Improved modal parameter estimation using exponential windowing and non-parametric instrumental variables techniques

In the present contribution, the applicability of improved non-parametric identification techniques in the field of modal analysis are investigated. Exponential windowing is applied during the signal processing step, reducing leakage effects as well as noise on the data. Two new approaches are validated using Monte Carlo simulations for which the poles and residues are estimated by applying a least squares estimator (LSCE -LSFD). In addition, an improved frequency response functions estimator, based on an instrumental variables approach, is integrated in the step prior to the parametric estimation. This allows for noise on both input and output measurements. The modal parameters together with their confidence intervals are derived by applying the frequency-domain Maximum Likelihood estimator. This is validated for an experimental case study.