Generalized Response Surface Model Updating Using Time Domain Data

Abstract

In finite-element (FE) model updating using response surface (RS) models as surrogate, the procedure of finding an appropriate design to build the RS models requires a number of trial-and-error approaches with different designs and subset models. To address this issue, a procedure is proposed in this paper to design and fit proper RS models in FE model updating problems. Also, formulation of the problem in an iterative format in time domain is proposed to extract more information from measured signals and compensate for the error present in the regressed models. This procedure is applicable to both linear and nonlinear models under static or dynamic analysis. The proposed methodology is applied to a numerical case study of a steel frame with global nonlinearity. Appropriate design and model order are successfully established and optimization in time performs well in all the simulated scenarios. Finally, the performance of this method in presence of measurement noise is compared with a method based on sensitivity analysis in terms of required time and accuracy.

A residual-based Gaussian process model framework for finite element model updating

“…Then the SIs S i and S Ti can be easily obtained by the expressions of Eqs. (9) and (10). In light of the SIs used to quantify the relative importance of structural parameters, it can be determined which parameters are chosen for the subsequent model updating process.…”

“…Fang and Perera [9] present a RSM-based FEMU scheme for damage identification using D-optimal design. Shahidi and Pakzad [10] develop an improved RSM that is applicable to both linear and nonlinear FEMU. As a powerful alternative to metamodel, Gaussian process model (GPM), also termed Kriging process, has been exponentially applied in a variety of engineering problems, including design optimization [11,12], uncertainty quantification [13][14][15][16], stochastic finite element analysis [17,18], global sensitivity analysis [19][20][21], to name but a few.…”

Assessment of the 2011 Virginia Earthquake Damage and Seismic Fragility Analysis of the Washington Monument

“…Model updating results using sensitivity method (normalized with respect to the initial set of model parameter values shown inTable 1)In the second calibration method, an FE-based error function is approximated and then minimized through numerical optimization algorithmsShahidi and Pakzad (2014a). In this method, first, polynomial response surface (RS) functions are trained to predict the response of the FE simulation in a pre-selected domain of model parameters.…”

Theory and Implementation of Finite Element Model Updating of the Structures Based on Time Domain Data

“…In this method, the modal parameters of the structure are identified first. And the accuracy of model updating is directly influenced by the identification accuracy of modal parameters [3]. Unfortunately, most of identification approaches may induce errors to some extent.…”