SYSTEM IDENTIFICATION APPROACH TO BOUNDARY CONDITION ESTIMATION USING EXPERIMENTAL SPATlAL DYNAMIC MODELING IN A BAYESlAN FRAMEWORK, A TEST CASE Scott A Vendemla

Computational models are commonplace in engineering gelds. The predictive ability of such models can bs improved by tuning them to experimental data. To that extent, the user must consider if the dynamics model possess the capacity to represent the dominant physics in the system. A pseudo dynamic stiffness finite element model will be used to rep-t the dynamics of the harmonically forced system. The method offers a direct solution for dynamic response ovar modal summation finite element models. Once a model is selected, the uncertainties in the model formulation must be recognized. and steps taken to reduce them. A system identification approach is used to deal with the model uncertainties through the use of experimental data. More specifically, a Bayesian formulation is used within a system identification framework The uncertainties in the parameters of the original model are quantified by means of prior distributions on the model parameters. The likelihood function for the observed experimental data is then be used to update these prior distributions, through Bayer theorem, to posterior distributions. Wii the posterior distributkns used in place of the prior distributions in the computational model, the model will be more representative of the actual system.