Wind Tunnel Testing for Buildings

Wind tunnel testing, the measurement on which Michael Stern’s slender tower work rests, is a black box to most practising structural engineers, and Roy Denoon, a wind engineer, wrote his guide in Structure magazine to give them the questions worth asking of a laboratory.¹ This entry takes seven of them in turn.

When is a test worth commissioning?

Denoon’s candid answer is that it depends. Codes can estimate structural loads and cladding pressures for ordinary towers, and the code values themselves came out of tunnel work and were written to cover the common run of buildings; their caveats then send the unusual building back to the tunnel as its path to compliance.¹ For a particularly slender tower, or one affected by local topography or by interference from its neighbours, he writes, code methods may not reach adequate reliability at all.¹ Pennoni, a firm that commissions tests rather than running them, draws its own line: it recommends testing to owners once a proposed structure passes ten storeys or has an unusual shape, because the tunnel returns design pressures that are usually lower than the code’s over large areas of the envelope, with savings in windows, walls, roofing, shear walls and foundations, and because it identifies motions occupants would feel and winds that would make a plaza unpleasant.²

What does the tunnel reproduce?

Natural wind near the ground is a gradient, slower at the pavement and faster aloft, and turbulent throughout. A tunnel of the boundary layer kind reproduces that profile instead of a uniform stream, and Pennoni describes the standard arrangement: an instrumented scale model of the building and all its surroundings on a turntable, rotated through every wind direction, with sensors reading both local pressures for the envelope and overall loads for the structure.² The New York Academy of Sciences describes the tunnel at the University of Western Ontario where 7 World Trade Center was tested in 2002: 128 feet long, 11 wide and 8 high, with a wind machine running from breezes to hurricane gusts past a 1 to 500 model roughly eighteen inches high.³

Figure

How a tower goes through a boundary layer test

1. 01
   Build the modelAn instrumented scale model of the building and every surrounding structure is made, at 1 to 240 in the Sarasota example.
2. 02
   Simulate the windThe tunnel reproduces the gradient and turbulence of natural wind near the ground instead of an even stream.
3. 03
   Turn the tableThe model sits on a turntable so every wind direction can be run.
4. 04
   Read the sensorsPressure taps give local design pressures for the envelope and, summed, the overall loads and responses for the structure.
5. 05
   Compare with codeResults are set against the code values, with the laboratory explaining any large difference either way.

What do the three test types measure?

Denoon separates three. The force balance method puts a lightweight rigid model onto a stiff base plate and reads the total forces there, a mechanical summation; the model is inexpensive and quick to modify, which is why it opens the shaping work while the architecture is still in flux.¹ Pressure integration, also run at high frequency, reads hundreds of taps across the surface at once and sums them numerically; its model doubles for the cladding pressure study, which makes it the common choice, though a very slender shaft may have too little cross section to extract enough tubes.¹ The aeroelastic model is different in kind: it reproduces, at scale, the real building’s stiffness, weight and damping and so measures aerodynamic damping, the one quantity the rigid tests cannot, which matters once the rigid tests hint at a large crosswind motion, since close to the shedding peak that damping goes negative and the building moves more than the rigid data would predict.¹

How much of the neighbourhood goes on the turntable?

A great deal: the neighbours alter the result. Denoon’s rule is a ring of neighbours out to 1,200 feet at the least, usually 1,600, at a model scale between 1 to 200 and 1 to 500 depending on the height of the subject.¹ The Academy’s account of 7 World Trade Center shows why: with no master plan yet for the site next door, the engineers ran three models, one with nothing rebuilt and two with neighbours at different heights and orientations, and made their assumptions conservatively.³ Informed Infrastructure describes a Manhattan team printing its tower and the surrounding half mile at 1 to 400, then spending three days in a Canadian tunnel reshaping the foam by hand, producing 19 workable shapes from which the apartment layouts picked one.⁶

Figure

Numbers a design team should expect

Model scale range for a tower, depending on its height

1 to 200 to 1 to 500

Radius of surrounding buildings modelled around the subject

1,200 to 1,600 feet

How closely two laboratories testing the same tower should agree

10 to 15 percent

Aspect ratio, height over width, above which across wind response usually governs

5 or 6 to 1

Can a computer replace the tunnel?

Not yet for a flexible slender tower. The Academy reported in 2006 that computational fluid dynamics let designers study airflow without a model but remained computationally heavy, highly dependent on assumptions and used mostly by the tunnel laboratories themselves; the SOM technical architect it quoted hoped to use it to spot check corners and parapets and then confirm with a physical test.³ Informed Infrastructure, a dozen years later, still found the simulations capable for cladding pressures but not evolved enough for the dynamic response of a slender building.⁶ The other route is data. NIST maintains a public repository of pressure coefficients measured in boundary layer tunnels, contributed first by the Western Ontario laboratory, to support an approach it names database assisted design, which uses measured pressures directly instead of the simplified coefficients in a standard.⁵

How are results checked?

Denoon tells engineers to begin by comparing loads and local pressures with code values and to expect the laboratory to have explained any large difference. Localised suction hot spots above code are common at architectural discontinuities; positive pressures well above code deserve a question unless channelling from upwind buildings explains them; loads far below the code’s 80 percent floor are a cue to ask more.¹ Two laboratories given the same tower ought to land within 10 to 15 percent of each other, and when they do not, the cause is usually the reading of the site’s wind climate, which matters since the dynamic part of the response climbs with at least the third power of wind speed.¹ Testing also extends to the storm types a code barely describes. The Guardian reported Florida International University running downbursts and hurricanes over replica Houston towers at its Wall of Wind facility, where the sideways suction in downbursts explained cladding losses a hurricane never caused.⁷

Figure

Reading a wind tunnel report

1. 01
   Set it against the codeLoads and local pressures are compared with code values, and the laboratory should already have explained any large gap.
2. 02
   Expect suction hot spotsLocal negative pressures above code are common where the architecture breaks, at discontinuities that spin off small vortices.
3. 03
   Question high positivesPositive pressures well above code deserve an explanation, unless channelling between upwind buildings accounts for them.
4. 04
   Question very low loadsLoads far under the code's 80 percent floor are a cue to ask about sheltering by neighbours that may not stay.
5. 05
   Check the wind climateTwo laboratories ought to match within 10 to 15 percent; when they do not, the reading of the local wind speeds is usually why.

Who owns the tunnel?

Usually a specialist consultancy. Peter Irwin, the chairman of one, wrote in Physics Today that the stepped form that confuses the wind on the Burj Khalifa was proven in a tunnel, and that for many of the tallest structures anywhere the across wind forces so measured are the largest they will ever meet.⁸ One architecture and engineering firm chose to build its own. SOM leased space in its Chicago studio for a 45 foot tunnel with fans at one end and an instrumented baseplate at the other, so that a concept could be tested in the earliest weeks rather than after the design was fixed; since 2015 it has been used on more than 40 projects and for the firm’s own research into efficient form.⁴ For most teams the question of who owns the tunnel has a simpler answer: the wind engineer the developer retains, early enough for the shape to still change, which is the subject of the next entry, on shaping the tower.