论文部分内容阅读
针对处理实际工业过程中提取的建模样本不纯而导致故障检测失效的问题,提出一种新的融合Fisher判别分析-可能性C-均值聚类(FDA-PCMC)的核主元分析(KPCA)故障检测算法.通过FDA特征提取、初分类和PCMC聚类相结合的方代来实现建模样本的有效分类和提纯,然后使用KPCA进行实时故障检测.对Tennessee Eastman(TE)过程的仿真研宄结果表明了该算法的可行性和有效性.
In order to solve the problem of fault detection failure caused by impure model samples extracted from actual industrial processes, a new KPCA method based on Fisher discriminant analysis-likelihood C-means clustering (FDA-PCMC) is proposed. ) Fault detection algorithm.Effective classification and purification of the modeling samples are achieved by the combination of FDA feature extraction, initial classification and PCMC clustering, and then real-time fault detection using KPCA.Study on the Tennessee Eastman (TE) process simulation宄 The result shows that the algorithm is feasible and effective.