Evaluating parallel bridges through wind tunnel studies and analysis
The plan put forward in the 1990s to build a second parallel structure to double the capacity of the 1950 Tacoma Narrows Bridge was undertaken with great care, and for good reason. When the original Tacoma Narrows Bridge opened in 1940, it provided a much anticipated link across Puget Sound. At the time, the elegant, light and narrow structure was the third longest suspension bridge in the world. However, just four months after its opening, sustained 42 mile-per-hour (68 kilometer-per-hour) winds led to severe aeroelastic flutter of the bridge, which soon ended in its spectacular collapse.
In the wake of the bridge failure, an extensive series of wind tunnel tests and theoretical studies at the University of Washington generated much new information about wind effects on long-span bridges. The lessons learned were incorporated into the design of a new suspension bridge that was constructed at the same site and opened a decade later, in 1950.
By the 1990s traffic had increased substantially, giving rise to the idea of boosting capacity by building a new bridge adjacent and parallel to the 1950 one, with each bridge carrying traffic in a single direction. The construction consortium responsible for the new bridge, Tacoma Narrows Constructors, contracted the design of this new bridge, as well as changes to the existing bridge to accommodate the changes of the traffic arrangement, to a joint venture of Parsons Transportation Group and Howard Needles Tammen and Bergendoff.
We were brought on board to investigate the aeroelastic stability of the proposed bridge, its potential aerodynamic effects on the 1950 bridge, and the effects of modifications that were planned for the 1950 bridge.
The planned close proximity of the new bridge to the 1950 bridge created a risk of aerodynamic interaction between the two structures. Also, in order to accommodate lane adjustments on the existing bridge, there was a proposal to cover a number of open gratings that had been included in the 1950 bridge for aerodynamic reasons. The designers were concerned that eliminating the openings might reduce the bridge’s aeroelastic stability to an unacceptable degree. We were engaged to evaluate the aerodynamics and provide detailed evidence about performance of the proposed design.
We adopted a four-part approach to this investigation.
1. Wind Climate and Exposure
One aspect of our work was to determine the wind conditions the bridge would experience. We needed to understand local wind speeds as well as the angles at which wind could be expected to strike the bridge.
Our analysis incorporated:
- wind climate data from three nearby airports with long historical wind speed records: Tacoma Narrows Airport, just 3 km away from the bridge; McChord Air Force Base, 19km southwest; and SeaTac International Airport, 27km northeast.
- the results of smoke tests performed onsite shortly after the 1940 collapse and other related research to establish test criteria for the angle of the wind hitting the bridge.
- analysis of local terrain and its likely effects on wind flows toward the bridge.
Our investigation concluded that the bridge deck should be designed to remain stable at wind speeds up to 118 miles (188km) per hour for all angles of attack up to 15 degrees. Earlier evidence had suggested that an angle of plus or minus 5 degrees would be sufficient but our analysis--and the consensus view that a conservative approach was important--called for a wider range. This first phase of work enabled us to establish parameters for tests of the bridge’s design.
2. Sectional Model Tests
We built sectional models of the existing and the proposed bridge at a scale of 1:50 to investigate the stability of the two deck sections against flutter and vortex shedding, both of which had been significant in the 1940 collapse. We tested both sections in smooth and turbulent wind flows; neither bridge deck exhibited any significant vortex-induced oscillations. Both bridges maintained positive aerodynamic damping up to very high wind speeds, which indicated good stability. In addition, the presence of the proposed bridge on the upwind side generally improved the stability of the existing bridge, and variations in the angle of attack did not cause the structures to violate stability criteria.
It is not possible to fully simulate wind turbulence effects on a sectional model, due to its limited length and its large size relative to the wind tunnel. However, we conducted a partial simulation of high-frequency turbulence on the sectional models to evaluate the sensitivity of their stability to such turbulence. We found that turbulence had beneficial effects on stability.
As expected, our tests showed that closing the deck vents on the 1950 bridge would result in the onset of flutter at lower wind speeds. But provided the frequency of the lowest torsional mode was high enough, the bridge would still meet the established flutter criteria. It therefore became important to be sure of the frequency of the lowest torsional mode of vibration; this led to our third phase of analysis: vibration measurements at full-scale on the 1950 bridge.
3. Full-Scale Testing of Dynamic Properties
Our full-scale tests included recording the structure’s response to ambient excitation by traffic, as well as and wind and forced-vibration tests, for which we used a large pendulum to excite the bridge in the absence of traffic. We placed six highly sensitive accelerometers at two stations, one at the mid-span of the bridge on the main span and a second 170m west. We also continuously monitored wind conditions during the field program, recording the response of the bridge to normal wind and traffic over one week.
The phase we found between the two accelerometer-equipped stations (one at the mid-main span and the other close to the span quarter) was close to zero, indicating that the identified torsional modes were symmetric. No other torsional modes were evident. We therefore concluded that the first torsional symmetric mode was at 0.33 Hz, and that the existing bridge was stable against flutter even with the air vents closed off.
4. Aeroelastic Testing to Capture Overall Structural Behaviour
Although the sectional model tests gave good indications that the two parallel decks would be free of vibration problems and have sufficiently high flutter speeds, for testing purposes it was necessary to simplify some phenomena. For example, in sectional tests, only one of the two deck models was moving dynamically at any given time. We also needed to investigate the effects of wind direction more fully. For these reasons, we built full aeroelastic models of both bridges and tested them side by side in the wind tunnel, placing accelerometers at the tops of the towers and strain gauges at important locations, such as the bases of the towers. We measured deck deflections using laser displacement transducers. The models were installed on a turntable on the wind tunnel floor, allowing us to rotate them relative to wind direction. Using 45 channels of instrumentation, we recorded wind speeds at various locations.
Taken together, the studies showed that despite their close proximity, the two bridges would not experience any major adverse aerodynamic interference effects. They also demonstrated that on the existing bridge it would be possible to close off the aerodynamic vents, originally put there to enhance aerodynamic stability, and still satisfy the applicable criteria. Our studies also generated detailed wind load distributions, which assisted the designers in developing safe and economical structural solutions. The new parallel bridge was successfully completed in 2007. Both bridges continue to serve the area’s transportation needs well and without incident.