Project Work A Generic Car Model - The Ahmed Body Computational Fluid Dynamics - TMMV08 Jonas Lantz Introduction You are to study a generic car model called the Ahmed model, which despite its simplicity is capable of replicating ow structures matching those generated around the C-pillar of a normal car. Your task is to perform a number of simulations and compare the results with experimental wind tunnel data. The geometry of the car is shown in Figure 1, with the ow coming from the left. Figure 1: The geometry of an Ahmed car model, with dimensions in mm. See the tutorial le blunt_body_geometry_creation.pdf for hints on how to create the CAD model. You can cut the computational domain in half due to symmetry, and make the total length 6 car model lengths (L). The inlet is positioned 1 L from the front of the model and the outlet is 4 L from the rear. The free-stream domain side is 1 L from the side of the car and the domain height is 1 L from the ground. For greater accuracy the domain should be extended (especially after the model), but to minimize computational time the domain is kept relatively small. If you want, you are of course free to investigate how the domain size inuences the results. Figure 2: The computational domain. 1
Experimental Data A number of wind-tunnel tests have been performed on the Ahmed body, and experimental data for ϕ = 25 and 35 are available at the course home page. The password to open the Excel le is TMMV082014cfd. The Reynolds number used in the experiments is 768000 (based on the height of the body) and the bulk inlet velocity is 40 m/s. The origin of the coordinate system is at x = 0 = end of car, y = 0 = symmetry plane, z = 0 = ground plane, see Figure 1. To simplify the comparison between experimental and numerical results, it is suggested that you choose the same coordinate system. An excellent example of how experimental and numerical data can be compared is shown in gure 3. Figure 3: Experimental data versus Large Eddy Simulation (LES). The gure shows the velocity proles in the wake behind the car. Taken from [5]. Task The examination of the project is a poster presentation and a written technical report. The poster is PASS/FAIL and the report gives 0-3-6-9 points. Create geometries for ϕ = 25 and 35 and set up the model based on the Reynolds number from the wind tunnel experiments. Before you compare your results with wind-tunnel data, make sure that your results are mesh independent and have reached the residual target (this is very important!). How you dene mesh independency should clearly be stated and shown in the report. There is a maximum limit of about 500 000 mesh nodes (about 1-1.5 million hex cells) due to licensing restrictions, but you will probably/hopefully not need that many cells. Use the SST-turbulence model and turn o heat transfer. Running the simulation in Local Parallel mode with 2 cores will decrease simulation time (see the manual). Poster Assignment - A Design Challenge Using the 35 -geometry, design your own spoiler and investigate how the drag and lift-coecients are aected. The shape of the spoiler is up to you to design, but it may not be longer than 0.1 car length (L), wider than 1 car width (W), and higher than 0.25 car height (H). You must place the spoiler inside an imaginary box with the dimensions 1.15L X 0.15L, 0W Y 0.5W (symmetry) and 0 Z 1.1H with the coordinate system dened in Figure 1. Place the spoiler at (at least) two dierent locations on the car. 2
What happens to the ow eld and why? How big is the dierence in drag coecient (c d ) and lift coecient (c l ) compared to a model without a spoiler (in raw number and percent)? Do you gain or lose driving performance? How do you dene driving performance? Can a small change in c d be accepted for a large change in c l and vice-versa? A way to quantify aerodynamic eciency for airplanes is the lift to drag ratio, can it be used in this context? Motivate the shape and location of your spoiler. Appendix A.1 explains how to calculate c d and c l in CFX. Technical Report - Dig Deeper! For 3 points (PASS), you need to investigate and discuss the following for both the ϕ = 25 and 35 models: Where do you need a ne mesh and where can the mesh be coarser? How many mesh elements did it take until the results were mesh independent? How do you dene mesh independency? What is your convergence criteria? Why? How did you investigate the accuracy? Compute the drag coecient c d and compare your values to values measured experimentally by Ahmed [1] (see gure 4). Do you get a good agreement? How well do your velocity proles agree with the ones measured in the wind tunnel? Investigate both the ow in the free stream and in the boundary layer for both the ϕ = 25 and 35 models. Here you need to create lines in CFX and export the results and postprocess them in e.g. Matlab or Excel. Figure 3 is a smart way of presenting lots of data in the same gure. You might need to normalize the velocities before plotting the results. Compare velocity proles in the free stream in one gure, and velocity proles in the boundary layer in another gure. Make sure the plots are large enough to show the ow details. The excel le with wind tunnel data contains both measurements extending far into the free stream and measurements that are close to the car surface. Describe the ow eld around the models, what are the main dierences between the ϕ = 25 and 35 models? Where and why? How does this aect the drag? What kind of vortex structures can be seen in the simulation? Are they the same as in the experiments? For a higher grade (6 or 9 points) you will need to do the above and dig deeper into a topic of your own choice: Study the cross-ow velocity eld around the car in the YZ-plane (2 velocity components). A number of experimental results in those planes are available and you are to compare them with your own CFD-results for the ϕ = 25 and 35 models. Where are the velocity elds similar and where are they dierent? Why? Discuss the shape of the ow eld. Compare your CFD results with experimental data and draw the back of the car in your gures (see Figure 6). Also, using the experimental data, compute the kinetic energy in the ow and compare with your own results. Quantify the similarities and dierences in a smart way. Normalization might be needed to increase visibility of the results. Relate your ndings to the results from other articles. Comment: No need to run any more simulations, but you will have to create a Matlab script (or similar) to plot both the CFD and experimental data in the same gure, see e.g. Figures 5 and 6. Focus in this assignment should be on understanding the ow and relating your results to other articles. During the ski season it is common to use a car roof box to store the skis in when driving to the ski resort. Design a roof box and add it to your ϕ = 35 model. The design criteria are that the minimum volume is 0.025 m 3 (+/- 2.5%), maximum outer sizes are 700x340x120 mm (length x width x height, remember symmetry!). The box is located 50 mm above the roof but you do not need to model the attachment to the car. You are free to modify the shape in 3
any way, as long as it fullls the design critera. Investigate the optimum location of the roof box: at the front, middle, or back of the roof. Do you gain or lose driving performance if the roof box is tilted a few degrees? Comment: You need to use your creativity when designing the roof box, but only a few more simulations are needed. Create a car with a 0 slant (giving the geometry the shape of a truck) and run a simulation to get the drag and lift coecient. Then, in the area 0L X 0.1L, 0W Y 0.5W (symmetry), 0H Z 1H behind the car create something that improves driving performance. Try to think beyond wings and airfoils (as you might have done in the poster assignment). Why does your add-on work? Comment: be creative! You must really motivate why your add-on works and why it should be incorporated onto a truck or bus. Create another car with the same size and place it at some distance behind the rst car. How does the rst car aect the second car? Is there a optimum distance between the cars? Use either the ϕ = 25 or 35 model. Investigate at least 5 dierent distances between the cars, preferably in the interval 0.1 to 2 x/l. You probably will have to extend the computational domain behind the cars. Comment: here you need to do fewer simulations, but instead analyze the results very carefully. There are several articles on this topic that might be of interest. Find the optimum ϕ-angle with respect to drag coecient, i.e. recreate Figure 4. Why is that angle the most optimal? What happens in the ow eld? Split the c d into pressure drag coecients c S, c B, c K as in the gure, and split the viscous drag coecient c R into c RS, c RB, and c RK. Investigate and comment at least 10 dierent angles and compare with Figure 4. Comment: a straightforward task, but it will take some time to do all the simulations. Also, the grading will be based on your discussion about how and why the ow eld changes with the angle. This is not as easy as it sounds. Note that only presenting your results will not automatically give a higher grade, you really need to discuss and explain why the results look like they do. 4
Examination Poster Presentation On the last lecture (12/3-2015 at 13.15 in T1) you are to present a scientic poster about your spoiler design and your key ndings. Email your poster before Wednesday 11/3-2015 12.00.00 to jonas.lantz@liu.se (preferably as soon as possible). The posters will be printed by us and brought to the lecture where they will be displayed on the wall. Be prepared to give a 2-3 minute talk about your poster (practice before!). The size of your poster should be A3, or 297x420 mm (width x height) and in.pdf format. Instructions on how to write a scientic poster can be found on: www.stanford.edu/group/blocklab/dos and donts of poster presentation.pdf http://lorien.ncl.ac.uk/ming/dept/tips/present/posters.htm Also, email (preferably in the same email as the poster) the values you obtained for c K *, c B *, c S *, c R * and c W at 25 and 35 (see Figure 4). We will compile the data and present all results (anonymously) at the poster presentation. Besides fame and glory, a prize is given to the students with the best poster. Grade for this part is PASS/FAIL and attendance is mandatory to pass this assignment. Report Write a full technical report including abstract, introduction, method, results, discussion, conclusion, and references. There is a word limit of maximum 2000 words, and if a report consisting of more words is submitted, only the rst 2000 words will be considered. If you are aiming for a higher grade (i.e. doing one of the 6-9 point assignments) the word limit is extended to 3000 words. The poster part of the project should not be included in the report. Verify your ndings with results from other articles and discuss your results carefully and thoroughly. Print your report and put it in the post box marked TMMV08 Reports in, in the C-corridor, entrance 15, no later than 21/3-2014 at 23.59. The reports will be handed out at the same location in a folder called TMMV08 Reports out. Use proper captions and legends in your gures, and if you use color in your report, remember to also print in color. Do not forget to include proper references. Before writing your report, please read the Rules Regarding Examination, found at http://www.student.liu.se/tenta/regler?l=en. Note that claiming that you did not know the rules is no excuse. To prevent plagiarism we are using the web-based system Urkund, which compares your report with sources on the Internet, other articles and your fellow students reports. Therefore, you must also email your report to jonla63.liu@analys.urkund.se (send the report as a.pdf,.doc, or.docx le), as well as handing in a printed version in the postbox marked TMMV08 reports in in the C-corridor, entrance 15. If you do not email a copy of your report it will not be graded. If you fail the assignment you will have one week to resubmit a new version and then only three points can be given. Grade for this assignment is either 3, 6, or 9 points which will then be added together with the points from your other assignments to a total score. Good luck! 5
Hints Create two Fluid Flow (CFX) boxes on the Project Schematic, one for 25 and one for 35 to be sure that you use the correct geometry when simulating. Name the dierent surfaces on the car, e.g car_front, car_side, etc. Set the Max. Iterations to a very high value to be sure that your solution always reaches the residual target. If the solution oscillates around a higher level, stop the solver and make appropriate changes. Which surfaces on the car give drag and lift force contributions? Dening a monitor of the drag and lift forces could help you judge convergence. Dene the expression in CFX-pre, and then goto Solver/Output Control/Monitor/. When comparing velocity proles, create lines and export the velocity and do the plotting in Matlab or Excel together with experimental data. To visualize cross-ow in a plane in CFD-post, choose Projection: Tangential in the vector options. When comparing two color gures you must have the same color scale in both gures. You can change the scale under Colour/Range:User Specified. When creating lines, increase the number of samples to 100 (or more) or use Line Type:Cut to get a better resolution. To create a plane with specied width and height, use an Iso Clip on that surface, with e.g Visibility Parameters Y <= 0.2, Z >= 0.3. What is the dierence between conservative and hybrid values when post-processing? Which should you use? There are numerous articles to aid you when interpreting the results, see the references for a sample of articles. Google Scholar will help you nd more. Literature 1. Ahmed, S.R., Ramm G., Some Salient Features of the Time-Averaged Ground Vehicle Wake. SAE Technical Paper 840300, 1984. 2. H. Lienhart, C. Stoots, S. Becker, Flow and Turbulence Structures in the Wake of a Simplied Car Model (Ahmed Model). 3. Spohn, A., Gillieron, P., Flow Separations Generated by a Simplied Geometry of an Automotive Vehicle. 4. Gillieron, P., Modelling of Stationary Three-Dimensional Separated Air Flows around an Ahmed Reference Model. 5. Hinterberger, C., Garcia-Villalba, M., and Rodi, W., Large eddy simulation of ow around the Ahmed body. 6. Howard, R. J. A., Pourquie, M., Large eddy simulation of an Ahmed reference model. Journal of Turbulence, Volume 3, Art. No. N12, 7. Gerardo Franck, et al. Numerical simulation of the ow around the Ahmed vehicle model. This is only a small number of the available literature. Use your favorite search engine for more articles regarding the Ahmed body. 6
A Appendix A.1 How to calculate c d in CFX The drag coecient c d is dened as: c d = F d 1 2 ρv2 A where F d is the drag force, ρ is the uid density, v 2 the velocity squared, and A a reference area. In this case the reference area will always be the projected area in the X-direction of the front of the car. In CFD-Post, goto Expressions and create a new expression similar to: force_x()@car_front/(0.5*(15[m/s])^2*1.185[kg/m^3]*abs(area_x()@car_front)) where force_x()@car_front gives you the force in X-direction on a surface called car_front, abs(area_x()@car_front)) is the projected area in the X-direction of the surface car_front, and 0.5*(15[m/s])^2*1.185[kg/m^3] is velocity and density of the uid (dynamic pressure). In this way you can calculate the drag coecient at dierent surfaces, just by changing the location after ()@. If you want, you can split the force into the two parts it consists of; an area integral of the pressure (i.e. pressure force) and an area integral of the wall shear (i.e. viscous/friction force). This means that force_x()@car_front is equivalent to areaint_x(pressure)@car_front + areaint(wall Shear X)@car_front which could be useful for some of the higher grade assignments. Note the dierent expressions for the area integrals. (1) A.2 How to calculate c l in CFX See previous section, but change force_x to force_z. The lift coecient is dened with the same reference area as the drag coecient, but the lift force may act on dierent surfaces. 7
Figure 4: Drag coecient versus ϕ angles, taken from Ahmed [1]. c K * is forebody drag coecient, c B * vertical surface coecient, c S * is slant surface drag coecient, c R * is friction drag coecient, and c W is the total drag coecient. Example: for ϕ = 20, c S * 0.09, c B * 0.18 0.09 = 0.09, c K * 0.20 0.18 = 0.02, and c R * 0.25 0.20 = 0.05 8
Figure 5: Cross ow velocity distribution at three downstream locations in the wake for ϕ = 25. 9
Figure 6: Cross ow velocity distribution for CFD and windtunnel data for ϕ = 25 and 35. Compare with Figure 5. 10