35 Replies Latest reply on Feb 6, 2015 4:07 PM by Steve Grossman

    Help with a race car CFD analysis

    Mark Cifuentes

      Hello,

       

      Recently I came across an accurate modle of a Porsche 944 and I desided to do a CFD analysis in order to calculate the Cd and Cl. Now I ran the simulation and got a 45% error on ther Cd (0.33 Claimed, 0.5 Measured). I relize that it wont be perfect because the model isn't 100% accurate, however, I would happy with at least a 10% error.

       

      During the simulation i set the initial air velocity to 160 km/h however, when the simulation finished the air velocity was 136 km/h. I dont understant how this happens because I set the velocity as a constant.

       

      From what I have read so far the mesh is KEY to have accurate results. So basically I need help setting up my mesh.

       

      Mesh Info:

      RefinementLevel
      Curvature Refinement Level4
      Tolerence Refinement Level3
      Refine Fluid Cells2
      Refine Partial Cells4
      Refine All Cells-
      Narrow Channels Refinement Level1
      Initial Mesh Level7

      If you need any more information just let me. I can also provide the actul CAD model and pictures if needed.

       

      I would greatly appreciate any information you guys can provide and thanks in advance.

       

      Thanks,

      Mark

        • Re: Help with a race car CFD analysis
          Jared Conway

          pretty complicated analysis

          consider your cad geometry's accuracy vs what was tested in the wind tunnel

          fluid properties...etc.

          did you have a moving floor...etc.

           

          what's your long term goal here? i might suggest starting with a simpler model like a wing shape that has a known answer and is simple enough to run and troubleshoot. once you have a good understanding of the software you can move to a more complicated problem.

           

          there are a bunch of really good articles about this type of analysis in the solidworks KB.

            • Re: Help with a race car CFD analysis
              Mark Cifuentes

              Thanks Jared for the responce. Basically my long term goal here is to develope an aero kit for the porsche 944 because I actually have one myself.

               

              Is there a way to make the wheels rotate at a given speed (X rads/s)? And as far a moving floor how do I do that? I have a pretty good idea as to how flow simulation creates the cells, and and good understanding of my way around the program itself.

               

              The geometry of the model is very close to what was actually tested in the wind tunnel, so the majority of the error comes from the meshing. In your experience what gives a more accutate result more fluid cells or more particle cells?

                • Re: Help with a race car CFD analysis
                  Jared Conway

                  check out boundary condition > all

                   

                  always better to have fluid or solid cells but with flow's technology, partials are nearly as accurate

                   

                  your limitation here is going to be computation time. i'd shoot for trends between designs than absolute accuracy. this isn't just flow simulation, this is all CFD. you'll see what i mean by trying a few smaller examples. you may also want to look at what you can do from a part to part perspective for initial design and then scale up to the full model once you've got some data.

                    • Re: Help with a race car CFD analysis
                      Mark Cifuentes

                      Hey jared sorry its been a while ive just been swamped with school.

                       

                      Like you suggested i went back and checked the boundary conditions and im getting a new error which I can't figure out. "Some of the components that were disabled are lost". Currently my model is in a separate configuration where it is just the body and wheels, with everything else suppressed. Have you ever encountered anything like this?

                       

                      thanks

                      • Re: Help with a race car CFD analysis
                        Steve Grossman

                        Hi Jared,

                         

                        I've been running NIKA/EFDPro/FLOWSIM for over ten years with excellent results, including for very demanding aero-optical distortion.

                         

                        But I'm just about to try the moving ground plane case for a landing airplane for the first time. The moving real wall BC setup is asking for a "face" selection. Does this mean I have to create a solid to represent the ground? (I'm sure others would also like to know.)

                         

                        Thanks,

                        Flyboy2160

                          • Re: Help with a race car CFD analysis
                            Steve Grossman

                            Well, I built a ground slab. But when I try to set its upper surface as a moving real wall, I keep getting a message that the model doesn't lie in the solid fluid interface - even if I fully enclose the slab with the domain. I also get an error message if I try to set the domain right on the upper moving surface of the slab. What's the trick? Thanks.

                              • Re: Help with a race car CFD analysis
                                Jared Conway

                                did you rebuild the flow study? if it wasn't rebuilt it won't be included.

                                but you should just have to sink the comp domain into the block and apply the real wall

                                  • Re: Help with a race car CFD analysis
                                    Steve Grossman

                                    Thank you for he reply Jared, but I got it running just minutes before you posted! LOL Have you ever seen that great, hysterical pseudo-scientific looking paper on Why You Should Procrastinate? If you want a good laugh, I'll send it along. I'll post a picture of my results here to encourage others doing cars with FLOWSIM. (I have FLOWSIM and SOLIDWORKS and ALGOR on the plane in foot high letters until I can get Sam Adams as a sponsr.....) Steve

                                      • Re: Help with a race car CFD analysis
                                        Steve Calkins

                                        Hey Steve, any chance of you posting the pic of your results as you mentioned?  I'm trying to do a similar analysis.  Not a car or plane, but a part on a moving conveyor belt.

                                          • Re: Help with a race car CFD analysis
                                            Jared Conway

                                            flow on a conveyor belt. that could be intersting. have you already checked out the tutorial on flow over a sphere in the tech documentation?

                                            • Re: Help with a race car CFD analysis
                                              Steve Grossman

                                              Steve, I didn't get an email notice that you wrote! I'll post my results in a few minutes.

                                              • Re: Help with a race car CFD analysis
                                                Steve Grossman

                                                Steve, I forgot about this and just noticed your request  - I didn't get an email.

                                                 

                                                This works. A summary of my experience:

                                                 

                                                -The biggest drawback is that this takes 4-10X as many fluid cells as a regular run! I don't know why. I've tried a couple of schemes to preclude this, to no avail. The pictures below show my latest scheme. The ground solid slab is exactly the same perimeter as the domain, with the upper surface of the slab at exactly the bottom of the domain. The aircraft wheels are slightly above the bottom of the domain. The domain is smaller than I normally use for such an aero analysis because the number of cells was killing me.

                                                 

                                                (Why I don't run with symmetry? A few months ago I discovered an anomaly with symmetry that has been confirmed by SW. Moderators, if you want to break this symmetry anomaly off into a separate thread, I'll give details there.)

                                                 

                                                -As shown below, you set this up by inserting a Boundary Condition, picking the surface, and making it a Real Wall with Moving as an option. This then allows you to choose the vector direction for the surface.

                                                 

                                                - You can run with the wheels below the slab top, but then you must have the domain bottom also below the slab top and below the bottom of the wheels.

                                                 

                                                -With the domain ending at the upper surface of the slab, the slab thickness doesn't matter.

                                                 

                                                -Because of the large number of cells, I set this up starting with the coarsest mesh possible and refine it with adpative meshing.

                                                 

                                                (The purpose of this run is to determine the aerodynamic torque of the aircraft about the main landing gear tire rotation point. This torque must be greater than the countering torque from the aircraft CG trying to keep the nose down. This is a check on my hand calculations, which show that
                                                I have sufficient elevator authority to rotate the aircraft.)

                                                 

                                                 

                                                ScreenHunter_ Feb. 03 05.27.jpg

                                                 

                                                ScreenHunter_ Feb. 03 05.34.jpgScreenHunter_ Feb. 03 05.38.jpg

                                                  • Re: Help with a race car CFD analysis
                                                    Jared Conway

                                                    could you post the details of the symmetry issue? i haven't seen that in my work

                                                      • Re: Help with a race car CFD analysis
                                                        Steve Grossman

                                                        I only used symmetry at home on my plane - work was full size due to asymmetry. After running symmetry for years, I decided to see how a widely used wing lift hand calculation called Schrenk's Approximation compared to CFD. I started with my usual half wing symmetry and noted a velocity anomaly at the centerline which was not there if I ran full span. SW confirmed this behavior a few months ago with the latest software (I'm running 2012.)

                                                         

                                                        This anomaly resulted in only a small discrepancy in the lift that would have been dismissed by most analysts under the bromide "well, you  know CFD can't get closer than 20-30%. But I had been using NIKA/EFD/FLOWSIM for15 years and seen it validated in a couple of very demanding cases with results much closer than this, so I wasn't going to let this go without asking SW.

                                                         

                                                        What's even stranger is that my full span models solve much quicker than the symmetry models even though they have twice as many cells! The solutions are much better "damped." In symmetry mode, some would take forever to settle down.

                                                         

                                                        The SW tech guy who looked into this said ~ "That's why we have the caveat in the manual about using symmetry only when you're sure it's OK."

                                                         

                                                        schrenk half span.jpg

                                                         

                                                        full span schrenk.jpg

                                                      • Re: Help with a race car CFD analysis
                                                        Jared Conway

                                                        nice analysis by the way

                                                         

                                                        what does a cut plot of the mesh look like?

                                                         

                                                        what kind of accuracy do you think you got?

                                                          • Re: Help with a race car CFD analysis
                                                            Steve Grossman

                                                            Thank you Jared.

                                                             

                                                            Here's a cut plot with the adaptive mesh. I was wondering if the moving floor would somehow create a layer of small "boundary layer" cells near it, but I don't see any.

                                                             

                                                            I have no idea how close this is: This software "won" a CFD shootout at work because of it's exceptional results with separated flow (<1% off the wind tunnel results!) and with aero-optical distortion, but I have had differences with traditional hand calc checks on my plane at home. The half span longitudinal stability CFD runs were ~10% different than the hand calcs. This went down to ~5% when I began running full span - but this is an important number. I'm shooting for 5% stability. Do I have the 5% or is it 0%?

                                                             

                                                            The separated flow results at high AOA at landing are still way off the hand calcs, ~20-30% This is really strange to me since it did so well at work.................??????

                                                             

                                                            ScreenHunter_ Feb. 03 12.22.jpg

                                        • Re: Help with a race car CFD analysis
                                          Daniel Alaniz

                                          I'm sorry Jared, I am not really following what you are asking?

                                          • Re: Help with a race car CFD analysis
                                            Daniel Alaniz

                                            Could you post the CAD model of the Porsche and some pictures of your simulation?  I would like to run it myself to see if I could replicate your issues.