摘要
Basedonthemultivariatecontinuoustimeautoregressive(CAR)model,thispaperpresentsanewtime-domainmodalidentificationmethodoflineartime-invariantsystemdrivenbytheuniformlymodulatedGaussianrandomexcitation.Themethodcanidentifythephysicalparametersofthesystemfromtheresponsedata.First,thestructuraldynamicequationistransformedintoacontinuoustimeautoregressivemodel(CAR)oforder3.Second,basedontheassumptionthattheuniformlymodulatedfunctionisapproximatelyequaltoaconstantmatrixinaveryshortperiodoftimeandonthepropertyofthestrongsolutionofthestochasticdifferentialequation,theuniformlymodulatedfunctionisidentifiedpiecewise.Twospecialsituationsarediscussed.Finally,byvirtueoftheGirsanovtheorem,weintroducealikelihoodfunction,whichisjustaconditionaldensityfunction.Maximizingthelikelihoodfunctiongivestheexactmaximumlikelihoodestimatorsofmodelparameters.Numericalresultsshowthatthemethodhashighprecisionandthecomputationiseffcient.
出版日期
2009年10月20日（中国期刊网平台首次上网日期，不代表论文的发表时间）
