The model was allowed to vibrate freely without any prescribed initial excitation using nine different wind tunnel fan frequencies (5, 10.5, 15, 20, 25, 26, 27, 27.5 and 28 Hz), increasing the resolution in the vicinity of the potential critical speed. The frequencies are labelled in the dataset as "f050", "f105" etc., with the values expressed in deci-hertzs to avoid the use of decimal points in the file naming convention. The corresponding wind speed time series were recorded by a Cobra probe located above of the testing section and stored under the labels "ucob" (along-wind), "vcob" (horizontal cross-wind) and "wcob" (vertical), while the average along-wind speed was also recorded by a Pitot tube ("Upit"). For each case, the response was recorded once steady-state conditions were reached, following each change in flow velocity, over 70 s intervals at a sampling frequency of 2000 Hz (fs_Hz).

The displacements were measured by three laser sensors measuring in millimeters, all of them located outside of the wind tunnel, where the support rig was placed. One laser pointed to the center of rotation of the deck on the right side of the wind tunnel ("laser_center"). The other two pointed to the support beam on the left side of the wind tunnel, one placed upstream ("laser_upstr") and downstream ("laser_downstr") of the center of rotation to allow the calculation of the model's rotation. The distance between the lasers and the center of rotation is stored in mm under the name "Arm_laser_mm".

Additionally, other quantities are stored in the dataset: "Temp" represents the average temperature of the air during the measurement in degrees Celsius, "rho" is the density of the air and "alpha" is the angle of attack of the bridge without wind (that is, before the measurement started). Unless said otherwise, all stored data uses SI units.

A separate set of measurements, obtained by imposing a prescribed initial displacement, was performed for the identification of the deck flutter derivatives (i.e., free-vibration tests). In this case, the wind tunnel fan frequencies used (2.5, 5.5, 9, 12, 14.5, 17, 19.5, 22, 25, 26, 27 and 27.5 Hz) are distributed homogeneously to describe the derivatives along all the domain. Since most oscillations are damped quickly, the recording time was reduced to 15 s. The rest of the parameters, including the location of the lasers, remain the same as in "Experiment 1: Direct flutter identification".

For each wind tunnel fan frequency, 3 different initial conditions were imposed: a bending vertical displacement (applied through a string attached to the center of the model), a torsional rotation (applied manually to the support bars at one of the sides of the wind tunnel) and a combined motion (applied through a string attached to the downstream box in the middle of the span). Each measurement was repeated 5 times (labelled "rep1", "rep2" etc.) to assess experimental repeatability and ensure result consistency.

The setup of this experiment is identical to "Experiment 1" of "Specimen 1", with the only difference being the cross-section geometry. Additionally, the last five fan frequencies were substituted by 22, 23, 24, 24.5 and 25 Hz, since the critical wind speed was lower in this case.

The setup of this experiment is identical to "Experiment 2" of "Specimen 1", with the only difference being the cross-section geometry. Additionally, the last four fan frequencies were substituted by 23, 24 and 24.5 Hz, since the critical wind speed was lower in this case.

The same measurements described in the "Experiment 1" of "Specimen 1" were carried out, with several modifications to adapt to the new flexible model.

First, the displacement of the deck was measured using laser sensors at five different locations. On the left side of the wind tunnel, the three original laser sensors used in "Specimen 1" and "Specimen 2" were placed at different locations of the support beams. Two sensors ("laser_upstr_A" and "laser_downstr_A") were placed symmetrically on the upwind box (labelled as box "A") at equal distances ("Arm_laser_mm") from its center of rotation. Another sensor ("laser_downstr_B") was positioned with the same offset on the downwind box (labelled as box "B"). On the opposite side of the wind tunnel, the vertical displacements at the centers of rotation of each box were recorded using four additional laser sensors with a reduced measuring range (i.e., 2.5 cm). Since the expected oscillations exceeded the range of the available lasers, two laser sensors were used for each measurement. For the upstream box, one laser was pointed to the center of rotation from above ("laser_center_upA"), while the second pointed to the center from below ("laser_center_downA"). The ranges of each laser were carefully calibrated so the deck remained within the operating range of at least one of the two sensors at all times, effectively extending the measurable displacement range. The final signal can then be re-constructed in post-processing steps. Analogously, the location of the center of the downstream box was recorded using two additional sensors ("laser_center_upB" and "laser_center_downB").

Second, the experiments were repeated not only for different wind speeds, but also for different combinations of connections between the boxes, ensuring different degrees of flexibility of the cross-section. A first test was carried out using rigid connections between the boxes, representing a value of the distortional stiffness of infinity ("Kdinf"). Then, a second test was done with only two hinged bar connections, providing the necessary kinematic constraint but no rigidity of the inner cell ("Kd0"). Finally, different different combinations of spring connections were added. Each spring connection was composed by two springs closing the inner cell diagonally (see image). Arranging a different number of spring connections, and using springs of different stiffness, it was possible to vary the value of the distortional stiffness “Kd”. The relation between the stiffness of the springs and the resulting value of "Kd" can be found in previous publications from the authors (https://doi.org/10.1016/j.awe.2026.100105). Using springs with a stiffness of 14 N/m, the model was equipped with 2 and 4 elastic connections, corresponding to the values "Kd0.095" and "Kd0.19", respectively. Then, 1, 2, 3 and 4 connections with springs of stiffness 280 N/m were tested ("Kd0.95", "Kd1.9", "Kd2.85" and "Kd3.8").

Since each combination of springs change the critical wind speed of the deck, the input fan frequencies and corresponding wind speeds vary from case to case. They are labelled "f123", were "123" is the frequency of the rotor in dHz, as explained in "Experiment 1" of "Specimen 1".

A different set of recordings imposing a certain initial displacement was made to calculate the flutter derivatives from the deck (free-vibration tests). Contrary to "Experiment 1" of this same specimen, the process was done for a single value of the distortional stiffness of the inner cell ("Kd0"), corresponding to the configuration with no spring connections at all.

As in "Specimen 1", "Experiment 2", the recording time was reduced to 15 s and the wind tunnel fan frequencies used were distributed homogeneously to describe the derivatives along all the domain, but they can also vary depending on the critical wind speed of the analyzed configuration. However, the laser configuration was kept the same as in "Specimen 3", "Experiment 1", in order to capture the distortion of the model and calculate the associated derivatives.

Once again, for each wind tunnel fan frequency, 3 different initial conditions were imposed, but they differ from the ones introduced in "Specimen 1". First, a bending vertical displacement (applied through a string attached to the center of the model). Second, a torsional rotation (applied through a string attached to the rotation axis of one box at the center of the span). Finally, the distortional mode was directly excited using an external shaker. In this case, one of the outer horizontal support bars was connected to the actuator via a string, and the excitation frequency was tuned to match the distortional natural frequency. Once the desired vibration amplitude was achieved, the string was cut by burning it, allowing the system to evolve freely under the imposed wind conditions. Each measurement was repeated 5 times (labelled "rep1", "rep2" etc.) to assess repeatability and ensure result consistency.

The setup of this experiment is similar to "Experiment 1" of "Specimen 3", with the only difference being the alternative mass distribution described above.

The setup of this experiment is identical to "Experiment 2" of "Specimen 3", but with a different mass distribution, as already described. Moreover, the current tests were carried out with a different connection configuration ("Kd1.9"), corresponding to 2 crosses of springs (of stiffness 280 N/m). This setup provides additional variability for the estimation of the flutter derivatives.

Additionally, the external shaker was not used for this experiment, since it was not possible to accurately capture the distortional mode of the model. A second torsional excitation was applied, connecting the string at the edge of one of the boxes instead of the rotation axis.

All the measurements carried out with this specimen were done without any wind speed, since the objective was to obtain more data about the structural properties of each box. First, the upstream box ("BoxA") was excited with 1/2 prescribed bending displacements and the process was repeated with a torsional disturbance. Each measurement only lasted 7 s, since it is enough to record the free-vibration of the box for a few seconds. Although both boxes are symmetric, the measurements were also carried out for the downstream box ("BoxB") to reduce the uncertainty of the results.

Since each box was tested individually, the laser sensors were moved between measurements. In each experiment, the excited box was instrumented with two wide-range lasers located at a distance "Arm_laser_mm" upstream and downstream from the center of rotation ("laser_upstr" and "laser_downstr", respectively). On the opposite side of the wind tunnel, two short-range laser sensors were placed above and below the center of rotation ("laser_center_up" and "laser_center_down", respectively), as already described in "Experiment 1" of "Specimen 3".