Validation and Verification of Nonlinear Structural Dynamic Models

The recent DNA-sponsored conference on ¿Verification and Validation of Nonlinear Structural Dynamic Codes,¿ September 1993, highlighted the need for reliable models as well as codes. At the present time there are no generally available software tools for utilizing experimental data to correct and refine models implemented on general-purpose codes such as DYNA3D. Nor are there tools available for evaluating the predictive accuracy of these models. This project will demonstrate the feasibility of using a generic algorithm for discrete optimization to eliminate systematic modeling error and make coarse parameter adjustments to obtain models suitable for continuous parameter estimation. Improved methods for defining response metrics and corresponding parameter sets based on the use of Principle Components and singular value decomposition will also be demonstrated. These advances will be used to enhance the effectiveness of Bayesian parameter estimation for ¿fine-tuning¿ the nonlinear models produced by discrete optimization. The use of a Bayesian estimator with improved response metrics and associated parameter sets will yield statistical estimates of residential model uncertainty for use in evaluating the predictive accuracy of the refined models. The proposed project will result in the development of new methods for verification and validation of nonlinear structural models. These methods may be used to reduce time-to-market and testing costs for a variety of Government and commercial applications including survivability and vulnerability analyses of military structures and crash worthiness studies of commercial automobiles. All nonlinear structural analyses will be able to employ the methods to increase model fidelity.