Proceedings of B25: 4th Conference of International Building Performance imulation Association, Hyderabad, India, Dec. 7-9, 25. COMPARION BETWEEN THE WIND TUNNEL EXPERIMANT AND CFD ANALYI INTENDED FOR APARTMENT HOU TUDY OF TUNDARD K-E MODEL AND DURBIN MODEL Kazuki Yamada, Yoshifusa Nishimura 2, Koji akai, Ryoichi Kajiya Meiji University, Kanagawa, Japan 2 HAKO Corporation, aitama, japan ABTRUCT In this study, it is a purpose to grasp the effect of natural ventilation in the apartment house. This study targeted the apartment house of 4 floors. First, we did the wind tunnel experiment that used the reproduction model, and measured the wind pressure coefficient. Next, we did CFD analysis that used k-ε model and Durbin model. We understood utility and problem of the CFD analysis by comparing the wind tunnel experiment. Consequently, analysis in windward was similar to the experiment. However, in leeward, the error margin was large. In addition, it has been understood that Durbin model is similar to the wind tunnel experiment than standard k-ε model. INTRODUCTION Natural ventilation is enumerated in one of the techniques for decreasing the cooling load of summer in the house. It is a purpose in this study to understand the effect of natural ventilation in the apartment house. In anamnestic study (e.g. Yamada et al., 24), it was confirmed that dwelling unit position and external wind direction influence the amount of ventilation. Therefore, we did the wind tunnel experiment and measured the wind pressure coefficient. The result indicated the difference by dwelling unit and the influence of external wind. In addition, we did CFD analysis. was used for standard k-ε model to the turbulent model. However, in analysis by standard k-ε model, there is a problem that the turbulence energy would be overestimated at peripheral windward corner section (e.g. Tominaga et al., 2). For that reason, it was also used Durbin model that has been improved its problem (e.g. Durbin, 996). To find out accuracy validation and verification of CFD analysis, we compared the wind tunnel experiment and CFD analysis, and standard k- ε model and Durbin model. THE WIND TUNNEL EXPERIMENT UMMARY The wind tunnel experiment used the Eiffel-type wind tunnel device of Osaka City University. The target dwelling unit for experiment is apartment house of 4 floors in Kanagawa Prefecture agamihara City. Experiment used the reproduction Figure Model photograph #3 Room A building #7 Room B building Figure 2 installation situation of the model and measurement points model of reduced scale /2. Model photograph is shown in Figure. A building reproduces balcony, pillar, and external stairs by using the PLA resin. On the other hand, B building made a rectangular model by using the acrylic board. Because, it has aimed at the measurement of A building in this study. The surrounding situation is not reproduced. Figure 2 shows the installation situation of the model and measurement points of wind pressure coefficient. The measurement points in perpendicular direction are position of the, 5, 8,, 3 and 4th floors. Experiment wind direction is 32 wind directions in which the range of -36 is divided into.25. The wind direction in the north was set to. The wind direction angle was set clockwise. - 87 -
Proceedings of B25: 4th Conference of International Building Performance imulation Association, Hyderabad, India, Dec. 7-9, 25. Inflow Wind B :.9(x).8(y) (z)hb A :.4(x).3(y) (z)hb θ 7Hb z y x 7Hb Figure 3 Calculation domain Table conditions software : House cowd, Tarbulence model : tandard k-ε model and Durbin model, Calculation domain : 7Hb(x) 7Hb(y) (z), Mesh size : 98(x) 98(y) 4(z), Algorithm : IMPLEC, cheme of converctive : Bounding Central(Momentum eq.), PLD(calar eq.), Reference length : Height of building(h =2.9cm),Reference speed : Inflow wind velocity in model vertex part(u =6.4m/s), Wall boundary condition : Generaraized Log Law, Outlet condition : Pressure boundary, Convergence criterion : Continuity eq.<.5e- 4(ave.), Wind pressure cofficient : calculate from static pressure 6 Measurement 6 Measurement 5 5 y = 3.286x 9.5376 u2 4 4 k3 y =.369x -.29 H/H 3 H/H 3 2 y = -2354x 2-76.9x + 23.993 k2 2 y = 36.4x 2-7.638x +.8267 u.8.6.4 - -.4 -.6 -.8 - Figure 4 Mesh layout y = 2E+7x 4.56 k..2.3.5.5 Turbulent energ(k) U/U Figure 5 Inflow velocity and turbulent energy #3 room #7 room.8.6.4 - -.4 3 53 83 -.6 7 57 87 3 33 43 -.8-7 37 47 22.5 45 67.5 9 2.5 35 57.5 8 22.5 225 247.5 27 292.5 35 337.5 36 22.5 45 67.5 9 2.5 35 57.5 8 22.5 225 247.5 27 292.5 35 337.5 36 3 53 83.8.6.4 3 33 43 - -.4 -.6 -.8-22.5 67.5 9 2.5 45 35 57.5 8 22.5 225 247.5 27 292.5 35 337.5 36 7 57 87.8.6.4 7 37 47 - -.4 -.6 -.8-22.5 67.5 9 2.5 45 35 57.5 8 22.5 225 247.5 27 292.5 35 337.5 36 Figure 6 measured by the wind tunnel experiment ANALYI UMMARY Calculation domain is shown in Figure 3, and mesh layout is shown in Figure 4. Calculation domain is 7Hb(x) 7Hb(y) (z), mesh was divided into 98(x) 98(y) 4(z). It was more finely divided the surroundings of A buildings. The size of the least mesh is.hb. This study analyzed the wind direction that corresponded to east-southeast (E), southeast (), south-southeast (), and south (). Therefore, building model reverses to x shaft and is arranged by the convenience of analysis software. As for building model, balcony etc. is not reproduced. Moreover, B building uses the model approximated like the stairs. The size of mesh of building model is.. software used the house code (e.g. akai et al., 25), and turbulent model used standard k-ε model and Durbin model. Other analysis conditions are shown in Table. In the study in the past, when Durbin model is used, numerical result is reported to be different by changing the value of calculation parameter α (e.g. Tominaga et al., 2). Then this study examined the most suitable α by calculating.-. of.. The inflow velocity and turbulent energy used measurements of the wind tunnel experiment (Figure 5). THE WIND TUNNEL EXPERIMENT REULT Result of the wind tunnel experiment is shown in Figure 6. Fig. shows wind pressure coefficient in corresponding of #3 and #7 room. The diagrammatic charts are made separately by and. In, the change is few when wind direction angle is 67.5-225. This reason is thought to be influence of B building. In, wind direction angle with the largest value is different. As a result, even if it is the same wind direction, it was confirmed that the wind pressure coefficient was different according to dwelling unit position. - 87 -
Proceedings of B25: 4th Conference of International Building Performance imulation Association, Hyderabad, India, Dec. 7-9, 25. tandard k-ε model Durbin model (α =.5) Figure 7 Flow-field -.3 -.3 -.3 -.3 -.3 -.35 -.35 -.35 -.35 -.5 -.5-5 -5 -.5 -.5-5 -5 -.5 -.5-5 -5 -.5 -.5-5 -5 -.3 -.3 -.3 -.3 -.3.5.5..7.3.3.4.4.3.3.4.4.3.3.4.4.3.3.4.4..5.5.6...5.5.6...6.5.5...6.5.5. α =. α =.9 α =.8 α =.7 -.35 -.35 -.35 -.3 -.5 -.5-5 -.3-5 -.3 -.35 -.4 -.5 -.5-5 -.3 -.35 -.5 -.5-5 -.35-5 -.3 -.3 -.3.5.7.3.3.4.4..5 α =.6 α =.5 α =.4 Figure 8 Distribution of each α ANALYI REULT.6.5. Flow-field Comparison Figure 7 shows flow-field by the analysis that uses standard k-ε model and Durbin model. Figure is x-y section in the vicinity of rooftop. wind direction is, and α of Durbin model is.5. For Durbin model, the separated flow was confirmed in rooftop. On the other hand, the separated flow on the rooftop was not able to be confirmed. Therefore, it was confirmed not to be able to reproduce the current of air distribution correctly by the analysis that used standard k-ε model. Distribution of each α Figure 8 shows analysis result by Durbin model. Figure showed the distribution of wind pressure coefficient in A building ( and ). Moreover, it to be easy compare it, development view reverses to x shaft and is written. Calculation wind direction is. α was changed into.-. and it.3..4.4.5.6.5.3..4.3..5.6.6.5 calculated. However, when α was.-.3, calculation was emanated. In south (windward side), it was confirmed that distribution changed little by little by α. On the other hand, in north (leeward) side, it was confirmed that it was more changeless than windward side. Then, in analysis on the windward side, it has been understood that it is important to change α according to the condition. COMPARION.4.3. α =.-.3 is emanated Distribution Figure 9 shows the distribution by the wind tunnel experiment and CFD analysis (standard k-ε model and Durbin model). Figure showed the distribution of wind pressure coefficient in A building ( and ). In this report, Root Mean quare Error (RM) of experiment and analysis on windward side was calculated, and, Durbin model shows only the result of α that RM is the smallest. First, it examines on the (windward side). In wind direction is, standard k-ε model has the difference - 872 -
Proceedings of B25: 4th Conference of International Building Performance imulation Association, Hyderabad, India, Dec. 7-9, 25. Wind direction The wind tunnel experiment tandard k-ε model Durbin model -.35 -.35 -.3 -.6 -.5 -.6 -.65 -.7 -.5 -.5-5 -.3 -.35 -.3 -.5 -.5-5 -.35 -.55 -.6.7 -.55.6.5-5.4 -.4.7.5 -.3 α=.4.4.7.5.6.6.5.4.3.3.4..7.5.6.4.3..4.3..5.6.6.5.4.3. -.6-5 -.35 -.45 -.5 - -.55 -.65-.7 -.5 -.5 - -.3 -.35 -.4 -.5 -.5 - -5 -.3 -.35 -.4-5.3.4.5.5.3.4.6.3.5.4 α=.4.3.6.7.4.5.5. -..3.4.7.8.5.5.9.6. -..3.6.7.4.5.5.8 -.5 -.3-5 -.35 -.45 -.5 -.5 -.. -.55. -.55.3 -.5 -.5-5 -.3 -.35 -.3-5..3 -.5 -.5-5 -.3 -.3 -.35. α=...4.5 -..4.5.6 -. -.3.4.5.3..3. -.3 -.4 -.55 -.55 E -.5-5 -.35 -.45 -.5 -.5 -. -.6 -.55 -.5 -.3 -.5 -.5 - -5 -.3-5 -.5 -.5-5-.3 -.35 -.4 -.45 -.4 α=. -. -. Figure 9 distribution by the wind tunnel experiment and CFD analysis with experiment in the vicinity of the rooftop. On the other hand, Durbin model shows that is nearer experiment than standard k-ε model. In, and E, both models became similar distribution, and it looks like experiment. econd, it examines on the (leeward side). In and, because it is uniform distribution with few changes in analysis, the difference is especially great compared with - -. - -. -. -.5 -.4 -.6 --.3 -.7 -.3 - -.6 -.4 -.5 -.7 - -.3 experiment. Neither nor E is corresponding to experiment. From these results, on the windward side, analysis that uses standard k-ε model and Durbin model is effective. However, distribution of standard k-ε model is difference in wind direction is. Yet on the leeward, regardless of wind direction, it was confirmed that there were a lot of problems. Therefore, the improvement was necessary. -. -. - 873 -
Proceedings of B25: 4th Conference of International Building Performance imulation Association, Hyderabad, India, Dec. 7-9, 25. Table 2 Root mean square error (RM) of experiment and analysis Windward Leeward Turbulence model E E tandard k-ε model.6.27.88.9 58 57 3.75 α=..38.92.63.6 7 49 47 α=.9.59.99.64.6 73 5 4.94 α=.8.76.3.68. 37 52 35.93 Durbin α=.7.83.7.7.7 7 5 38.96 model α=.6.67.3.8.3 5 5 47 4 α=.5.37.84.76.32 56 68 52 α=.4.6.68.69.7 87 95 6 53 α=.3~. Emanation Emanation.8.6.4.8.6.4.8.6.4 α=.4.8.6.4 α=.4 - - - - -.4 -.4 -.4 -.4 -.6 -.6 -.6 -.6 -.8 -.8 -.8 -.8 - - -.8 -.6 -.4 -.4.6.8.8.6.4 - - -.8 -.6 -.4 -.4.6.8.8.6.4 E -.8.6.4 - -.8 -.6 -.4 -.4.6.8 α=. -.8.6.4 - -.8 -.6 -.4 -.4.6.8 E α=. - - - - -.4 -.4 -.4 -.4 -.6 -.6 -.6 -.6 -.8 -.8 -.8 -.8 - - -.8 -.6 -.4 -.4.6.8 tandard k-ε model Durbin model Figure Correlation diagram of experiment and analysis Distribution of each α Table 2 shows RM of experiment and analysis. Red-letter in Tab is a case where RM is minimized in each wind direction and. On the windward, in all wind direction and all α, error of Durbin model is smaller than of standard k-ε model. In analysis by Durbin model, it was confirmed that error changed depending on value of α. α with the smallest error was.4 in and, and it was. in and E. Therefore, it has understood that it was important to decide the best α according to wind direction. On the other hand, on the leeward, error of Durbin model was smaller than of standard k-ε model in and. However, in and, standard k-ε model was smaller than Durbin model. Correlation diagram of experiment and analysis is shown in Figure. On the windward(c>.), it is roughly corresponding excluding E. On the leeward(c<.), difference is great in all wind direction and both turbulence model. CONCLUION In this study, it is a purpose to grasp the effect of natural ventilation in the apartment house and to improve the predictive accuracy of CFD analysis. The result of wind tunnel experiment, it was confirmed that the wind pressure coefficient was different according to dwelling unit position and - - -.8 -.6 -.4 -.4.6.8 - - -.8 -.6 -.4 -.4.6.8 wind direction. The result of comparing the wind tunnel experiment with CFD analysis(standard k-ε model and Durbin model) is shown below. ) In analysis of windward, Durbin model is nearer the wind tunnel experiment than of standard k-ε model. However, in leeward of analysis, both turbulent models did not coincide with the wind tunnel experiment. 2) When Durbin model is handled to the turbulent model, it is important to change α according to wind direction. REFERENCE Yamada, Nishimura, akai: A tudy on the Grasp of the Natural Ventilation Performance in Apartment, Proceedings of J / JWEA joint Conference, pp. 67-7, 24 Tominaga, Mochida, Murakami, awaki: Comparison of Performance of Various Revised k-ε Models Applied to CFD of Flowfield around a High-Rise Building, J. Archit. Plan. Environ. Eng., No.556, pp. 47-54, 22 Durbin,P.A.:On the k-ε stagnation point anomaly, Int.J.Heat and Fluid Flow, vol.7, 89-9, 996 akai, Kubo, Ishihara: A study on the Estimate of the Natural Ventilation Effect (Part.4), Architecture Institute of Japan Research Report. Kyushu Branch, pp. 43-46, 25 - - -.8 -.6 -.4 -.4.6.8-874 -