 Wysłany: Nie 4:09, 08 Maj 2011 Temat postu: Abercrombie and Fitch Feature extraction based on |
|
| Feature extraction based on independent component analysis of complex neural network fault diagnosis
USA20LuisBAlmeida. ICAoflinearandnonlinearmixtureasedonmutualin {ormation. ISTandINESC a 1D, Lis-bon,Abercrombie and Fitch, PortugalMulti-NeuralNetworksforFaultsDiagnosisBasedonICAFeatureExtractionYangShixiJiaoWeidongWuZhaotong (DepartmentofMechanicalEngineering. ZhejiangUniversity, Hangzhou,oakley norge, 310027, China) Abstract: Artificialneuralnetwork (ANN). especiallytheself-organizingmap (SOM) basedonunsupervisedlearningisakindofexcellentmethodforpatternsclusteringandrecognition. Independentcomponentanslysis (ICA) isapowe, rfultoolforanalyzingnongaussiandata. InICA. theFASTICAbasedonfixed-pointiterationisakindofANNalgorithmwithhigheffi-ciency. whichisspeciallyappropriatetofeatureextractionofmultivariatedataatrealtime. Inthispaper, theFASTICAisfirstlyproposedforfeatureextractionofdifferentmechanicalpatterns (includingnormal, gearfaultandloosefoundation) · followedbycertaintypicalANN (forexampleRBFNorSOM),Abercrombie Fitch uk, whichimplementsthefinalclassification. BymeansofICAandthefurtherfeatureextractionstrategybasedonresidualmutualinformation (RMI),ed hardy lippis, higherthansecondorderfeaturesembeddedinmulti-channelvibrationmeasurementscanbecapturedeffectively. Thus, mechanicalfaultpatternscanberecog-nizedcorrectly. TheresultsfromcontrastexperimentsshowedthatthecompoundICA-SOMclassifiercanbeconstructedinsimplerway,oakley brasil, andclassifyvariousfaultpatternsatconsiderableaccuracy, bothofwhichimplyitsgreatpotentialinhealthcon-ditionmonitoringofmachines. Keywords: faultdiagnosis; neuralnetwork; independentcomponentanalysis first author : Yang Shixi , male, associate professor , PhD . January 1968 students. Tel : (0571) 87951924; E-mail: yangsx @ zju. edu. cn 相关的主题文章： oakley norge Jiaxing food packaging export volume abercrombie et fitch Large bags at home and abroad Franklin and Marshall uk Packaging industry 's fir |
|
