资源描述：
《A New Method for the Identification of Hammerstein Model》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、A1PergamonPSOOO5@0+BPAHALaMNK-AAKeyWord&Identification;neuralnets;nonlinearcontrolsystems;processcontrol;recursiveestimation;time-seriesanalysis.Abstract—Anewmethodfortheidentificationofthenonlineartheunknowncoefficientsal,,an,6.,...,bm,andcl,...,CL.Ha
2、mmersteinmodel,consistingofastaticnonlinearpartinTheunknowncoefficientsaredeterminedbytransformingcascadewithalineardynamicpart,isintroduced.Thestatictheproblemintomulti-input–single-output(MISO)linearnonlinearpartismodeledbyamultilayerfeedforwardneura
3、lidentificationproblem.Theinputstothesystemwillbenetwork(MFNN),andthelinearpartismodeledbyanautoreg-u(t),uz(t),...,uL(t).Themaindrawbacksofthisapproacharetheressivemovingaverage(ARMA)model.Arecursivealgorithmisassumptionthatthenonlinearityisofapolynomi
4、alformandthedevelopedforestimatingtheweightsoftheMFNNandtheincreaseofthenumberofinputsinthelinearidentificationparametersofARMAmodel.Simulationexamplesareincludedproblem.Ithasbeenshownin(Gallman,1975)thatifthetoillustratetheperformanceofthepropo8edmeth
5、od.(Q1997nonlinearityisnotapolynomialandtheinputisnotGaussian,ElsevierScienceLtd.thesealgorithmsdonotconverge.Differentnonparametricapproaches(GreblickiandPawlak,1.Introduction1986,1989)havebeenusedfortheHammersteinmodelidenti-TheHammersteinmodelshowni
6、nFig.1consistsofastaticfication.Mostofthenonparametricmethodsusekernelsregres-nonlinearpartinserieswithalineardynamicpart.Thesignalssionestimatestoidentifythenonlinearity.Theidentificationofbetweenthenonlinearandlinearpartsareinaccessibletothenonlinear
7、ityisdoneseparatelyfromtheidentificationofthemeasurements.TheidentificationoftheHammersteinmodellinearpart.Twopossiblestructures,Figs.2and3,canbeusedinvolvesestimatingboththenonlinearandlinearpartsfromthetodescribemulti-input–multi-output(MIMO)Hammerst
8、eininput–outputmeasurements.Thesingfe-input-single-outputmodeldependingonwhetherthenonlinearitiesareseparateor(S1S0)Hammersteinmodelhasbeensuccessfullyusedtomodelcombined(Eskinatetat.,1991),Noefficientmethodiscurrentlyalargeclassofnonli
