10 Replies Latest reply on Feb 14, 2014 6:33 PM by Jerry Steiger

    System Hardware RAM for Flow Simulation !!!

    Feras Khouj

      I was running a flow simulation for my snior project ( Designing of a hypersonic windtunnel) and one of the parts of the desgin is the pebble bed heater which is the most difficult part in the design.

      I tried to run flow simulation but I ran out of memory as the case is complicated and has too much details and parts I guess.

      the pebbles amount about 5000 pcs inside the heater.

       

      I'm willing to upgrade my memory to 16GB and I wish it would be suffeicient for this simulation but I have some concern about the timing.

      my current RAM is one slot of 8GB XMS Memory kit for dual channel systems, 1600MHz.

       

       

      My question is do is it better to get a new set of RAM 2X8GB with higher frequency or should I just get another 8GB slot to run this simulation.

      how much the software rely on the RAM during this simulation.

        • Re: System Hardware RAM for Flow Simulation !!!
          Shaun Densberger

          I'd just get another stick of what you are currently using.

           

          Out of curiosity, are you trying to simulate hypersonic flow with SolidWorks?

            • Re: System Hardware RAM for Flow Simulation !!!
              Feras Khouj

              Thanks for the answer but for your question it sounds stupid to say yes.

              but I'm new to these kind of simulation specially solidworks and still learning.

              is it possible to do this hypersonic simlation or not ? and what do you recommend?

               

              Thanks

                • Re: System Hardware RAM for Flow Simulation !!!
                  Shaun Densberger

                  To be honest, I'm not sure if SW "has" the ability to do hypersonic flow. I'm not a CFD guy, and I don't have anything above some basic experience with SW Flow. That being said, I'd be surprised if SW Flow could do hypersonic flow (and do it accurately). I think Fluent can do hypersonic flow, but it's probably worth it to do some Google digging.

                   

                  Is this project for a BS degree? If so, then I'd talk with your advsior(s) about using SW Flow for this and the potential shortcomings; they could very well not care as along as you document it and talk about the effects. If it's for a MS degree and you plane on publishing, then it'd probably be best to use a more acceptable code for hypersonic flow (whatever that may be).

              • Re: System Hardware RAM for Flow Simulation !!!
                Art Radomski

                It all depends on your model and the complexity.  I have run some complex models that have used up to 24 of 30 GB.  Without seeing your model I cannot make a recomendation as to how to refine it so it will run for you.

                Another place you can get help is the CFD forums.

                 

                Regards

                Art

                  • Re: System Hardware RAM for Flow Simulation !!!
                    Feras Khouj

                    Actually I have two cases here. one of them is the Conv-Div Nozzle. this one I want to simulate the hypersonic flow on it.

                    here is a screenshot of it

                     

                    Screenshot 2014-02-12 21.33.21.png

                     

                    The other one is the pebble bed heater. the flow is slow but the pressure and temperature is very high. P=200bar, T=1100K

                     

                    also here is screenshot for it

                    Screenshot 2014-02-12 21.34.49.png

                     

                    it is complicated and has so many parts inside it which could consume my whole memory and that's why I would like to upgrade

                      • Re: System Hardware RAM for Flow Simulation !!!
                        David Paulson

                        I think that you may be including much more detail into your analysis than is needed and is overdemanding on the memory requirement and will also adversely effect solver time.  You don't usually need nuts and bolts in a flow simulation.  The bigger issue is that you should model the particle bed as a porous medium.  If you have difficulty defining the paramters for the porous medium, you could run a secondary CFD analysis of the pebbles on a much smalller scale to determine that. 

                         

                        I never run a CFD analysis on an assembly that is finalized for manufacturing.  I always simplify the model as much as possible parsing all features and parts that do not effect flow.

                          • Re: System Hardware RAM for Flow Simulation !!!
                            Feras Khouj

                            You are right about the bolts and nuts but these bolts are few and doesn't affect too much on the simulation even if it is unneeded for the simulation.

                            For the porous medium I heard about it from an engineer that they use it in ANSYS but I don't know how to do it in solid works flow simulation. For sure it will reduce the solving time by hours but as I have said I'm new to these kind of simulations.

                            I'm forced to do it as it is one of the requirement in the senior project although the CFD course in the department is elective course.

                              • Re: System Hardware RAM for Flow Simulation !!!
                                Jared Conway

                                Flow simulation has porous medium as well and that problem is solvable if setup properly.

                                 

                                I do not that flow can handle hypersonic but before we go there, let's be sure we are talking the same thing. What effects are you wanting to consider? What are you expecting to happen? What Mach number?

                                 

                                On that problem it looks like you could quarter it. Which might limit your ram need.

                                 

                                For how much ram, look at the process, windows will use disk when ram isn't available, look at max ram needed for your problem. Your best perf will come when you have enough ram for your problem. And ideal a bit more because windows won't allocate one to one.

                                 

                                As for what kind of ram, go with whatever you can afford and that is compatible. Ram speed won't give you an order of magnitude change in solve time.

                                • Re: System Hardware RAM for Flow Simulation !!!
                                  Jerry Steiger

                                  Feras,

                                   

                                  I think David was speaking metaphorically when he said nuts and bolts. What he meant was too much detail. As you note, it is probably the individual balls that are causing the system to be too complex.

                                   

                                  I suspect it would be best to first approximate your pebble bed as a porous medium. You should probably start out with a quarter symmetry model or even a 2D (axisymmetric) model, even if your actual design is not exactly symmetric. Once you think you understand the behavior of the system you could try a symmetric model with the pebbles.

                                   

                                  Jerry S.