Learning Forum

[mohalinasre2001]

Hello Everyone,

I am a master's degree student working on my thesis. I am running a 3D transient simulation with coupled solver and IDDES turbulence model. The problem is that I am struggling in choosing the approriate CFL number found in Solution Controls. It's 200 by default, but it gives wrong power coefficient. I tried numbers like 1, 0.7, 0.5 and I realised the following: the more I decrease the CFL number, the larger the power coefficient gets. So I am confused on what to choose as my CFL, because for each Tip Speed Ratio it's different. i tried tunning it for each TSR, but I feel I am doing it wrong because I am trying to validate the setup vs. experimental data and I should not change settings for each TSR.

I would appreciate if someone can help me get around this because I am running out of time. I am willing to include the name of whoever helps me finds a solution in my thesis Acknowledgements section.

Thank you in advace!

 

[mohalinasre2001]

Side Notes:

• y+ values for the blades are below 1. 
• Mesh quality is decent

[Essence]

Hello,

You need not add my name in your thesis. It's totally fine.

What is the equation of power coefficients you are using? Are you sure that CFL number is the only one which is impacting the results? There can be other factors which may be affecting the results.

[mohalinasre2001]

Hello!

I am using the moment coefficient defined in report definitions. I am averaging it over a complete turbine rotation, and then multiplying it by TSR to get power coefficient (Cp=Cm* TSR). So basically CFL in solution controls is affecting the moment coefficient, which in turn affects the power coefficient.

And yes after trying several simulations while changing CFL number only, the moment coefficient was changing such that: increasing CFL makes Cm-avg smaller, while decreasing it makes it larger!

Thank you for your help!

[Essence]

CFL/Courant number in Solution controls is used to define the stability of the residuals. How do the residuals look? Are they converging for each time step?

[mohalinasre2001]

Yes they are converging well. I set maximum iterations per timestep to 60. The continuity residual falls below a limit of 10^-3 while the rest fall below 10^-5 without a problem. 

[mohalinasre2001]

This is why I am confused! But I feel like 60 iterations is a high number for residuals to converge.. It depends on the TSR, for hight TSRs (1, 1.25), it takes around 40 iterations, while at low TSRs (0.25, 0.5) it takes about 60+ iterations to converge.

 

[mohalinasre2001]

I made this graph to show you how increasing CFL slightly moves the moment coefficient curve downwards, which in result leads to decrease the average moment coefficient. Each curve is taken from a seperate identical simulation with different CFL as shown. They are all taken at the same time (5th revolution of the VAWT).

[Essence]

Thanks for your response.

Could you please use lower time steps (use equations to determine the time step)? You can also reduce the number of iterations to 20 - 25 maybe.

[mohalinasre2001]

I did a timestep sensitivity study, I tried azimuthal increments of 0.6 deg and 1.2 deg for turbine revolutions, and they gave the same moment coefficient curve. I even tried lower timesteps (0.1 and 0.25 deg) but I got a totally wrong solution. So I don't think lower timesteps will help. 
Based on the paper I am referencing in my thesis (Experimental and numerical investigation of a three-dimensional
 vertical-axis wind turbine with variable-pitch), they used 1.2 deg increments, and it worked fine with LES turbulence model.

[Essence]

Thanks for trying out the suggestions. Could you share the screenshot of the viscous models panel? Please be careful about the reference values too. Since they are used to determine the moment coefficients. And moreover, please check if the moment report definitions and that from the "Forces" in Results section match or not.

[mohalinasre2001]

Concerning the reference values, they are all correct (area = height x diameter of turbine) (length = radius of turbine) 

And for the turbulence model, please find attached a screenshot showing the panel:

But I didn't quite understand this "please check if the moment report definitions and that from the "Forces" in Results section match or not." 

Thank you so much for your help!

[mohalinasre2001]

Hello again!

Is there anything else I should do?

Thank you!

 

[Essence]

From the graphs you have shared, it seems that they are quite close to each other. What percentage difference are you aiming for? From my persepective, 2 - 5% should be fine for most cases. But at the end it depends on the specific applications and the end user.

I would also recommend you to carry out mesh independence test. You can use Interpolation tool in Fluent for quickening the process.

[Essence]

Additionally, may I know whether the cell pass time is equivalent to time step size? Since you are simulating IDDES, did you use Data sampling? Generally, you don't need 60 iterations per time step. This indicates that something in the flow hasn't been resolved yet.

[mohalinasre2001]

I did a mesh independence test and I went with the medium mesh containing 11 million elements. 

For the timestep, I am using a 1.2 degrees azimuth increment, and then I calculate the timestep using: t=1.2*PI/(180*omega)  where omega is the rotational speed (rad/s)

For example, for TSR=1 (omega = 20 rad/s), the timestep is 0.0010471 sec. While for small TSR like 0.25, the timestep is 0.004188 sec.

Concerning the cell pass time, I don't know how to calculate it, but I think you meant cell size/flow speed.

Smallest element length is 0.07 cm, and the flow speed is 8 m/s. Which gives a cell pass time of 0.0015 sec roughly.

Does this create any problem is resolving the flow? 

(And no, I didn't do data sampling.)

Thank you so much!

[Essence]

Please ensure the time step is equal to or lesser than the cell pass time. And by the way, I am referring to smallest cell pass time in your entire domain.

 

But I didn't quite understand this "please check if the moment report definitions and that from the "Forces" in Results section match or not." 

What I meant is, since you are calculating Cm (Moment Coefficient) from Fluent, check for only the Moment which is available in Results > Forces > Moments and the one which is also available in Report Definitions > New > Force Report > Moment. This way, you can cross check both the values for any discrepancy.

[Essence]

And one more thing. Can you let me know what is maximum percentage difference you are okay with, for the values of Cm for different CFL numbers?

[mohalinasre2001]

My problem is that I can't have the cell pass time smaller than the timestep for all tip speed ratios.

Also, my main problem is why does my moment coefficient change in the first place after changing the CFL in solution controls?  And what should I choose for each TSR? Should I aim at increasing CFL or decreasing it? 
I am trying a CFL sensitivity study now. I tried 3 values (0.7,1,1.5). I am thinking of choosing the one with % difference < 5%. But is that how I should choose the CFL or are there some guidelines. And is it normal for CFL to affect my solution this much where changing it slightly can change the average moment coefficient by more than 15%?

Thank you so much!