Premium
Variability of updated finite element models and their predictions consistent with vibration measurements
Author(s) -
Papadimitriou Costas,
Ntotsios Evangelos,
Giagopoulos Dimitrios,
Natsiavas Sotirios
Publication year - 2012
Publication title -
structural control and health monitoring
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.587
H-Index - 62
eISSN - 1545-2263
pISSN - 1545-2255
DOI - 10.1002/stc.453
Subject(s) - pareto principle , discretization , finite element method , modal , identification (biology) , modal analysis , mathematical optimization , mathematics , computer science , engineering , structural engineering , mathematical analysis , chemistry , botany , polymer chemistry , biology
SUMMARY A case study on a small‐scale laboratory vehicle frame is used to investigate the variability of the updated finite element (FE) models that arises from model and measurement errors and demonstrate the effect of this variability on response predictions. Conventional weighted modal residuals and recently introduced multi‐objective identification methods for structural model updating are used to provide the entire spectrum of Pareto optimal FE models consistent with the measured modal data. Similarities and differences between the two model updating methods are explored and the advantages of the multi‐objective identification methods are emphasized. A significant variability in Pareto optimal models is observed, which is indicative of the uncertainty in the updated FE models. The dependence of the variability of the Pareto models on the information contained in the measured data and the size of model and measurement errors is explored by varying the number of measured modes, number of sensors, FE mesh discretization sizes, and number of model parameters. The effectiveness of the updated Pareto optimal models and their predictive capabilities are assessed. Frequency response functions and fatigue lifetime predictions are used as example of structural performance variables in order to demonstrate the variability in the response predictions that arises from the variability in the Pareto optimal models. A large variability in the response predictions is observed that cannot be ignored in decisions based on updated FE models. The multi‐objective optimization method provides the general framework for properly accounting for model uncertainty in model‐based response predictions consistent with measured data. Copyright © 2011 John Wiley & Sons, Ltd.