Damping uncertainty due to noise and exponential windowing

Estimation of damping levels in structures is required in many applications and the theoretical damping predictions in many cases are either difficult or not very reliable. Therefore, determination of damping levels based on experimental data is employed quite often. However, in some cases, experimental approach may not be able to provide definite answers either in the sense that the identified damping level may exhibit high level of uncertainty. In experimental approach, the most widely used methods for damping determination require vibration spectrums or the Frequency Response Function(s) which are obtained by Fourier transformation of the time-domain data. During this process, it is often necessary, especially for lightly damped structures, to modify the time-domain data by using exponential windowing so as to minimise the leakage effect in the spectrum. The so-called numerical damping artificially added by this process can be subtracted later in order to obtain the correct damping level. However, for lightly damped structures, the artificially introduced numerical damping can be significantly greater than the actual damping level. This inevitably brings the accuracy and the reliability issues, especially when the data are contaminated by noise. This paper addresses damping uncertainty in frequency-domain estimation when (i) the data are contaminated by noise and (ii) numerical damping via exponential windowing is introduced during the signal processing phase of the spectrum estimation. Some numerical simulations are performed first in order to assess the adverse effects of noise on damping estimations and resulting damping uncertainty is examined as a function of noise level in the data. Then, damping uncertainty due to the use of exponential windowing is investigated using experimental data. A relationship between damping uncertainty and the level of added numerical damping is presented when the so-called Line-Fit method is used for damping estimation from measured Frequency Response Functions.

► This paper addresses damping uncertainty due to noise and exponential windowing.
► A relationship between damping uncertainty and signal-to-noise ratio is presented.
► The effect of exponential windowing is investigated using experimental data.
► Damping uncertainty becomes significant when artificial damping is high.
► Results can be used to estimate damping uncertainty due to exponential windowing.