基于改进核局部保持投影的故障检测方法研究  

RESEARCH ON FAULT DETECTION METHOD BASED ON MODIFIED KERNEL LOCALITY PRESERVING PROJECTIONS

作　　者：张学磊 徐进学 祝朋艳 Zhang Xuelei;Xu Jinxue;Zhu Pengyan(Marine Electrical Engineering College,Dalian Maritime University,Dalian 116026,Liaoning,China)

机构地区：[1]大连海事大学船舶电气工程学院,辽宁大连116026

出　　处：《计算机应用与软件》2022年第5期115-124,131,共11页Computer Applications and Software

摘　　要：针对传统核局部保持投影方法存在不能全面地提取数据特征和故障检测率低的局限性,提出一种基于改进核局部保持投影(MKLPP)的故障检测方法。将核主元分析方法融入到核局部保持投影中,使得到的数据特征不仅包括原始数据的局部信息,而且包含数据的全局信息;引入特征向量缩放,使得低维数据变化波动更加稳定,提高故障检测率。针对MKLPP方法对微小故障不敏感的问题,将多元指数加权移动平均(MEWMA)运用到MKLPP中,提出一种MEWMA-MKLPP故障检测方法。对上述两种方法分别构造T^(2)和SPE统计量进行故障检测。采用TE过程数据进行仿真实验,实验结果表明所提方法可以取得较好的检测效果。In view of the limitations of traditional kernel locality preserving projections,which cannot fully extract data features and has a low fault detection rate,a fault detection method based on modified kernel locality preserving projections(MKLPP)is proposed.The kernel principal component analysis method was integrated into the kernel locality preserving projections,so that the acquired data characteristics could not only include the local information of the original data,but also include the global information of the data.The feature vector scaling was introduced to make the fluctuation of low-dimensional data more stable and improve the fault detection rate.In order to solve the problem that MKLPP method was not sensitive to incipient faults,a new method for fault detection of MEWMA-MKLPP was proposed by applying multivariate exponentially weighted moving average(MEWMA)to MKLPP.The above two methods respectively constructed T^(2) and SPE statistics for fault detection.TE process data was used in the simulation experiment.The experimental results show that the proposed methods can achieve better detection effect.

关 键 词：核局部保持投影 核主元分析 多元指数加权移动平均 微小故障 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]