SPECTRAL AND MODAL PARAMETER-ESTIMATION FROM OUTPUT-ONLY MEASUREMENTS

A number of approaches that estimate spectra, natural frequencies and dampings from the response to an unknown random input-the autocorrelat ion, random decrement, maximum entropy and two-stage least squares met hods-were compared with traditional frequency response function estima tes using both a Monte Carlo simulation and real data. It was found th at the Autocorrelation method combined with the Correlation Fit is, ar guably, the best method to use if sufficient data are available. All o f the methods gave less accurate estimates if the assumption of a 'whi te' random input was invalid.