Computationally Efficient Time-Domain Finite-Element Model Updating for Joint Input-Parameter Estimation Based on a Simplified Response Sensitivity Calculation Approach

Performing time-domain output-only finite-element (FE) model updating to estimate unknown input load time histories is usually hindered by the high computational cost required for calculating FE response sensitivities with respect to the inputs. This study presents a new and computationally efficient approach for time-domain output-only FE model updating. The new approach is based on the idea that the sensitivity of the FE response with respect to the input excitation time history can be calculated using the impulse response function for linear elastic or mildly nonlinear systems. In this approach, the impulse response function of the model is calculated and shifted in time to calculate the FE response sensitivities with respect to the discrete values of uncertain input excitation time histories. This procedure is notably less complicated in terms of implementation and considerably faster in terms of run time compared to the available methods for calculating the FE response sensitivities. To validate the applicability of calculating sensitivities using the impulse response function method (IRFM), it is incorporated within a Bayesian inference framework and is utilized for joint input-parameter estimation using experimental data. The data are collected during a series of forced-vibration experiments carried out on a pair of full-scale decommissioned prestressed bridge I-girders. The model updating results obtained using the IRFM are compared to those obtained using the finite-difference method (FDM). The comparison shows similar accuracies of the two methods in estimating the input and model parameters, whereas the IRFM significantly enhances the computational efficiency compared to the FDM. The results indicate the success of the IRFM in advancing the performance of time-domain output-only Bayesian FE model updating.