1. Introduction
The increasing frequency of wind-induced disasters caused by climate change is placing greater demands on the safety of buildings. Double-deck truss girder suspension bridges are widely used in cross-river and cross-sea engineering due to their excellent spanning and transportation capabilities [
1]. However, as the span increases, the structural stiffness, mass, and damping of suspension bridges decrease, leading to light flexibility and increased sensitivity to wind-induced vibrations [
2]. This problem is particularly prominent in the complex and variable wind environment, and large-span bridges often suffer from wind-induced vibration problems such as vortex-induced vibration (VIV), buffeting vibration, and galloping vibration [
3,
4]. Among them, VIV has become a primary concern [
5]. Many bridges have already suffered severe VIV, such as the Tacoma Narrow Bridge [
6], the Humen Bridge [
7], the Kesssock Bridge [
8], the Fred Hartman Bridge [
9], and the Xihoumen Bridge [
10]. Therefore, as a crucial infrastructure, ensuring the wind stability of bridges is of utmost importance.
VIV is a type of amplitude-limiting vibration with both forced and self-excited characteristics [
11]. The incoming flow generates periodic vortex shedding on the surface of the bridge structure, resulting in a periodic vortex force. When the frequency of vortex shedding is close to the intrinsic frequency of the girder, it triggers VIV in the bridge structure [
12,
13]. Unlike fluttering and buffeting, VIV does not cause destructive damage to the bridge [
14]. However, since it occurs at low wind speeds, persists for prolonged periods, and has a high probability of occurrence, long-term VIV can impact the comfort of traffic and pedestrians and can reduce the durability of the bridge [
15,
16,
17]. Truss girders, unlike Π-girders and box girders, consist of numerous discrete components, with web members that change along the bridge axis, creating complex aerodynamic disturbances [
18]. This leads to the generation of multi-frequency and multi-size vortex shedding after different members, making the evaluation of VIV performance more challenging. Therefore, analyzing VIV and studying measures to suppress vibration in large-span double-deck truss girder suspension bridges is of particular importance.
Aerodynamic measures modifying the cross-sectional shape of components or adding appendages that alter airflow patterns (e.g., stabilizer plates, deflector plates, and wind fairings) represent the most commonly employed method for suppressing bridge VIV [
19]. This approach involves making small, cost-effective, highly reliable structural changes, making it the preferred method for VIV suppression [
20,
21,
22,
23]. Additional aerodynamic measures can be more effective at suppressing VIV but require more materials and can negatively impact bridge aesthetics. Optimizing bridge aerodynamic performance by changing the structural shape, such as adjusting the railing ventilation rate or modifying the dimension and location of the maintenance track, can address these issues and better align with the demands of sustainable development, even though it may result in slightly reduced vibration suppression. Furthermore, VIV can also be suppressed through structural and mechanical measures [
24,
25,
26,
27]. Structural measures involve enhancing the stiffness or mass of the structure and increasing internal or external constraints, which can carry a higher cost. Mechanical measures entail incorporating various dampers to suppress vibration, which can lead to a more complex structural design. Therefore, these measures are more suitable for special cases where aerodynamic measures are ineffective.
The VIV of double-deck truss girders has received relatively little attention. Only Yao et al. [
28] have studied the effect and mechanism of fairing and deflector plates on truss girder VIV. Fang et al. [
29] have investigated the VIV mechanism of a truss girder through wind tunnel tests and numerical simulations. Previous studies on box girders and Π-beams have demonstrated that modifying the railing design can effectively suppress VIV. Nagao et al. [
30] found that both the installation location and shape of the railing can significantly affect bridge VIV. Zhan et al. [
31] found that installing wavy railings on the upper surface edge of the box girder can entirely suppress vertical and torsional VIV. Xin et al. [
32] discovered that inclined railings could completely suppress both vertical and torsional VIV by significantly reducing the spanwise correlation of the streamwise velocity in the wake. Ge et al. [
33] and Zhao et al. [
34] demonstrated that installing suppressor plates at the top of the railing or removing the maintenance rails under the girder significantly suppressed the VIV of steel box girders through wind tunnel tests. Li et al. [
35] and Cui et al. [
36] found that reducing the railing ventilation rate can significantly suppress VIV in box girders. In conclusion, these studies demonstrate that adjusting the railing ventilation rate can effectively suppress bridge VIV. However, there is a lack of research conducted on the railings of double-deck truss girders. Additionally, the VIV of double-deck truss girders suffers from insufficient research data, incomplete understanding of vibration mechanisms, imperfect design specifications, and scarce field monitoring data. This research aims to investigate the impact of railing ventilation rate on truss girder VIV, which will fill in the shortcomings of existing research and provide valuable insights for designing truss girder suspension bridges.
This research investigates the VIV characteristics of the Huangjuetuo Bridge, currently under construction in China, along with the impact of railing ventilation rates and the associated mechanisms. Wind tunnel tests and computational fluid dynamics (CFD) simulations are conducted to achieve these objectives.
Section 2 focuses on calculating the structural dynamic characteristics of the bridge using the finite element method. In
Section 3, wind tunnel tests are carried out to measure the VIV response of the sectional model under different railing ventilation rates during the completion stage and the VIV response of the construction stage cross-section.
Section 4 investigates the mechanism of VIV occurrence and the influence of railing using numerical simulation.
Figure 1 provides a visual representation of the technology roadmap employed in this research.
4. Vortex Shedding Characteristics by Numerical Simulations
With the advancement of computer technology, CFD simulation has become increasingly efficient and accurate. CFD simulation offers the ability to visualize fluid flow characteristics, to handle complex disturbance problems, and to explain the causes of various fluid phenomena. Consequently, it has been widely used in the study of wind-induced vibration of bridges [
50,
51]. Due to the difficulty of establishing an accurate 3D model of the truss girder, this research employed ANSYS Fluent 15.0 [
52] to conduct 2D CFD simulations. A 2D CFD model with a scaling ratio of 1:55, identical to the experimental sectional model, was established. Although the truss girder section was simplified to some extent and discontinuous members such as stiffening ribs were neglected, the equivalent total area of wind action in the vertical direction and the flow field disturbance by the horizontal members were consistent with the experimental model. The simplification was appropriate and improved the calculation efficiency without compromising the accuracy of the results, which is still an effective method to investigate the VIV of bridges [
53,
54].
4.1. Solution Setting and Mesh Division
The CFD simulation utilized transient calculations, with the corresponding mathematical model being the Reynolds time-averaged Navier–Stokes equation. The
SST k-ω model served as the turbulence model, while the Coupled algorithm was used for pressure–velocity coupling [
55]. The second-order windward interpolation was employed as the discrete format, and the computational residual was controlled to 1 × 10
−5. A sufficiently small time step was necessary to capture the more minor non-constant flow field characteristics and to ensure calculation accuracy. However, excessively small time steps would significantly reduce calculation efficiency. After conducting trial calculations, a time step of 1 × 10
−4 s was selected to meet the Coulomb number requirement [
28,
29].
The computational domain mesh was generated using the preprocessing software ICEM 15.0, with the computational domain taking the form of an overlapping external and internal region [
56]. The external region was 15B long and 15D high, with the inlet boundary being the velocity inlet located 5B from the center of the section and the outlet boundary being the pressure outlet located 10B from the center. The upper and lower boundaries were considered free-slip symmetric boundaries, positioned 7.5D from the center of the section. A structured mesh was employed, and the external region was large enough to minimize the boundary’s impact on the internal region’s flow field. The internal region was 2B long and 2D high, using an unstructured mesh. The outer wall of the truss girder section and accessory members were treated as smooth wall boundaries.
Figure 14 shows the boundary conditions of the computational domain. The number of meshes significantly influenced the calculation results, with excessively sparse meshes affecting calculation accuracy and overly dense meshes significantly increasing the calculation volume and reducing calculation efficiency. This research set the maximum mesh size in the internal region to 6 × 10
−3 mm. To meet the computational needs of the
SST k-ω turbulence model, a boundary layer mesh was positioned at the wall surface. The boundary layer consisted of 12 layers with a linear growth rate of 1.2 for the section and accessory components, with the height of the first layer taken as 5 × 10
−5 mm to ensure a smooth transition of the near-wall mesh [
57]. The total number of meshes was 0.22 million, which was sufficient after the trial calculation [
28,
29]. Furthermore, the Y-plus values of the truss girder section were all less than one, indicating that the mesh division met the requirements of the
SST k-ω model [
58].
Figure 15 displays the computational domain mesh division.
4.2. Accuracy Verification of Calculation Results
Truss girders have a continuously changing cross-section with the length direction of the bridge, resulting mainly from the web member position changes. To investigate the impact of web member position on the calculation results, three cross-sections were selected for CFD simulations, with web members located at 0.25D, 0.5D, and 0.75D from the upper chord, as shown in
Figure 16. The simulations were performed using wind angles of attack of +3° and +5°, and a wind speed of 2.23 m/s (equivalent to an actual wind speed of 10.0 m/s, consistent with the wind speed used in the force tests conducted in the wind tunnel). The aerodynamic coefficients were averaged over 5 s after the calculation converged, and the results are presented in
Table 5. The findings indicate that the differences between the aerodynamic coefficients of the three cross-sections are within 5%. Moreover, the web member position does not produce significant differences in the calculation results. Consequently, subsequent calculations were conducted using 2-2 cross-sections.
To validate the model’s reliability, five different wind angles of attack were used to calculate the aerodynamic coefficients at a wind speed of 2.23 m/s (equivalent to an actual wind speed of 10.0 m/s, consistent with the wind speed used in the force tests conducted in the wind tunnel). The calculated results were compared with the wind tunnel test results, and the findings are presented in
Table 6.
Table 6 shows that the maximum error between the simulation and test values is 9.14%, while the minimum error is 0.34%. The drag coefficient has an average error of 3.51%, the lift coefficient has an average error of 3.85%, and the moment coefficient has an average error of 6.15%. The overall average error is 4.50%, indicating that the simulation results have a small error. Therefore, the chosen algorithm and the mesh division of the CFD simulation are feasible and can be used to analyze the flow characteristics near the girder cross-section.
4.3. Numerical Analysis of VIV Performance
Figure 17 and
Figure 18 depict the static flow around vortex distribution at the peak response point of the vertical VIV at +3° and +5° wind angles of attack for the cross-section during the construction and completion stages. The construction stage section demonstrates good aerodynamic performance, with the incoming flow producing a vortex only behind the lower chord. The remaining section does not produce a vortex of sufficient strength and scope. Hence, the wake cannot produce a sufficiently sizeable vortex-induced force, leading to the absence of the VIV phenomenon in the construction stage section. However, the presence of the railing, maintenance vehicle track, and sidewalk corbel causes non-negligible vortices near these accessory structures, which continue to shed to the wake. As a result, the intensity and range of wake vortices are significantly higher than those of the construction stage section. Among the vortex shedding of various frequencies and sizes, one vortex with a shedding frequency similar to the natural frequency of the truss girder drives the VIV at +3° and +5° wind angles of attack in the completion stage section.
In order to find these vortices that cause VIV in the truss girder, 29 measurement points were positioned at various locations close to the girder to record wind speed time-history curves. The placement of the measurement points is depicted in
Figure 19. Specifically, measurement points P1–P13 were situated around the upper deck plate, while P14 and P15 were positioned behind the windward and leeward diagonal bracing members, respectively. Measurement points P16–P29 encircled the lower deck, with P23 and P29 behind the maintenance track.
Table 7 presents the characteristic frequencies of wind speed time-history curves recorded at measurement points and their corresponding Strouhal numbers.
The Strouhal number of measurement points P1–P6 and P13 is 0.51, consistent with the Strouhal number of 0.47 observed at the peak response point of vertical VIV during wind tunnel tests. Similarly, the Strouhal numbers of measurement point P22 are 0.31 and 0.64, corresponding to the Strouhal numbers of 0.36 and 0.64 for minor and major torsional VIV in the tests, respectively. The above phenomenon indicates that the vortex shedding in the wake area of the upper deck or above the upper deck may cause the vertical VIV of the truss girder, while the vortex shedding in the wake area of the lower deck may cause the torsional VIV. Therefore, optimizing the sections in these areas should be prioritized when considering aerodynamic measures to suppress VIV in truss girder bridges.
4.4. Influence Mechanism of the Railing Ventilation Rate
Wind tunnel tests revealed that decreasing the pedestrian railing ventilation rate suppressed the torsional VIV response of the truss girder significantly. The wind speed time-history curve recorded by measurement point P22 indicates a frequency close to the first-order torsional fundamental frequency of the truss girder, suggesting that vortex shedding in the wake region of the lower deck may dominate the torsional VIV. Reducing the railing ventilation rate may disrupt this region’s original pattern of vortex shedding. In order to investigate the mechanism behind this vibration suppression, numerical simulations were conducted on the girder cross-section using various pedestrian railing ventilation rates, and the resulting flow fields were analyzed. The simulations were conducted with a wind angle of attack of +5° and a wind speed of 16.67 m/s (equivalent to an actual wind speed of 16.67 m/s, consistent with the major torsional VIV peak response point wind speed in the wind tunnel tests). This working condition produced the most significant torsional VIV in the wind tunnel tests.
Figure 20 displays the variation in vorticity contours in the wake region of the lower deck in a complete vortex shedding cycle.
Reducing the pedestrian railing ventilation rate alters the vortex shedding mode in the wake region of the lower bridge deck significantly: the vortex intensity at the top of the railing decreases markedly; the blocking effect of the railing increases, and the vortices generated by the airflow passing through the gap of the railing were gradually decreased to zero; the vortex on the sidewalk surface also decreases significantly with the reduction in the gap of the railing until it vanishes when using an airtight railing; and the slowed airflow above the lower bridge deck reduces the pressure difference on the upper and lower sides of the bridge deck, leading to a marked reduction in the strength of the vortex behind the lower chord. Overall, the decrease in the railing ventilation rate reduces the vortex intensity and quantity in the wake region of the lower bridge deck, resulting in weak vortex shedding.
Figure 21 displays the frequency spectrum analysis of wind speed time histories detected at the measurement point P22. The vortex shedding frequency at this point did not change significantly, consistent with wind tunnel tests, but the amplitude decreased considerably, indicating a marked reduction in vortex intensity. When the pedestrian railing ventilation rate was zero, the amplitude of the peak corresponding to the minor torsional VIV was almost zero, possibly due to the significant reduction in vortex because of the inability of flow to pass through the railing, resulting in the nearly complete disappearance of minor torsional VIV in wind tunnel tests. The torsional VIV of the truss girder at low wind speeds may be related to vortex shedding caused by flow passing through the railing gap and vortex shedding at the top of the railing. Furthermore, the torsional VIV at high wind speeds may be related to vortex shedding caused by flow being blocked by the lower chord. By reducing the railing ventilation rate, the vortex intensity and quantity in the wake region of the lower bridge deck decrease significantly, achieving the goal of suppressing VIV. Currently, Chinese design codes provide for the minimum height, railing form, and crashworthiness of bridge railings, but there are no recommendations for aerodynamic performance [
45,
59]. The findings of this investigation suggest prioritizing railings with a low ventilation rate during the design of railings for double-deck truss girder suspension bridges to mitigate potential VIV issues. However, it is necessary to further investigate whether reducing the railing ventilation rate causes other wind-induced vibration problems.