Mechanical systems and signal processing vol:15 issue:4 pages:775-788
This paper discusses the parametric identification of multiple-input-multiple-output (MIMO) rotor-bearing systems in the frequency domain on the basis of the maximum likelihood estimator (MLE). An identification procedure, considering noise on both inputs and outputs (errors-in variables model), is developed for real as well as for complex modal testing. Quantitative measures for the uncertainties (covariances) of the estimated parameters were to be obtained. It is shown that the MLE solution of real and complex modal analysis is fully equivalent. It follows that, in case of complex modal testing, reduction of the system is not allowed in general when noise information is taken into account. A comparison between random and broadband periodic (multisine) excitations is made. Measurements on a rotor test rig demonstrated that multisines yield more accurate estimates of the modal parameters (as compared to random excitation). (C) 2001 Academic Press.