Cable-Stayed Bridge Model Updating Using a Novel Approach for Deflection Influence Line Identification: Theoretical Basis and Field Validation

The finite element model developed during the design stage often lacks the ability to accurately reflect the actual operational conditions of a bridge. To establish a finite element model suitable for high-precision analysis of cable-stayed bridges, we propose a finite element model update that leverages the measured deflection influence line and the GA-BP network. Obtaining high-precision influence lines is a prerequisite for model updating; however, to effectively eliminate the dynamic components in the vehicle-induced bridge response, an innovative method in which E-VMD is combined with Tikhonov regularization is proposed. Based on this, the quasistatic influence line of the bridge can be accurately restored. The GA-BP network is subsequently used to obtain the updated parameters to construct the regression prediction method, and the deflection influence line is the target parameter. The effectiveness of the proposed method is verified using a numerical simulation and an engineering example. The results indicate that the power effect rejection of E-VMD outperforms that of the EMD, VMD, and PE-VMD methods and achieves a relative error of only 2.77% compared with those from the theoretical simulations. Furthermore, its identification error remains below 4% even when the vehicle speed increases to 120[Formula: see text]km/h. The relative error at the deflection influence line of the corrected model’s control section significantly decreases from 57.2% to 14.1%. The gray relational coefficient increases to 0.9076. The accuracy of the corrected finite element model is significantly enhanced, and this model can more closely replicate the actual operational state of the bridge.