NEW CETINA BRIDGE (Omiš, Croatia)

From vehicle safety to structural response: why an extensive campaign was needed

The wind tunnel campaign was carried out to support the design of the new bridge over the Cetina River in Croatia, a steel girder bridge characterized by a main span crossing a deep valley and directly connected to two tunnels at its ends. In such a context, traffic safety under strong winds represents a critical design issue, especially for northerly winds that can be significantly accelerated by the local orography. To mitigate this problem, the Client proposed the installation of a tall protective wind barrier along the deck.
The experimental programme was therefore conceived to assess, within a single and coherent framework, the aerodynamic loads acting on the bridge, the aeroelastic behaviour of the deck, and the actual effectiveness of the barrier in sheltering vehicles.

Definition and role of the sectional model

A sectional-model approach was selected as the most appropriate tool to investigate both steady and unsteady aerodynamic phenomena governing the bridge response. The sectional model was designed at a geometric scale of 1:65, chosen as the best compromise between blockage constraints in the test chamber and the need to reproduce in sufficient detail the deck geometry, parapets and wind barrier.

A key aspect of the experimental strategy was the definition of a single, modular sectional model, capable of serving all phases of the campaign. Two representative deck cross-sections were identified as critical by the designers: the midspan section (x/L = 0.5) and the one-quarter-span section (x/L = 0.25), which mainly differ in the lower geometry of the steel box girder. Instead of building separate models, the sectional model was conceived so that the lower aluminium plate could be replaced, allowing both cross-sections to be reproduced while preserving the same upper geometry, sidewalks, parapets and barrier system. This ensured full consistency between static, aeroelastic and flow-field tests.

The model was predominantly manufactured in aluminium to guarantee rigidity during static force measurements and a well-defined mass distribution for aeroelastic tests. Secondary elements such as sidewalks, pedestrian parapets and guardrails were reproduced using high-resolution 3D printing. Sharp edges and geometric discontinuities were accurately modelled, as they govern flow separation and dominate the aerodynamic behaviour of bluff bridge decks. End plates were employed to suppress three-dimensional edge effects and enforce quasi-two-dimensional flow conditions, consistent with sectional modelling assumptions.

The same sectional model was used for both static and dynamic tests, ensuring that aerodynamic coefficients, aeroelastic stability and flow-field measurements could be interpreted in a fully coherent manner, without ambiguities associated with different model geometries.

Wind barrier and configuration variants

The sectional model was equipped with a laser-cut aluminium wind barrier reproducing the geometry proposed for the real bridge, including height, porosity and position relative to the deck. Since even minor geometric details may influence the flow, an initial sensitivity study was carried out on the barrier thickness. This allowed verifying that the observed aerodynamic effects were governed by the overall barrier layout rather than by secondary construction details, and ensured that subsequent results could be confidently attributed to the presence or absence of the barrier itself.

In addition to the reference configuration with wind normal to the bridge axis, the experimental programme also addressed skew wind conditions, which are unavoidable in a site characterized by complex terrain and flow channeling. For this purpose, a dedicated solution was adopted: yawed appendices were designed to be directly mounted at the ends of the sectional model, allowing tests at a yaw angle of 45° with respect to the incoming flow (see Figure below). This solution preserved full compatibility with both static and aeroelastic rigs and allowed direct comparison between aligned and skew wind conditions without modifying the core model or its support systems.

Static aerodynamic tests and load characterization

The first phase of the experimental campaign focused on the determination of static aerodynamic force coefficients acting on the bridge deck. Tests were performed for different deck cross-sections, with and without the wind barrier, for wind coming from both North and South, and under smooth and turbulent flow conditions. A broad range of angles of attack was investigated to identify critical trends in the aerodynamic response.

The static results clearly showed that the installation of the wind barrier leads to a significant increase in aerodynamic drag, particularly when the barrier is directly exposed to the incoming northerly wind. The largest load increments were observed for the deck configuration with the greatest cross-flow dimension. These results quantified the structural cost associated with the barrier in terms of increased wind loads, highlighting the trade-off between traffic protection and structural demand.

Aeroelastic tests and dynamic stability

Following the static characterization, the campaign moved to aeroelastic testing to assess the susceptibility of the bridge to wind-induced vibrations and possible dynamic instabilities. In these tests, the sectional model was mounted on an aeroelastic rig in which the aeroelastic setup consists of two shear-type frames supporting the sectional model; stiffness is assigned by leaf springs, thus constraining the model to move only vertically.

Particular care was devoted to the dynamic identification of the system. Free-decay tests in still air were systematically performed whenever a configuration change occurred (e.g. installation or removal of the barrier, reassembly of the setup, continuation of tests on different days), allowing accurate estimation of vibration frequency and damping. This was essential to control the Scruton number and ensure meaningful comparison with full-scale conditions.

The aeroelastic results revealed critical behaviour. With the wind barrier installed and exposed to northerly winds, the bridge exhibited a violent transverse instability at wind speeds significantly lower than the assumed design reference. The onset of this instability coincided with vortex-resonance conditions, indicating a combined mechanism involving vortex-induced vibration and galloping. Importantly, such behaviour could not be reliably predicted based solely on quasi-steady static coefficients.

Even without the barrier, the bridge was found to be unstable for a small positive angle of attack, again at wind speeds below the design reference. In this case, the growth of oscillation amplitude was less abrupt, but still led to unacceptable vibration levels. These findings highlighted the strong aerodynamic sensitivity of the deck configuration and demonstrated that the barrier, while beneficial for traffic safety, can exacerbate dynamic instability issues unless additional mitigation measures are adopted.

Buffeting response assessment

The static force coefficients and aerodynamic derivative 𝐻1 information obtained experimentally were subsequently employed in buffeting analyses to estimate the bridge response to turbulent wind. The calculations confirmed that the increased aerodynamic loads associated with the barrier lead to higher predicted vibration levels. More importantly, they supported the experimental evidence that the barrier-equipped configuration has a stronger tendency toward instability-like behaviour, while also emphasizing the limitations of a purely quasi-steady approach in capturing unsteady aerodynamic effects observed in the wind tunnel.

Flow-field measurements: hot-wire anemometry and PIV

To directly assess traffic safety, the experimental campaign included a dedicated investigation of the flow field behind the wind barrier. Measurements were performed using both hot-wire anemometry and Particle Image Velocimetry (PIV), providing complementary information and mutual validation of the results. Tests were carried out with and without the barrier, for wind coming from North and South, and for aligned and skew flow conditions.

The flow-field measurements demonstrated that the wind barrier produces a clear reduction of the mean wind velocity over the roadway, particularly in the height range relevant for high-sided vehicles. This confirmed the effectiveness of the barrier in sheltering traffic. At the same time, both hot-wire and PIV results revealed a complex wake structure behind the barrier, characterized by significant velocity fluctuations and non-uniform flow patterns. The comparison between smooth and turbulent oncoming flow showed that the sheltering effect persists under realistic turbulence levels, while differences in velocity fluctuations become more pronounced.

Vertical profiles of mean velocity and velocity fluctuations provided a quantitative description of the flow environment experienced by vehicles, allowing the benefits of the barrier to be weighed against the increased aerodynamic complexity introduced by its wake.