摘要
针对SVDD方法在训练阶段计算量大、训练时间久的问题,提出了基于局部保持投影支持向量数据描述(LPP-SVDD)的故障检测方法.结合LPP处理线性降维和SVDD在异常点检测的优势,使用LPP算法对原始数据进行维数约减,对降维后的数据采用SVDD算法建立监控模型,在最大程度保留数据局部结构特性的同时达到数据维数约减的目的,从而降低SVDD的计算量,缩短建模及检测时间.通过数值例子和半导体工艺过程进行仿真研究,对比LPP、k NN、SVDD、LPP-SVDD方法,验证所提方法的性能.结果证实了LPP-SVDD不仅具有准确的检测能力,而且具有较高的检测效率.
Aiming at the problem of large amount of calculation and long training time in the training phase of SVDD method,a fault detection method based on local preserving projection support vector data description(LPP-SVDD)is proposed.Combined with the advantages of LPP processing linear dimensionality reduction and SVDD in outlier detection,LPP algorithm is used to reduce the dimension of the original data,and SVDD algorithm is used to establish the monitoring model for the reduced-dimensional data,which preserves the local structural characteristics of the data,so as to retain the local structure characteristics of the data to the greatest extent and reduce the dimension of the data,so as to reduce the amount of calculation of SVDD and reduce the modeling and detection time.The performance of the proposed algorithm is verified by comparing LPP,k NN,SVDD and LPP-SVDD algorithms.It is proved that LPP-SVDD not only has accurate detection capability but also has high detection efficiency.
作者
谢彦红
薛志强
李元
XIE Yan-hong;XUE Zhi-qiang;LI Yuan(Shenyang University of Chemical Technology,Shenyang 110142,China)
出处
《沈阳化工大学学报》
CAS
2022年第1期60-68,共9页
Journal of Shenyang University of Chemical Technology
基金
国家自然科学基金资助项目(61673279,61490701)
辽宁省教育厅重点实验室基础研究项目(LZ2015059)
辽宁省教育厅一般项目(L2015432)。
关键词
维数约减
局部保持投影
支持向量数据描述
半导体工艺过程
故障检测
dimensionality reduction
locality preserving projection
support vector data description
semiconductor process
fault detection
