欢迎来到天天文库
浏览记录
ID:52175858
大小:6.58 MB
页数:8页
时间:2020-03-23
《基于小波包分解与主流形识别的非线性降噪.pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第37卷第9期仪器仪表学报Vol37No92016年9月ChineseJournalofScientificInstrumentSep.2016基于小波包分解与主流形识别的非线性降噪苏祖强,萧红,张毅,罗久飞(重庆邮电大学先进制造工程学院重庆400065)摘要:为解决工程实际中强噪声、非线性且频率成分复杂的振动信号降噪问题,提出了基于小波包分解和主流形识别的非线性降噪方法。采用小波包分解将原始振动信号正交无遗漏地分解到各频带范围内,根据各子频带中信噪空间分布,分别采用相应参数对小波包分解系数进行相空间重构;采用局部切空间排列(localt
2、angentspacealignment,LTSA)主流形识别方法在高维相空间中实现信号与噪音的分离,并重构出降噪后的一维小波包分解系数,最后进行小波包分解重构得到降噪后的振动信号。通过仿真实验和实例应用对本文所提方法的有效性进行了验证,试验结果表明本文方法具有良好的非线性降噪能力。关键词:小波包分解;相空间重构;流形学习;非线性降噪中图分类号:TH165.3TN911.2文献标识码:A国家标准学科分类代码:510.40Nonlinearnoisereductionmethodbasedonwaveletpacketdecompositiona
3、ndprinciplemanifoldlearningSuZuqiang,XiaoHong,ZhangYi,LuoJiufei(SchoolofAdvancedManufacturingEngineering,ChongqingUniversityofPostsandTelecommunications,Chongqing400065,China)Abstract:Nonlinearnoisereductionmethodbasedonwaveletpacketdecompositionandprinciplemanifoldlearningi
4、sproposed,aimingtoreducethenonlinearnoiseofvibrationsignalswithcomplexcomponentsinthepracticalengineeringapplication.Firstly,thecollectedvibrationsignalsareorthogonallydecomposedintoseveralsubfrequencybandsbywaveletpacketdecomposition,andthewaveletpacketdecompositioncoeffic
5、ientsarereconstructedintoahighdimensionalphasespace.Here,theparametersofphasereconstructionareselectedaccordingtothedistributionofsignalandnoiseineachsubfrequencybandandthenprinciplemanifoldlearningbylocaltangentspacealignment(LTSA)isperformedtoseparatethesignalandnoiseinh
6、ighdimensionalphasespace.Waveletpacketdecompositioncoefficientsarereconstructedbackintoonedimensionalseries.Atlast,thevibrationsignalafternoisereductionisobtainedbythewaveletpacketreconstruction.Theeffectivenessoftheproposedmethodissimulated.Theexperimentandpracticalapplic
7、ation,andtheexperimentalresultsdemonstratetheexcellentnonlinearnoisereductioncapacityoftheproposedmethod.Keywords:waveletpacketdecomposition;phasereconstruction;manifoldlearning;nonlinearnoisereduction[3]齿轮箱冲击特征提取;王广斌等人提出分形维和局部[4]1引言切空间均值重构的非线性降噪方法;梁霖等人在采用局部切空间(localtangents
8、pacealignment,LTSA)进行冲相空间重构是一种有效的非线性时间序列分析方击故障特征提取时,采用峭度和偏斜度来确定相空间重[1][5]法,
此文档下载收益归作者所有