Abstract
A framework is developed for the determination of the prediction accuracy of a model for ground-borne structural vibration in buildings due to dynamic loadings on a nearby underground railway tunnel. For a model of such a large scale and complex problem, the determination of the accuracy is a highly difficult task. Buildings are large and complex structures and, mostly, only simplified models restricted to the main structural parts can reasonably be envisaged. As a result, significant modelling errors may be present in the model. Furthermore, model parameter uncertainties are inevitable. To incorporate the parameter uncertainties and the modelling errors into the computational analysis, the non-parametric probabilistic approach, recently proposed by Soize, is adopted. This approach constructs a non-parametric probabilistic model associated with the prediction model. The dispersion parameters of this probabilistic model are determined based on experimental information on the model output. For these dispersion parameters, the uncertainty on the output of the probabilistic model is an estimate of the precision of the prediction model. The approach is illustrated with a case history. A prediction model for the transmission of vibrations from a shallow cut-andcover tunnel to a six storey reinforced concrete frame structure in Paris is considered. A highly simplified model for the frame structure is hereby used. The work described here was carried out under the auspices of the CONVURT project sponsored by the European Community.