ASHRAE 4709-2004 CFD-Based Parametric Study of Ventilation and Diesel Exhaust in Locomotive Facilities《机车设施RP-1191的通风及柴油机排气 基于CFD参数研究》.pdf

《ASHRAE 4709-2004 CFD-Based Parametric Study of Ventilation and Diesel Exhaust in Locomotive Facilities《机车设施RP-1191的通风及柴油机排气 基于CFD参数研究》.pdf》由会员分享，可在线阅读，更多相关《ASHRAE 4709-2004 CFD-Based Parametric Study of Ventilation and Diesel Exhaust in Locomotive Facilities《机车设施RP-1191的通风及柴油机排气 基于CFD参数研究》.pdf（11页珍藏版）》请在麦多课文档分享上搜索。

1、4709 (RP-1191) CFD-Based Parametric Study of Ventilation Locomotive Faci I ities and Diesel Exhaust in Liangcai Tan, Ph.D. Member ASHRAE Amy Musser, Ph.D., P.E. Member ASHRAE ABSTRACT A CFD-based parametric study was carried out for a prototype general exhaust system widely used for ventilation and

2、control of diesel exhaust in large enclosed locomotive facilities. The parameters of fan flow rate, ceiling height, fan spacing, and locomotive position relative to the fan were varied between two values representing the low and high ends of their expected ranges. A set of simulations was set up usi

3、ng factorial experimental theov, which allows variables and interactions having a significant efect on an outcome to be identlfed. Two outcome variables were considered: the maxi- mum time-averaged concentrations at the breathingplane for occupants standing on thefloor and on aplatform. Regression w

4、as then used to obtain an equation to predict these maximum concentrations as a function of the independent parameters. Using these equations, a design procedure that could be used by designers to maintain a target concentration limitfor a crit- ical contaminant has been developed and demonstrated.

5、INTRODUCTION Currently available design guidance for ventilation of enclosed locomotive facilities has been developed primarily based on practical experience. However, little formal research exists to validate or invalidate the effectiveness of these venti- lation quantities or to provide a framewor

6、k for adjusting them as indoor sources of contaminants change or published design indoor concentration limits for levels of airborne contami- nants change. Computational fluid dynamics (CFD) modeling provides a fast and economical way to obtain the type of infor- mation needed to conduct a more deta

7、iled parametric study of the effects of indoor contaminant sources and design concen- tration targets on system requirements. The computational approach, verified by comparison with field data collected in operating facilities (Tan and Musser 2003), uses a public domain large eddy simulation program

8、 to solve for the indoor airflow and contaminant trans- port. A parametric study is then performed for a prototype general exhaust system, varying the exhaust fan flow rate and important geometric dimensions. To permit flexibility, contaminant concentrations are defined so that they can be scaled to

9、 the actual generation rate. The outcome of the para- metric study is measured in terms of maximum steady-state contaminant concentration at breathing planes for a person standing on the floor or on a platform. Parameters and their combinations that have the largest influence on this outcome are ide

10、ntified, and regression analysis is used to develop simplified predictive algebraic equations based on design parameters. The end result is a set of handbook-appropriate equations that can be used in system design. MODEL DEVELOPMENT Software The NIST (National Institute of Standards and Technol- ogy

11、) Fire Dynamics Simulator (FDS) public domain large eddy simulation software was used for this study (McGrattan et al. 2002). It has been available to the public at no cost for several years. It was created for modeling building fires but has also been validated for a variety of indoor air quality m

12、odeling scenarios (Musser et al. 2001). This code uses large eddy simulation (LES) to model turbulence. Rather than deal with turbulence through time averaging, LES allows the direct calculation of large-scale turbulent motion. The LES method then addresses smaller-scale effects using the Smagorinsk

13、y Liangcai Tan is a lecturer and Amy Musser is assistant professor in the Architectural Engineering Program, University of Nebraska-Lincoln, Omaha, Neb. 02004 ASHRAE. 167 1 c ttttt : I ! i Q I Figure I Section view of locomotive and general exhaust system. Sub-Grid Scale model (Smagorinsky 1963), wh

14、ich requires only one empirical coeficient. This allows direct, time-depen- dent calculation of large-scale motion, which is usually most important for fire-related flows. The addition of a subgrid scale model then allows the diffusion due to motion that is smaller than the computational grid to be

15、approximated using an empirical relationship, resulting in less computational time than would be needed to directly calculate all aspects of the flow. The solution of the LES simulations is inherently both three-dimensional and time-dependent. To obtain a steady- state result, the simulation output

16、must be time averaged over a sufficient period to cancel the effects of turbulent fluctua- tions. This is ideal for solving transient problems but can increase the time required for steady-state problems such as those involved in this study. However, this program was selected because of the licensin

17、g flexibility it offers. Verification with Field Data Field data collected from three facilities with general exhaust systems were used for verification of the modeling approach (Tan and Musser 2003). Predictions of the LES model and a commercial software package using a k-epsilon turbulence model w

18、ere compared with field measurements of temperature, carbon monoxide, and oxides of nitrogen. Veloc- ity profiles predicted by the two models were also compared. The facilities chosen for the field study included parameters in the range studied here. The results of that effort showed reasonable agre

19、ement between the two CFD models and the field data collected. I tttttt Q tttttt Q If II Figure 2 Elevation view of locomotive and general exhaust system. CFD TEST PLAN Goals of CFD Analysis The objectives of the CFD portion of this research were to define a prototypical general exhaust system, iden

20、ti9 param- eters relevant to its performance, and develop a first order correlation to predict system performance as a function of these parameters. A set of CFD simulations making up a formal parametric experiment was performed. Analysis of these simulations then produced first order correlations t

21、o describe the effects of relevant parameters on the performance of the general exhaust system. Schematic General Exhaust System The general exhaust system, shown in Figures 1 and 2, has exhaust inlets located at the ceiling. A section of an enclosure containing a single track was modeled with a sol

22、id, adiabatic floor and ceiling. Symmetry boundaries are used between tracks so the simulation results will be valid for multi-track facilities that may have more than one locomotive operating at once. Consistent with the facilities that were field measured, most of the makeup air comes from open do

23、ors on either end of the enclosure. Therefore, the end walls are modeled with large open doors that are represented as passive openings for makeup air to enter. Track to track spacing was set to a fixed value of 25 ft (7.6 m) based on the sites that were visited for the field measurements. The lengt

24、h ofthe facility modeled was 180 ft (54.9 m). Ambient temperature was set to 90F for all simulations. The locomotive exhaust stack temperature and flow rate, as well as other temperature and flow boundary conditions at the locomotive, were specified as fixed values typical of the locomotives encount

25、ered during the field measurements. For the general exhaust system case, the ceiling height, exhaust inlet spacing, fan flow rate, and the locomotive offset position from the inlets were varied. A single, scalable value 168 ASHRAE Transactions: Research Quantity Ceiling height, ft (m) Fan spacing, f

26、i (m) I X 1 20and60 1 (6.1 and 18.3) Symbol Values Z 20 and 45 (6.1 and 13.71 Fan flow (total in ACH) width, ft (m) Emissions (conc. at exhaust) Locomotive exhaust tempera- ture Exhaust duct inlet length and Q 5 and 12 ach L 5 (1.5) C I mg/m3 (scaleable) T 350F (1 77C) correlations can then be devel

27、oped based on the test results. The validiiy of these linear correlations can be checked by generating “center points”, in which all parameters are set to the mean of the two values initially used. It is important to choose typical high and low expected values of each parameter in a 2k test plan bec

28、ause the effect of each parameter is linearized. This prevents a parameter that only impacts the outcome for unrealistically high or low values from erroneously appearing to be significant over the entire range investigated. On the other hand, it is necessary to choose a wide enough interval that th

29、e analysis will identify all of the parameters that are significant for most realistic cases. Tables 1 and 2 show the high and low values chosen for each of the parameters investigated, as well as the fixed values used for parameters that were not varied. These values were selected to represent a ty

30、pical range for each parameter based on observations from the field measurements and discussion with professionals in the field. As Table 1 shows, four independent parameters are vaned for simulation of the general exhaust system: ceiling height, fan spacing, fan flow rate, and locomotive offset pos

31、ition. To reflect current design guidance, fan flow rate has been expressed in air changes per hour (the fan flow rate i facility volume). Locomotive offset position is modeled at two loca- tions, under and between fans. So that it can be expressed as a continuous variable, the offset position (P) i

32、s defined as Locomotive exhaust flow rate Track to track spacing, ft (m) Ambient temperature Locomotive offset position (1) p=- 2d Factorial Experiments X F 2,000 ch (944 L/s) Y 25 (7.6) Ta 90 “F (32C) P 0-1 (Under and between fans) Factorial design theory (Montgomery 2001) is used to design sets of

33、 experiments to assess the effects of parameters and their interactions on a result using a minimum number of experiments. This was an important consideration for this research because of the time and computational resources required for each CFD simulation. When actual physical experiments are perf

34、ormed, it is common to replicate each experiment in the factorial set a number of times to assess the uncertainty associated with the measurement techniques used. The primary difference in applying factorial design theory to a computational analysis is that random experimental errors are not an issu

35、e, and replica- tion is therefore unnecessary. The general factorial design theory discussed here could be applied to the design of either a set of experiments or a set of computational simulations. When the independent variables in a factorial experiment are continuous, each can be tested at any nu

36、mber of values. When the goal is to identi3 important parameters and inter- actions, it is typical to select only two values of each. This is called a 2k factorial experiment, since the total number of combinations that can be generated by two values of each of k parameters, and, hence, the number o

37、f experiments required, is 2k. The analysis of a factorial experiment identifies individ- ual parameters and interactions that have a significant effect on the dependent parameter (also referred to as the response variable) over the range of values considered. First order where d = locomotive exhaus

38、t stack distance from the nearest exhaust fan, x = fanspacing. This definition allows the offset position (P) to assume a nondimensional value of 1 when the locomotive stack is located between fans and a value of O when the stack is located directly under a fan. The strength of factorial experimenta

39、l design is that it allows, not only individual variables, but their interactions to be analyzed with a minimum number of simulations. For example, one might expect exhaust fan flow rate and ceiling height to individually influence indoor contaminant concen- trations. However, these two parameters m

40、ay also interact, meaning that a change in both parameters may have an impact that is greater than (or less than) would be expected by analyz- ing each of these quantities individually. A factorial design makes it possible to generate handbook-appropriate algebraic equations that predict a dependent

41、 variable based on the simu- lations. Sometimes it is possible to group together these interac- tions in the analysis in such a way that only half the total number of simulations needs to be performed. This is called a “one-half fraction” factorial experiment. This can sometimes be done when very fe

42、w of the parameters tested interact with one another to significantly affect the value of the response ASHRAE Transactions: Research 169 Table 2. Sequential Listing of CFD Simulations Performed - General Exhaust System Simulation number 1 Parameter Value: High (+) or Low (-) Ceiling height (Z) Fan s

43、pacing (X) Fan flow rate (Q) Locomotive offset (P) I I I 1 2 2 + + + + 4 5 6 1- I I + + + + + + 7 8 9 1 , 1 J + + + + + i- + 10 Il 12 13 14 15 16 variable, and it reduces the total number of simulations for each case to eight. Since computational results are not influ- enced by the order in which th

44、ey are conducted, we completed an initial set of eight simulations and compared the magnitude of combination effects to main effects. In this case, the inter- actions were found to be significant, so we completed the remaining eight simulations for each case to complete the full factorial experiment

45、, Table 2 shows a listing ofthe sixteen (z4) simulation combinations that were performed for the general exhaust system. The effects of the parameters and their combinations must be defined in terms of dependent variables (or “response vari- ables”) that quantify the outcome that is being produced.

46、Here, the desired outcome is the removal of exhaust contaminants by the ventilation system. Specifically, occupant exposure should be reduced. Two concentration measures were used for eval- uating the outcome of the simulations. These were the maxi- mum steady-state (time-averaged) concentrations at

47、 heights of 5.5 ft and 9.5 ft (1.7 m and 2.9 m) above the floor. These loca- tions represent the breathing zones for occupants standing on the floor and a four-foot platform, respectively. + + + + + + + + + + + + + + + PARAMETRIC STUDY RESULTS Analysis of Factorial Experiments A comprehensive treatm

48、ent of the analysis of factorial experiments can be found in Montgomery (2001). This section covers only on those aspects pertinent to the problem at hand, focusing on the regression process. A 24 factorial experiment with four independent param- eters, A, B, C and D, will have four primary effects

49、that corre- spond to each of the independent parameters. There will also be six two-factor interactions (AB, AC, AD, BC, BD, and CD), four three-factor interactions (ABC, ABD, ACD, BCD), and one four-factor interaction (ABCD). Effects are defined as the average change in the response variable (the dependent variable) as each factor is changed from its low to high value. For example, the main effect associated with the parameter A is estimated by 1 -(O) .f ad- bd+ ab -cd+ ac- hc + abc 8 -d+ a- h + abd- c + acd- bcd + I = -( A and the two-factor effect AB is estimated by where (O) represe