ASHRAE OR-05-8-4-2005 Numerical Simulation of Airflow and Airborne Pathogen Transport in Aircraft Cabins - Part 1 Numerical Simulation of the Flow Field《气流和在飞机舱空中传播的病原的数值模拟-第1部分 流场.pdf

《ASHRAE OR-05-8-4-2005 Numerical Simulation of Airflow and Airborne Pathogen Transport in Aircraft Cabins - Part 1 Numerical Simulation of the Flow Field《气流和在飞机舱空中传播的病原的数值模拟-第1部分 流场.pdf》由会员分享，可在线阅读，更多相关《ASHRAE OR-05-8-4-2005 Numerical Simulation of Airflow and Airborne Pathogen Transport in Aircraft Cabins - Part 1 Numerical Simulation of the Flow Field《气流和在飞机舱空中传播的病原的数值模拟-第1部分 流场.pdf（9页珍藏版）》请在麦多课文档分享上搜索。

1、OR-05-8-4 Numerical Simulation of Airflow and Airborne Pathogen Transport in Aircraft Cabins-Part i: Numerical Simulation of the Flow Field C.-H. Lin, PhD, PE Member ASHRAE Member ASHRAE K.H. Dunn J.L. Topmiller R.H. Horstman, PE ABSTRACT An initial study to develop a numerical tool using compu- tat

2、ional fluid dynamics (CFD) methodsfor investigating the potential of disease transmission in commercial aircraft is completed. To gain insight of the general airflow pattern, a detailed CFD model of a small section in the passenger cabin of a B 767-300passenger cabin was built and a Reynolds-aver- a

3、ged Navier-Stokes (RANS) simulation was performed. By comparing with the available test data, the RANS simulation substantially underpredicted the turbulence intensiy, espe- cially in and around the breathingzone. A separate large eddy simulation (LES) was conducted to obtain a more realistic turbul

4、ent energy transport in a generic cabin model. The LES- predicted turbulence level is in fairly good agreement with the test data. Based on the LES results, the kund E equations used in the RANS simulation were modijed by using a special user subroutine. A RANS simulation with adjusted turbulence wa

5、s then employed to simulate the dispersion ofairbornepathogen in the detailed passenger cabin model. These adjustments allow for the simulation ofdisease transmission using less than I/l O0 of the computing hardware resources required for an equivalent LES of airflow and particle transport. INTRODUC

6、TION Applications of computational fluid dynamics (CFD) in studying airflow and heat transfer in ventilated rooms were incepted nearly three decades ago (Nielsen 1974). Scalar species transport was later added to address indoor air quality issues by researchers (Murakami et al. 1988; Horstman 1988;

7、Chen et al. 1990). Haghighat et al. (1989, 1990, 1992) have expanded the domain of interest to a building of multiple compartments. CFD has since beenused to evaluate the indoor M.F. Ahlers L.M. Sedgwick, PE J.S. Bennett, PhD Associate Member ASHRAE S. Wirogo, PhD environment of various types of bui

8、ldings, as reported by Chen and Srebric (2001). Numerical studies dedicated to the contaminant transport in hospital operating rooms have been conducted by Lo (1997). To improve occupant thermal comfort, CFD has been used to modiSr and/or optimize the air ventilation system in automotives (Lin et al

9、. 1992) and commercial airplanes (Aboosaidi et al. 1991; Baker et al. CFD techniques used in this study varied. Baker et al. (2000) analyzed complex aircraft interiors using a laminar flow simulation. Like most reported work, Mizuno and Warf- ield (1992) and Aboosaidi et al. (1991) both applied the

10、Reynolds-averaged Navier-Stokes (RANS) approach to look at velocity fields but did not address the species transport issues. To accurately predict the turbulence levels in room airflow, Emmerich and McGrattan (1998) and Zhang and Chen (2000) have used a large eddy simulation (LES) tech- nique. Due t

11、o the relatively large physical dimensions involved in air ventilation flows, direct numerical simulation (DNS) is still prohibitively expensive to pursue with the currently available computing resources. Airplane cabin airflow has the characteristics of very high turbulence levels with transitional

12、 Reynolds numbers. It is not always the case; however, a majority of known turbulence models used in UNS simulations underpredict the turbulence levels to various degrees (Jin and Braza 1994; Robinson and Hassan 1997). Therefore, the greatest hurdle in accurately predicting the airborne pathogen dif

13、fusion lies in realizing the very large turbulence levels that occur in aircraft cabins. Since diffusion is dominated by turbulence, an accurate prediction of turbulence is required. 2000). Chao-Hsin Lin, Raymond H. Horstman, and Leigh M. Sedgwick are associate technical fellows and Mark E Ahlers is

14、 a lead engineer at Boeing Commercial Airplanes Group, Seattle, Wash. Kevin H. Dum is an environmental engineer, Jennifer L. Topmiller is a mechanical engineer, and James S. Bennett is a service fellow at the National Institute for Occupational Safety and Health, Cincinnati, Ohio. Sutikno Wirogo is

15、a support engineer at Fluent, Inc., Lebanon, NH. 02005 ASHRAE. 755 The objective of this part of our study is to provide a realistic simulation of the flow field in an aircraft cabin using CFD. The B767-300 was chosen as the representative airplane cabin. The method used focused on the implementatio

16、n of a commercially available code, with adjustments made to the predicted diffusion to more accurately match test and LES data. LES was used as a predictive tool for turbulence levels by comparison to a relatively scarce set of test data of cabin airflow. However, LES modeling of a passenger cabin

17、is not practical due to intensive computing requirements. For example, an LES model for one seat row of a passenger cabin of a B767-300 airplane would require 1000 gigabytes of RAM. In fact, even grids built for RANS models are barely within the available resource limits for just two seat rows of a

18、B767-300 passenger cabin. For the turbulence study, we were unable to build an LES model of a real airplane cabin because of the aforementioned resource constraints. Instead, a simplified geometry was conceived that retains the transitional nature of the flow but with orthogonal geometry that is mor

19、e amenable to the development of a highly detailed grid. The rationale is that if the turbulence levels from the simplified geometry LES model match those measured in the airplane cabin, then model adjustments could be made on a more complete set of data available from the LES results. GENERAL AIRFL

20、OW PATTERN IN A 8767-300 PASSENGER CABIN To obtain the general airflow pattern in a B767-300 passenger cabin, a three-dimensional CFD model of a B767- 300 cabin section (38.7 in. long) was built for this transient RANS simulation, as shown in Figure 1. In order to preserve the geometric fidelity of

21、the model and to keep the modeling time reasonable, hexahedral elements (424,704 cells) are used in the nozzle section and tetrahedral elements (2,229,013 cells) in the cabin section. The ground conditions (Pstaiic = 14.7 psia, Tinlet = 5 1F) are specified for the simulation. The total air inflow is

22、 94.9 ch. Velocity inlet boundary conditions are imposed at the four nozzle inlets and static boundary conditions at the six return air grills. An assumption was made that axial flow is negligible and a symmetrical boundary condition was set at the FWD and AFT faces of the model. Seven numerical pro

23、bes are placed at the locations indi- cated, as detailed by Lin et al. (200 i), to monitor air movement across the center plane in the cabin section. To study the unsteadiness of airflow movement in the cabin, a steady-state flow field was needed as the initial condi- tion for the subsequent transie

24、nt simulation. The RANS equa- tions are solved using a commercial flow solver. Simultaneously, the turbulence-caused closure problem is addressed by solving the equations of the renormalization group k-e (RNG k-e) model. Note that second-order schemes in space are necessary to obtain better accuracy

25、 for the steady- state solution in this study. Figure 2 illustrates the flow field of the steady-state solution and shows that there are two large counter-rotating recirculation zones in the cabin, located around the passen- ger head height level at each aisle way. As shown in Figure 2, the regions

26、where the air velocity exceeds 100 ft/min are outside the scale for the plot and, therefore, are white in color. Note that the overall flow pattern is not symmetrical with respect to the cabin cross section even though the geom- etry and the boundary conditions are symmetrical. This asymmetrical flo

27、w pattern is the result of inherent unsteadi- ness that characterizes this type of flow regime and geome- try. As shown in Figure 3, the airflow at the nozzle section also supports the observation mentioned above. Before pass- Figure I The B767-300 cabin and nozzle model. Pat Figure 2 Steady-state f

28、low field (velocity magnitude in $/ min). 756 ASHRAE Transactions: Symposia 990 0.W gsP 300 -+ Figure 3 Nozzle section airjlow (velocity magnitude in ft/ min). ing through the 0.05-in. gap, the airflow in the nozzle section is symmetrical, as shown in Figure 3. However, the airflow is separated from

29、 different walls after the gap in each of the nozzle sections. Assuming the turbulence is isotropic in this study, the fluctuating velocity component can be calculated as v = E, where k is the turbulent kinetic energy. The turbulence intensity is obtained by dividing the velocity fluc- tuation by it

30、s mean velocity magnitude. In addition, the turbulence length scale, 1 = (v ) /E, where E is the turbu- lence dissipation rate, is calculated. Substantial turbulence intensity (50% to 150%) is observed around the region of the two large recirculation zones, which is consistent with the measurements

31、reported by Jones (2000). The length scale of the large eddies is about 0.5 - 1.0 ft. The transient simulation is performed with a constant time step, At = 0.001 seconds. As previously mentioned, the lateral air movement across the symmetric plane occurs when the y-component of the air velocity at t

32、he numerical probes changes its sign (Le., from negative to positive or vice versa). Due to the small time step and the size of the grid, we were only able to complete a time period of about 15 seconds. The swing motion across the symmetric plane is observed at one of the numerical probes at t = 13.

33、88 seconds, which is consistent with the estimation of the turbulence time scale previously mentioned. At t = 14.33 seconds, as shown in Figure 4, the airflow pattern is developing into a more symmetrical pattern in the cabin, especially at the interfaces where the nozzle sections meet the cabin. .3

34、 THE LES RESULTS OF THE SIMPLIFIED CABIN MODEL With RANS, the movement of large eddies in the domain of concern (which is essential for the transport of airborne pathogens) would not be resolved, as reported by Bjorn and 100 80 80 70 30 20 10 O Figure 4 TransientJowpeld at t = 14.33 seconds (velocit

35、y magnitude in ftimin). Nielsen (1998). To capture the unsteadiness of the airflow in aircraft cabins, other CFD techniques such as DNS or LES are required. As mentioned earlier, LES is not practical, given the avail- able computing resources, for generating flow predictions of a model as large and

36、complex as a B767-300 passenger cabin. For the B767-300 model, with an inlet Reynolds number (Rei) of 31417, an inhibitory mesh size on the order of (Rei)94, i.e., -1.3 x 10” cells, is needed to do a DNS (Mathieu and Scott, 2000). The mesh requirements for an LES are approximately one order of magni

37、tude less (about lo9 cells) than for the DNS (Fluent 1998). Because the smaller mesh size requirement for the LES is still far beyond the capability of available comput- ing resources, a simplified cabin model was built to study the airflow in the cabin using this method. As shown in Figure 5, the s

38、implified cabin has a single slot inlet representing the interface between a nozzle and a cabin. The incoming air is set at a velocity magnitude of 2 fts with a slot width of 2.1 in. for a 2x slot Reynolds number of about 2500. This is typical for most airplanes supply nozzles. By preserving this tr

39、ansitional Reynolds number, the expectation was that the flow instability present in real cabins would also be present in the simplified model. Another important feature is the cabin scale. The seven foot dimension is representative of half of an airplane cabin cross section. The overhead storage bi

40、ns, which are normally non-rect- angular in an airplane cabin, are modeled as rectangles for two reasons. First, they provide a consistent location for inlet jet separation and, second, a rectangular geometry is compatible with the most accurate meshing scheme. Non-rectangular bins can add uncontrol

41、led instability as the jet separation point moves about. Non-orthogonal surfaces also require unneces- sary numerical approximations and wall function changes that may affect stability. Note that an orthogonal geometry is also compatible with more flow solvers, especially the ones using Cartesian me

42、shes. ASHRAE Transactions: Symposia 757 l d Figure 5 The simplified cabin model with Jive numerical probes on the centerplane. Figure 6 Temporal variation of monitored (V/ using a commercial code: three-dimensional LES, t = O to I I3 seconds (velocity magnitude inft/min). Two three-dimensional simpl

43、ified cabin models were built to capture the large eddy motion, which is essential to predicting disease transmission via airborne routes. The velocity fluctuations observed from the original three-dimen- sional simplified cabin model (735,000 hexahedral cells) range beween 20% and 30% of the mean a

44、ir velocity after adding the contribution from the minor sub-grid motion. It was decided that the original model needed to be refined. Based on the Reynolds number at the inlet, Re = 2225, the Kolmogorov length scale, q, is 0.003 ft. Simulating the airflow in this case using a DNS requires Re9?4 = 3

45、4 million cells. To do LES for this case, the refined mesh has the grid spacing of 5 - 20 in all directions and consists of 2.55 million hexahedral cells. A time step, At = 0.05 seconds, was selected based on the Kolmogorov time scale of the inlet z = 0.059 seconds. As shown in Figure 5, the simplif

46、ied model has all of the key dimensions of a 7-ft long aircraft cabin (corresponding to the length of two seat rows with half of the cross section). Five numerical probes, denoted as ?lower-left,? ?upper-left,? ?middle,? ?upper-right,? and ?lower-right? in clockwise order, are placed on the center p

47、lane to monitor the flow motion, as shown in Figure 5. At those points, the instantaneous velocity magnitude of the large eddy motion, Ifl = 14 + IV?LEs, is recorded at every time step. Note that the third component of the instantaneous velocity magnitude due to the sub-grid scale motion (SGS), IV?I

48、SGS, is not included due to its negligible contribution found in this study. The time history of the instantaneous velociy magnitude at these locations was recorded and is shown in Figure 6. These signals were verified by a parallel LES simulation using the (kaeiSePinSea(edAiea m, 1 Figure 7 Typical

49、 measured air velocity magnitudes in an airplane cabin seated area using hot-wire anemometer (velocity magnitude in jlimin). NIST-developed fire dynamics simulator (FDS), as detailed by Lin et al. (2001). Furthermore, at the five-point temporal, the predicted velocity fluctuations, from both the three-dimen- sional LES and the FDS LES, have shown the same turbulence level compared with the one-point hot-wire data, as shown in Figure 7. The three-dimensional LES results at t= 1 13 seconds are shown in Figure 8, and more temporal instances of the predicted flow field are available