Error in Two phase (condensation) modeling
I am modeling direct contact condensation in nozzle-tank arrangement.
The tank is filled with water at 30 degree Celsius and the nozzle is used to inject saturated steam at 2 bar absolute. I am using thermal phase change model. I have tried many things but the solver still give error. It runs normally for more than 300 iterations converging slowly, and out of nowhere sudden peaks in residuals happens and solver gives error. I have tried everything in the FAQ, but still no success. I have used a variety of meshes, very small timescale, played with boundary conditions, initial conditions and what not. PS, single phase simulation converges just fine in the same arrangement. I could really use some help regarding pin pointing the source of the error. http://imagizer.imageshack.us/v2/xq90/661/l9yprh.jpg |
Is this steady state or transient?
This sounds like a tricky simulation to me - compressible flow with phase change. I would run the simulation generating a results file every timestep/coeff loop (or at least the ones just before it crashes). Have a look at what happens around the time it crashes - does some steam first hit the outlet boundary? Does some water backflow into the nozzle? Does a water vapour bubble collapse? |
Quote:
I switched from opening boundary condition to Outlet boundary condition and now the solver doesn't crash, but doesn't converge either. The rate of convergence is 1 for all equations for a 1000 iterations now. I should think of it as a step in the right direction, right? |
Quote:
Firstly - did you read my previous post? Unless you have a look at what is happening you are just guessing. Secondly - it is quite likely this flow is transient and no steady state answer is possible. The FAQ talks about this. Finally - if you want us to help you you need to provide more information. Please post your CCL and some images of the flow you are getting and the mesh you are using. |
1 Attachment(s)
You were right, thanks. I gave Steam volume fraction at Opening/outlet to be zero as boundary condition. When steam reaches the boundary, the solver crashes, or at least that is my interpretation. Because I removed that condition and the error disappeared.
Now my solver is running but the solution neither converges nor diverges. Code:
LIBRARY: |
Code:
MULTIPHASE MODELS: |
5 Attachment(s)
I have used different meshes but right now I am using this one. It is automatically generated mesh by CFX.
The result file gives these values for Steam Volume fraction and Steam Mach number. Isn't the mach number supposed to be highest at the outlet of nozzle? Attachment 40351Attachment 40352Attachment 40353 Attachment 40354Attachment 40355 |
The shock wave can occur in the divergent section in some flow regimes.
You do not need the long pipe leading to the nozzle. You could put your inlet boundary much closer to the nozzle and save lots of mesh. Given that this simulation appears to have: * compressible flow * Shock waves, sonic flow * free surface * phase change This is a very difficult model and I would expect a lot of difficulties in getting this to converge. This is going to take a CFD expert quite a while to get working. Unless you are an expert in CFD I fear you have taken on too advanced a model. This model is going to be too hard to fix over the forum. |
Quote:
This is my for Masters Project so I am stuck with it anyway. I will try different approaches, and will ask the experts of the forum whenever I get jammed. Questions arise in mind during the course of a study, and I will ask them rather than asking to solve the entire problem. |
If are doing a masters on it then you have time to become an expert in the topic. So you have some time to get it right.
I would recommend a staged approach: 1) incompressible, single phase flow (you have already done this) 2) compressible flow, single phase flow 3) compressible water vapour flow, single phase 4) incompressible free surface multi phase, simple fluids (air and water- if this is a free surface simulation) Once you can successfully do all these only then would I consider combining them. And don't just get them to converge, do some sensitivity studies to show that your results are accurate. |
This is not a free surface simulation. And thanks for moral support Glenn. I really appreciate it.
I will start doing exactly that. I have two questions for you Sir, 1. What type of mesh do you think should be adequate for this purpose? 2. Generally, to simulate a real life bigger tank, on the sides of the the tank in the simulation is given Opening boundary conditions. My question is, buoyancy being ON, hydrostatic pressure in play, can I give the static pressure at opening BC to be atmospheric pressure? This opening BC being at the sides and/or at the bottom of the tank. |
Tip:
1) Read the CFX theory guide as to which sources you are solving for. 2) Find the derivative of those sources and supply the information to CFX as a linearisation coefficient for all the sources you have. 3) The residuals should behave better with a linearisation coefficient . Read the theory guide on "linearization coefficient" and read up Patanker's book on numerical heat transfer. It has a good section on source term linearisation. |
Quote:
Well you must be right, because I didn't get anything you said. Going to research on everything you recommended. I really appreciate the help Sir. |
Quote:
http://www.cfd-online.com/Wiki/Sourc..._linearization |
3 Attachment(s)
I used a ridiculously coarse mesh and my solution converged, in terms of residuals. The residuals dropped all the way to E-05. The domain imbalances found were away from the region of interest.
The mach number is maximum at the tip of the nozzle as one would expect. I put a moniter point at the tip of the nozzle for Steam Mach number and the value is stable. I refined the mesh and again the residuals started to repeat a trend. I know this solution is not to be trusted, but I do want to know why this happens generally. I would appreciate any help. Attachment 40367 Attachment 40368Attachment 40369 |
There is not much point theorising into the details of a coarse mesh simulation. The coarse mesh will mean the result is not accurate, so I see little point in thinking too much about its details. Refine the mesh to a point where the results are trustworthy and then think about the result it tells you.
The answer to your problem is probably something to do with the exit shock moving about in the divergent section, and the location of this shock moves with mesh refinement. But as I said, I see little point in analysing this in too much detail as it is wrong to begin with. |
All times are GMT -4. The time now is 23:05. |