Evolutionary Updating of Model Structure and Parameters

University Of Houston, Houston TX

Investigators

Abstract

The validation of structural analytical models in practice is mainly based on comparing both experimental static and dynamic data with model predictions. The objective of this project is to develop an algorithmic framework suitable for distributed processing which updates both the model structure (constitutive laws, modeling idealizations, discretization) and model parameters (material parameters) to produce a test validated model capable of predicting the response/character of the structure when subjected to a wide range of inputs and design changes. The algorithmic framework is based on the Genetic Algorithm (GA) optimization paradigm. The GA paradigm loosely implements Darwin's "survival of the fittest"concepts and attempts to evolve better solutions to a given problem statement. The GA paradigm is being used because it does not require sensitivity information (which doesn't exist for model structure changes) and can seamlesslyhandle both continuous and discrete design variables. Several GA formulations are being developed which address the computational efficiency of the algorithm based on what experimental data is available and the model structure and parameter changes that are allowed. Even with advances in computational power, model updating can be applied to realistic problems only if a suitable approximation theory relating changes in vibration properties to changes in model structure and parameters is developed. An iterative structural dynamics modification theory is under development which provides accurate approximations in a computationally efficient manner. The major impact of this project is that upon completion a framework for developing test verified structural models which can be used with confidence in model-based simulations will be established. The ability of engineers to rely on model-based simulations will have a profound impact in a wide range of areas, including product design/development, product safety evaluation, national security and infrastructure assessment.