Improved modal parameter estimation for lowly damped systems using non-parametric exponential windowing techniques

In practice, the estimation of frequency response functions (FRFs) is often complicated by the influence of noise on the measured data as well as by spectral leakage in the case of arbitrary (non-periodic) excitation. Accurate modal parameter identification starting from FRF data acquired in the presence of measurement noise is possible by considering FRF estimators developed in an errors-in-variables framework. However, given a finite measurement time, effects of leakage can still hamper the non-parametric processing when non-periodical (arbitrary) excitation signals are used. In modal analysis, an exponential time-window is often applied, for reducing both the effects of leakage and measurement noise, in the case of a time-limited excitation such as for pulse or burst random signals. In this paper, the applicability of using of a non-parametric exponential time-window in the pre-processing of data, acquired with arbitrary excitation, is now studied for improving the modal parameter estimation of lowly damped systems. In the end, for their dynamics research, engineers are often interested in obtaining the modal parameters for reasons of physical interpretation and applications of numerical model updating and modification prediction. A comparison of different approaches, including other methods suggested in the literature, demonstrates the effectiveness for improving modal parameter estimation. An experimental case study for an engine supporting subframe of a car studies the practical applicability of the most important methods derived in this work.