The simulation of non-synoptic effects for wind damage studies

The Seventh International Colloquium on Bluff Body Aerodynamics and Applications (BBAA7) Shanghai, China; September 2-6, 2012 The simulation of non-synoptic effects for wind damage studies Matthew Haines a , Mark Sterling a , Andrew Quinn a a School of Civil Engineering, The University of Birmingham, UK ABSTRACT: This paper presents a brief analysis of the pressure distribution on a physical model building (scale 1:1000) arising as a result of the passing of a non stationary gust front. The gust front is generated using the University of Birmingham’s "downburst" simulator. Issues and challenges relating to the simulation of such pressures in a flow field akin to a moving downburst are briefly discussed. There is tentative evidence to suggest that there are differences in the pressure distribution arising as a result from a non stationary gust front compared to boundary layer flow. However, these differences are far from conclusive. KEYWORDS:thunderstorm, downburst 1 INTRODUCTION It is now an accepted fact that the disruption and economic losses arising as a result of extreme storms are increasing at a significant rate ABI (2005). There is also tentative evidence to suggest that these storms are increasing in frequency and magnitude due primarily to climate change effects, although it is acknowledged that such evidence is far from conclusive (Kasperki, 1998). Any increases in magnitude and frequency of extreme storms are likely to result in serious damage to the urban infrastructure, the world economy and society as a whole. In European terms, it is predicted that by 2080, there will be an increase in "wind-related insured losses from extreme European storms by at least....e25-30bn" (ABI, 2005) . However, it is perhaps worth noting that these estimates do not take into account society’s increasing exposure to extreme storms, due to growing populations, wealthier populations and increasing assets at risk, e.g., ".. if Hurricane Andrew had hit Florida in 2002 rather than 1992, the losses would have been doubled, due to increase coastal development and rising asset values." (ABI, 2005). Within the last few years, thunderstorm downburst type events have received considerable interest (Chay and Letchford (2002a), Lin and Savory (2006), McConville et al. (2009)). Figure 1a illustrates that severe thunderstorms can produce a streamwise velocity distribution which differs from the typical boundary layer flow. In order to assess the potential damage that could be caused by these extreme wind events a number of studies have examined the wind loading present during an extreme event, for example Chay and Letchford (2002a) studied the flow around a simple cube. Mason (2009) examined the effect of surface roughness on the flow characteristics and noted the importance of possible Reynolds number effects. Sterling et al. (2011) discussed the multiple scales that could be used when interpreting physical simulations and as a resulted noted that it can be difficult to ensure that the scaling requirements of ISO2009, i.e., "for thunderstorms...wind methods should be used that adequately generate the spatial and temporal variation of wind speeds and turbulence characteristics within such systems" occur in simulations. Work by Sterling and co-workers (McConville et al. (2009); Sterling et al. (2011)) has sought to examine such events using physical experiments at relatively large scale in order to try and minimise the problems associated with scaling and it is with such experiments that the current work is concerned. 1276 Velocity (m/s) Synoptic wind and downburst velocity- time comparison -500 40 35 30 25 20 15 10 5 0 Synoptic wind Downburst wind 0 500 1000 Time (s) (a) Streamwise vertical profile (b) Synoptic and downburst velocity time history Figure 1: (a) A schematic illustration of the mean streamwise velocity profile corresponding to a ’typical’ downburst and a typical boundary layer wind. (Lin and Savory, 2006). (b) A comparison of a synoptic and downburst wind velocity time history. 1.1 Flow around simple structures Figure 2 illustrates the distribution of pressure coefficient (C p ) over the central portion of a cubic structure arising as a result of a typical boundary layer flow normal to a cube. The pressure coefficient is defined as: Cp = p − pre f 1 2 2 ρre f Vre f (1) where p is the pressure in the location being measured, pre f is a reference pressure taken outside of the flow, ρre f is the density of the air and Vre f is typically taken to be the mean velocity at an equivalent height of 10m above the ground. The flow around a simple cube in a boundary layer flow has been studied in a number of wind tunnel and full scale experiments including but not limited to Baines (1963), Castro and Robins (1977), Hunt (1982) and Hölscher and Niemann (1998). As illustrated in figure 2 the (mean) pressure distribution on the windward wall of a cube in boundary layer flow has a positive mean C p value, with a slight increase as the leading edge of the cube is approached. There is then a sharp reduction in C p associated with flow separation on the leading edge of the cube, resulting in a region of high negative pressure which gradually reduces as the windward edge is approached. Flow reattachment then occurs and towards the leeward edge of the roof and on the leeward wall there is a slight negative mean C p value. 1277 The Seventh International Colloquium on Bluff Body Aerodynamics and Applications (BBAA7) Shanghai, China; September 2-6, 2012 Figure 2: Vertical central section mean pressure coefficients with the wind normal to one face (0o ). From Richards et al. (2001) There have also been a number of full scale tests on cubic structures. Richards et al. (2001) and Hoxey and Richards (2002) examined the flow structure around a full scale simple cube in synoptic wind flow conditions in a wind tunnel and a full-scale in a rural sight located in Bedfordshire, England. Chay and Letchford (2002a), Chay and Letchford (2002b) and Mason et al. (2009) have previously looked at the flow around a simple cube in a simulated downburst outflow. Chay and Letchford (2002a) used a stationary wall jet in order to simulate the impinging nature of the flow but did not create the ring vortex structure observed by Fujita (1985) in a full scale downburst. X At a distance of D = 1 (where X is the distance from the centre of the impingement and D is the diameter of the nozzle), the general trend of the pressure coefficients over the windward wall and leeward wall of the building were found to resemble cubes placed in boundary layer flow (albeit with differences in magnitude). Noticeable differences over the roof were found. However, it should be stressed that the results varied with respect to different X/D values. Chay and Letchford (2002b) used a translating jet and more successfully produced the pressure field seen in a downburst. However because of the non-stationarity of the data a direct comparison to previous boundary layer flow tests could not be made. For low translating speeds no downburst gust front was formed and the profile of the event closely matched the stationary jet case. However when translating speeds were increased a gust front was formed and previous assumptions regarding the quasi-steady nature of the jet broke down. As a result using the standard pressure coefficient definition seen in equation (1) was not appropriate. Instead the following definition of pressure coefficients was used which took into account the translating velocity component of the flow: CPE (t) = P(t) − PAT MOS 1 2 2 ρVEAV ES, X (2) D 2 VEAV X is the smoothed eaves height velocity and it was calculated from Vre f by fitting a ES, D polynomial to the time history. Chay and Letchford (2002b) also looked at taking into account the static pressure variation caused by the passage of the translating jet. This gave the only the wind induced forces on the structure, internal pressures were assumed to reach equilibrium with their surroundings. The 1278 following equation was used: CPE (t) = P(t) − PSTAT IC, X D (3) 1 2 2 ρVEAV ES, X D Where PSTAT IC, X was modelled using a study of PSTAT IC in the translating jet. However this D had little effect on the values of the pressure coefficients for the largest magnitude pressues and suctions compared with equation (2). The above treatment resulted in the transient wall jet event closely matching the pattern of pressure coefficients found in boundary layer flow for the windward and near the leading edge of the roof. However the magnitude of the pressure and suction coefficients were found to be far greater in the translating flow than the stationary wall jet case. It was also found that the surface pressures on the cube were well correlated for the translating wall jet flow. Since these experiments there has been little in the way of examination of downburst flow around simple or complex structures in a physical simulation. There has however been more recent interest in using numerical modelling to look at wind loading on more complex structures. Chen and Letchford (2004) used a numerical model based upon empirical models of downbursts and then examined the impact on models of a cantilevered tower. The examination of downburst loading on transmission lines has also attracted interest within the numerical modelling community because of the destructive impact downburst events have on such structures. Shehata et al. (2005), Shehata and Damatty (2007), Shehata and Damatty (2008) and Fu et al. (2010) looked at transmission line failure with reasonable success at simulating the failure of a section of transmission line which had already failed. There have also been a limited number of full scale measurements. Lombardo (2009) examined the response of the Texas Tech. Wind Engineering Research Field Laboratory (WERFL) in downburst winds. 2 THE BIRMINGHAM SIMULATOR The Birmingham downburst simulator consists of nine axial flow fans each with a cross sectional area of 0.85m2 used to direct air downwards into a transition section which narrows to a circular cross section of diameter 1m. The latter is suspended 2m above the floor. This opening can be closed with the use of a series of flaps and opened suddenly to create a pulse of air which then travels to the ground and spreads out horizontally as a vortex, in a similar manner to downburst outflow (illustrated in figure 3). Further information can be found in McConville et al. (2007) and McConville et al. (2009). In order to simulate the translating aspect of a downburst two tracks were positioned on the ground and a translating rig built which passed under the simulator, the opening mechanism was automated to release when the rig was at a specific distance from the centre of impingement. 1279 The Seventh International Colloquium on Bluff Body Aerodynamics and Applications (BBAA7) Shanghai, China; September 2-6, 2012 (a) NOAA photo library Image (b) The Birmingham downburst ID: nssl0106 simulator Figure 3: A visual comparison between an actual downburst event and a physical simulation. Since the experiments undertaken by McConville et al. (2007; 2009) a number of modifications have been made. These include the addition of a translating platform with false floor to enable different surface roughness effects to be studied as well as making it far simpler to examine the velocity flow field of the simulator. 3 DATA COLLECTION METHODS A 244 × 98 × 104mm model was pressure tapped down the centreline so that the pressure field around the centre of the simulated downburst could be studied. The arrangement of the pressure taps can be seen in figure 4. Figure 4: The arrangement of pressure taps on the model building The location of the pressure taps is defined in terms of distance from the ground on the windward face measured around a line extending over the central portion of the building. For example, a tap on the windward face at the base would have a position of 0mm while a tap on the base of the leeward face would have a position of 592mm. In what follows the location of the taps have been expressed in terms of a normalised distance, i.e., the distance from the base of the building at the leading edge divided by 592mm. Hence, the location of the taps shown in figure 4 are: tapping 2 (0.0338), 4 (0.0676), 8 (0.2061), 10 (0.3784), 12 (0.4037), 13 (0.4206), 15 (0.4747), 16 (0.5000), 17 (0.5253), 19 (0.5794), 22 (0.6149) and 24 (0.7872). Hjemfelt (1988) found that for the JAWS project data that the maximum wind speed occurred at a height of 80m and a distance of 1.5km from the centre of impingement. Using these values a very rough scaling parameter for height can be found. In the simulator peak wind speed occurred at a height of 20mm, which gives a (vertical) length scaling of 1 : 4000. 1280 However there is an inconsistency in length scaling if the downburst diameter is examined. The average downburst diameter found by Hjemfelt (1988) was 1.8km, the diameter of the downburst simulator was 1m, giving a length scaling of 1 : 1800. This inconsistency highlights one of the difficulties in simulating downburst winds, it is exceedingly difficult to produce a flow field which matches a real world situation. Therefore the results below should be treated with caution, they may not be entirely representative of a full scale downburst wind striking a cube at full scale. 4 RESULTS Preliminary investigations of the wind induced pressure field on a simple structure indicated that the use of a translating platform raised an important issue in terms of the wind loading experienced by the building, i.e., once the release mechanism had been fired the fans continued to rotate (due to the inertial of the fans it was not possible to stop rotation instantly). Hence, as the building moved through the jet a positive pressure field around the entire building was created and as such the flow field was considered not to be representative of a ’typical’ downburst. Hence, in order to general a more appropriate simulation it was decided the building would remain stationary. Figure 5 illustrates the pressure coefficient time series for three pressure taps located around the building arising as a result of a passing gust front associated with the ring vortex. The mechanism for generating the gust front is outlined in McConville et al. (2009). The building X is located at a normalised distance of D = 1.4 from the centre of the impingement. The pressure coefficient is calculated using equation (1) but with Vre f taken as the peak velocity associated with the gust front at an equivalent height of 10m above the ground. In order to smooth the data an ensemble average of 5 velocity times series were used and a 50 point moving average time wind implemented (in keeping with McConville (2008)). This yielded a value of Vre f equal to 15ms−1 . 2 4 6 8 10 12 14 16 18 20 Time(s) (a) Cp time history at tapping 8 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 −0.1 −0.2 0 Pressure Time History Cp 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 −0.1 −0.2 0 Pressure Time History Cp C p Pressure Time History 2 4 6 8 10 12 14 16 18 20 Time(s) (b) Cp time history at tapping 16 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 −0.1 −0.2 0 2 4 6 8 10 12 14 16 18 20 Time(s) (c) Cp time history at tapping 24 Figure 5: Pressure time histories at the centres of the windward,roof and leeward faces Figure 5a is illustrative of the general trend in pressure with respect to time for all of the taps on the windward face. The only difference between the windward face pressure taps is the magnitude of the pressure coefficients, which is greater at the lowest height of 20mm and reduces with height. At the centre of the roof (figure 5b) there is a slight reduction in C p values, possibly indicative of flow separation occurring on the edge between the windward face and the roof. A comparison between figures 5 and 2 illustrates the differences between the C p values of boundary layer and downburst flow. It would appear that flow separation is far less obvious in the downburst case than for boundary layer flow. As a result of this the recovery on the leeward side of the building back to atmospheric pressure is not as noticeable as the boundary layer case. However, it needs to be appreciated that the pressure coefficients in figures 2 and 5 are not directly comparable but the comparison is somewhat instructive. 1281 The Seventh International Colloquium on Bluff Body Aerodynamics and Applications (BBAA7) Shanghai, China; September 2-6, 2012 0.9 Leeward 0.8 roof edge 0.7 Roof 0.6 face 0.5 Windward 0.4 roof edge Leeward 0.3 face Windward 0.2 face 0.1 0 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Normalised Building Height Cp values at time 9.656 s Cp Cp values at time 8.066 s Cp Cp Cp values at time 7.746 s 0.9 Leeward 0.8 roof edge 0.7 Roof 0.6 face 0.5 Windward 0.4 roof edge Leeward 0.3 face Windward 0.2 face 0.1 0 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Normalised Building Height 0.9 Leeward 0.8 roof edge 0.7 Roof 0.6 face 0.5 Windward 0.4 roof edge Leeward 0.3 face Windward 0.2 face 0.1 0 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Normalised Building Height (a) Cp variation at different heights (b) Cp variation at different heights (c) Cp variation at different heights time 7.746s time 8.066s time 9.656s Figure 6: Variation of Cp at different vortex locations (times) Figure 6 illustrates the values of C p across the various faces of the model at different phases of the vortex interaction with the building. In order to estimate the location of the vortex with respect to time, a crude form of PIV was used (see figure 3b). A general trend across all phases of the interaction is the reduction with height of C p coefficients on the windward face of the building. The exact trend varies depending on the degree of interaction of the building with the vortex: figure 6a illustrates the time at which the vortex is just impacting on the front face of the building. The vortex at this stage has fully developed but has yet to interact wholly with the building, hence relatively constant values of C p over a proportion of the windward face can be observed. In figure 6b the vortex is fully interacting with the building and it is hypothesised that the vertical velocity profile seen in figure 1a will have developed against the windward face. Compared to figure 6a there is a slight decrease in the value of pressure coefficients on the roof of the building. In both their is a reduction on the roof-leeward edge of the building before rising again on the leeward face of the building and staying roughly on the leeward face. After the vortex has passed the values of the pressure coefficients on the windward face reduce to around the values seen as the vortex began to interact, only with a trend more similar to that seen when the vortex was interacting. The pressure coefficients on the roof and leeward face are roughly constant with a slight drop on the windward-roof edge and a slight raise on the roof-leeward edge, which is in contrast to the reduction seen at other points of the vortex interaction. 5 CONCLUSION From a relatively low resolution study of the pressure field around a modelled high rise structure in a downburst flow it would appear that there are some differences between the established C p loadings seen in a boundary layer flow and the simulated downburst used in this study. One of the most notable differences is the effect of the large (non-stationary) gust front. The pressure time histories on the model building showed a general trend of having a large positive C p value on the windward face of the building with a slight negative C p value on the roof and leeward faces. The largest C p values were seen when the vortex was interacting most strongly with the building with the weakest seen after the vortex had passed over the building. 5 REFERENCES ABI, 2005: Financial risks of climate change. ABI, URL http://www.climatewise.org.uk/storage/610/ financial_risks_of_climate_change.pdf, [Online; downloaded 19th March 2010]. Baines, W., 1963: Effects of velocity distribution on wind loads and flow patterns on a building. Symposium No. 16, Wind effects on Buildings and Structures, England, Wind effects on Buildings and Structures. Castro, I. and A. Robins, 1977: The flow around a surface mounted cube in uniform and turbulent streams. Journal of fluid mechanics, 79(2), 307–335. Chay, M. and C. Letchford, 2002a: Pressure distributions on a cube in a simulated thunderstorm downburst, part a: stationary downburst observations. Journal of wind engineering and industrial Aerodynamics, 90, 711–732. Chay, M. and C. Letchford, 2002b: Pressure distributions on a cube in a simulated thunderstorm downburst, part b: moving downburst observations. Journal of wind1282 engineering and industrial Aerodynamics, 90, 733–753. Chen, L. and C. Letchford, 2004: A deterministic-stochastic hybrid model of downbursts and its impact on a cantilevered structure. Engineering Structures, 26, 619–629. Fu, D., F. Yang, Q. Li, and J. Yang, 2010: Simulations for tower collapses of 500kv zhengxiang transmission line induced by the downburst. 2010 international conference for power transmission, China Electric Power Research Institute, Qinghe, Beijing, China, CSEE and ISEE/PES. Fujita, T., 1985: Downburst: Microburst and macroburst. University of Chicago Press, IL, pp. 122. Hjemfelt, M. R., 1988: Structure and life cycle of microburst outflows observed in colarado. Journal of Applied meteorology, 27, 900–927. Hölscher, N. and H.-J. Niemann, 1998: Towards quality assurance of for wind tunnel tests: a comparastive testing program of the windtechnologische gesellschaft. Journal of wind engineering and industrial dynamics, 74-76, 599–608. Hoxey, R. P. and P. Richards, 2002: A six metre cube in an atmospheric boundary layer flow, part one full scale and wind tunnel results. Journal of wind and structures, 5(2-4), 165–176. Hunt, A., 1982: Wind-tunnel measurements of surface pressures on cubic building models at several scales. Journal of wind engineering and industrial dynamics, 10, 137–163. Kasperki, M., 1998: Climate change and design wind load concepts. Wind and structures, 1(2), 145–160. Lin, W. and E. Savory, 2006: Large-scale quasi steady modelling of a downburst outflow using a slot jet. Wind and structures, 9, 419–440. Lombardo, F., 2009: Analysis and interpretation of thunderstorm wind flow and its effects on a bluff body. Ph.D. thesis, Texas Technical University, [PhD thesis available from Atmospheric Science group, Texas Techincal University]. Mason, M., 2009: Simulation of downburst wind fields. Ph.D. thesis, University of Sydney, 265 pp., [Available from University of Sydney, School of Civil Engineering]. Mason, M., D. James, and C. Letchford, 2009: Wind loading on a cube subject to a pulsed wall jet simulation of a stationary downburst. Wind and structures, 12(1), 77–88. McConville, A., 2008: The physical simulation of thunderstorm downbursts. Ph.D. thesis, University of Birmingham, 236 pp., [Available from University of Birmingham, Department of Civil Engineering, Edgbaston Birmingham, B15 2TT ]. McConville, A., A. Sterling, and C. Baker, 2009: The physical simulation of thunderstorm downdrafts using an impinging jet. Wind and structures, 12(2), 133–149. McConville, A., M. Sterling, and C. Baker, 2007: An introduction to the scaling issues associated with the physical simulation of thunderstorm downbursts. In Proceedings of the 12th International Conference on Wind Engineering, Cairns Australia, ICWE, 1431–1438. Richards, P., R. Hoxey, and L. Short, 2001: Wind pressures on a 6m cube. Journal of wind engineering and industrial dynamics, 89, 1553–1564. Shehata, A. and A. E. Damatty, 2007: Behaviour of guyed transmission line structures under downburst wind loading. Journal of Winds and Structures, 10, 249–269. Shehata, A. and A. E. Damatty, 2008: Failure analysis of a transmission tower during a microburst. Journal of Winds and Structures, 11(3), 193–208. Shehata, A., A. E. Damatty, and E. Savory, 2005: Finite element modelling of transmission line under downburst wind loading. Journal of Finite Element in Analysis and Design, 71–89. Sterling, M., C. Baker, M. Haines, and A. Quinn, 2011: Scaling a thunderstorm downburst simulator. In Proceedings of the 13th International Conference on Wind Engineering, RAI, Amsterdam, The Netherlands, ICWE. 1283