11 Replies Latest reply on May 20, 2015 6:46 AM by Thomas Annet

    Weird Stuff in Flow Simulation

    Jeff Ly



      I ran a thermal analysis in SW Flow Simulation. It was an internal flow with solid conduction. My model included 2 TECs to pump heat out of the chamber to the ambient. I also used 2 heat sinks with fans (heat sink simulation features). Air was fluid. Everything started out at high initial temperature. Internal fluid was to be cooled down to a target temperature. Steady state analysis was Ok and the results made sense. But, when I attempted to run time-dependent analysis, the temperature exploded to really high value. I really have no clue to why this happens. I really appreciate anyone give me a few pointers to figure this out.


      Thanks so much.



        • Re: Weird Stuff in Flow Simulation
          Pratham Singh

          if u upload your project files, that might be useful

          • Re: Weird Stuff in Flow Simulation
            Thomas Annet

            Hello everyone,


            I am experiencing exactly the same problem, In a casing, multiple air entries at environment pressure and two fans in OUTLET stream (pumping the heat out of the casing). Thermal analysis on multiples batteries inside the casing. In steady state everything seems to work well even if it is not relevant regarding the real test (discharge of the batteries after 12minutes). When running the time dependant analysis, the temperature reaches 50000°C and the speed of the air flow is around 10000m/s.

            Trying exaclty the same test BUT with the fans in INLET  configuration (blowing fresh air to the casing), everything worked well and I reached coherent values for temperature and speed.


            Is there something I am missing ? Is it possible that the problem comes from the meshing ?





              • Re: Weird Stuff in Flow Simulation
                Jared Conway

                if it works with steady, it should work with transient. so either time step is too small, mesh is too coarse or there is a problem with the outlet fan.


                To rule out a software issue, I would setup a really simple analysis with the outlet flow rate and see what happens.

                  • Re: Weird Stuff in Flow Simulation
                    Thomas Annet



                    Thank you for the short time answer. Indeed it seems that the time step was too big. Since I wanted a large physical time (1760 seconds) I thought that a 0.1s time step could be enough. Now with a smaller time step it seems to have acceptable speed values

                    However the calculation time is really long since the model is big (0.5m*0.5*0.25 casing) and parts are small (1mm minimum thickness ). I am trying to use the flow freezing option in order to reduce calculation time. Does this time step looks good to you ?


                    From what I red in the link that Andrei Popov gave, I could not find out how to sort the right time step criteria regarding the physical time. It is related to the cell size and the speed but shouldn't the physical time be taken into account ?

                    Using the CFL number I find a time step of about 0.0001s, is this really a time step that could fit to my problem ? The calculation time would be enormous.


                    Thank you again for your advice,



                      • Re: Weird Stuff in Flow Simulation
                        Jared Conway

                        "Does this time step looks good to you ?"


                        if th results look good, the time step looks good


                        you won't know until you try it

                        • Re: Weird Stuff in Flow Simulation
                          Andrei Popov

                          Hi Thomas,


                          I am quoting from the link:

                          A rule of thumb is to set time step< deltax/u

                          where deltax is the smallest cell size and u is the velocity.


                          if your smallest cell is 1 mm and your expected flow speed in that area is, lets say 100 m/s then your time step should be dt =  1 mm / 100000 mm/s = 0.00001s, otherwise you will have spikes in the fluid velocities and numerical instability in the flow solving.


                          This is correct for explicit schemes but solver take cares of it. In case of implicit solver we can go for higher time steps (>100 times). This results in faster solution.


                          So I would go with a time step of 0.001s since Flow Simulation is using implicit schemes. But even so you have to take into account the minimum cell size, which is the most critical in narrow channels, where also there is the highest speed. Solidworks documentation recommends to have a minimum 5 cells per narrow channel. So it depends on your narrow channels dimensions. Of course if you want a very high accuracy then you can have as many as 10 or 20 but at the start of an analysis I would go with minimum recommendations, with the coarsest mesh acceptable.

                          Here is a good paper about meshing in Flow Simulation :


                            • Re: Weird Stuff in Flow Simulation
                              Thomas Annet

                              Hi Jared and Andrei,


                              The results looks good so the time step must be good :-)


                              As a first approach I used a 0.01s time step and the results seem to be close to what I expected (we ran some bench tests with the real batteries).

                              I need now to refine my model and to lauch a 0.001s time step over the weekend to compare results to the 0.01s time step.

                              I will look for the minimum cell size and the narrow channels, but the max speed is only about 10m/s and the narrowest channel is 5mm.


                              Thank you so much again guys, I learnt a lot of things on the subect thanks to you.


                              Best regards