Quantification of Uncertainty in Structural Dynamic Models

Structural dynamic models are used to simulate the performance of structures in a dynamic environment. The predictive accuracy of a model has been defined as the accuracy of predicted structural response under conditions for which the structure has not been tested. This paper addresses the quantification of modeling uncertainty, not including the additional uncertainties that may be introduced by the dynamic environment itself. Uncertainty is quantified relative to measured quantities (i.e., experimentally derived modal frequencies and displacements). Although these "measurements" are known to be uncertain themselves, they are taken as the "truth" reference, so modeling uncertainty by definition includes experimental uncertainty as well as parametric uncertainty and the uncertainty of model form (the equations of motion). This paper shows how this modeling uncertainty can be derived by comparing analysis and test modes of generically similar structures and thereafter be used to evaluate the accuracy of numerical simulations based on prior analysis and test experience. Practical examples are given.