代价敏感核主元分析及其在故障诊断中的应用

来源期刊:中南大学学报(自然科学版)2013年第6期

论文作者:唐勇波 桂卫华 彭涛

文章页码:2324 - 2330

关键词:核主元分析;代价敏感;混沌粒子群算法;阈值调整;故障诊断

Key words:kernel principal component analysis; cost-sensitive; chaos particle swarm optimization; threshold adjusting; fault diagnosis

摘    要:针对传统核主元分析没有考虑误分类代价的差别、对故障工况不敏感等问题,提出代价敏感核主元分析方法。该方法将代价敏感机制引入核主元分析,以误分类代价最小化为目标,设计最佳阈值调整方法获取最佳阈值,并采用混沌粒子群算法对核参数进行优化,最后利用SPE(squared prediction error)统计量诊断新样本类别。研究结果表明:该方法能有效地降低误分类代价,具有故障敏感性和诊断准确率高以及泛化能力强等特点。

Abstract: Cost-sensitive kernel principal component analysis(CS-KPCA) was proposed. Cost-sensitive mechanism was firstly introduced into kernel principal component analysis aiming at the problems that traditional kernel principal component analysis doesn’t consider the misclassification cost, and is insensitive to fault condition. CS-KPCA aimed to minimize misclassification cost. The best threshold adjusting method was designed to get the best threshold and the chaos particle swarm optimization(CPSO) algorithm was adopted to optimize the kernel parameters of kernel principal component analysis. At last, class labels of new instances were diagnosed by using squared prediction error(SPE) statistic. The results show that the proposed method can reduce misclassification cost effectively with high fault sensitivity, diagnosis accuracy and strong ability of generalization.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号