17 Replies Latest reply on Nov 15, 2016 2:31 PM by Michael Henderson

    Ground Effect Simulation Values Way off

    Michael Henderson

      Hi

       

      I'm currently attempting to validate a CFD set up with a NACA 0012 by comparing it with the data from the following paper

      http://eprints.soton.ac.uk/51083/1/51083.pdf

      Unfortunately my initial values that I'm getting at 0deg AoA and 0.3 h/c clearance are incomparable to the data from the paper. I'm relatively new to CFD so I'm not sure if my set up fully correct or my mesh may need further refinement, although it's approximately 900,000 currently. I've attached images of my following set up to summarise, I've used:

      -Surface Goals X & Y for Lift and Drag on aerofoil surface and end of wing

      -Comp Domain splits through aerofoil, therefore force has been multiplied by 2

      -Ground is real wall with velocity = air flow

      -Aerofoil surface is treat as real wall

       

      Values obtained for 0deg 0.3h/c, Cl = -0.03, Cd = 0.01

      Values trying to obtain Cl = -0.1, Cd = 0.0075

       

      Capture.PNG

       

      Any feedback will be appreciated

      Thanks!

        • Re: Ground Effect Simulation Values Way off
          Gian Flavio Violi

          Michael,

           

          Best advice I can give you right now is to do MAX local mesh refinement in everypart of the airfoil and floor. If you can upload the model so I can actually have a look at the set-up I may be able to give you some extra pointers as I've worked with CFD airfoils validations before.

           

          One thing to note though. SW FlowSim uses k-epsilon turbulence model, which isn't recommended for airfoils charaterization since you can simulate the effect of boundary layer detachment (flow separation, airfoil stall/stagnation) at all. But when the angle of attack is 0 degrees you shouldn't be facing major issues, FlowSim limitations are shown when you exceed the maximum angle of attack limit.

           

          Let me know how I can help you further.

           

          Regards,

          GF

            • Re: Ground Effect Simulation Values Way off
              Bill McEachern

              GF,

              The information above is miss-leading. Flow sim is better than most codes, where almost all codes are not are particularly good,  in predicting high angle of attack lift and drag. It is known to be better than most at getting separated regions at least in the ball park due to the in solution automatic remeshing. Where it is deficient is in the low angle of attack range due to the K-epsilon model as you note. The lift prediction will be pretty good as B'layers are transparent to pressure, the drag will typically be quite a bit higher. There is paper that was published, exact reference escapes me at the moment but I might be able to find it if it is of interest.

                • Re: Ground Effect Simulation Values Way off
                  Gian Flavio Violi

                  As someone who has made aerodynamic CFD for prototypes and validated with wind tunnel testing (both as academic and applied experiments), I respectfully disagree in some points you mention bill.

                   

                  First of all: yes, you are gonna know when flow separation is gonna occur. Thas awesome about FlowSim, a lot of codes can't even preddict that, but the fact that other codes suck worse doesn't make it suitable for this kind of excersice, specially when you are comparing to an academic research. Also once flow separations occurs, results are gonna be all over the place, this is a K-epsilon model limitation from the start. 

                   

                  From my experimentation, Flow was actually pretty good predicting CD & CL for a aerodynamic test we did with a Formula SAE & BAJA SAE prototypes back when I was in the university, but only when Flow separation didn't occur, mostly with low attack angles. Back in that experiment we also used ANSYS CFX with K-epsilon turbulence model (again, no detachment means great results, detachment is all over the place) & also Shear Stress Transport model (SST for short, great results with & without detachment). When I mean 'great results' I mean less than 5% error compared to wind tunnel testing. Sadly, I can't share that information with you guys due to a NDA, sorry.

                   

                  It is good to note that MOST of the testing conditions on the wind tunnel where made with very very high Reynolds numbers, which would have kept separated regions to a minimum. Thats probably why correlations between CFD & WTT were quite good.

                   

                  Conclusion: Yes, Flow is actually competitively good (even great for certain applications); but it has some limitations for this application in particular, the fact that Flow is better than most other out there doesn't make it OK. Specially when you wanna match the results to a WTT academic research.

                      • Re: Ground Effect Simulation Values Way off
                        Gian Flavio Violi

                        No link provided. I can only see a plain not hyperlinked "here is the paper" message.

                        • Re: Ground Effect Simulation Values Way off
                          Gian Flavio Violi

                          1-119NW28 wrote:

                           

                          did you read the paper?

                          "This study has also shown certain limitations of CFD codes that implement the RANS equations as they apply to this problem. The dynamics of the turbulent fluid flow around the flippers in this study were exceedingly complex and included such factors as low Reynolds-number effects and detached, coherent vortex interaction."

                           

                          "3) For simulations at high   in the stall region, characterized by flow detachment, SFS is the recommended code due to superior CL predictions and slightly better CD predictions (although CD predictions for both codes were problematic in the stall region, so neither code did particularly well)."

                           

                          I find this statement to be in-line with my clarification. Maybe my first statement lacked some contex to avoid been misleading, but it sure wasn't dead wrong.

                            • Re: Ground Effect Simulation Values Way off
                              Bill McEachern

                              You wrote initially:

                              " SW FlowSim uses k-epsilon turbulence model, which isn't recommended for airfoils charaterization since you can simulate the effect of boundary layer detachment (flow separation, airfoil stall/stagnation) at all. But when the angle of attack is 0 degrees you shouldn't be facing major issues, FlowSim limitations are shown when you exceed the maximum angle of attack limit."

                               

                              I don't think that is what you just said above though. In any event,there is enough here that the less knowledgeable can appreciate the level of complication present in using these tools if not wisely then at least with the appropriate level of caution.

                                • Re: Ground Effect Simulation Values Way off
                                  Gian Flavio Violi

                                  "I find this statement to be in-line with my clarification. Maybe my first statement lacked some contex to avoid been misleading, but it sure wasn't dead wrong."

                                   

                                  English might be my second language, but i think that is crystal clear clarification means second statement, not what I initially wrote. Also, I don't know if its something language related but what I actually said its consistent with the paper. It might be overstated (depends on your personal accuracy treshold) but isn't untrue.

                                   

                                  This particular statement "which isn't recommended for airfoils charaterization since you can simulate the effect of boundary layer detachment (flow separation, airfoil stall/stagnation) at all. But when the angle of attack is 0 degrees you shouldn't be facing major issues, FlowSim limitations are shown when you exceed the maximum angle of attack limit." is consistent with: "(although CD predictions for both codes were problematic in the stall region, so neither code did particularly well)" and the fact that SW wasn't bad at low angle of attacks, just not as good as good as Star CCM+. The language in which we (me and those who wrote the paper) present our conclusions certainly is different, they are more 'politacally correct' and I just have a lower treshold to say 'this is acceptable'. (honestly, this feels like the ZTG post's all over again).

                                   

                                  Also, looking at the documentation I have specifically for airfoil characterization (taken from the diploma I'm getting in simulation technology) k-epsilon model its still not recommended. Which its consistent with what I've read for about 6 years now and, from MY PERSONAL EXPERIENCE doing CFD + WTT, I honestly would recommend using Shear Stress Transport model if available.

                                   

                                  Regards,

                                  GF

                                   

                                  Note: For those who don't know, Star CCM+ is a widely accepted CFD package. World Champions Formula 1 cars have been designed with it (ie: Fernando Alonso's R25 & R26). Allthough, from what I've heard from a couple friends working in F1, most teams use ANSYS CFX with SST or Reynolds Stress. Those who use Star CCM+ use Reynolds Stress model.

                            • Re: Ground Effect Simulation Values Way off
                              Michael Henderson

                              Gian,

                               

                              Couldn't figure out how to upload onto this post so I've put the set up on my one drive, found here.

                              OneDrive

                               

                              Thanks

                          • Re: Ground Effect Simulation Values Way off
                            Michael Henderson

                            Hi Gian,

                            I'll try and get a copy of the set up, to you by Tuesday, when I get access to the computer. My approach at the moment has been to use the equidistant mesh refinement to have the mesh be refined at all surfaces of the aerofoil, I've also tried refining the global mesh to be more dense as it approaches the ground. I used a similar approach for the 2D simulation which provided me with somewhat similar lift coefficients, especially with the lower angle of attacks. However, the drag results didn't necessarily take into consideration any of the effects at the ends of the wings, hence the attempt to simulate in 3D.

                             

                            Thanks

                             

                            Michael