Why does velocity increase unexpectedly in density based solver for compressible flow
I am trying to simulate a mixing process of Helium jet in atmospheric air. He is coming out from a tube of 4.6 mm inner dia with a Re number of 2500 where the air is considered stagnant.
To set the model, I use an inlet velocity BC with a velocity of 66.5 m/s for He jet inflow in the domain. I also set a zero pressure outlet BC at a sufficient distance from the inlet to allow enough space for mixing. The model is in a steady state. To capture the mixing, I use the species transport model available in Fluent. It looks like the model works fine. However, I found the velocity goes unexpectedly high (~1200 m/s max) in the domain when the solution is converged. If anyone has any suggestions, it would be highly appreciated. |
Wouldn't the zero pressure outlet cause the air in the duct to accelerate out?
The setup seems to be basically a rocket exhaust nozzle in space vacuum. High exhaust velocity makes sense for such a rocket nozzle. Though, I'm not sure about 1200 m/s. Am I misunderstanding something?:) |
aerosayan,
Thanks for your reply. Zero Pressure outlet: that's a valid point. Though the velocity profile looks like rocket propulsion, the physics of my case is simple - helium is coming from a tube like a jet and being mixed into the air. To my understanding, the velocity at the outlet should be much smaller than the inlet velocity as the flow decelerates as it gets diffused to air. In that case, could you please suggest any other BC type for the outlet? |
If I understand, the model is for full compressible flows, isn't it?
With such a high velocity the flow is supersonic and the outlet condition does not enter in the computation :confused: What about the other inflow conditions? Being subsonic you need to fix all but one boundary condition. Have you verified the other variable? |
the model is for full compressible flows, isn't it?
Yes The inlet velocity is only 66.5 m/s (M ~ 0.2), and I am not sure why it is becoming supersonic as it moves forward. I am not sure if the inflow condition is the crucial factor. Have you verified the other variable? I am not sure what you meant. I made sure the model setup was okay except the BCs as the results are not expected. |
Quote:
First of all, your simulation should be unsteady in a deterministic sense, thus are you using a RANS formulation? In compressible flows you need to prescribe the inlet condition for the full set of equations, that is density, momentum and energy. For subsonic flows you have to let a condition coming from the interior, you cannot prescribe all conditions in terms of Dirichlet. Hence, what about the prescription of the other variables? |
Quote:
If you could write a bit detail, it would be really helpful for me. |
Quote:
You need to know the theory of the chararacteristic for hyperbolic equations. As example, consider a 2D case, you have to consider the conditions for 4 variables at the inlet, that is density, x-velocity, y-velocity, total energy. For subsonic flow the characteristic line related to the eigenvalue u-a will come from the interior towards the inlet section. Therefore you have to set one variable free to come from the interior. Have a look to the theory in some textbook. |
Quote:
Are you using pressure based solver or density based solver? |
Quote:
The title of the post is about density based solver ... |
Quote:
I recently implemented two phase comparessible flow model and did some similar calculations like injecting a gas etc and I did not observe this type of behaviour. But I did see this type of behaviour with pressure based solver. So i find it bit strange the density based solver shows this. |
Quote:
I would not use an inlet velocity boundary condition. Pressure inlet or mass flow inlet should be the proper one. |
Could you give more details on the relevant settings? 2D/3D/AXI? What are all the material settings? All the other boundary conditions? Maybe a sketch of the mesh/domain? Convective scheme? Gradient? Turbulent or laminar?
Also, to which extent is this high speed region relevant? Does it seem as a physical result of the used boundary conditions or some numerical artifact? Maybe a picture would help as well. Finally, qualify converged results. |
All times are GMT -4. The time now is 17:02. |