Mechanical systems and signal processing vol:15 issue:4 pages:759-773
This paper discusses the non-parametric identification if MIMO rotor-bearing systems in the frequency domain for real as well as for complex modal testing. The FRFs are estimated on the basis of the maximum likelihood estimator, considering noise on both inputs and outputs (error-in-variables model). It is demonstrated that multisine excitations with random phases yield improved FRF measurements as compared to random excitations. Moreover, quantitative uncertainties (covariances of the FRFs) are obtained allowing us to quantify the quality of the measurements. The experiments are performed on a rotor test rig and the evaluation is made for both real and complex modal analysis. (C) 2001 Academic Press.