5 Replies Latest reply on Jun 14, 2013 4:53 PM by Jared Conway

    Verifying Accuracy of CFD Results

    Tom Quinlan

           For a school project, I am going to be running simulations comparing drag forces against a car model with varying external accesories. Before I move to complex geometries, I am trying to prove the validity of the CFD results with known values. Since there is no way to use hand calculation/known drag force coefficents on anything but the most basic of shapes, thats what I have started with.


           I have been using a smooth sphere, with varying diameters and flow velocities. I am calcuating reynolds number based on these parameters, and also the drag force coefficent based on the drag force experianced by the model. I have plotted the different Cd Vs. Re values on a graph, and have been comparig them to text book values for smooth spheres. I am attaching an excel with the data I have gathererd so far and a plot of accepted values. My results are quite different than what they should be and I am trying to figure out what parameters in my model/calculation error which could be casuing this error.


      Current Setup:


      V= 0.1 to 30 m/s

      D= .0625 to 1 m

      Boundary condition of a real wall with 0 roughness

      Air at 293K,  1.2041 kg/m3,  0% humidity,  0.1% turbulence

      Computational Domain Size: 5D downstream, 2D upstream, 1.5D Surrounding


      Surface Goal: normal force in direction of flow

        • Re: Verifying Accuracy of CFD Results
          Jared Conway

          Hi Tom, you're in luck, you problem is in the validation examples that are included in flow. Have you checked them out? They are in the technical reference.


          Also, have you taken a look at the solidworks KB? there are a bunch of really good articles that will help you set your expectation about the results for drag/lift and recommendations.


          Also, this topic was discussed recently. Check this forum post out:



          To get really accurate results, you will need a lot of cells and you may need a transient analysis depending on the reynolds number. Also the computational domain will need to be appropriately sized. All are discussed in the tutorial/validation problem in the software documentation.


          a couple suggestions:

          1. for your simulation and your physical/hand calcs, write down all the assumptions, know how the differences may affect your comparison.

          2. variance testing, check comp domain, mesh and setup parameter changes and how they affect the results

          3. outputs. you've chosen normal force, why not force? double check the definition in the help.

            • Re: Verifying Accuracy of CFD Results
              Tom Quinlan

              Thanks for the response Jared, very helpful.


              I do not have access to the knowledge base becuase my SW licence is an educational version, but I did read through the linked artical and there is deffinitly some helpful information in there. I was unable to find the validation examples, (i have SW 2012 SP4), I see them for FEA simulations but nothing for flow analysis.


              From what im reading I think my computational domain was ok, but my mesh was far too large. I just tired increasing the global mesh size quite a bit, im running about 1.2m fluid cells now. At the larger mesh size I had previously tried expanding my computational domain, but saw very little differences so I reduced it back to the original proportions. For now I have the convergence set to 0.1 N on the normal force calculation. I have tried both nomal/axial forces and have only seen very minor differences, and my professor advised using the normal force.


              My system has an i7@4.2ghz and 24gb of ram, so I feel like it should be powerful enough to avoid a parametric study, but I may be wrong. The study im currently running has gotten through 50 iterations and I am waiting on convergence.

                • Re: Verifying Accuracy of CFD Results
                  Jared Conway

                  the documentation is in the install directory > lang > english > docs


                  regarding comp domain and element size, take a look at the last post with teodor. to match his 2d, i had to have a HUGE comp domain and high mesh refinement. my comp was running on a computer iwth 32GB of ram, in 2D and it still took about 8hrs to start converging towards a solution.