20 Replies Latest reply on Apr 2, 2014 3:23 PM by Jared Conway

    modeling steady state gas concentration for dense gases

    Dan Hofstetter

      I was wondering how to correctly model a tank full of gas that has a higher density than air:  I've got a below-ground tank full of hydrogen sulfide gas, and air up above.  If I set up a steady-state simulation, eventually the entire system reaches equilibrium with a mixture of H2S and air filling the entire computational domain.  I was expecting the heavier gas to settle to the bottom of the tank over time - can flow simulation model this?  I tried changing the temperature of the H2S gas to 55F (local initial condition) while keeping the rest of the computational domain at 68F, and it seemed to help a little bit at the beginning, but overall it still doesn't behave like I was expecting.  I've tried with and without pressure potential checked, and gravity is turned on.  Are the densities of the two gases too similar for this type of settling to occur?

        • Re: modeling steady state gas concentration for dense gases
          Bill McEachern

          do you have gravity on and the pressure potential flag set?

          • Re: modeling steady state gas concentration for dense gases
            Jared Conway

            what does this model and analysis setup look like

             

            you mention you expect it to settle over time but then you're doing a steady state analysis?

             

            are you using local initial conditions for defining one domain vs another?

             

            did you check the solidworks kb? i thuoght i saw something like this awhile ago.

             

            are these your own materials? have you tried fluids that come with the software?

              • Re: modeling steady state gas concentration for dense gases
                Dan Hofstetter

                I built a 100' x 40' tank that is 8' deep, placed underneath a building of the same size that has two open endwalls.  Gravity is turned on, pressure potential is on, and the default values for pressure and temperature are being used.

                 

                I decided to run this as steady-state because I'm not concerned with how long it takes to reach equilibrium, I just want to see the equilibrium state.  Is the steady-state solution the same as the infinite time solution?

                 

                I'm using a local initial condition for a fluid body in the tank.  Pressure potential is also turned on in the local initial condition inside the tank.  The tank has 100 ppm of hydrogen sulfide gas and 999,900 ppm of air.

                 

                 

                I searched the KB but didn't see anything related to dense or heavy gases settling in air.

                 

                The H2S gas was my own material - I just tried CO2 which has similar properties but comes stock in the engineering database, and got similar results with the CO2 mixing with air in the building above the tank.

                 

                I just watched some dry ice in a jar that was half-full of water - the CO2 doesn't stick around forever, so maybe this isn't reasonable behavior to expect from a simulation.  I'll try increasing the density of the H2S gas by 10x just to see if that has any effect.

                 

                Thanks,

                 

                Dan

                  • Re: modeling steady state gas concentration for dense gases
                    Bill McEachern

                    put in a goal to monitor the concentration of interst to see how it varies on its way to steady state - it will be similar to a transient study but time is not tracked specifically and see if it makes sense.

                    • Re: modeling steady state gas concentration for dense gases
                      Jared Conway

                      picture would be worth a thousand words here. i don't really know if i understand "two open end walls" or how the rest of your bcs are applied

                       

                      are you also expecting the fluids to unmix and settle in different places? (from the concentrations you have to start)

                        • Re: modeling steady state gas concentration for dense gases
                          Dan Hofstetter

                          I attached a copy of a simple model to the original post.  It has no barn, so no airflow above the tank.  The tank is smaller so that the simulation can be run much faster.  It's a rectangular tank with nine outlet slots.  I disabled the pit fan, and added a contaminant gas that is a copy of air, with 1,000x the molecular mass (named "1000x Air").  I ran this as a quick and dirty transient study to see how the gas would act in the simulation.  The inside airspace of the tank has a local initial condition with a concentration of "1000x Air" = 100 ppm.  The rest of the computational domain is pure air.

                           

                          Should the heavier gas stay inside the tank?  It diffuses up and out over time.  Am I doing something wrong?  Or is this just normal behavior for diffusion of gases?

                           

                          Thanks,

                           

                          Dan

                            • Re: modeling steady state gas concentration for dense gases
                              Jared Conway

                              runnning your studies now, but here's my initial thoughts:

                              1. you have manual time step, that is recipe for disaster sometimes

                              2. your mesh is pretty coarse, my understanding of diffusion in flow simulation is that it requires a pretty refined mesh

                              3. you have interferences in your model, for flow only this is sometimes ok, but generally don't recommend having them in your model

                              4. difusion is based both on density and Cp. Article directly from the KB. so your fluid should be different because you changed the density but the Cp would be the same.

                               

                              i'm not an expert in this by any means but what are your expected results? do you have something with a known solution to use as a baseline?

                               

                              i notice you're faking an external analysis, have you tried with an actual external analysis?

                              what about a steady state? does it look as expected?

                              what happens if you use a tracer study instead of this method or an initial pocket of the gas vs it being spread out? does it behave differently?

                               

                              Solution Id:

                              S-020225

                              Category:

                              FAQ

                              Question:

                              Can COSMOSFloWorks analyze the mixture of several gases (for example diffusion of noxious fumes in air, etc.)?

                              Keyword:

                               

                              Visible to Customer:

                              Yes

                              Created On:

                              2/5/2008

                              Updated On:

                              8/23/2013

                              Answer:

                              COSMOSFloWorks can analyze the mixing of up to 10 different gasses. COSMOSFloWorks uses the assumption that Lewis number is equal to 1 (Le=D/a=1, here D – diffusion coefficient, a - thermal diffusivity). This assumption denotes from the physical point of view that the heat and diffusion processes are similar. Therefore the user doesn’t need to set the diffusion coefficient. Thermal diffusivity is connected with coefficient of heat conductivity a=lambda/Ro/Cp (lambda - heat conductivity coefficient, Ro - density, Cp - heat capacity). The user has an influence on diffusion by means of lambda.

                                • Re: modeling steady state gas concentration for dense gases
                                  Jared Conway

                                  also was surprised there was no incremental saving enabled and that you had flow freezing enabled.

                                   

                                  have you tested without flow freezing?

                                   

                                  how are you evaluating how things are behaving without looking at the time effects?

                                    • Re: modeling steady state gas concentration for dense gases
                                      Dan Hofstetter

                                      Sorry, that was a quick-and-dirty model that I modified so that I could upload something for discussion.  I must have saved it after manually enabling flow freezing while watching the solver monitor.  I normally would have flow freezing disabled - I enabled it during one particular run after the velocity field converged to speed up the concentration calculations.

                                       

                                      I turned off incremental saving to save storage space - I was sitting in front of the solver monitor while running this simulation.  I usually use a combination of tabular and periodic saving when the simulation results are important.  I just want to see if the overall trends seem correct for this simulation.  If it looks like the results are reasonable for this small tank, I'll take a harder look at a bigger tank with saving enabled.

                                       

                                      I've been messing around with time step size while trying to match the accuracy of temporal results from experimental data.  Another CFD package, Phoenics, seems to reach convergence at the end of every time step, which results in pretty decent temporal accuracy for the concentration decay process.  The solver in Flow Simulation seems to be completely different in that regard, and there is no option for setting the number of iterations per time step.  The closest I can come is to force a very small time step for the first 10 seconds, then transition to a larger time step afterwards.  It seems like the concentration calculation "runs away" (the concentration decays too rapidly) if I keep the very small time step for the duration of the calculation.  I was messing around with a formula definition for time step size:  { 1 - exp(-t) } seemed to work pretty well for the first 10 seconds of simulation.

                                      • Re: modeling steady state gas concentration for dense gases
                                        Jared Conway

                                        and out of curiosity, did you change the default turbulence parameters?

                                      • Re: modeling steady state gas concentration for dense gases
                                        Dan Hofstetter

                                        Jared Conway wrote:

                                         

                                        3. you have interferences in your model, for flow only this is sometimes ok, but generally don't recommend having them in your model

                                         

                                        What do you mean by interferences?

                                         

                                         

                                        Jared Conway wrote:

                                         

                                        i notice you're faking an external analysis, have you tried with an actual external analysis?

                                         

                                        I changed this model to an external analysis once, but the default computational domain size is much larger than I want, and I need to be able to control that automatically when building the model from a website where I input dimensions and flow rates.  The internal model with the "fake" external volume allows me to define the distance outside of the model using a design table.  I couldn't see a way to control the size of the external computational domain without manually going in and changing the domain extents.  The external analysis would make wind a lot easier to deal with since I could use ambient conditions to define that.

                              • Re: modeling steady state gas concentration for dense gases
                                Jared Conway

                                can you post a video of your results?

                                 

                                in 2014 sp2.0, with your study and incremental saving enabled, the 100 PPM of heavy gas drops to the bottom.