[tarkesdora]

The objective of my CFD problem is to determine velocity at a particular region of interest (Picture attached) in the domain. It is a 3-D periodic model of nitrogen gas flow in a hollow cylindrical shaped domain. The model corresponds to Precision Glass Molding machine where nitrogen gas at 20 C flows with in the domain for cooling of glass.

Software used:
ANSYS (14.5) Workbench for Model
ICEM CFD for meshing
Fluent for Solving

Inlet velocity: 43.4 ms-1
Outflow conditions: Not known

ANSYS workbench files are available in the drop box folder in following link. Periodic boundary conditions and inlet condition are set in the model
https://www.dropbox.com/sh/ftjtpu9d1...Mq47xBIUhlFo-a

Challenge in the problem:
The hollow cylindrical domain is large abut 300 mm and in the inlets are very very small 4 X 0.3 mm on the surface of the cylinder. T0 my understanding a dense mesh is required at the inlets for capturing the behaviour of jet.

Issue is:

1. With this type of mesh, solution does not converge for laminar/
turbulent analysis (velocity <0.1 and continuity <0.01). Parameters in FLUENT considered were default values. It should converge in the order of 10^-6 for a reliable solution. How is the quality of the mesh for the model ? What kind mesh pattern would be desirable for convergence?

2. The jet at the inlet is laminar for a duct of 4*0.3 mm. However, once the jet gets into the domain how to know whether the flow is laminar or turbulent inside the domain? What kind of analysis has to be done to know the flow is laminar/ turbulent?

Kindly could you help me in solving the problem? I would be highly thankful to you for all the help and support.


Note: I am a new person working on a CFD problem with zero experience and also I am new to the CFD community. I work on FE and structural analysis. However for the problem a CFD analysis is required. I am unable to proceed in appropriate direction for an acceptable solution.

[Martin Hegedus]

In regards to incompressible gas dynamics analysis and CFD, it's all about the Reynolds number. If the Reynolds number is low enough, then an analysis can be done without a turbulence model. As the Reynolds number increases, a turbulence model is required to to capture the effects of the small vortices which can not be modeled (because they are too small relative to the entire domain). So if you can refine your grid enough to capture the small vortices, then a turbulence model is not required. It's also important to understand the difference between unsteady laminar and turbulent. Conceptually, they may seem the same, but in regards to CFD, they are not. When one zooms in enough, everything is laminar. For example, smoke rising from a cigarette. People will say the flow is initially laminar and then becomes turbulent. That's a "human" distinction. In regards to CFD, the Reynolds number is very low, so initially the flow is laminar and then the flow becomes laminar unsteady.

In regards to your problem, it almost sounds like your problem is unsteady. So your solution process must take that into account.

Metrics to determine if your flow is turbulent or not are the Reynolds number, what is out there in literature, and ones own experience. You can also refine the grid. If vortices keep getting smaller and smaller, and you are not getting grid convergence, then your flow is probably turbulent.

[tarkesdora]

Graph.png

Picture1.jpg


In the model already I am having 1.4e6 cells. Do you feel further refinement of the model is necessary ?

Initially I had done a laminar steady analysis. THe graph for convergence is attached.

I have done a laminar unsteady analysis. With time step 0.0001, no of time steps 50 and Maxitre/time step 200
Remaining all the paremeters of solution controls are default. The solution converges atleast to 3rd order for all time step. Thanks a lot martin for giving a good understanding into. Could you kindly suggest some reference which suggest these fundamentals. It will be a great help

[Martin Hegedus]

In general, I recommend that a grid refinement study always be done.

Unfortunately I don't know of a practical guide to solving problems with CFD. One would think something would have been written, but I have never come across one.

[tarkesdora]

Grid refinement definitely I will do thaat. Now I was just thinking how to explain the physics of flow with proper examples and references