5m ago

2 Views

1 Downloads

543.50 KB

7 Pages

Transcription

ARTICLE IN PRESSBuilding and Environment 42 (2007) 4079–4085www.elsevier.com/locate/buildenvA comparison between CFD and Network models for predictingwind-driven ventilation in buildingsOmar S. Asfour , Mohamed B. GadiSchool of the Built Environment, The University of Nottingham, University Park, Nottingham NG7 2RD, UKReceived 23 August 2006; received in revised form 12 October 2006; accepted 1 November 2006AbstractThe aim of this paper is to compare the implementation of computational ﬂuid dynamics (CFD) and Network models for airﬂow rateestimation in buildings. The CFD software used is Fluent 5.5. Comparison between the predicted and simulated airﬂow rate is suggested as avalidation method of the implemented CFD code, while the common practice is to compare CFD outputs to wind tunnel or full-scalemeasurements. This could be useful for studies that have no access to laboratory or full-scale testing facilities. Results obtained from testing anumber of cases have been compared and analysed, considering normal and oblique wind directions. The comparison held between mathematicaland CFD results generally showed a good agreement, which seems to justify the use of CFD code for predicting natural ventilation in buildings.r 2007 Elsevier Ltd. All rights reserved.Keywords: Natural ventilation; CFD; Network model; Validation1. IntroductionApplication of computational ﬂuid dynamics (CFD) wasoriginally introduced for industrial applications, but todayit has also become a common tool for assessing buildingventilation and environmental performance. As describedby CIBSE [1], CFD is ‘‘a very powerful technique’’ inpredicting air movement and characteristics. CFD model isbased on the concept of dividing the solution domain intosub-zones. Then, for each zone, the mass, momentum, andenergy conservation equations are solved, utilising theprocessing power of computers. This helps to performcalculations more easily and, in comparison with naturalventilation mathematical models, gives more detailedresults. For example, CFD codes are used to predictairﬂow rate, air velocity and temperature, and airﬂowpatterns inside and around buildings. Many softwaresbased on CFD codes have been developed like, Fluent,Phoenics, and Flovent.In the last few years, an intensive work has been doneusing CFD. Comparisons of CFD results with wind tunneltests have shown good agreement. However, some studiesare limited to the use of CFD modelling, and have noaccess to experimental testing facilities. In this case, it iscrucial to validate the implemented CFD code prior to theproposed modelling study in order to avoid producing anymisleading results. This is because CFD is considered to bea sophisticated modelling technique. One possible option isto compare CFD results to natural ventilation mathematical models, which is the scope of this paper. These modelsvary from simple to complicated models. The commonlyused Network mathematical model has been found to beappropriate for this study, where calculation process isreasonable.Both implemented CFD and Network models are brieﬂyexplained below. This followed by the results obtainedfrom both models regarding airﬂow rate estimation in thecase of normal and oblique wind direction.2. Network mathematical model Corresponding author. Tel.: 44 115 9513118; fax: 44 115 9513159.E-mail addresses: laxosa@nottingham.ac.uk, o.asfour@hotmail.com(O.S. Asfour), mohamed.gadi@nottingham.ac.uk (M.B. Gadi).A ﬂow network is used here to represent different airﬂowpaths in buildings. This network is represented by a group0360-1323/ - see front matter r 2007 Elsevier Ltd. All rights araa.com

ARTICLE IN PRESSO.S. Asfour, M.B. Gadi / Building and Environment 42 (2007) 4079–40854080of nodes, simulating building openings, and a group oflines, simulating ﬂow paths. Thus, it has more ﬂexibility inanalysing natural ventilation problems. Bahadori andHaghighat [2] explained the implementation of this model.Assuming a multi-zone problem, i.e., a building with Nzones, a number of N equations will be established in theNetwork model. Each equation is dependant on theprevious one.Thus, these equations are solved iteratively, where thesummation of mass ﬂow rates should be zero. In the case ofwind-induced ventilation, the knowledge of internal andexternal pressure coefﬁcients for each opening is requiredto ﬁnd out pressure difference across this opening. Thesecoefﬁcients are usually obtained experimentally or fromstandard pressure coefﬁcients data. The following equationis commonly used in predicting airﬂow rate induced bywind pressure ﬃﬃﬃﬃﬃ Qn ¼ Aeff(1)2Dp r ,where Qn is airﬂow rate through an opening n (m3/s), Aeff isthe effective area of this opening, DP is pressure differenceacross it (Pa), and r is air density (kg/m3). Pressuredifference mentioned in the previous equation can beestimated using DP ¼ 0:5rV 2 C pn C pi ,(2)where V is wind velocity at datum level (m/s), Cpn ispressure coefﬁcient at opening n, and Cpi is pressurecoefﬁcient inside the space. Substituting Eq. (1) into Eq.(2), we get 1 2Qn ¼ Aeff V C pn C pi C pn C pi .(3)Considering the Law of Mass Conservation, this equationcan be rewritten asNX 1 2Aeff V C pn C pi C pn C pi ¼ 0:(4)n¼1Therefore, it is in possible to estimate internal pressurecoefﬁcient using Eq. (4) and then airﬂow rate using Eq. (3).To do so, the knowledge of Cpn is essential, as discussed inSection 4.3. CFD codeFluent 5.5 program, one of the most widely usedcommercial CFD codes, has been used in this study.Gambit 1.3 program, which is a pre-processor program,has been used to deﬁne the building geometry. In fact,using two-dimensional modelling in room ventilationproblems does not give realistic simulation of airﬂow, asit does not consider some phenomena that determineairﬂow characteristics, such as airﬂow separation overbuilding sharp edges. Thus, the use of the three-dimensional modeling has been chosen in this study. Thedisadvantage, however, is that this method signiﬁcantlyincreases the required time for solution convergence.Boundary and continuum types have been also deﬁned inGambit program, where velocity inlet and outﬂowboundary types have been used for the solutiondomain. In order to predict airﬂow behaviour in themodelled buildings, CFD calculation mesh has beengenerated, which replaces the air inside and around thebuilding. The idea of this mesh is to divide the solutiondomain into small cells, which are used to predict airﬂowbehaviour using computer processing ability. This can beachieved using different types of mesh, like hexagonal ortetrahedral.For these cells, or volumes in the case of threedimensional modelling, Fluent 5.5 software can numerically solve the three basic conservation equations of mass,momentum and energy in an iterative manner. However,wind-induced ventilation is believed to be more effective inhot climates, when compared with stack-induced one. Thisis because of the relatively lower difference between indoorand outdoor temperatures, which is the main factoraffecting stack ventilation [3]. Therefore, energy settingshave been set off in this study.To process these calculations, an appropriate mathematical model has to be applied on the solution domain. Mostof airﬂow problems in buildings consider airﬂow to beturbulent. Thus, deﬁnition of turbulent model is requiredto help solving the transport equations. Awbi [4] mentioned that the standard k–e Model is believed to be themost used and developed turbulence model. This model ismost likely to predict reasonable results for airﬂow studiesin buildings. Turbulence characteristics have been speciﬁedusing the Turbulence Intensity and Length Scale option,which is recommended in room airﬂow problems, anddepends on the Reynolds number and the inlet size of thecase. Segregated solver has been used, where the fundamental equations are solved sequentially or segregatedfrom each other. The solution reaches the end when theconvergence criteria are met.The modelled prototype represents a room with an inletand an outlet. In Fluent 5.5 program, it is required todeﬁne air velocity magnitude at this inlet. Air velocitymagnitude varies with height. Thus, it is crucial toestimate air velocity at building height, as will bediscussed in Section 5. Rotating the building modelinside the three-dimensional solution domain, whichrepresents the ambient air, can be used to simulate windangle. It is important to note that Guage pressure shouldbe set to 0. This is because the only pressure acts is theatmospheric pressure. The solution progresses in the formof several iterations, which can be monitored by the user.During this iteration process, the residual information ofthe velocity, continuity, and the turbulence parameters ofthe viscous model are continuously updated. Convergenceis achieved when a sufﬁcient error tolerance, deﬁned by theuser, is reached. Many outputs in different presentationmethods can be obtained from Fluent 5.5 software. Thisincludes airﬂow rates and contours of air velocitymagnitudes.www.manaraa.com

ARTICLE IN PRESSO.S. Asfour, M.B. Gadi / Building and Environment 42 (2007) 4079–408540814. Prediction of airﬂow rate using Network mathematicalmodelIt is possible to implement this model for any openingknowing its area and wind static pressure coefﬁcient at thisopening. Prediction process here is divided into four stages: speciﬁcation of pressure coefﬁcient data;speciﬁcation of building conﬁgurations;calculations of internal pressure coefﬁcient;calculations of airﬂow rate.Pressure coefﬁcients data that will be used here are thoserecommended by Liddament [5]. These data have beenproduced in wind tunnels for some common buildingconﬁgurations, and are considered more developed thanthe data mentioned in the British Standards [6]. This is interms of testing more wind angles and considering theeffect of terrain nature. Cases proposed here have nearlythe same volume, but different aspect ratios. Aspect ratiosof 1:1, 1:2 and 2:1 are considered, as depicted in Fig. 1.This ﬁgure also shows values of pressure coefﬁcient, with areference number indicating wall number. Initially, calculations have been performed considering a relatively lowand high reference wind velocities, namely 1 and 5 m/s, andtwo wind angles, namely 01 and 451. However, it has beenobserved that discrepancy percentage is not sensitive towind velocity magnitude, as the percentage is nearly thesame in the case of same building geometry and differentapproaching wind speeds. This result supports whatLiddament [5] has mentioned that pressure coefﬁcient isnormally assumed to be independent of wind speed but notdirection. Thus, the following investigation will be conﬁnedto velocity magnitude of 1 m/s.It is important to note that at oblique wind direction,effective area of the windows is less. By knowing that windangle is 451, window effective width can be simplycalculated using Right Triangle Trigonometry. Thus,window effective area can be estimated by multiplyingwindow width by its height. This is 1.4 2 ¼ 2.8 m2.Internal pressure coefﬁcients can be estimated from Eq.(4). As the above-illustrated models have only two zones:indoor and outdoor, this equation can be simpliﬁed as C pn C pi 1 2 þ C pðnþ1Þ C pi 1 2 ¼ 0.(5)It is required to know air velocity at building height toimplement Network model for airﬂow rate estimation. Thisis possible using the following common equation [7]:aV ¼ V r cH ,(6)where V is wind speed at datum level (m/s), Vr is referencewind speed (obtained from meteorological data) (m/s), H isthe height of the building, c is parameter relating windspeed to terrain nature (0.68 in the open country terrain),and a is an exponent relating wind speed to the heightabove the ground (0.17 in the open country terrain).Knowing different building heights, as illustrated inFig. 1. Illustration of the cases tested in this study.Table 1Airﬂow rate prediction, using the Network modelCaseOpeningAeff .250.125 0.1 0.025 0.2 0.275Qn (m3/s)Qn (kg/s)2.40 2.402.37 2.372.66 2.661.53 1.531.52 1.521.75 1.752.94 2.942.90 2.903.26 3.261.88 1.881.87 1.872.14 2.14Table 1, wind velocity at building height has beenestimated and implemented in Eq. (3) to estimate airﬂowrate. Results are shown in Table 1.www.manaraa.com

ARTICLE IN PRESS4082O.S. Asfour, M.B. Gadi / Building and Environment 42 (2007) 4079–4085This airﬂow rate is given in m3/s. It is possible to convertit to mass ﬂow rate, in kg/s, by multiplying it by air densityin order to facilitate the comparison with CFD results.Value of air density should match the one under whichpressure coefﬁcient data were estimated. As there is noindication of this value in the related resource, this valuewill be assumed to be 1.225 kg/m3, which is the defaultvalue used in Fluent 5.5 program and nearly the same ofthe standard air density value, i.e., 1.2 kg/m3.5. Prediction of airﬂow rate using CFD modellingUsing Fluent 5.5 program, it is possible to estimateairﬂow rate through building openings. Modelling processincludes the following four stages:Fig. 3. Illustration of the calculated and simulated wind velocity proﬁle(for case 1, as an example). Table 2Air velocities for the different sub-inlets used in CFD modelling, asestimated from Eq. (6) drawing of building models in Gambit program;generating and exporting the calculation mesh to Fluent5.5 program;deﬁnition of solution code;computing airﬂow rate.As discussed in Section 3, building models are simulatedthree-dimensionally in order to obtain more realisticairﬂow pattern. This explains the signiﬁcant differencesbetween results obtained using both two and threedimensional modelling in the early stage of this study.Thus, it is important to simulate the ambient air around thebuilding. To do so, the modelled room will be placed insidea three-dimensional box, as shown in Fig. 2.By rotating the room inside this box, it is possible tosimulate oblique wind direction. However, this requires alarger domain to ensure solution convergence. Forexample, in the case of 01 wind direction, the building isplaced in a 30 m 30 m 20 m box. In 451 wind direction,the extra sheer stresses, resulted by the oblique buildingwalls, cause reversed ﬂows to occur. This leads the solutionto diverge, and makes it essential to use a larger domainsize. This size has been gradually increased until anacceptable size of 50 m 50 m 20 m was achieved (Fig. 3).In ventilation modelling, it is important to consider windspeed variation with height, due to the frictional effect ofthe ground. It is possible to do so using Eq. (6). Windvelocity proﬁle can be deﬁned along the velocity-inlet, asillustrated in Fig. 2, using ‘User Deﬁned Function’ optionin Fluent 5.5 software. However, application of this optionFig. 2. Simulation of the ambient air around buildings.Height above the ground (m)Sub-inlet numberWind velocity (m/s)Cases 1 and 4 (room height is 5 m)511021532040.8941.0061.0781.132Cases 2,3,5 and 6 (room height is 4 m)41821231642050.8610.9681.0371.0891.132in three-dimensional simulation required some advancedC programming. It is possible, as an approximationmethod, to divide the velocity inlet into many sub-inlets.Each sub-inlet will have a different air velocity magnitudedepending on its height above the ground. This method hasbeen found to be useful and good results have beenobtained, as has been concluded from this study. To allowfor the estimation of airﬂow rate, it is recommended to splitthe solution domain into the following zones: room walls,ambient air, room interior, and openings volumes. The lastthree volumes were deﬁned as ﬂuid continuum. The beneﬁtof this arrangement is that Fluent 5.5 program will deﬁneopenings surfaces separately, which facilitates airﬂowcomputing process (Table 2).The use of three-dimensional simulation leads to thinkabout the resulting mesh sizes and the required processorcapacity. For example, the use of 0.2 m mesh spacing forthe entire domain, 30 cm 30 m 20 m in cases 1–3, wasbeyond computer memory and speed in this study. Acommon solution here is to create a hierarchy in mesh sizeto be ﬁne inside the building and larger around it. In anycase, a trial-and-error process is recommended to ﬁnd outthe most appropriate mesh conﬁguration. Rough meshescan be suitable and sufﬁcient in many CFD simulationwww.manaraa.com

ARTICLE IN PRESSO.S. Asfour, M.B. Gadi / Building and Environment 42 (2007) 4079–4085cases [8]. Generally, the following mesh conﬁguration hasbeen found to be acceptable: Meshing scheme: hex-map or hex-submap. Mesh elements here are hexahedral, which is applicable to theshape and topological characteristics of the modelledcases. In case of oblique wind, the use of tetrahedralmesh has been found to give more accurate results, asdiscussed in the following section.Mesh node spacing: in case of normal wind direction,this spacing is 0.6 m. In the case of oblique winddirection, solution domain was divided into two zones:room interior, with 0.5 m spacing in the hex-map mesh,and room exterior, with 1 m spacing in the tetrahedralmesh.Once the mesh is ready, it can be exported to Fluent 5.5program in order to perform the calculations, which maytake several hours depending on the size and complexity ofeach individual case. Fig. 4 illustrates the resulting airﬂowpattern for these cases, presented by contours of velocitymagnitudes over a 2-m height section. Generally, the4083observed airﬂow pattern, in all the cases, has been found tobe reasonable for such sharp-edge geometries. Forexample, when wind reaches the windward face, a highpressure zone is formed there. This pressure pushes airinside, around, and over the building. Some standardfeatures can also be observed. This includes airﬂowseparation over building sharp edges. This phenomenonusually occurs when airﬂow layers hit a sharp edge of thebuilding and thus lose their momentum. After somedistance, the separated airﬂow joins its original streamagain in a point called the reattachment point.After achieving solution convergence, it is possible toobtain the mass ﬂow rate directly from the software,utilising the Surface Integrals option for the relevant inletsurface. Results obtained for the different cases areillustrated in Table 3.6. Comparison between airﬂow rate prediction usingNetwork and CFD modelsTable 4 shows a comparison between airﬂow ratepredicted by the Network and CFD models. The negativeFig. 4. Airﬂow pattern, presented by contours of velocity magnitudes, for the cases modelled in this study.www.manaraa.com

ARTICLE IN PRESSO.S. Asfour, M.B. Gadi / Building and Environment 42 (2007) 4079–40854084Table 3Airﬂow rate prediction, using CFD modelCaseOpeningQn (kg/s)CaseOpeningQn (kg/s)11234123.19 3.192.98 2.983.30 3.3041234121.78 1.781.69 1.691.92 1.922356Table 4Discrepancy percentage between estimated and modelled airﬂow rate forcases 1–6CaseOpeningQn (kg/s)(CFD)Qn (kg/s)(Network)11234121234123.19 3.192.98 2.983.30 3.301.78 1.781.69 1.691.92 1.922.94 2.942.90 2.903.26 3.261.88 1.881.87 1.872.14 2.1423456Discrepancy(%)7.82.71.2Fig. 5. Skew observed in the hex-map and sub-map meshes used in thecase of 451 wind direction.Table 5Discrepancy percentage between estimated and modelled airﬂow rate inthe case of oblique wind direction, after changing mesh typeCaseOpeningQn (kg/s)(CFD)Qn (kg/s)(Network)Discrepancy(%)41234121.84 1.841.83 1.832.26 2.261.88 1.881.87 1.872.14 2.14 2.2 5.6 10.75 11.5sign indicates that the modelled airﬂow rate is less than thecalculated one. In general, it shows that a good agreementhas been achieved. Discrepancy percentage observed isusually acceptable in airﬂow rate prediction, which is givenas a snapshot and measured in kg/s.However, discrepancy percentage, in general, is higher inthe case of oblique wind direction. CFD code used here isthe same of that one used with the normal wind direction.However, it has been found that the use of hex-map andsub-amp meshes in the oblique wind direction results in aless mesh quality. This is because tilted walls resulted inhigh angular skew between the edges of mesh cells, about0.7, which is considered high. For example, an excellentmesh has a skew less than 0.25 [9] (Fig. 5).Therefore, mesh type of the ambient air zone has beenchanged to the tetrahedral mesh, which has more ﬂexibilityin meshing such geometries. Hex-map mesh is still used forroom interior. Table 5 shows a recalculation of thecomparison held between results obtained from bothmathematical and CFD models in the case of obliquewind direction. Change of mesh type seems to have a goodeffect, as discrepancy percentage has been signiﬁcantlyreduced in all the three cases.The differences observed between airﬂow rates predictedby the Network and CFD models at both wind directionscan be justiﬁed by many reasons. One of them is theapproximation method used in simulating air velocity6 2.25.3proﬁle, as explained in Section 5. This is because windinduced airﬂow rate is dependant on wind velocity. As thesquare of air velocity is used in the estimation of thispressure difference, any error in air velocity results in alarger error in airﬂow rate value.Another reason can be the approximation of themathematical procedure used. This is, on one hand,because the used wind pressure coefﬁcient values areaveraged over the whole speciﬁed building face, and nota speciﬁed position on it. Liddament [5] highlighted thispoint and told that accurate evaluation of this parameter(i.e. pressure coefﬁcient) is one of the most difﬁcult aspectsof air inﬁltration modelling. In addition, these data weregenerated in wind tunnel experiments, where air density,and therefore pressure, is affected by air temperature,which is not the case in the isothermal CFD simulationcarried out. On the other hand, air inﬁltration model usedcontains many assumptions to enable the estimation ofairﬂow rate through a reasonable mathematical process.On the opposite, CFD considers the different values of airpressure on the opening, and calculate airﬂow rate as asummation their product.One more reason is related to the pressure coefﬁcientdata used in the case of oblique wind direction. Airpressure distribution around a solid model changes if it isprovided with openings. In the case of 45o wind incidence,www.manaraa.com

ARTICLE IN PRESSO.S. Asfour, M.B. Gadi / Building and Environment 42 (2007) 4079–40854085of the building two windward faces, values of pressurecoefﬁcient for the two windward faces in Fluent 5.5 havebeen found to be different, i.e., 0.28 for the face that hasthe window, and 0.32 for the other one. This can benoted in Fig. 6, showing contours of static pressure.7. ConclusionThis paper has compared the use of CFD and Networkmodels for wind-induced ventilation prediction in buildings. It compares the calculated airﬂow rate using themathematical Network model and the modelled one usingFluent 5.5 program. Many cases with a variety in buildinggeometries and wind directions have been considered.Results obtained support the use of the proposed CFDcode for wind-induced natural ventilation in buildings, as agood agreement has been achieved. This can be recommended as a validation method for studies that have noaccess to laboratory or full-scale testing facilities. Thisstudy has also revealed that the chose of mesh type andsize, in addition to the domain size are critical parametersin three-dimensional CFD modelling.Fig. 6. Contours of pressure magnitude (Pa) showing different pressuredistributions in the modelled cases and the standard solid ones.building model has two windward faces. In the case of asolid model, average pressure coefﬁcient over these windward faces is the same for the square cases, and has arelatively small difference in the rectangular one. In thecase of placing an opening at one of these two windwardfaces, it is expected that air pressure distribution willchange, and there will be no more balance at itsdistribution on these windward faces. This is becausewindward with solid geometry receives more wind deﬂection on it. On the opposite, windward with an openingreceives less wind deﬂection, and therefore wind pressurewill be less too. For example, pressure coefﬁcient values onthe windward faces in case of 451 wind direction and asquare building form are the same, i.e. 0.35, as presentedin the standard data used in this study. This is true for solidmodels. However, in the case of placing an opening in anyReferences[1] CIBSE. Natural ventilation in non-domestic buildings. London:CIBSE; 1997.[2] Bahadori M, Haghighat F. Passive cooling in hot arid regions indeveloping countries by employing domed roofs and reducing thetemperature of internal spaces. Building and Environment1985;20(2):103–13.[3] Chow WK. Wind-induced indoor-air ﬂow in a high-rise buildingadjacent to a vertical wall. Applied Energy 2004;77:225–34.[4] Awbi H. Ventilation of buildings, second ed. London: Spon Press;2003.[5] Liddament MW. Air inﬁltration calculation techniques—an application guide. Bracknell. Coventry: The Air Inﬁltration and VentilationCentre; 1986.[6] British Standards Institution. BS 5925: code of practice for ventilationprinciples and designing for natural ventilation. London: BSI; 1991.[7] CIBSE., ﬁfth ed. CIBSE guide, vol. A. London: CIBSE; 1988.[8] Kindamgen J, Krauss G, Depecker P. Effects of roof shapes on windinduced air motion inside buildings. Building and Environment1997;32(1):1–11.[9] Fluent Inc. Gambit 1.3 program help menu; 1988.www.manaraa.com

The CFD software used i s Fluent 5.5. Comparison between the predicted and simulated airﬂow rate is suggested as a validation method of the implemented CFD code, while the common practice is to compare CFD outputs to wind tunnel or full-scale . Both implemented CFD and Network models are brieﬂy explained below. This followed by the .

Related Documents: