This publication proposes a discussion of the general problem of validating numerical models for nonlinear, transient dynamics. The predictive quality of a numerical model is generally assessed by comparing the computed response to test data. If the correlation is not satisfactory, an inverse problem must be formulated and solved to identify the sources of discrepancy between test and analysis data. Some of the most recent work summarized in this publication has focused on developing test-analysis correlation and inverse problem solving capabilities for nonlinear vibrations. Among the difficulties encountered, we cite the necessity to satisfy continuity of the response when several finite element optimization are successively carried out and the need to propagate variability throughout the optimization of the model's parameters. After a brief discussion of the formulation of inverse problems for nonlinear dynamics, the general principles which, we believe, should guide future developments of inverse problem solving are discussed. In particular, it is proposed to replace the resolution of an inverse problem with multiple forward, stochastic problems. The issue of defining an adequate metrics for test-analysis correlation is also addressed. Our approach is illustrated using data from a nonlinear vibration testbed and an impact experiment both conducted at Los Alamos National Laboratory in support of the advanced strategic computing initiative and our code validation and verification program.
PDF
download
878987797
028-85220240
