Developments in Wind Testing of Building Envelope Systems Peter Irwin, PEng; Arindam G. Chowdhury, Tuan-Chun Fu, Roy Liu-Marques, and Ioannis Zisis Dept. of Civil & Environmental Engineering and International Hurricane Research Center Florida International University, College of Engineering and Computing 1055 West Flagler Street, EC 3740, Miami, FL 33174 Phone: 305-348-4883 • Fax: 305-348-2802 • E-mail: peter.irwin@fiu.edu S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 I rwi n et a l . • 5 9 ABSTR ACT Building envelopes are currently designed using the wind load provisions of building codes and standards or through special wind tunnel tests. Mock-up tests such as those of ASTM are then applied to representative portions of the envelope to test the robustness of the system. These approaches have their limitations in that testing full-scale mock-ups often excludes important local aerodynamic effects, while wind tunnel testing at small scale lacks resolution and may introduce scale effects. The new Wall of Wind facility at Florida International University enables more faithful replication of the true aerodynamic effects and helps to bridge this gap. SPEAKER S Peter Irwin — Dept. of Civil & Environmental Engineering and International Hurricane Research Center Dr. Peter Irwin is recognized as one of the top wind consultants in the world when it comes to the design of large structures, having been a wind consultant and wind-tunneltested structures such as Burj Khalifa (the world’s tallest building), Taipei 101, and the Petronas Towers, as well as many other tall buildings, stadiums, and long-span bridges. He has over 38 years of experience in wind engineering, including eight years as president of RWDI, the well-known wind engineering company. He has received prestigious awards for his contributions to wind engineering research and been recognized for his numerous publications. NON PRE SENTING COAUTHOR S Arindam G. Cowdhury — Florida International University Tuan-Chun Fu — Florida International University Roy Liu-Marques — Florida International University Ioann is Zisis — Florida International University 6 0 • I rwi n et a l . S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 ABSTRACT Building envelopes are designed using the wind load provisions of building codes and standards and, in the case of large buildings, wind tunnel tests. Mock-up tests such as those of ASTM are applied to representative portions of the envelope to test the robustness of the system. The level of reliability provided by these methods is only truly tested when an extreme storm event occurs such as a hurricane or local storm. The results are not always favorable, indicating that there are still things to be learned. For example, static testing of fullscale mock-ups necessarily excludes important aerodynamic effects of the building. Additionally, testing at a small scale lacks resolution and may introduce scale effects. Also, some types of building component, e.g., small items such as sun shades, solar panels, fins, tile systems, roof pavers etc., are not incorporated in the existing methodologies. This paper describes research work at Florida International University (FIU) that bridges this gap. Using the new 12-fan Wall of Wind facility at FIU, completed in 2012, new testing and analysis methods that more faithfully replicate the true physics of wind loading on components have been developed and will be described. INTRODUCTION Performance of the building envelope is critical to the performance of the building as a whole. It has been stated that about 90 percent of the value of insurance claims from windstorms comes from loss of the contents and damage to the interiors of buildings. These are the result of failure of the building envelope to stand up to wind forces and prevent water infiltration. The building envelope has a tough job to do and requires care in engineering design, manufacture, installation, and maintenance. This paper is concerned with defining wind loading, testing the envelope in realistic wind loading, and testing the combined effects of rain and wind. ASCE 7 Wind Pressure Calculations One of the challenges of designing for wind is that the behavior of wind around buildings is extremely complex. Committees that write building codes and standards do their best to define wind loading in simple terms, typically dividing the building envelope into neat, rectangular-shaped zones within which a constant pressure coefficient is specified, but it has to be recognized that this is highly simplified compared to reality. The example of codified local wind load provisions on a low-rise monoslope roof, taken from ASCE 7, is depicted in Figure 1. The rectangular zones marked as 3 have the highest suctions. For the 140- by 100- ft. building shown in Figure 2, these zones have side lengths 2a = 20 ft. (4.8 m) and 4a = 40 ft. (9.6 m). Any component within these zones near to the corners needs to be designed using the highest curve for pressure coefficient GCp. How realistic is this simplified approach? The following describes the actual wind effects that occur on a building envelope in contrast to the simple uniform pressure zones defined by ASCE 7. The aerodynamic reason for high corner suctions is the formation of conical vortices at the corner as illustrated in Figure 2. They cause large suction variations under the vortices as depicted by the bell-shaped curves in Figure 3. The bell-shaped curves have their greatest central suction near the corner and are very narrow in this region; but as distance from the corner increases, the suction reduces and the width of the bell shape grows larger. The effect of these Developments in Wind Testing of Building Envelope Systems S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 I rwi n et a l . • 6 1 Figure 1 – Zones of wind pressure coefficient on monoslope roof from ASCE 7-10. Figure 2 – Vortices that cause high suctions on roofs’ corners. suction distributions on the roof will depend on the type of roof system being used, but it is unlikely to be well represented by a uniform pressure distribution. The different effects on various roof systems will be described later. It should also be noted that along with the high suctions from the vortices, there are also high velocities passing over the surface as the flow rotates rapidly about the vortex center. The vortex is analogous to a small tornado with its axis approximately horizontal and with very high velocities near the vortex core. Thus, there are not only high suctions tending to lift roofing material, but also high tangential air speeds immediately adjacent to the roof surface, which are prone to penetrating under the edges of tiles and shingles and lifting them. To fully replicate these wind effects on a roof in a test, it is very important to generate these vortices as part of the test. However, current standard test methodology does not do this. As will be described later—for multilayer building envelopes, such as rain-screen walls, roof pavers, and vented energy-efficient walls—it is not just high suctions or positive pressures that are important to loading of the envelope, but also the spatial gradients of these pressures. Testing that simply applies uniform static pressures across such systems is equivalent to ignoring gradient effects and will not represent the true state of affairs. To be truly representative, the pressure gradients across the wall or roof must also be generated in the tests. ASCE 7 Wind Tunnel Testing So we have compared the simplified ASCE calculation for wind pressures to actual wind effects. However, where does boundarylayer wind tunnel testing fit into the picture? Typically, this type of testing is done on large projects such as tall buildings, stadiums, arenas, convention centers, etc. The model scales are in the range of 1:200 to 1:500; and at these scales, a fairly good simulation of the planetary boundary layer on the earth’s surface can be economically achieved in appropriately designed boundary-layer wind tunnels. The tests can determine overall wind loading and even the main features of local cladding pressures, including the effects of corner vortices. The loads coming out of the tests are equivalent to code loads but are much more accurate and specific to the project, since they include the effects of its unique shape, unique surroundings, and local climatology. However, while they provide good information on wind pressures (and even gradients of wind pressures), they do not test the ability of any given building envelope system to withstand those loads. Also, the tests are at small scale, and there is evidence that the central suctions in corner vortices may be subject to scale effects (Reynolds number effects in aerodynamic terminology, see Irwin 2009). To evaluate the ability of the real envelope to withstand the local pressures and pressure gradients, and to avoid scale effects, tests of the fullscale system are needed in flow conditions as close to the real situation as possible. The challenge of testing at full or large scale is that it is difficult to simulate the full-scale wind field, including all the scales of turbulence that are present in the real wind. The planetary boundary layer ranges in depth from a few hundred feet to several thousand, and the size of range of turbulent eddies within the boundary layer near the ground can range up to several hundred feet. It is simply not possible to economically generate the full range of turbulence scales in standard test facilities. However, in recent years, research has indicated that what most affects the local aerodynamics on a roof or a wall for any 6 2 • I rwi n et a l . S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 Figure 3 – Suction variation on roof under corner vortices. Figure 4 – Wall of Wind schematic. given wind direction is the small-scale turbulence— i.e., turbulent eddies that are of similar size to the widths of vortices and shear layers generated at building corners and edges. As long as one can include sufficient intensity of small-scale turbulence, then a good representation of the real aerodynamics and its effect on the envelope system can be obtained. The influence of the missing large-scale turbulence can be accounted for by interpreting the test as being representative of a gust with a certain speed and direction. This principle is used at FIU’s Wall of Wind facility, which will now be described. Wall of Wind at Florida International University The Wall of Wind at FIU is illustrated schematically in Figure 4. It consists of twelve 6-ft.-diameter fans arranged in two tiers with six fans to each tier. The fans, in a circular arc configuration, are powered by a variable frequency drive system with a total of 8,000 HP, and each tier can be controlled separately. They blow into a contraction section that is then followed by a 20-ft.- wide by 14-ft.-high flow conditioning section. In the flow conditioning section, spires and floor roughness are used to tailor the mean velocity profile, turbulence intensity, and integral scale of the turbulence to desired values. The mean velocity and turbulence intensity are measured at the test specimen location using fastresponse Cobra probes (probes with multiple pressure taps connected directly to miniature pressure transducers, which can measure instantaneous flow velocity and angle). The test specimen is located on a 16-ft.-diameter turntable, downwind of the end of the flow conditioning section. The Wall of Wind is contained inside a hangar-type building with sliding doors at each end. Air is sucked in at one end and blown out of the other end. Figure 5 shows views of the intake end and flow conditioning spires and roughness. Figure 6 shows an example of a roof paver system set up in the working section downwind of the flow conditioning section. Beyond the working section, the flow passes out of the building into an open field where any debris that may come loose from the test specimen can land. Without the spires and roughness, wind speeds of over 155 mph have been generated. With the spires and roughness present, somewhat higher speeds have been recorded at the top of the test section and lower speeds near the bottom. The design and development of the whole system has been described by Aly et al. (2011). It included computational fluid dynamics (CFD) studies and the construction of a small-scale model at 1:15 scale. Attached to the spires is a water injection system to simulate rain with wind. The water droplets are ejected from an array of nozzles on the down-wind side of the spires, and rates of up to 8 in. per hour can be generated. S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 I rwi n et a l . • 6 3 Figure 5A and 5B – Wall of Wind fan intake (left) and spire/roughness flow management system (below). Figure 6 – Example of roof paver system under test. Figure 7 – Wind profile in the Wall of Wind compared with Category 5 hurricane profiles. Flow Conditions The flow conditions in the Wall of Wind can be tailored to represent a variety of situations. In Figure 7, an example is shown of the mean velocity profile, averaged over a minute, in miles per hour. In this figure, the vertical axis (z) is the height above ground. However, in some applications, z may be measured relative to some other reference level rather than ground. Although averaged over a minute, the profile is virtually the same for any averaging time longer than about three seconds at the speeds shown. Fluctuations on time scales less than three seconds are effectively simulated by the turbulence generated in the flow conditioning section. Thus, for a full-scale test, the Wall of Wind, in essence, simulates what happens during the passage of a three-second gust. Figure 7 also shows the gust profiles at the dividing line where a Category 4 hurricane becomes Category 5. It can be seen that the Wall of Wind profile is in line with the open terrain profile and well in excess of that for suburban terrain (suburban terrain corresponds to Exposure B of ASCE 7). To simulate gusts coming from different directions, the test specimen can be rotated on the turntable or even pitched to simulate small variations of gust velocities from horizontal. Often, the worst test angles can be selected in advance, based on experience. For example, for rectangular flat roofs, a horizontal wind at a quartering angle (i.e., 450 to a roof edge) creates the strongest corner vortices and will provide a robust test of many roof systems. In situations where the worst-case direction is difficult to determine in advance, it is also possible to rotate through a range of directions in order to seek out the highest loading conditions, using pressure and/or load cell measurements to determine loads. The ability of the Wall of Wind to simulate small-scale turbulence is important. Many of the most critical features of wind loading are created by the flow separating from sharp edges into shear layers, forming a separation bubble underneath and reattaching to the building surfaces, as illustrated in Figure 8, or originating at corners and forming vortices as already described. The shear layers typically have widths from a few inches to a few feet (see Figure 8). The mixing within these shear layers is strongly affected by turbulent eddies of similar size to the shear layer width, and it is found that tests in smooth flow (illustrated in the upper part of Figure 8) produce a separation bubble that is too long compared with the real situation where turbulence is present. Also, the streamlines above the roof have too little curvature. The small-scale turbulence enhances the mixing in the shear layer, which entrains more air from the separation bubble, causing it to shrink in length and generate streamlines with higher curvature. The shape of the separation bubble affects the peak suctions near the leading edge of the building, higher suctions being possible with the reduced length of the bubble and the more highly curved streamlines over the roof. The widths of corner vortices are also in the same range as the widths of shear layers, and they are likewise affected by the small-scale turbulence. Therefore, even if it is not possible to simulate the large, turbulent eddies of natural wind in wind-testing facilities, most of the important effects of turbulence will be reflected in the tests, provided the energy in the small-scale turbulence eddies is correctly simulated. The large-scale eddies can be treated as equivalent to changes in the mean-flow velocity and direction, since they do not interact directly with the important local aerodynamic features such as shear layers and corner vortices. The concept of treating large-scale turbulence in this way is referred to as the “quasi-steady” assumption. While the importance of flow turbulence has been known from wind tunnel and full-scale measurements, the current standard mock-up tests do not take proper account of it. The Wall of Wind provides the ability to do this. One question then is: How much of the turbulence do we need to simulate in the tests in order to be sure we include its most important effects? To answer this requires thinking of the turbulence as having eddies with a range of sizes or wavelengths. A big 6 4 • I rwi n et a l . S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 Figure 8 – Effect of turbulence on length of separation bubble. Figure 9 – Example of simulation of power spectrum of turbulence in the Wall of Wind. Full turbulence intensity su / U = 0.28, partial turbulence intensity su / U = 0.08. eddy (i.e., one with a long wavelength) takes a long time to pass an observer and is, therefore, associated with low frequencies ( f ), whereas a small eddy goes by quickly and is associated with high frequencies. The amount of turbulence energy per unit of frequency is called the power spectrum S ( f ), and if S is summed over all frequencies, it is equal to the variance ( s 2 u ) of the turbulence velocity fluctuations in the direction of the wind. It can be plotted in form fS/U2 versus wavelength U/f, as shown in Figure 9, where U = mean wind velocity. In Figure 9, the solid curve illustrates the typical spectrum in the real wind, and it extends over wavelengths up to thousands of feet. The broken curve illustrates what can be achieved in the Wall of Wind using the spire and roughness technique and with a turbulence integral scale of about 1/160th of that in the full spectrum. The good alignment of the solid and broken curves in the shaded range of wavelengths indicates that a good simulation of the turbulence energy can be achieved for wavelengths up to about 10 ft., which covers the important range for interacting with shear layers and corner vortices. Comparison With Field Tests As part of the development and validation of the partial turbulence simulation approach used with the Wall of Wind, comparisons have been made with field measurements on the Texas Tech University (TTU) instrumented low-rise building (Levitan and Mehta, 1992). Figure 10 shows a 1:6-scale model of the TTU building in the Wall of Wind, and Figure 11 shows an example of mean and peak pressure coefficients derived from the Wall of Wind tests for a selected wind direction, the coefficients being based on mean dynamic pressure at roof level. The results are compared in Figure 11 with the field data at 16 pressure taps, whose locations are shown in the roof plan. Techniques proposed by Sadek and Simiu (2002), Simiu (2009), and Fu et al. (2012), were used in the extrapolation of measured pressure coefficients to full-scale equivalent peak pressure coefficients, treating large-scale turbulence as quasi-steady fluctuations. The level of agreement is generally very good in the most important areas (i.e., those where the loading is highest), giving confidence in the partial turbulence simulation approach. Similar levels of agreement have been achieved comparing Wall of Wind tests on a scale model with the full-scale data of Richards and Hoxey (2012) for a 6-m3 building in open terrain. Pressure Gradient Effects There are a number of useful applications of the Wall of Wind, but it is especially useful for envelope systems that are sensitive to pressure gradient effects, since it can simulate those very well. Figure 6 shows the example of roof pavers being tested. Building codes currently do not provide much guidance on the design of pavers for wind, although there have been a number of research papers on the topic. Pavers are one example of a multilayer building envelope that is particularly sensitive—not just to pressures, but also to pressure gradients. Figure 12 (from Irwin, 2012) illustrates the pressures pU and pL on the upper and lower surfaces, respectively, of a paver system. The pressure on the underside is dictated by the topside pressures at the edges of pavers. On a paver (AB) sitting under the high suction created by a corner vortex (see the bell-shaped uplift pressures of Figure 3), the suction at its edges can be much less on average than in the middle of the paver, and it is the edge suctions that are transmitted to the underneath. Thus, a high net uplift is created under a corner vortex. But the impact of this depends a lot on the size of the paver relative to the width of the corner vortex. If the paver is much larger than the width of the vortex, then the impact is reduced, since only a small fraction of the paver area is affected by the high suction. Also, if the paver is much smaller than the width of the vortex, then—even if it is sitting in a high-suction zone—the pressure equalization effect of the gaps at S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 I rwi n et a l . • 6 5 Figure 10 – Large-scale (1:6) model of TTU building in the Wall of Wind. Figure 11 – Comparison of pressure coefficients derived from the WOW with fullscale data from the TTU building. its edges substantially reduces the difference in pressure between top and bottom surfaces. However, if the paver and vortex widths are similar, the net uplift will be at a maximum. Figure 12 shows how linking pavers together by strapping (or other means) reduces the ability of wind to lift the paver AB. To lift it requires a lift force equal not only to the weight of paver AB, but also to a good fraction of the added weight of its neighbors, which are not subject to much uplift. Similar gradient effects impact the performance of a variety of envelope systems, such as standing-seam metal roofs, various vented curtain-wall systems, roof shingles and tiles, and PV solar panels. CONCLUSIONS The performance of building envelope systems depends to a significant extent on accurate knowledge of wind loads and the ability to test parts of the envelope under realistic wind and rain conditions. Existing static, uniform-pressure test methods help to screen envelope systems for weaknesses but do not reflect the full complexity of real wind loading. In the Wall of Wind facility at FIU, a new tool is now available for creating close-to-real wind conditions, with and without rain. By simulating the small-scale turbulence that interacts with shear layers and vortices, the important aerodynamic effects of wind are included, and the fact that large-scale turbulence can be handled by treating it as equivalent to slow changes in oncoming flow speed and direction. Comparisons of wind loads derived from the Wall of Wind using the partial turbulencesimulation method with field measurements at full scale show very good agreement, thus validating the usefulness of this approach. Looking to the future, the Wall of Wind will enable the development of more realistic test procedures for building envelope systems, as well as providing what may be regarded as a calibration of existing, more simplified test protocols. ACKNOWLEDGMENTS The authors are grateful to Dr. Douglas Smith of TTU for supplying the detailed full-scale field data obtained from the TTU low-rise test building. REFERENCES A.M. Aly, A.G. Chowdhury, and G. Bitsuamlak, 2011, “Wind Profile Management and Blockage Assessment for a New 12-Fan Wallof- Wind Facility at FIU,” Wind and Structures, Vol. 14, No. 4, (2011) 285-300. T-C Fu, A.M. Aly, A.G. Chowdhury, G. Bitsuamlak, D. Yeo, and E. Simiu, (2012). “A Proposed Technique for Determining Aerodynamic Pressures on Residential Homes.” Wind and Structures, 15(1), pp. 27-41. P.A. Irwin, 2009, “Wind Engineering Research Needs, Building Codes, and Project-Specific Studies,” keynote lecture, Proceedings of the 11th Americas Conference on Wind Engineering, June 22-26, 2009, Puerto Rico. P.A. Irwin, C. Dragoiescu, M.D. Cicci, and G.P. Thompson, 2012, “Wind Tunnel Model Studies of Aerodynamic Lifting of Roof Pavers,” Proceedings of the ATC/SEI Conference on Advances in Hurricane Engineering, Miami, FL, October 24-26, 2012. M.L. Levitan and K.C. Mehta, 1992, “Texas Tech Field Experiments for Wind Loads Part 1: Building and Pressure Measuring System.” Journal of Wind Engineering and Industrial Aerodynamics, Volume 43, Issues 1–3, Pages 1565–1576. P.J. Richards and R.P. Hoxey, 2012, “Pressures on a Cubic Building – Parts I and II,” Journal of Wind Engineering and Industrial Aerodynamics, 102 (2012) 72-96. F. Sadek and E. Simiu, 2002, “Peak Non- Gaussian Wind Effects for Database- Assisted Low-Rise Building Design,” ASCE Journal of Engineering Mechanics, May 2002, pp. 530-539. E. Simiu, 2009, “Wind-Loading Codification in the Americas: Fundamentals for a Renewal,” keynote lecture, Proceedings of the 11th Americas Conference on Wind Engineering, June 22-26, 2009, Puerto Rico. 6 6 • I rwi n et a l . S y m p o s i u m o n B u i l d i n g E n v e l o p e T e c h n o l o g y • No v e m be r 2 0 1 3 Figure 12 – Uplift mechanism on roof pavers.
