针对大规模样本集的核主成分分析(KPCA)存在计算代价巨大的问题,提出一种新的KPCA快速算法.该算法通过施行改进初始中心选择策略的K-均值聚类算法划分样本集,然后选取每个分类的中心作为样本集建立KPCA模型.将该方法应用于TE(Tennessee Eastman)过程的故障诊断,与基于全体样本的KPCA进行比较.实验结果表明,二者的诊断效果相当,但是新的方法在计算上所耗费的时间更少.%For a large-scale sample data set, there is a problem that the computational cost is huge for kernel principal component analysis (KPCA). To solve this, a novel fast algorithm of KPCA is proposed. The algorithm partitions the sample set by k-means clustering algorithm which makes the improvement on selection strategy of initial centres, and then selects the centres of every classification as the sample set to build up KPCA model. The proposed method is applied to the fault diagnosis of Tennessee Eastman (TE) process and is compared with the all samples-based KPCA. Experimental result shows that the diagnosis effect of the two methods is almost the same, but the new one has less time spent on computation.
展开▼
