Fluids

Fluids

Material changing composition without reasons in multiphase simulation

    • michele rocchi
      Subscriber

      Hello, I’m running a simulation of a bubble column in CFX. The domain is set up as multiphase, WATER and AIR. WATER is the continuous phase and AIR is the disperse phase.


      Both materials are defined as variable composition mixture; the material list for AIR is O2 and N2 and the material list for WATER is H2O and dissolved_O2.O2 and N2 are set as ideal gas. H2O and dissolved_O2 are set as liquid using the default property for water.


      More in detail the material AIR is defined as: material group>user, mixture property>ideal mixture, thermodynamic property>equation of state>ideal mixture.


      The boundary are:


      An opening at the top of the column.


      An inlet in the bottom region, where both AIR and WATER are introduced in the domain thought the throat of a jet ejector.


      An outlet at the very bottom; the bulk mass flow rate is set in order to match the mass of water entering the domain trough the jet ejector.


      The AIR that enter the domain is roughly 0.21 O2 and 0.79 N2 in mass fraction. The AIR composition at the opening on the top is set exactly in the same way, 0.21/0.79.


      Mass transfer is right now set to none. My problem is that the mixture composition of AIR change throughout the domain with no reason at all, going from 0.00/1.00 to 1.00/0.00 O2/N2.


      This happen completely at random: sometimes the composition remain constant at  0.21/0.79 for thousands of time step, sometimes some cell in the mesh change composition inexplicably, and once, after changing  convergence control>length scale option from conservative to aggressive in the solver manager, in the entire domain the AIR changed in composition almost entirely to 1.00/0.00 O2/N2.


      I hope someone can give me some explanation because now are few weeks that I’m tiring to understand the reason of that.


      In the attached picture it can be seen a region close to the column wall where the mol fraction of
      O2 in the material AIR change without explanation from 0.21 (the expected value) to something close to 1.  
      Only a quarter of the column is simulated.


      I thank in advance anyone that will be so kind to provide his opinion.

    • DrAmine
      Ansys Employee

      Hi,


      Possible reason might be a strong backflow from the opening. Do not forget to enable buoyancy and set the density to the one of the continious phase. Moreover check if walls are created at your outlet. Moreover check for deep convergence in every single time step (by judging the variations of some Monitor points and flow imbalances). For post-processing do not forget to limit the visualization to the areas /regions where gas is present


       


      A.

    • michele rocchi
      Subscriber

      Dear abenhadj,


      I really apriciated your advice.


      Actually the reference density was set to the one of the disperse phase. Thank you for that.


      Walls are created some times at the outlet at the bottom of the column (20%-30%), the fluid that tend to "back-flow" is the disperse phase AIR. I don't have idea why AIR should back-flow from there; since the boundary is set as outlet I can't define the "fluid values" as for an opening. From a physical point of view, the column of liquid above the drainage hole should produce more than enough hydro static pressure to provide the required liquid flow at this boundary.
       


      <>


      Do you mean that I should not care about the composition of the disperse phase in the locations where the disperse phase is not present?


      <>


      I will monitor the flow at all boundary to see if this happens before the inexplicable change of composition.


       


      I'm really grateful for your help.


      KR


      Michele 


    • michele rocchi
      Subscriber

      I star again the simulation. in the picture on the right are evident some mesh elements  where the oxygen concentration is close to 1 as indicated by their color.


      However these elements are external to the iso-volume defined for AIR volume fraction larger than 1e-6 (picture on the left).


      So probably I may assume that the gas is not present in the region outside the green iso-volume.


    • DrAmine
      Ansys Employee

      Hi,


      Is the whole column filled with water or do you have a continuous puffer in the top (free surface). Regarding the walls which are created at the outlet perhaps is related to the pressure distribution.


       A.

    • michele rocchi
      Subscriber

      Dear abenhadj,


      There is a free surface at the top.


      I can avoid  the free surface calculation (which I'm not interested in) if I use degassing condition for the opening on the top.


      In that case I'm not allowed to set a pressure for the boundary. And I think this means that pressure in the domaine would be undetermined since there would be no other boundaries where I impose the pressure.


      BR


      Michele

    • DrAmine
      Ansys Employee

      Hi,


       


      In that case then I would rather use the density of lighter phase as operating density and to initialize the pressure field with the correct pressure taking into account the hydro-static head (which is reduced by reference density time height time acceleration). Accordingly you have to set correct pressure definitions for your pressurized b.c's. 


       


      In case you work with degassing you still have an outlet and this would prescribe the pressure level. If you do not have any pressurized boundaries you can prescribe the pressure level information under Solver Control.


       


      A.

    • michele rocchi
      Subscriber

      Dear abenhadj,


      I really appreciate your help thank you very much for the time.


      I'm still using free surface and I put back the buoyancy reference density to the one of the lighter phase.


      Now I add a short piece of conduit at the very bottom and I set the boundary to opening so that I specified the pressure and the fluid values: WATER volume fraction to 1.00 and AIR volume fraction to 0.00 with component details for AIR to 0.233 O2 mass fraction and 0.767 N2 mass fraction. (before the drainage was set as outlet and only the bulk mass flow rate was specified)


      Regarding the fluid-dynamic I think the solution is pretty much converged. At least by seeing the gas volume fraction distribution in the column that remain stable over the time. The absolute pressure at the drainage, on the sparger and at the opening on the top (101325 Pa) are all three stable as well. Same situation for mass flow of WATER and AIR on the sparger at the drainage  boundary and on the opening at the top of the column.


      Checking the molar fraction of O2 at the sparger is 0.21 as it should be. 


            


      But again the domain does whatever it wants in terms of composition, I really don't understand and I'm a little bit contrariated about this. Now is quite some time that I can't get ahead of this. Below some screenshot in chronological order from left to right and from up to down.


      You can see the molar fraction of O2 going from roughly 0.21 (at least in the region where air is present) to 0.00, then to values very close to 1.00, and then again some pockets at O2 molar fraction of 0.00 appears. I don't understand where that AIR with that composition come from (now all the opening are set as: component details for AIR to 0.233 O2 mass fraction and 0.767 N2 mass fraction).



      Additional information that may be problematicin my setting are:


      -the disperse phase is compressible (AIR is defined as: material group>user, mixture property>ideal mixture, thermodynamic property>equation of state>ideal mixture) 


      -At the sparger  the volume fraction of water and air are set as zero gradient. This boundary is set as inlet>fluid dependent and then I specified separately the mass flow of AIR and WATER.


      KR


      Michele


          


       


       


       


       

    • michele rocchi
      Subscriber

      Below are reported 2 plots for the O2 molar fraction in the AIR at the sparger location (areaAve(Air.O2.Molar Fraction)@Sparger) before and after the "compositional catastrophe". At the same time the color mapp on the sparger throat remain constant and fixed at 0.21 (first figure in the previous post low-right corner).



      BR


      Michele

    • DrAmine
      Ansys Employee

      Hi,


      Please create an expression where the mass fraction of oxygen is multiplied by the air volume fraction and define it as a variable for post-processing. 


      Specifying zero gradients is generally used for fully developed flows where the gradient of volume fraction normal to boundary is zero.


      A.


       


      A.


       


       


       

    • michele rocchi
      Subscriber

      Dear abenhadj,


      Here  some sequential screenshots showing the product of AIR volume fraction multiplied by the O2 mass fraction in the air  (Air.O2.Mass Fraction*Air.volume Fraction). On the very right are reported the AIR volume fractions corresponding to the first and to the last result.


      Where the dispersion of air in water is present it is clearly visible a slightly change in the shade of the blue: from dark (first 2 screenshot) to lighter blue (the rest). Going from ~Air.O2.Mass Fraction*Air.volume Fraction=0.233*0.15=0.035 to  ~1.00*0.15=0.15. On the top, where the AIR volume fraction is 1.00 it is clear the change in color from roughly 0.233 to 0.00 and then to 1.00.  


       


       


         


       


      I will calculate a the volume fraction at the sparger using the absolute pressure given by the solver at the sparger throat location and set the corresponding values rather than using the zero gradient option. I will also run a transient simulation hoping to see only 0.233 everywhere.


       


      BR 


      Michele 

    • michele rocchi
      Subscriber

        Other additional information regarding my setting in default domain>fluid specific>component models models are:


      - AIR>N2 constraint (I assume the liquid is saturated in N2 and this species is not interested by interphase transport)


      - AIR>O2 transport equation (but right now the mass transfer option is set to none) 


      -WATER>H2O constraint


      -WATER>O2_sol transport equation ( AIR|WATER is a fluid pair and O2|O2_sol is in component pair details but right now the mass transfer option is set to none)


      What if fluid specific>component models instead of constraint I specifie myself an algebraic equation for N2 mass fraction of the kind 1-Air.O2.Mass Fraction?


       


      Thank You for your time abenhadj     

    • DrAmine
      Ansys Employee

      Hi,


      From the pictures it looks plausible in almost all parts. Bu If I look in the volume rendering screenshots I see in the top of column 100% O2 which is wrong if you are describing proper opening boundary conditions and you are working with proper density definition. How did you patch the top part of your column? Did you patch a air composition here?


       


       


      I would moreover work on deeper convergence. The settings regarding components are okay. I think the results are fair enough. Perhaps one of the community member might add a comment here because as an ANSYS Employee I am not able to add more than I have already done without looking into the case directly


       


      A.


       


       

    • michele rocchi
      Subscriber

      Dear abenhadj,


      May you explain how this can be plausible. I don't understand I'm sorry. I really don't get why the composition of AIR change completely at random when I imposed 0.233 O2 mass fraction at the only inlet boundary present in my domain. And when I monitor the area average AIR composition at the sparger boundary the molar fraction of O2 instead of remaining fixed at 0.21 it vary at random. As shown in the plot.   


      Could you please check those picture. Because as it is I can't do anything with a simulation where the material composition change at random when the mass transfer is turned off and the simulation is steady State.


      Actually I realized right now that the images get compressed when uploaded and is really not possible to read anything


       


        


      KR


      Mauro

    • Rob
      Ansys Employee

      Hi Mauro,


      As abenhadj has noted, as ANSYS staff we can't be too specific with answers, and you may have seen us use forms of "unfortunately this question is the beyond the scope for ANSYS employees to answer on a public forum, so hopefully other student community users can chime in" elsewhere. 


      Having said that, the results do look odd if you have all of the phase material composition set at the sparger and top boundary: the gas volume fraction looks fairly sensible. So, check your settings, and also look at convergence as I can't think of a scenario where the gas would suddenly lose all of one component in normal operation. 


      Good luck with the project. Have you spoken to your supervisor?


      Cheers,
      Rob


       


       


       


       


       


       

    • michele rocchi
      Subscriber

      Dear Rob and Abenhadj,


      I will talk with my supervisor next week.


      For now I would like to gain a little insight in the main problem I had and still have:


      The change  in composition of a material without an apparent reason.


      I think all the other small issues are much less problematic. It is obvious that I can't use a simulation where the composition of the domain vary over the time when all the boundaries are set in steady conditions.


      After setting up a completely new simulation (from zero) in CFX-pre I have run a steady simulation and a transient. Both simulation are ok in terms of pressures and flows at the boundaries and at the monitor points. After reaching a more or less stable condition then again with no reasons the composition of the dispersed phase start to change both in the transient and in the steady state simulation. As I said the O2 molar fraction in the dispersed phase should be everywhere 0.21. This is the case for around 2000 step but than it change.


      The only things I changed myself during the simulation were the flows before the1500th step (see link 3, 4, 5), in order to partially empty the domain from the continuous phase and to gradually (in stepwise manner) reach the final velocity at the sparger location. 


      Transient sim: disperse phase composition at the sparger location and at the top opening


      Steady sim: disperse phase composition at the sparger location and at the top opening


      Transient sim: absolute pressure at the drainage, top opening and sparger location


      Transient sim: reactor mass content


      Transient sim: continuous phase mass flow at the drainage, top opening and sparger location


      I attach a link  to a video showing the change in composition of the disperse phase: . The video is relative to the steps between 2000 and 3500, the physical length is 1.46 minutes, and it is played 3 times.


      O2 molar fraction in the disperse phase: video


      (https://vimeo.com/281457092)


      It is clear that  the composition of the disperse phase has the correct value of 0.21 molar fraction in all the domain but then some element positioned below the sparger location change to 1.00 and others to 0.00; after the inconsistency propagates to the entire domain. 


      Is there any reason for that? Any clue on what can be?


       


      BR


       


      Mauro 

    • DrAmine
      Ansys Employee

      Hi Mauro,


      as ANSYS staff we can't be too specific with answers and I highly recommend that you contact your supervisor. and talk with your ASC I recommend checking the imbalances in the solver run. Check the diffusivity value of your O2 gas component and use the density of the continuous phase instead of the dispersed phase (pressure almost hydrostatic with the exception of the gas layer which you can get rid of at first trial).


       


       

    • michele rocchi
      Subscriber

      Dear abenhadj,


      Right now in fluid specific model>component model, O2 is set as transport equation but kinematic diffusivity  and turbulent flux closure  are unchecked and the corresponding cells are empty.


      This is the setting for both


      -O2 in the dispersed phase material-mixture


      -O2-solute in the continuous phase material-mixture


      Does this make sense?


      BR 


      Mauro

Viewing 17 reply threads
  • You must be logged in to reply to this topic.