New Developments in Commercial Wind Engineering

The primary intent of this invited presentation at the International Workshop on Wind Engineering and Sciences is to discuss new developments in commercial wind engineering, to pose some questions that may (hopefully) lead to some conversations between attendees, and perhaps to speculate on the future of our peculiar specialty. Many of us gather every four years at the International Conference on Wind Engineering in order to discuss the minutia of wind engineering developments, but what is happening to our field in a holistic sense? What could we do to make our work more useful to the public or our clients? Some items presented herein are minor and can easily be dismissed as just my idiosyncratic view of consulting, engineering and ethics. However, some issues are definitely more serious and need to be tackled by consulting wind engineers, and to some extent researchers, in the near future. Whilst attempting to develop some points of consequence, nay controversy, perhaps we can have some fun too. 1 MODEL CONSTRUCTION In recent years the use of stereolithography (SLA) to build wind-tunnel pressure models has largely superseded the traditional, machined Plexiglas pressure model. As architectural designs become more complex, the ability to generate the dual-curvature shapes using programs like AutoCAD and SolidWorks allows the pressure tap paths to be incorporated into the design before the laser-induced growth of the model commences in the stereolithography vat (Figure 1). There is some skill in knowing the best way to design the pressure model components for useable pressure path lengths, strength in construction and optimal material volume, but the competitive cost of this technique means that about 80% of pressure models are now built using this method at CPP Inc. An example of a finished, curved, SLA, pressure model is shown in Figure 7. The ability to make very thin shells for lightweight high-frequency force balance (HFFB) models is also a useful advantage for this model construction technique (see Figure 8). 2 VELOCITY MEASUREMENTS The measurement of mean and peak wind speeds in the wind tunnel has traditionally been done with the hot-wire or hotfilm anemometer; the former being more responsive, but less robust. This well established technology (Schubauer and Klebanoff, 1946; Sandborn, 1972 and 1981) shown in Figure 2 now has some competition from the highly-responsive, multi-hole pressure probes, such as the Cobra Probe in Figure 2: The hot-film anemometer has traditionally been used to collect turbulent velocity data in the wind tunnel. Figure 1: The laser booth and operational console of a stereolithography machine. Figures 3, 4 and 6. These devices typically have between 4 and 13 holes in the head along with the miniature pressure transducers (yielding good frequency response), and can measure speed and direction over angular ranges that incorporate some level of reverse flow in separated regions. The ability to respond to the reverse flow condition depends to some degree on the number of holes (and so transducers) in the probe design. The software that comes with the probe evaluates the many pressure signals to yield a time-series of flow speed and direction. Figure 5 shows profiles developed near, and downwind from, a 90-degree step. In recent years many commercial laboratories have discovered the usefulness of the multi-hole pressure probe, even in highly turbulent flows. The wind-engineering community is still evaluating this relatively new technology, but it is likely to be the way of the future for mean and gust velocity measurements in the wind tunnel. Laser velocimetry (Figure 7) has been used routinely in fluid mechanics for at least two decades, but it is now starting to appear in the commercial side of wind engineering to help clients with particular problems. Figure 7 shows this device being used to define the wind speeds and directions around an open lattice structure at the newly refurbished Pennsylvania Railway Station in New York City. These fine measurements can be made at the location of the lattice or open structure without interfering with flow that was being measured. Laser velocimetry is usually the more expensive option, but with a good Figure 3: Four-hole Cobra probe in use at HKUST (after Dr. Peter Hitchcock). Figure 6: Schematic of the four-hole Cobra Probe (after Turbulent Flow Instrumentation Pty. Ltd.). Figure 5: A composite photograph of flow visualization over a vertical escarpment and the vectorized mean velocity profiles measured by a multi-hole pressure probe. The scales of the flow visualization and profile images are roughly the same (after Noriaki Hosoya, CPP Inc.). Figure 4: Close-up of a four-hole probe (after Turbulent Flow Instrumentation Pty. Ltd.). understanding of the winds around the lattice and knowledge of the member shapes the structural wind loads can be better assessed by the structural engineer. 3 FORCE MEASUREMENTS Even with the popularity of simpler and cheaper aerodynamic models (both the high-frequency force balance technique and the simultaneous pressure approach) to assess dynamic structural loads there is still the occasional unconventional project that requires a fuller exploration of the nonlinear relationship between the structural response and the forcing function via an aeroelastic study. Two interesting examples of this “Rolls Royce” analogue solution to the differential equations of motion are the Titan V Launch Vehicle (Figure 8) prior to liftoff and the architecturally decorative Houston Arches (Figure 9). The potential wind loads during the critical moments prior to the launch of any orbital vehicle may vary greatly with the arrival of an unexpected front or thunderstorm. In this study these load probabilities were assessed for a variety of meteorological conditions, positions of the Mobile Service Tower and fuel masses in the vehicle. The last condition provided a challenge for the aeroelastic model construction, particularly when there was no fuel load. The mass scaling parameters in this condition dictated that a very light thin shell be built. A variety of approaches were tried, including stereolithography and spun carbon fibre. With some experimentation a very thin payload shell was built using the finer limits of the stereolithography machine. Of course, the traditional issues of surface roughness and Reynolds Number for these circular cylinders came into play as well. A roughened, black surface in the payload area can be seen in Figure 8. Figure 7: Laser velocimetry in the Meteorological Wind Tunnel at Colorado State University. Figure 9: Aeroelastic model of the Galleria Arches in Houston, Texas. Figure 8: Aeroelastic model of the Titan V Launch Vehicle with the Mobile Service Tower backed away to the rear. A more Earth-bound, but equally interesting aeroelastic study was that of the Houston Galleria Arches in Figure 9. This public art spanning a major thoroughfare in commercial Houston had an interesting aerodynamically interactive response that required aeroelastic modeling. The aeroelastic models, fitted with very small accelerometers, responded to their own vortex shedding as well as the turbulence flowing off the upwind arch. Seven modes were effectively reproduced with this aeroelastic model. The final result was an elegant, fullscale, span of two 600 mm stainless steel tubes across the six-lane road in Houston, Texas. The vast majority of buildings do not require the elegance of the aeroelastic approach to assess useful design wind loads, and so these projects may be evaluated using an aerodynamic model. In essence, this technique seeks to obtain the external loading (base-moment time series) on a given building shape via a light, stiff model in the wind tunnel, after which the dynamic response may be calculated in the time and/or frequency domain for any desired combination of mass, stiffness, damping ratio and wind speed. The structural engineer finds this methodology valuable since revised dynamic properties may be applied to the base-moment spectra or time-series data without returning to the wind tunnel, provided that the external building shape remains unchanged. This encourages a more economic and iterative design scenario for the structural engineer. Many attendees will be fully familiar with this approach, but those who wish to read more should read papers on the topic by Boggs (1992) and many others in the windengineering literature. However, what is relatively new in wind-tunnel testing is the availability of cheap pressure transducers. As a consequence, many laboratories can apply 500 to 1000 transducers to a pressure model and collect pressure timeseries data, essentially simultaneously, over the entire building. To obtain the same base moment data as the force balance one needs to assign tributary areas, and moment arms to the axes for each of the taps – effectively a substantial accounting problem. From that point on the data-reduction is almost identical to the high-frequency force balance technique. The obvious advantage to this approach is only the pressure model needs to be built (Figure 10), and the lightweight balsawood (typically) force-balance model is not needed. There are, however, less obvious advantages. The high-frequency force balance theory is dependent upon linear mode shapes in bending, whereas in reality the building may have a mode shape with some curvature. This is even more of a concern for torsion, which should be approximately linear with height in the full scale but is constant with height on the force balance. Correction factors for these two criticisms of the high-frequency force balance are available in the literature, but the simultaneous pressure approach offers a Figure 10: Simultaneous pressure data collection applied to a midrise condominium in Miami (510 taps). Figure 11: New Miami Air Traffic Control Tower with the old tower in the background. way to accommodate these mode-shape issues via weighting the pressure data according to the true mode shapes of the full-scale structure. For long, lowrise buildings (Figure 10) the high-frequency force balance will generate base moments contaminated by roof uplift pressures at the building extremities, well removed from the axis of rotation. The structural engineer does not want this impacting the horizontal loads on each floor. Those roof uplift forces are accommodated elsewhere in his design. For tall buildings (Figure 11 is an extreme example), with a relatively small footprint, this effect is imperceptible. The simultaneous pressure technique removes this problem since the experiment can be designed to take simultaneous data from wall taps only. This observation is fortuitous since it results in a useful and practical demarcation between times the highfrequency force balance is preferred over the simultaneous pressure approach. Tall building models tend to have a small internal volume, for pressure tubing, and so the force balance is preferred on that pragmatic basis. Conversely, the squat buildings do not lend themselves to the force balance and they have plenty of volume for tubing. Figure 12: Base-moment comparison between data taken using the high-frequency force balance and simultaneous pressures on 1:100 models of the new Miami Air Traffic Control Tower in Figure 11. The obvious question any structural engineer would ask is “do both techniques result in the same design loads?” Additionally, the wind engineer would like to know how many taps are needed to generate reliable design data. At CPP we have compared data collected using both the high-frequency force balance and simultaneous pressure for a variety of building shapes and surroundings. Those studies have suggested a relative insensitivity to the actual number of taps used – a sufficient number to capture the cladding data appears to be more than adequate for the integrated structural loads. The AWES Quality Assurance Manual (2001) also has some guidance of the number of taps needed. By way of example, Figure 12 shows the comparison of a 1:100 balsawood highfrequency force balance model of the new Miami Air Traffic Control Tower (ATCT) with a 1:100 pressure model of the same tower (note that 1000 k-ft = 1.356 MNm). This is a tall slender structure in an open upwind environment that would typically not be studied by the simultaneous pressure technique. However, it is useful in exploring how many taps are needed for such a simple prismatic shape. There were about 190 taps in the pressure model (all that could be placed inside the ATCT stem) for simultaneous pressure data collection. The relatively regular shape of the ATCT and lack of interfering structures resulted in a good data match even with these few number of taps. Similar comparisons with more complex buildings in more complex surroundings have produced comparable results with about 500 to 700 taps. Figure 12 also shows data for a 1:225 model of the same ATCT tested at an earlier time. An observant reader will notice that easterly flows generate somewhat different peak base moments (My) over a range of azimuths for the two 1:100 studies when compared to the 1:225 study. This is due to a 45-degree alteration in the understanding of the orientation of the existing ATCT upwind (Figure 11) between the 1:225 and 1:100 tests. This came about from better surrounding photographs in the second study when the orientation of the four legs of the existing ATCT became apparent. Many comparisons have been made between these two approaches in more complex urban environments. Figure 13 shows an extreme example, with comparably tall buildings very close to the subject building. The mean and peak base moment coefficient data are compared in Figure 14 and the spectral responses are in Figure 15. In this case only 290 taps Figure 13: Thirty-storey Florida condominium, with tall proximate neighbours, used to compare balsawood HFFB (shown) data with the simultaneous pressure technique. Figure 14: Mean and peak base moment coefficients about the x and y axes using both techniques. on the pressure model of the tower were used. The result is fair, but the southerly flow impacting the My base moment indicates an underestimation on the mean and peak base moments, probably due to the low number of taps used. Data like this have been used to suggest a lower bound to the number of taps needed. Interestingly, the fluctuating component of the load (distance between the mean and peak loads in Figure 14 and the sample spectra taken from the 200 degree load case in Figure 15) is in good agreement between the two techniques. These data, and other in-house studies, have led CPP to use between 400 and 700 taps in the typical simultaneous pressures study of a new midrise building in a complex cityscape. 4 NEW CLIENTS AND SERVICES The arena of new developments in commercial wind engineering is influenced greatly by the needs of the client. One growing new area is forensic wind engineering with client driven needs as varied as glass or louvre failures in relatively modest winds to court cases concerning deaths resulting from wind-induced crane failures. The wind tunnel can produce crucial and convincing data for the legal fraternity in these areas of science and engineering. Some clients wish to explore their own product research in the wind tunnel. Vortex fences and spoilers (Cochran Cermak and English, 1995; Cochran, 2004; Banks Sarkar Wu and Meroney, 2001) have been applied to critical-use buildings in hurricane areas of the United States. In a similar manner new designs for vertical and horizontal axis wind turbines (Figure 17) are more commonly being investigated on a consulting basis in the wind tunnel and in the field. Small-scale terrain models are often used to assess the best approach conditions for a building sited in, or adjacent to, complex terrain. Those measured profile data are then Figure 17: Full-scale wind turbine design is used to confirm model studies in the wind tunnel and initial CFD input. Figure 16: Spoiler on the edge of the TTU Building in Lubbock, Texas (after Banks Sarkar Wu and Meroney, 2001). Figure 15: Mx and My spectra from the HFFB and 290-tap simultaneous pressures at 200 degrees. used to define the appropriate profile upwind of the subject-building turntable at, say, 1:400 or 1:500. For these small-scale studies Meroney (1980) suggested a model scale limit of about 1:6000 while Bowen (2003) suggests about 1:5000. The latter discussion lists a thought-provoking array of shortcomings associated with these small-scale physical terrain models. Obviously as the scale reduces the more significant turbulence wavelengths fall prey to viscous dissipation. Is this a serious concern for the designer of the experiment? Is the omission of the Coriolis-induced Ekman spiral a shortcoming of consequence as the modeled area increases? When is the loss of Coriolis forces acceptable in complex terrain flows? When is ignoring possible full-scale variation in atmospheric stability diminishing the value of the physical model study? Strong winds, perhaps? When is the loss of gravity waves and the consequent asymmetry on either side of a mountain range, created by stability in the atmosphere, acceptable in the modeling process? The use of a stepped model is commonly used to replicate, or perhaps artificially exaggerate, the true surface roughness at these small scales. This may be reasonable for gross profile assessment as in Figure 18, but what if data closer to the surface are needed? Should the steps be smoothed out and the Reynolds Number mismatch of several orders of magnitude be accepted? Large-area flows like these might now be better modeled using nested, mesoscale, numerical models that have their origins in the field of atmospheric science. CPP is currently comparing profile wind data from 1:4000 terrain models with those generated numerically. The numerical runs may be performed with neutral stability and no Coriolis forces to replicate the conditions in the wind tunnel (for initial comparative purposes) and then, if satisfactory, run again with these pieces of physics turned on for a truer picture of flows over complex terrain. It may be that once this approach is validated the use of small-scale physical models may be used far less. Perhaps this will be the first practical use of Computational Wind Engineering (CWE), rather than the dubious solutions to flows around buildings in an urban environment with all the inherent turbulence modeling concerns. Another trend in consulting wind engineering, which seems likely to continue, is the combination of complex architecture (Figure 19) and reduced real costs of a typical windtunnel study. This has caused many mid to lowrise buildings to be tested for cladding and structural loads. It is not uncommon for buildings in the eight to twelve-storey range to be put in the wind tunnel. Even a few exotic, expensive (20 M$), single-storey homes have been tested in the wind tunnel, although this is not a common client. As more condominium developers realize the benefits for their design this trend will continue, particularly in hurricane prone areas. For residential towers some developers actually use the pressure model and a flow-visualization DVD of the testing in their display unit as a selling point, emphasized to potential purchasers. Consulting in wind engineering will occasionally expose you to a structural engineering client who appears to be somewhat out of his/her depth as the wind-tunnel study evolves. They really need to be helped, but it can be painful! Clues might include being asked to spell “eigenvalue” or “eigenvector” to him/her over the telephone so that he/she may look it up in the STAAD or ETABs manual. Perhaps when you are asked to explain what “torsion” is on a 50-storey building you should be concerned. If you are asked how much it would Figure 18: Separated flow over Central caused by southerly winds over Victoria Peak in Hong Kong, China (1:4000). “cost” to reduce the loads prior to issuing the Final Report you suspect this is an ethically challenged client. On the rare occasion that the structural loads are larger that the relevant code the engineer may suggest filing the wind-tunnel data and just “going with the code”. Panic! How do we subtly ease them in the right direction? At what point may our engineering ethics cause us to consider stronger action? If so, what? The peer-review process will eventually “educate” most of these ethically challenged clients, but not all of them. One of mine actually ended up with prison time when he colluded with a building inspector to ignore a “minor problem”. Figure 19: Complex geometries, such the Denver Art Museum, are immediate candidates for a windtunnel study. However, even more conventional midrise structures, such as shown in Figures 10 and 13, are routinely tested for the economic benefits that acrue when compared to a code-based design.