An earlier paper [I] presented results from a two and a half year NASA SBIR project to develop methods for evaluating the predictive accuracy of structural dynamic models. The present paper describes how modeling uncertainties derived from comparisons of prior modal analysis and test data are propagated through the equations of motion to predict response uncertainties, andlor to infer design parameter uncertainties. Three alternative methods for uncertainty propagation are presented. They provide measures of response uncertainty in terms of possibility intervals, probability intervals and probability distributions.

No abstract

A general method is presented for using experimental data to verify math models of “mixed” dynamic systems. The term “mixed” is used to suggest applicability to combined systems which may include interactive mechanical, hydraulic, electrical, and conceivably other types of components. Automatic matrix generating procedures are employed to facilitate the modeling of passive networks (e.g., hydraulic, electrical). These procedures are augmented by direct matrix input which can be used to complement the network model. The problem of model verification is treated in two parts; verification of the basic configuration of the model and determination of the parameter values associated with that configuration are addressed sequentially. Statistical parameter estimation is employed to identify selected parameter values, recognizing varying degrees of uncertainty with regard to both experimental data and analytical results. An example problem, involving a coupled hydraulic-mechanical system, is included to demonstrate application of the method.

No abstract

No abstract