Wind Barriers Shield Bridge Traffic but Shake the Deck, Wind Tunnel Tests Show

Wind barriers along the decks of long-span bridges are meant to keep crosswinds from shoving trucks and buses into dangerous sideslips or rollovers, but a new wind tunnel study shows that the same barriers that calm the air for traffic can shake the bridge itself. Researchers at Yokohama National University tested two box-girder sections derived from the Trans-Tokyo Bridge at 1:100 scale in a closed-circuit wind tunnel with a maximum wind speed of 40 meters per second and a test section measuring 1.8 by 1.8 meters. The two sections differed in side ratio, the ratio of deck width to girder depth: a thick girder with a side ratio of 3.44 and a thin girder with a side ratio of 4.98. The team, led by Phan Duc Tam with Anousit Vilaivong, Lam Trong Hoang and Vo Duy Hung, published the work in Advances in Bridge Engineering.

The experimental program was deliberately broad. Rigid sectional models mounted on stiff supports and instrumented with a high-precision load cell measured aerodynamic force coefficients at wind speeds of 4 and 7 meters per second, across angles of attack from minus 10 to plus 10 degrees in one-degree increments. Spring-mounted sectional models with two degrees of freedom underwent free-vibration tests for vortex-induced vibration, or VIV, and flutter, with model properties matched to similarity requirements for mass, mass moment of inertia, reduced frequency and damping ratio. Fixed models carried an X-type hot-wire anemometer that recorded wind speed and direction simultaneously, sampling for 30 seconds at 1,000 hertz under a uniform 4-meter-per-second flow, with measurement points stacked up to a prototype height of 12 meters above the deck along traffic-lane centers, handrail positions and the deck midline.

The barrier variables were just as systematic. Prototype barriers 3 meters tall were tested with holes arranged in vertical and horizontal orientations, at porosities of 25, 50 and 70 percent, where porosity is the percentage of open area divided by the total barrier surface. Installation method was compared by mounting barriers only above the handrail versus extending them over the full handrail height. Mitigation configurations included a 50-percent-porosity barrier with non-uniform horizontal spacing, wider openings near the bottom and narrower ones near the top, designated 50-H-DS, plus single and double flap additions, and for the thin girder a reduced barrier height of 2.5 meters.

On the force side, wind barriers generally increased drag, with the 25-percent-porosity barriers producing the highest drag coefficients because of flow blockage, while the 50-percent design delivered drag values around 1.3 times those of the bare deck. Lift coefficients rose at negative angles of attack, most strongly for the 25-percent barriers, but every configuration showed a negative lift slope, and the galloping stability index remained positive in all cases, indicating no risk of galloping for the thick girder. Partial installation above the handrail produced slightly higher drag at negative angles, especially for the 25-percent barriers, and the double-flap configuration produced the largest increase in the moment coefficient. The 50-H-DS barrier had drag coefficients nearly identical to the standard 50-percent barrier, suggesting that porosity, not hole spacing, governs drag.

The aeroelastic results told a more complicated story. As porosity decreased, vertical VIV amplitudes grew noticeably for both girders. Relative to the bare deck at zero degrees angle of attack, the average peak vertical response increased by roughly 3.3, 1.8 and 1.3 times for porosities of 25, 50 and 70 percent. At plus 3 degrees, the penalties became far larger, at approximately 6.2, 4.0 and 3.4 times the bare-deck response, confirming that barriers do their worst damage to vertical VIV when the deck faces a positive attack angle. Barriers also shifted the wind-speed range over which VIV develops: for the thick girder the onset moved from 18 to 22 meters per second at zero degrees, and the thin girder behaved similarly, with onset moving from 16 to 20 meters per second at zero degrees and from 18 to 22 meters per second at plus 3 degrees.

Flutter margins also narrowed as barriers went up. For the thick girder at zero degrees, the critical flutter speed fell from about 110 meters per second for the bare deck to 98, 94 and 82 meters per second for porosities of 70, 50 and 25 percent. The thin girder followed the same trend, dropping from roughly 170 to 135, 120 and 80 meters per second. The one exception came at plus 3 degrees for the thick girder, where barriers slightly raised the critical velocity from around 76 meters per second to above 81 meters per second. Torsional behavior was more forgiving: no torsional VIV was observed for any barrier configuration on either girder within the tested wind-speed range, and torsional flutter occurred only at wind speeds exceeding approximately 80 meters per second, with the 50-percent barrier suppressing torsional VIV most effectively.

The flow-field measurements explained why the thin girder is the more vulnerable of the two. Across all configurations, wind speeds within the 0-to-4-meter height range that encompasses most road vehicles were consistently higher over the thin girder than over the thick girder. Normalized mean wind speed reached unity at heights above about 4 meters for the thick girder but at around 2 meters for the thin one, and peak wind speeds in the windward lanes reached up to three times the approaching flow over the thin girder, compared with roughly twice the incoming wind for the thick girder. Flow visualization showed the thick girder producing compact velocity vectors with limited vertical dispersion, a broader wake and higher turbulence intensity near the surface, while the thin girder allowed flow reattachment and a narrower wake. Installing the 50-percent barrier transformed the thin girder’s flow completely, eliminating reattachment and shifting the flow acceleration region upward, a change the authors link to the limited effectiveness of mitigation measures on that section.

The barriers did deliver real shielding. Barriers with 25 and 50 percent porosity significantly reduced both mean and peak wind speeds in the critical 0-to-4-meter band. In the first traffic lane, mean wind speed fell by roughly a factor of three for the thick girder and two for the thin girder, while peak speeds dropped by about four times and 3.2 times respectively. For the thick girder, shielding was nearly identical at both porosity levels, whereas the 25-percent barrier gave slightly stronger reduction over the thin girder. Not every mitigation helped, though: widening the lower perforation spacing and adding a single flap raised wind speeds below about 2.5 meters in the windward lanes of both girders, potentially affecting light and medium-height vehicles, and reducing the barrier height to 2.5 meters on the thin girder gave slightly weaker shielding below 2 meters than the 3.0-meter barrier.

The clearest practical finding emerged from the trade-off analysis between shielding and vibration. The 50-H-DS barrier, with its wider openings near the bottom, consistently produced smaller peak vertical VIV amplitudes than other configurations. At zero degrees angle of attack it yielded a peak VIV amplitude of 182.84 millimeters for the thick girder, well below the 386.53 millimeters of the standard 50-H-3.0-meter barrier and the 237 millimeters of the bare deck. At plus 3 degrees it produced 575.33 millimeters against 787.69 millimeters for the standard barrier. Notably, with the 50-H-DS barrier installed, results were nearly identical for both side ratios, which the authors suggest reflects reduced vortex-shedding coherence near the deck surface where permeability is increased. The authors also note that a configuration is considered acceptable when vertical VIV amplitude is comparable to or lower than the bare-deck case and the critical flutter wind speed remains above the bridge-specific threshold of 67.7 meters per second.

Overall, the study concludes that porosity is the dominant parameter governing the trade-off between wind shielding and aeroelastic stability: lower porosity enhances wind attenuation but worsens VIV and flutter penalties, while higher porosity eases aeroelastic effects at the cost of weaker shielding. The thin girder proved more sensitive to barrier-induced aerodynamic effects, experiencing higher near-deck wind speeds and larger vertical VIV amplitudes, and mitigation strategies that worked well for the thick girder offered only limited benefit for the thin one. Among all configurations examined, the 50-percent-porosity barrier, especially combined with the non-uniform permeability distribution of the 50-H-DS design, provided the most balanced performance, achieving meaningful wind-speed reduction while maintaining acceptable aeroelastic behavior, and the authors recommend it as a practical compromise for bridge regions most exposed to crosswinds, such as mid-span. The findings are limited to the two sectional geometries tested at model scale in uniform flow, and the authors note that Reynolds-number effects were assumed secondary following common practice in bridge aerodynamic studies.

Subject of Research: Wind tunnel testing of wind barrier configurations on box girder bridge aerodynamics and aeroelastic response

Article Title: Wind tunnel study of the effect of wind barrier configurations on box girder aerodynamics and flow characteristics with different side ratios

Article References: Tam, P. D., Vilaivong, A., Hoang, L. T., & Hung, V. D. (2026). Wind tunnel study of the effect of wind barrier configurations on box girder aerodynamics and flow characteristics with different side ratios. Advances in Bridge Engineering, 7(1), Article 29. https://doi.org/10.1186/s43251-026-00218-3

Image Credits: AI Generated

DOI: 10.1186/s43251-026-00218-3

Keywords: wind barriers, box girder, bridge aerodynamics, vortex-induced vibration, flutter, wind tunnel testing, porosity, side ratio, crosswinds, vehicle stability, aeroelasticity, long-span bridges