基于核理论均衡聚类和模糊支持向量机的模拟电路诊断方法

来源期刊:中南大学学报(自然科学版)2011年第z1期

论文作者:唐静 秦亮 史贤俊 肖支才

文章页码:108 - 113

关键词:核函数;K-均值聚类;模拟电路;故障诊断;模糊支持向量机

Key words:kernel function; K-means clustering method; analog circuit; fault diagnosis; fuzzy support vector machine

摘    要:核方法通过非线性映射将原始数据嵌入到高维特征空间,然后进行线性分析和处理,为基于知识的数据分析带来新的方法和模式。本文针对传统的K-均值聚类方法无法解决故障特征数据维数高、在故障样本交叠严重时多分类性能较差的问题。在电路故障特征数据预处理阶段,提出了一种类互均衡核K-均值法进行聚类,不仅克服了传统方法中分配不均或漏分问题,而且解决了特征数据维数高带来的奇异性问题,还有效提高了故障样本交叠时的多分类聚类性能;设计了一种核密度函数用于模糊支持向量机实现多故障的分类,给出了最优带宽的求解方法。将该方法应用于国际标准电路中的CTSV滤波器电路故障诊断中,结果表明:该方法能突出不同故障的特性,具有很高的故障识别率,具有广阔的工程应用前景。

Abstract: A new methods and models for knowledge-based data analysis is brought by kernel method. It uses nonlinear mapping, the original data is embedded into a high dimensional feature space and linear analysis of the data and processing. Based on the fact that the traditional K-means clustering methods cannot solve the fault characteristics of the problem of high data dimension, and overlap in a serious fault samples, the method of classification performance is poor, in the circuit fault characteristic data pre-processing stage, the cross-equalization of nuclear cluster K-means method is presented, which not only overcomes the uneven distribution of the traditional method or the issue of data leakage points, but also solves the high dimensionality of feature data to bring the singularity, also effectively improves the fault samples overlapping clustering performance for the multi-classification. A kernel density function for fuzzy support vector machine is designed to achieve a classification of multiple faults. At the same time, the optimal method is given to solve the bandwidth. The method is applied in the circuit CTSV international standard filter circuit fault diagnosis, the results show that this method has higher recognition rate of failure, and has a good application prospect.

相关论文

  • 暂无!

相关知识点

  • 暂无!

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