Bayesian estimation of a dynamic structure's response

doi:10.1016/j.jsv.2009.08.032

2010-01-01

PRE

Giorgio Calanni Fraccone *

Vitali Volovoi

Massimo Ruzzene

DOI：10.1016/j.jsv.2009.08.032

原文链接

原文求助

摘要

En 中文

This study focuses on the estimation of uncertainty associated with the stress/strain prediction procedures from dynamic test data of structural systems. An accurate prediction of the maximum response levels for physical components during in-field operating conditions is essential for evaluating their performance and life characteristics, as well as for investigating their behavior in light of system design and reliability assessment. Stress/strain inference for a dynamic system is based on the combination of experimental data and results from the analytical/numerical model of the component under consideration. Both modeling challenges and testing limitations contribute to the introduction of various sources of uncertainty within the given estimation procedure with consequent reduced confidence in the predicted response. The objective of this work is to quantify the uncertainties present in the current response estimation process by means of a Bayesian-network representation of the modeling process which allows for a rigorous synthesis of modeling assumptions and information from experimental data, as it takes into account the multi-directional nature of uncertainty propagation. More specifically, the focus is on the residual uncertainty associated with the system's inferred response, and its dependence upon the amount of test data being included in the estimation analysis. Both discrete and linear Gaussian networks were investigated with a focus on their training accuracy and performance in the presence of nonlinear relationships among the physical quantities, weak cause-effect nodal links, as well as different sensitivity levels with respect to infused evidence. (C) 2008 Elsevier Ltd. All rights reserved.

Keyword:

SYSTEM-IDENTIFICATION

PROBABILISTIC APPROACH

NONLINEAR-SYSTEM

UPDATING MODELS

NEURAL-NETWORKS

PHASE-I

UNCERTAINTY

BENCHMARK