PSI - Issue 84

Laura Dieci et al. / Procedia Structural Integrity 84 (2026) 591–598

595

(a)

(b)

Fig. 4: FE model in Strand7: (a) deck model with the reference static load configuration, and (b) detail of the Gerber saddle.

4. Numerical modelling The finite element model of the bridge, shown in Fig. 4a, is developed in Strand7 to represent the three-dimensional deck. It includes the five main girders, twelve transverse beams, slabs, and Gerber saddles, progressively assembled to capture the overall structural behavior. All elements are modeled based on the geometry obtained from the laser scanner survey, using 2D plate elements with assigned mechanical properties and ensuring a realistic representation of the deck response. Special attention is given to the Gerber saddles, accurately modeled according to the surveyed geometry, using an intermediate connection element as shown in the detail of Fig.4b. The bidimensional connection element has been assigned the same elastic molduls and thickness of the main girders, as preliminary analyses showed that these properties provided the best agreement with the numerical results. The slab is modeled as continuous with the main deck and an increased density is assigned to account for non-structural weight, including pavement and guardrails. Cantilever slabs with a width of 1.5 m were included to account for the presence of sidewalks. A higher mass was assigned to these elements, consistent with the greater thicknesses identified during the geometric survey. To accurately represent the torsional behavior, the secondary lower slabs, located over a portion of spans near piers, are also introduced in the model. Piers and abutments are not explicitly modeled; instead, their structural influence is captured through deck supports. Vertical rollers are employed at each support, with longitudinal stability ensured by fixing the x-translation at a single node to avoid over-constraining the system. Given that the boundary conditions of these equivalent elements significantly dictate the dynamic response, preliminary sensitivity analyses were conducted. The results demonstrate that the highest correlation with numerical data is achieved by applying roller supports at the pier bases and roller supports coupled with translational springs at the abutments. The reference static load configuration was implemented by applying concentrated forces at selected deck nodes, distributed along the central span according to the adopted loading scheme, as presented in Fig. 4a. 5. FE model calibration This section outlines the procedure used to calibrate the FE model, enabling it to accurately capture both the static and dynamic behavior observed during the experimental tests, and presents the main results. 5.1. Calibration procedure The validation of the finite element model is carried out by adjusting a limited set of uncertain mechanical parameters in order to minimize the discrepancy between numerical predictions and the available reference quantities. The calibration is performed using the DE-S algorithm, a surrogate-assisted evolutionary strategy specifically developed for computationally demanding optimization problems. This algorithm combines the robustness of Differential Evolution with the computational efficiency of a second-order surrogate approximation of the objective function, allowing an effective exploration of the parameter space while reducing the number of full FE model analysis required. Details on the optimization algorithm can be found in Vincenzi et al. (2017). Three parameters are selected as variables to be tuned during calibration: the elastic modulus of the main girders , the equivalent density of the concrete slab and the longitudinal translational stiffness of the abutment bearings