On the parameter sensitivity in structural parameter identification using Eigensystem Realization Algorithm for a MW-size wind turbine

Abstract

Accurate estimation of structural parameters such as the natural frequencies, modal shapes and damping ratios is important for both numerical modelling and structural health monitoring, and ambient vibration analysis is usually used for the estimation. Although the Eigensystem Realization Algorithm (ERA) is widely used, the results are known to depend on the selection of the calculation parameters. In this study, first the sensitivity of the calculation parameters of the NExT-ERA method for the full-scale measurement data of a 2,4MW wind turbine is studied to provide reference for practical application of the method. Small initial model order resulted in unstable estimation of the damping ratios, while higher values gave robust results. The optimum value for the initial model order was 400 to 600, which was much larger than the recommendation in the literature as four times the model order. The optimal values for the correlation length parameter are 3500 to 4000 for the 1st modes, and 100 to 350 for the 2nd modes, which corresponds to the length of the free decay process seen in the cross- correlation function. Higher number of points used for FFT gave lower damping ratios for the 1st mode in both fore-aft and side-side directions. The effect of the number of the FFT points is smaller in 2nd mode, which may be due to difference of the length of data required to capture the characteristics of the system. The structural parameters estimated with NExT-ERA method agreed well with the results of the previously conducted forced vibration test except for the fore-aft 1st mode damping ratio, of which value was especially small compared to other modes.