摘要
针对旋转机械高维故障数据集中特征属性冗余导致的故障分类困难问题,提出一种基于局部Fisher主成分判别分析(local Fisher principal component discriminant analysis,简称LFPCDA)的故障数据集降维算法。首先,利用Laplacian得分算法过滤高维故障特征集中冗余特征,并将主成分计算融入局部Fisher判别分析(local Fisher discriminant analysis,简称LFDA)中,自适应地选取出最能反映故障本质的主成分来构成投影矩阵,得到低维特征子集;其次,将低维特征子集输入K近邻(K-nearest neighbor,简称KNN)分类器中进行故障模式辨识;最后,使用双跨转子实验台模拟的转子故障数据集对所提算法进行验证,并与其他几种典型降维算法进行对比。结果表明:所提算法可剔除高维故障数据集中冗余信息,且保留特征的主要成分,使得故障类别之间的差异性更加突出,从而达到提高故障模式识别准确率的效果。所提算法可为转子故障智能决策技术提供数据降维处理的理论参考依据。
To address the difficulty of fault classification caused by redundant feature attributes in highdimensional fault dataset of rotating machinery,a local Fisher principal component discriminant analysis(LFPCDA)method for dimensionality reduction of fault data sets is proposed.First,the Laplacian score algorithm is used for filtering the redundant features in the high-dimensional fault feature set,and principal component calculation is integrated into the local Fisher discriminant analysis.The principal components that best reflect the fault nature are adaptively selected to form the projection matrix,so as to obtain the lowdimensional feature subset.Second,the low-dimensional feature subset is fed into the K-nearest neighbor for fault mode identification.Finally,the rotor fault data set simulated by a double-span rotor test bench is used for verifying the proposed algorithm and compare it with other typical dimensionality reduction algorithms.The results show that the proposed algorithm has the function of eliminating redundant information in highdimensional fault data sets and retaining the main components of features,thereby making the differences between fault categories more prominent.Thus,the accuracy of fault pattern recognition can be improved.This algorithm can provide theoretical reference for data dimensionality reduction processing of intelligent decisionmaking technology of rotor fault.
作者
陈芳军
赵荣珍
邓林峰
CHEN Fangjun;ZHAO Rongzhen;DENG Linfeng(School of Mechanical and Electrical Engineering,Lanzhou University of Technology Lanzhou,730050,China;School of Intelligent Manufacturing and Electrical Engineering,Guangzhou Institute of Science and TechnologyGuangdong,510540,China)
出处
《振动.测试与诊断》
北大核心
2025年第6期1151-1156,1274,共7页
Journal of Vibration,Measurement & Diagnosis
基金
国家自然科学基金资助项目(62241308,51675253)
甘肃省优秀研究生“创新之星”资助项目(2022CXZX-415)。
关键词
故障诊断
局部Fisher主成分判别分析
主成分计算
降维
fault diagnosis
local Fisher discriminant analysis
principal component calculation
dimensionality reduction
