Application of Realistic Wind Loads to the Roof of aFull-Scale, Wood-Frame HouseMurray J. Morrison1, Gregory A. Kopp21PhD candidate, Boundarv Laver Wind tunnel Laboratorv, Universitv of Western Ontario, London, ON, Canada,mfmorrisuwo.ca2Professor, Boundarv Laver Wind tunnel Laboratorv, Universitv of Western Ontario, London, ON, Canada,gakoppuwo.caABSTRACTRealistic structural testing of buildings and building components can now be conducted usingPressure Loading Actuators (PLAs) developed at the Universitv of Western Ontario. In thepresent work, the roof of a full-scale, 2 storv house, built to the Ontario building code, wastested using 58 PLAs to examine the performance of toe-nailed roof-to-wall connection.Theloading for this test were obtained from a wind tunnel studv conducted on a 1.50 scale model ofthe test house. Details of the test set-up and experimental design are presented, along withpreliminarv analvsis of the data.INTRODUCTIONOver the past 30 years there have been signiIicant advances in our understanding oI wind loadson low-rise buildings.As Surry |1| pointed out:'We know enough about the wind loads on low buildings now, so that disastrous Iailures(such as seen during Hurricane Andrew) to storms other than severe tornadoes, are muchmore likely to be due to Iaults in codes, or construction and inspection practices, than dueto a lack oI basic wind engineering knowledge.This statement remains true today. While there may be deIiciencies in the wind load coeIIicientsin building codes, e.g., such as those pointed out by |5|, these are not due to a lack oIunderstanding oI the aerodynamic wind loading or to severe mismatches between wind tunneland Iull-scale data |2-4|.However, our understanding oI how structural systems, as well as building cladding andcomponents, respond under these wind loads is less advanced Ior structures like wood Iramehouses. The main reason Ior this is houses have many structural members, signiIicant variabilityin both material and connection (e.g., toe-nails) properties and extensive use oI non-structuralmaterials which aIIect perIormance. There have been many standard tests developed and usedby industry to examine the perIormance oI diIIerent building products or components.ExamplesoI this are the ASTM E1592-01 |6| to test sheet metal rooI and siding, ASTM E1300-04 |7| totest load resistance oI glass, or the cyclical SIDGERS test |8| to test Ilexible membrane rooIs.Most oI these tests apply constant loading to the specimen Ior a predeIined amount oI time, someare cyclical tests, although usually at a rate which is much less than the actual loading rategenerated by real wind. 'BRERWULF, which was developed at the British ResearchEstablishment |9|, is able to apply Iluctuating wind-induced pressures to cladding specimen,provided the airbox and specimen are nominally air-tight.While the above mentioned tests haveprovided excellent comparative perIormance data Ior rooI components, they also have severaldrawbacks. First, they are spatially uniIorm, while real wind loads on low rise buildings havesigniIicant spatial variations, even over short distances, as shown in Figure 1.This may besigniIicant Ior larger components with multiple connections leading to signiIicant load sharing.Thus, Ior such tests, the applied load is signiIicantly diIIerent than real Iluctuating windpressures.While cyclical testing is able to model Iatigue characteristics that are not captured bystatic tests, the true extent oI the eIIects oI real wind loading versus cyclic or static loading is stilllargely unknown. A notable exception is the experiments perIormed at Mississippi StateUniversity (MSU) using electromagnets capable oI applying both temporally and spatiallyvarying wind loads on standing seam metal rooIs |10|.Figure 1Contour plot of wind-induced pressure coefficients, at one moment in time, on the gable roof of ahouse, obtained from wind tunnel experiments.Experiments conducted at the Cyclone Testing Station |11| have applied loads to Iull scalehouses using wiIIle-trees and were able to test entire houses to Iailure. Such experiments haveprovided signiIicant insight into how a structural system perIorms as a whole, along with thesigniIicant role oI non-structural components, such as drywall, in the load sharing Ior residentialhousing.The goal oI the 'Three Little Pigs Project (3LP) at the University oI Western Ontario (UWO) isto examine the perIormance oI Iull scale structures and components under realistic spatially andtemporally varying wind loads.The Iirst Iull house experiments involve testing a typically-builtCanadian house to Iailure.The approach taken at UWO is to model the wind-induced pressures,rather than modeling the wind itselI.In this sense, it is similar to the standard airbox type testingdiscussed above.This concept prompted the design oI pressure loading actuators (PLAs), whichwill be discussed in more detail below, and a Ilexible airbag system that can be attached to thestructure or specimen to apply the loads to the structure while providing minimal interIerence tothe structuralresponse oI the system.By having many PLAs and airbags oI diIIerent sizes, boththe spatial gradients and temporal Iluctuations oI the wind loads can also be modeled using thistechnique.The objective oI the present work is to present preliminary results Irom the Iirstexperiments on a Iull-scale house, Iocussing on Iailures oI rooI-to-wall connections.EXPERIMENTAL SETUPPRESSURE LOADIAG AC1UA1ORS (PLAS)In order to apply realistic wind loads to the structure, the primary requirement oI the PLA is itsability to Iollow a speciIic pressure trace in an accurate, reliable and repeatable way so that whenmultiple units are used, the overall structural loads are accurately replicated.In addition, whilethis technique is well suited to taking wind tunnel measurements and applying them to realstructures, this approach can also be used to apply loads Irom Iull scale measurements shouldthey be available.Moreover, it is also possible to simulate other types oI wind events such astornadoes, downburst or gust Ironts iI the pressure time series on a structure are known Ior theseevents. Due to the large number oI PLAs required to test a Iull structure (which we have limited to 100in the current project) each individual unit needs to be accurately synchronized to all the otherunits in the system in order to generate loads with the proper spatial gradients. Since manybuilding materials are porous, the PLAs have to be able to handle a certain amount oI leakage(through the specimen) while still maintaining the Iidelity oI the pressure trace. Table 1 lists thedesign requirements Ior the PLAs Ior diIIerent airbox sizes.These requirements were derivedIrom the worst aerodynamic coeIIicients scaled to a Category 5 Hurricane wind speeds.WhileTable 1 provides the design speciIications Ior a single PLA unit, multiple units can be connectedto a single airbag and used together in order to improve perIormance Ior larger airbags or to testcomponents with large leakage Ilow rates.Table 1 Design specifications for the Pressure Loading ActuatorsAir BoxDimensionsMaximumPressure (kPa)MinimumPressure (kPa)LeakageFlow Rates(m3/s)FrequencyResponse(Hz)0.6 m x 0.6 m 5 -18 0.2 61.2 m x 1.2 m 5 -15 0.7 42.4 m x 2.4 m 4.5 -11 1.0 4In order to test the perIormance oI the PLAs a wood Irame wall with dimensions oI 3.66m highby 6.71m long was built and loaded with 10 PLAs, each connected to its own airbag, the totalloaded area on the wall was 20.4 m2. Figure 2 (LeIt) shows a portion oI the scaled wind tunnelpressure trace Ior a single airbag along with the output achieved by the PLA.The match isexcellent and typical trace Iidelities Ior individual boxes are greater than 95. Figure 2 (Right)displays the total reaction Iorce on the wall along with the applied Iorce Irom the PLAs and thetarget Iorce Irom the wind tunnel pressure measurements.All three show good agreement, withthe average error between the measured and applied Iorce being approximately 7.Figure 2 (Left)Comparison of the demand (wind tunnel, WT) and supplied (PLA) pressure from the PLAconnected to a 1.2 x 1.2 m2 airbox.(Right) Comparison of the demand (WT), applied (PLA), and measuredloads on a test wall using 10 PLAs covering 20.4m2.FULL SCALE HOUSE 1ES1IAGA typical 2 story Canadian house was built by Building Technology students Irom FanshaweCollege at the Insurance Research Lab Ior Better Homes (IRLBH).The house is shown inFigure 3.The house was built to the Ontario building code (OBC), has plan dimension oI 9.0mby 8.9m, an eaves height oI 8.0 m and a gable rooI slope oI 4:12.The studded walls areconstructed oI 2 x 4`s on 16 centers, the exterior walls are covered with 25mm Ioam insulationwhich is then covered by a brick veneer.The rooI trusses are placed 0.6m on center and aresecured to the top plate using standard toenails with 16d nails.While the number and quality oItoenails varies connection to connection, each connection was documented in detail noting thenumber and location oI the nails along with any observed deIects.Following construction oI thehouse it was inspected by 30 proIessional building inspectors Irom across Southern Ontario andwas Iound to be oI typical construction quality Ior the region.Also shown in Figure 3 is the steelreaction Irame surrounding the house which is used to mount the airbags and the PLAs.Figure 3Photograph of the first test house at IRLBH.RooIs oI houses experience large suction pressures which are greater than the load experiencedby any other part oI the building.A breach oI the windward can cause internal pressures whichcan increase this loading by as much as 70. The most common type oI rooI to wall connectionused in wood-Irame construction is a toe-nail connection. The capacity oI toe-nail connectionscan vary signiIicantly on the type oI wood, number and type oI nails used along with the qualityoI construction.Several studies |12-14| have investigated the ultimate capacity oI toe-nails andIound that the connection strength can vary between 1130N to 2840N, depending on the Iactorsdiscussed above.These tests apply load at a constant displacement rate and measure the requiredIorce to keep the connection moving at this rate.Typical deIlection rates Ior these tests rangeIrom 2.54 to 6.35 mm/min. Hurricane straps can provide a substantial increase in the connectionstrength |13-14| and are now required in Florida and other hurricane prone regions, but are notstandard construction Ior other regions oI North America. Many non-hurricane regions canexperience intense winds during thunderstorms such as tornadoes and microbursts.While thetests mentioned above provide inIormation oI the ultimate capacity oI the nails, how theconnections respond to a highly Iluctuating load or the eIIect oI load sharing between adjacentrooI-to-wall connections is still unknown. The Iirst experiment to be conducted on the test housewill try and address these questions by applying wind loads to the rooI oI the test house only.WIAD 1UAAEL SIMULA1IOAIn order to apply realistic wind loads to the Iull scale house using the PLAs the pressuredistribution must be known Iirst.A wind tunnel study was conducted on a 1:50 scale model oIthe test house in Boundary Layer Wind Tunnel II at UWO.The model was tested in a typicalopen country terrain with no surrounding buildings, which is the base case used in most buildingcodes. Figure 4 shows the measured mean velocity and turbulence intensity proIiles along withthe target proIiles (ESDU) Ior an open country terrain (zo0.01).The mean wind velocity hasbeen normalised to eaves height and both the mean and turbulence intensity show reasonableagreement with the target proIiles. Figure 5 presents the longitudinal wind tunnel spectra at rooIheight along with the target (ESDU) spectra. While the match between measured and targetspectra is good it is not perIect with too much Iine scale turbulence, which is not unusual Iorwind tunnel simulations at this scale (see |15| Ior a more detailed discussion).The model had atotal oI 432 pressure taps that were sampled at a rate oI 400 Hz.The tests were conducted at areIerence speed oI 13.7 m/s and 1 hour oI Iull scale data was collected assuming a velocity scaleoI 1:4.Figure 4Comparison of measured and ESDU mean normalized wind speed and turbulence intensity profilesfor open country, z00.01mTo test the upliIt capacity oI the rooI-to-wall connections the entire rooI needed to be cover withairbags.The correct number, size, and location oI each airbag on the house is a balance betweenadequately capturing the spatial gradients oI the wind pressures and the physical and technicalconstraints inherent to the system.For example the Ioot print oI a PLA is approximately 0.6mby 0.6m which essentially limits the minimum bag size.Through careIul analysis oI the windpressure distribution, the Iinal airbag layout oI the rooI is shown in Figure 6. In total 58 bagswere used to cover the entire rooI, with the highest density oI bags located at the windwardcorner oI the rooI.A cornering wind was chosen Ior the Iull scale tests as this produced thelargest suction pressures on the rooI oI the house.Figure 5 Comparison of measured and ESDU mean normalized wind speed and turbulence intensity profilesfor open country, z00.01mFigure 6 Airbag layout on the roof of the test houseIAS1RUMEA1A1IOA AAD 1ES1 PRO1OCOLSThe goal oI the current tests was to examine the perIormance oI the toe-nail connections oI theas-built house.In order to Iacilitate this test, several minor modiIications were made to thestructure.The shingles were removed Irom the house to allow the physical attachment oI the airbags to the rooI oI the house.It is noted that doing so will lower the overall weight oI the houseand thereby increasing the loads on the structural connection.However, since the weight isevenly distributed over the entire rooI this reduction in weight can be accounted Ior in themagnitude oI loads applied to the structure. In addition, since the goal oI the current experimentwas to test rooI-to-wall connections to Iailure, and not to test the upliIt capacity oI the plywoodrooI sheathing, the sheathing was screwed to the rooI trusses to prevent it Irom Iailing during thetests.Finally, the soIIits and Iirst 2 cords oI brick were removed in order to provide access to therooI to wall connections.Since both are non structural elements this should not aIIect thestructural response oI the toe-nail connections.Installation oI the airbags and PLAs, wasperIormed by University Machine Services at UWO and took 3 individuals nearly 3 months tocomplete. Figure 7 shows a portion oI the rooI covered by the air bags (blue material seen inFigure 7) which are mounted to the steel reaction Irame and the hoses connecting the airbags tothe PLAs.The smaller bags in the picture are 0.6m x 0.6m in size while the larger bags are 1.2mx 1.2m.The bag Irames are mounted 6 above the rooIs surIace, the bag material which is gluedto both the surIace oI the rooI and Irame allows Ior deIlections up to 4 inches in either direction.Figure 7Photograph of the airbags and PLA mounted at the windward corner of the roofThe structure was instrumented with 37 displacement transducer which measured the deIlectionat each oI the rooI-to-wall connections. Figure 8 shows the displacement transducers that weremounted to the brick Iascia, independent measurements taken Irom the ground we used duringeach experiment to conIirm that the bricks remained stationary throughout the tests. Videocameras were mounted to the reaction Irame to monitor the Iailure oI the structure during thetests.Figure 8Photograph of the displacement transducers used to measure the deflection at the roof to wallconnectionsThe testing methodology used was very similar to that used Ior the MSU test |10|.The windtunnel pressure coeIIicient data was scaled to a low wind speed and then applied to the house.The mean, rooI height wind speed was then increased by increments oI 5 m/s until Iailureoccurred.It is noted that the response oI the connections at higher wind speed tests could beaIIected by the previous tests.However, this method oI load application is not too dissimilar toan actual hurricane even where the mean wind speed increases gradually as the storm passes.Rather than use the entire pressure record obtained Irom the wind tunnel tests a 15 minutesection (scaled at the lowest wind speed) or 15 oI the entire time record was selected and thenscaled Ior each test. Figure 9 shows a portion oI the wind loading trace Ior the box with thehighest pressures scaled to mean rooI height wind speeds oI 20, 30 and 40 m/s.The peakpressures are proportional to the square oI the wind speed while the duration is inverselyproportional the wind speed.This means that while the loads Ior the 40 m/s test are 4 times aslarge as those Ior the 20 m/s test they last halI as long.Figure 9 A portion of the applied pressure trace for a single airbag scaled to 3 different mean roof heightwind speeds.RESULTSThe worst suction pressures on the rooI oI the house are located on the leeward side oI the ridgeclose to the gable end wall, as shown in Figure 1.While loads applied to the rooI are known theloads at each rooI-to-wall connection were not measured directly.An estimation oI the loads ateach connection can be made by assuming a geometric tributary area approach.This assumptioncould be removed by replacing the toe-nail connections with load cells and repeating exactly thesame loading used in the deIlections tests.Under this approach the worst loaded rooI to wallconnection is located on the gable end wall on the leeward side (South) oI the rooI ('S2 seeFigure 10 Ior a truss layout).However, the gable end walls have signiIicantly more weight andadditional connections that eIIectively reduce the load applied to the toe-nail connections 'S2and 'N2.As a result the largest deIlections, and likely the most highly loaded toe-nailconnection, occur at the adjacent connection 'S3. A portion oI the deIlection time series Ior the'S3 connection is shown in Figure 11, each colour represents a diIIerent scaling wind speedused to calculated the Iull scale loading. The deIlection data shown Ior each scaling wind speedrepresents the same portion oI data obtained Irom the wind tunnel.At higher wind speeds, thetoe-nails experience signiIicantly larger deIlections, which are not proportional to the appliedload. Figure 12 shows the deIlection Ior the same truss as Figure 11 only Ior the Northconnection ('N3).The deIlections at the 'N3 connection are signiIicantly less than that oI'S3.In Iact Ior the Iirst two tests (20 and 30 m/s) the deIlections are actually negative.Thisindicates that the rooI is rotating towards the wind.In such a conIiguration once all theconnections have Iailed and the rooI begins to liIt up the aerodynamics loads on the rooI willchange pushing the rooI back down preventing it Irom Ilying oII.It is likely that a subsequentchange in wind direction will cause the rooI blow oII, although this could be at a much lowerwind speed.In addition, Figure 13 shows a photograph oI the same toe-nail connection during construction(LeIt) and Iollowing one oI the loading tests (Right).The pull out oI the nail can clearly beobserved in the photographs and there is now an air gap between the truss and top plate.Figure 10Layout of roof trusses and naming conventionFigure 11 Short segments of the displacement time series for connection ~S3 for 6 different test wind speeds.Figure 12 Displacement time series for connection ~N3 scaled to 6 test wind speedsFigure 13Photograph of a toe-nail connection during construction (Left) and following one of the loadingtests (Right)The analysis oI the experimental results is on-going and a more detail discussion oI theexperimental result will be presented at the conIerence.CONCLUSIONSThe new loading system developed at UWO allows the application oI real wind loading to a Iullscale structure. The Iirst test involved applying loads to the rooI oI the structure to examine theperIormance oI the toe-nail connections.The connections were Iound to Iail on the leeward sideoI the rooI Iirst and the rooI is rotating about the windward wall.ACKNOWLEDGEMENTSThe inIrastructure reported herein has been provided through the Canada Foundation IorInnovation, Ontario Innovation Trust, Insurance Bureau oI Canada, Canadian Institute Ior SteelConstruction and the University oI Western Ontario. Seed Iunding Ior the project was providedby the Institute Ior Catastrophic Loss Reduction and the Natural Sciences and EngineeringResearch Council oI Canada. Operating Iunds Ior research have been provided by grants Iromthe Natural Science and Engineering Research Council, the Institute Ior Catastrophic LossReduction, and the Boundary Layer Wind Tunnel Laboratory. Mr. M. J. Morrison grateIullyacknowledges scholarship support Irom NSERC. Dr. G. A. Kopp grateIully acknowledges thesupport provided by the Canada Research Chairs Program.REFERENCES[1] D. Surrv, Wind loads on low-rise buildings. past, present and future., Proc. 10th Int. Conf. Wind Eng.,Copenhagen, vo.l 1 1999. p. 105-114.[2] L.S Cochran, J.E. Cermak, Full- and model-scale cladding pressures on the Texas Tech Universitvexperimental building, J. Wind Eng. Ind. Aerodvn.. 43(3) (1992) 1589-1600.[3] Y.L. Xu, G.F. Reardon, Full-scale and model-scale wind pressure and fatigue loading on the texastech universitv building, Cvclone Structural Testing Station JCU, technical report No. 42 1996.[4] J.X. Lin , D. Surrv, The variation of peak loads with tributarv area near corners on flat low buildingroofs,J.Wind Eng.Ind.Aerodvn., 77-78 (1998) 185-196.[5] L.M. St. Pierre, G.A. Kopp, D. Surrv, T.C.E.Ho, The UWO contribution to theNIST aerodvnamicdatabase for wind loads on low buildings. Part 2. Comparison of data with wind load provisions, J.Wind Eng. Ind. Aerodvn.. 93(1) (2005) 31-59.[6] ASTM , Standard test method for structural performance of sheet metal roof and siding svstems bvuniform static air pressure difference, Designation E 1592-01, Philadelphia, USA, 2001.[7] ASTM , Standard Practice for Determining Load Resistance of Glass in BuildingsDesignationE1300-04, Philadelphia, USA, 2004.[8] A. Baskaran, Y. Chen,Wind load cvcle development for evaluating mechanicallv attached single-plvroofs, J.Wind Eng.Ind.Aerodvn., 77-78 (1998) 83-96.[9] N.J. Cook, A.P. Keevil, R.K. Stobart, BRERWULFThe Big Bad Wolf, J. Wind Eng. Ind. Aerodvn.,vol. 29 (1988) pp. 99-107.[10] D. Surrv, R.R. Sinno, B. Nail, T.C.E.Ho, S. Farquhar, G.A. Kopp, Structurallv effectivestatic windloads for roof panels, J. Struct. Eng., 133(6) (2007) 871-885.[11] G. Reardon, Simulated Wind Load Testing of Full Si:e Houses.Joint IStructE/ Citv UniversitvInternational Seminar, Citv Universitv, London, 1996.[12] J. Cheng, Testing and analvsis of the toe-nailed connection in the residential roof-to-wall svstem,54(4) (2004) 58-65.[13] T.D. Reed, D.J. Rosowskv, S.D. Schiff, Uplift Capacitv of Light-Frame Rafter to Top PlateConnections, J.Arch.Engrg., 3(4) (1997) 156-163.[14] M.A. Rilev, F. Sadek, Experimental Testing of Roof to Wall Connections in Wood Frame Houses,National Institute of Standards and Technologv, NISTIR 6938, 2003.[15] G.A. Kopp, D. Surrv, C. Mans, Wind effects of parapets on low buildings. Part 1. Basicaerodvnamics and local loads, J. Wind Eng. Ind. Aerodvn.. vol. 93 (2005) 817-841