Does more sophisticated modeling reduce model uncertainty? A case study on vibration predictions

P.H. Waarts and M.S. de Wit
TNO Building and Construction Research, Delft, The Netherlands

In this paper, the reliability of vibration predictions in civil engineering is quantified. Emphasis is laid on the vibration predictions for road- and rail traffic and vibrations from building activities such as (sheet)pile driving. Several kinds of prediction techniques were investigated: expert opinion, very simple empirical models, dedicated models, and FEM. The prediction techniques were applied by four different institutes which are leading in the Dutch practice in vibration prediction: TNO, GeoDelft, Delft University of Technology and Holland Railconsult. Predictions were generated for a variety of characteristic situations and were compared with measurements. Besides the total uncertainty, which can be derived out of the difference between prediction and measurement, a break-down of uncertainty sources was made.

Key words: Vibration predictions, civil engineering, uncertainty analysis, model uncertainty, expert judgment, model reliability