Probabilistic model updating and variability assessment in blast event simulations.

摘要

In this Dissertation a methodology is developed and validated for model update and variability assessment in blast event simulations. The new algorithms are based on extracting the principal components from the response matrix of the simulation results, and on creating metamodels for the principal components. The Kriging method is used for developing the metamodels. Fast running models are created from the metamodels of the principal components. The fast running models allow completing multiple analyses very fast since they eliminate the need for performing actual simulations. Multiple simulations are required by the model update and the variability assessment processes. Thus, by replacing the actual simulations with the fast running models, it is possible to perform model update and variability assessment in practical applications that use simulation models with a very large number of degrees of freedom. The new methodologies are validated for Naval structures and for blast simulations related to ground vehicles. A simulation model and test data for an UNDEX event are employed in the Naval application. An Eulerian solver and a non-linear finite element simulation model along with test results for a target structure subjected to the blast load from a buried charge are used in the application related to ground vehicles. In both cases the model update and variability assessment processes are validated. The two validations demonstrate the value of the new research and the ability to handle real simulation models used is practical applications.