Uncertainty propagation in modal analysis

Errors in predicted modal behavior can result from modeling errors or actual variations of the properties due to flight conditions, change in payload or manufacturing tolerances. Launcher models used by EADS-ST for modal analysis studies have a nominal definition and adaptations for each flight. Experimental modal analysis and model updating are used to validate the nominal model giving very satisfactory results for the considered Ariane 5 applications. Given a test validated model, one then seeks to exploit an uncertainty characterization on physical parameters to predict uncertainties on modal properties. These predictions are then used for dynamic response validation and control design. Predictions are obtained using reanalysis techniques thus allowing very fast estimates within a design hypercube while having an error evaluation strategy that is detailed. Model reduction is obtained using the multi-model approach and possible choices in implementation are discussed. The results are illustrated on a model of upper cryogenic stage of Ariane 5.