Some Issues in the Identification of Structural Systems for Control and Response Prediction

Abstract

A general framework is presented for the identification of nonlinear structural systems for control and response prediction in the presence of uncertainties. First, a review is presented of a general Bayesian statistical framework for parameter estimation which is applicable to both parametric and nonparametric models, linear or nonlinear, and which explicitly treats the uncertainties arising from the .aforementioned causes. Then, an examination is made of a hybrid form for nonlinear modeling in which a parametric model is used to describe the linear part of the structural model and a nonparametric approach is used to describe the nonlinear part. A review of some methods for identification of the parametric linear part is given, which is followed by an examination of several non parametric approaches for modeling the nonlinear part, including the restoring force surface method and neural networks.

Additional details