Algorithms for identification of the nonlinear behavior of structures
The purpose of this thesis is to develop identification algorithms that are versatile enough to be applied to geometrically and materially complex structures, are able to identify multiple parameters as well as multiple forces, can reconstruct complete events from incomplete data, and are applicable to linear/nonlinear, static/dynamic problems alike. This is achieved by having the core algorithms use sensitivity responses combined with Newton-Raphson iterations as the way to access the Finite Element Method as a process external to the inverse programming. As a consequence of being FEM based, once the model is fully specified (by having identified all unknown parameters and forces), the complete analysis is then easily accomplished. The time sliding feature of the force identification algorithms is very useful because it allows incorporating adaptive features that are essential for solving highly path-dependent problems in a robust manner. ^ The algorithms are tested on nonlinear dynamic problems involving large displacements, large strains and elastic-plastic material behavior. The synthetic data used are time traces from sensors such as accelerometers and strain gages, and whole-field image data as come from Moiré or holography. ^
Major Professor: James F. Doyle, Purdue University.