Calibration of the Damping Dispersion Parameter in the Non-parametric Probabilistic Approach

Response predictions in structural dynamics are in general very sensitive to random uncertainties associated with the underlying predictive mathematical model. The non-parametric probabilistic model provides the possibility to capture both data and model uncertainties. The uncertainties are introduced at a global level and controlled through one dispersion parameter each, for the mass, damping and stiffness matrix. In the present paper the role of the damping dispersion parameter is investigated, for the case in which the non-parametric model is calibrated with respect to an existing parametric model. It is shown that in the non-parametric model the influence of the damping dispersion on the scatter in the frequency response is relatively small, if a calibration criterion based on matrix norms is used. A novel method for calibrating the non-parametric model is proposed, which enforces the same scatter of the FRF at the first resonance frequency, in the parametric and the non-parametric model. A case study involving a satellite FE-model, shows that with this approach, the FRF scatter of the non-parametric model reaches a level similar to that of the parametric model used for its calibration.