Dear Mladen Bošnjaković,

In am opinion, the residual shown in the picture is too far from convergence. Model, Boundary conditions, Time step size will play a vital role.

K-e standard model is sufficient, I think no need to change RSM or k-omega. Problem indicated that choose more reasonable time step size and BC. 

In last, are you sure about mesh quality and mesh aspect ratio ?

Regards

Ali

Correct specification of values of k, epsilon (or omega) at inflow and outflow boundaries is critical for turbulent flow simulations. The turbulent viscosity ratio error in your simulation indicates that excessive amount of (turbulent) viscosity is getting generated in the simulation.

If a heat exchanger is being simulated, your should use the "turbulent intensity and hydraulic diameter" option in FLUENT for specifying k and epsilon (or omega) at the inflow and outflow boundaries. Turbulent intensity can be in the range 2-5% whilst hydraulic diameter would be equal to the diameter of the inlet to the heat-exchanger (assuming circular cross-section).

I have written about specification of turbulence parameters (k and epsilon) in detail in chapter six of my master's thesis. Here's a link to my thesis. I hope your problem gets resolved. All the best!

Thesis Development of a Finite Volume Solver for Turbulent Multiphase Flow

Muhammad Furqan Ali,

data about mesh is attached. Propose reasonable time step size and BC (where to define them).

regards

Dear, Shaswat Anurag Saincher

is below OK?

where i can specifying k and epsilon at the inflow and outflow boundaries?

Regards

Dear, Shaswat Anurag Saincher

result is same ...

Dear Mladen

if you did what was suggested till now and you got nothing for convergency improvement, please check this point as well. As you know governing equations containing Navier, k-e, continuity and energy are coupled and nonlinear. It means if there is no control on updated value at each iteration, it may drastically change. from my experience, under relaxation factor also plays an important role in convergency specially in complicated geometry and B.C. Change it to a lower amount and check the trend of convergency. please let me know if it works.

Dear Yahya,

before i change under relaxation factor what is your proposal for k-epsilom model?

Standard, RNG or Realizable?

Dear Mladen Bošnjaković,

I am concerned about the large residual of epsilon (turbulence dissipation rate) or omega (specific dissipation rate) at the beginning of the simulation. Can you tell me the range of values of k and epsilon at the start of the simulation??? You can check these values from display >>>> contours for instance.

The warning of turbulent viscosity ratio indicates that the turbulent viscosity (rho*cmu*(k^2/epsilon)) is coming out to be very large at the beginning of the simulation....this further indicates that k is very large (or epsilon is too small) at the beginning of the simulation. For stability and preventing excessive viscosity generation at the start of the simulation, one should initialize from a dissipative field that is epsilon>>k. 

If the turbulence specification option for intensity and hydraulic diameter is yielding large viscosity ratio, try initializing turbulence at inlets and outlets in terms of intensity and viscosity ratio. 

You can initially set intensity in the range 2-5% and turbulent viscosity ratio as unity (assuming that molecular and turbulent diffusion are initially equal). Kindly check if this resolves the problem. If not, then there might be some issue with the mesh....such as skewness. Have you ensured that the mesh does not contain a large number of highly skewed elements???

Dear Mladen

I have used Realizable mode and it gave me good results close to experimental data. this model also exhibits superior performance for flows involving rotation, boundary layers under strong adverse pressure gradients, separation, and recirculation. moreover, in every measure of comparison, Realizable k-ɛ demonstrates a superior ability to capture the mean flow of the complex structures.

Dear Yahya,

I set epsilon=100 and k=1.

At beginning it looks good till iteration nr 7, but after that diverge ...

Under-relaksation faktors are as is shown on picture ...

Dear Shaswat,

I make some small changes in mesh ...

Now it looks much better ... You are right, mesh is very important ..

I hope this time will be convergence ...

Dear Mladen Bošnjaković,

I am happy to see that the solution has stabilized. Hope that your problem gets solved.

Best regards.

Program is still working ...

Now it looks like in picture ...

please try relaxation factor for kinetic and momentum 0.4 to 0.45 (bellow 0.5). unrealistic reversed flow in some nodes occur mostly when 

1- there are drastic changes in the value of dependent variable

2- boundary conditions are not satisfied correctly

Program is finished.

835 iterations ...

It would be better ...

 There are probably plenty of reserves to improve the parameters of the mesh...

I started working with Fluent prior 40 days. Almost everything I learned, I learned from tutorials and forums and Reasgate.

Since I working on a doctoral dissertation expected me a lot of work that is scheduled to end in 2017. I will certainly have a lot of questions, and with your help I will overcome problems ...

Therefore all thank to you very much for your help.

Below I will list the parameters of the previous case for those who follow this issue...

....parameters are:

Mesh:

- ortogonal quality. 0,84

- skewness: 0,26

- Corvature on

Smoothing: high

max face size: 0,37 mm

max. size: 1 mm

Inflation: 12 layers

first layer: 0,05 mm

Solution:

Steady, model k-epsilon realizable, enhanced wall treatment

Boundary c.: Air entry (ulaz zraka):

Turbulence: specified method K-epsilon

Turb. kin. energy 1, Turb. diss. rate 40

Air outlet(izlaz zraka): Intensity and viscosity ratio

Backflow Tur. intensity 2

Backflow turb. Visc. r. 10

Solution method: SIMPLE

Under relax. factors:

pressure-0,5 density-0,9 body f.-0,9 Turb. kin.e.-0,45  Turb. visc.-0,7  energy-0,9

Solution intialization: Standard