CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   Specified Blend factor =1 Vs Central Difference (https://www.cfd-online.com/Forums/cfx/25759-specified-blend-factor-1-vs-central-difference.html)

Kushagra May 1, 2008 17:31

Specified Blend factor =1 Vs Central Difference
 
From the CFX-doucment (Theory guide Release 11.0) page 289, the 'Specified blend factor =1' and the 'Central Difference' schemes for the advection term seem to be second-order-accurate. I am wondering what level of improvement should I expect to see in my results if I change one over the other?

Could some one help me getting the answer please? I am doing LES simulation and I know that the 'central difference' is recommended. But I am still interested in knowing the difference if I use 'specified blend factor=1' scheme.

Thanks, Kushagra

Glenn Horrocks May 1, 2008 18:34

Re: Specified Blend factor =1 Vs Central Differenc
 
Hi,

Central differencing is second order accurate but has stability problems in that it has no numerical dissipation. Hybrid with blend=1 uses a second order upwinding scheme which does have some numerical dissipation which increases stability.

Central differencing is sometimes used in LES and DNS applications so the dissipation is purely handled by the sub grid model. If you use an upwinding scheme then differencing adds some additional dissipation.

My PhD thesis did a LES study where I used a second order upwinding scheme to give the majority of the grid scale energy dissipation (It is chapter 6 where I do the study). http://hdl.handle.net/2100/248

Glenn Horrocks

Rogerio Fernandes Brito May 1, 2008 20:29

Re: Specified Blend factor =1 Vs Central Differenc
 
Hi Mr. Kushagra, take a look on this text:

Solver Control Setting:

The next icon of interest, , opens up the Solver Control window. Unlike other programs that have a ton of possible settings for solver controls, CFX is quite simple. The solver itself is quite robust so you don't have to worry about the multitude of options presented in other codes. Click on the Solver Control icon to open up the window shown below in Fig. 5. Let's begin with the Advection Scheme. It has three possible settings: High Resolution, Upwind and Specified Blend Factor. The upwind scheme is a simple and robust convective scheme but can be very diffusive. It can be used to start up difficult problems but in general should be avoided. The Specified Blend Factor allows the user to enter a value between 0 (upwind) and 1 (2nd Order), this will fix the order of the scheme. The High Resolution option will attempt to set the order of the scheme as high as possible automatically while keeping the solution bounded everywhere. The High Resolution scheme is what I would recommend for users. I will go into more details about the advection schemes in future articles but for the most part, stick to the High Resolution scheme. Under Convergence Control, use the Auto Timescale for Timescale Control and also leave the value of 100 for Max Iterations. Keep the Length Scale Option set to Conservative. Change the Residual Type from RMS to MAX and leave the Residual Target at 1.E-04. A residual value of 1.E-04 for the RMS setting is good for obtaining a faster solution and to get an idea of how the flow is developing. Setting the value of 1.E-04 for the MAX value gives a tighter convergence and is pretty good for extracting quantitative results for most problems. Click Ok to accept the settings. The next icon, , is useful for saving information during the solution of a model. You can save as much or as little information as you want and you also have several options to control when you save the information. We will skip this icon for this example. Step 7: Saving the Simulation The final icon on the horizontal row of icons, , is for writing out the solver file. The files that are written are MixBox.def, MixBox.cfx and the MixBox.gtm. The MixBox.cfx file contains the information entered during the CFX-Pre session and the MixBox.gtm contains the grid information. Both of these are combined to form the MixBox.def file. The *.def file is used in CFX-Solver. If you are trying to save space on your computer, you can delete the MixBox.cfx and MixBox.gtm files and just keep the MixBox.def file. Click on the icon to get the Write Solver File window as shown in Fig. 6. The name of the model is automatically used in the File name: input window. You can change the name if you want. Under Operation, the setting is such that the solver manager will automatically start up and use the new MixBox.def file. Generally, I click on the Quit CFX-Pre check box to close the preprocessor but you can keep it open if you wish. You are now done! Click on the OK button to write out the solver file and shut down the preprocessor (if you selected the quit option for CFXPre). You may get a series of menus so just click on the Save & Quit button. The CFX-Solver Manager window will open up as shown in Fig. 7. All you have to do is to click on the Start Run button and you will be on your way to solving the model. You can rest assured that if you need more options, such as distributed processing and want. Under Operation, the setting is such that the solver manager will automatically start up and use the new MixBox.def file. Generally, I click on the Quit CFX-Pre check box to close the preprocessor but you can keep it open if you wish. You are now done! Click on the OK button to write out the solver file and shut down the preprocessor (if you selected the quit option for CFXPre). You may get a series of menus so just click on the Save & Quit button. The CFX-Solver Manager window will open up as shown in Fig. 7. All you have to do is to click on the Start Run button and you will be on your way to solving the model. You can rest assured that if you need more options, such as distributed processing and memory resource control those are easily at your disposal. I will go into the details of setting some of the advanced options in the solver manager in future articles so if have an interest in using this option then you won't want to miss some of the future articles.

Source: December 15, 2006, The Focus Issue, 54.

Kushagra May 2, 2008 12:05

Re: Specified Blend factor =1 Vs Central Differenc
 
Thanks so much Glenn and Rogerio. This was really very helpful. Regards, Kushagra

Rogerio Fernandes Brito May 2, 2008 13:14

Re: Specified Blend factor =1 Vs Central Differenc
 
You´re welcome!


All times are GMT -4. The time now is 19:53.