Then, the partial derivative of the kernel function is used to obtain the contribution rate of each original variable in the KPCA monitoring to the statistic T2 and SPE, and the fault sources can be identified by using the contribution of each variable to the monitoring statistic. The above method is applied to the TE process. Simulation results show that this method can not only effectively identify the fault, but also improve the fault detection and identification speed.