Grid independence

hbf · July 20, 2009, 09:58

Hello everybody,

I'm performing (or a least trying to perform) a grid independence study.

Are the results with the highest amount of cells the most "accurate" ones? Does the shape of the cells influence this rule of thumb?

If I increase the number of cells in one direction, it is possible that I will get a aspect-ratio minimun at the point where the cells have similar dimensions with the other two directions. By refining further from this point, the aspect ratio will increase. Which one is the "accurate" solution??

I'd appreciate if someone could help me with this since I'm a bit lost when interpreting this results.

Thanks!

ak6g08 · July 20, 2009, 11:31

hi,

first thing you should do is have a uniform mesh, and start increasing the cell number by a factor each time (uniformly), do this until the solution stops changing by a lot (5%)...the quantity you should monitor depends on what type of flow it is really...but I suppose as a start you could just have a look at the value of say maximum velocity,temperature (if applicable), and maybe pressure...if it is an isotropic problem or homogenous along a certain plane then look at the mean velocity (along with other mean values) along that plane and see how they change according to cell number, it is probably better if you look at how a statistic changes rather than a single value.

You should then refine in the areas of interest...boundary layers, areas where you expect things like recirculation, sharp changes in pressure/velocity etc...as far as aspect ratio is concerned...if it is a real issue, FLUENT will tell you when you do the grid check (the check will fail), always check your grid before running a simulation (look in the manual to see how to do this)...and as a rule of thumb the value you monitor (according to grid) should not vary by more than 5% from one grid choice to the next.

I hope that helps

hbf · July 21, 2009, 05:50

Thank you very much for the thorough answer.
If I plot the resulting pressure drop, I always get a minimum more or less where I also get an aspect-ratio minumum. For coarser as well as for finer grids I get higher pressure drops. If I apply the 5% rule, I will get two possible grids, one coarser, one finer. This is the reason why I'm not so sure which one I should use.

Farhat · July 21, 2009, 06:42

If you have experimental data you can easily verify because such results will be grid free and what you want. Also for grid independence start from very coarse grid and use first order equations, i.e keep things simple.

July 20, 2009, 09:58	Grid independence	#1
hbf New Member Join Date: Mar 2009 Posts: 5 Rep Power: 17	Hello everybody, I'm performing (or a least trying to perform) a grid independence study. Are the results with the highest amount of cells the most "accurate" ones? Does the shape of the cells influence this rule of thumb? If I increase the number of cells in one direction, it is possible that I will get a aspect-ratio minimun at the point where the cells have similar dimensions with the other two directions. By refining further from this point, the aspect ratio will increase. Which one is the "accurate" solution?? I'd appreciate if someone could help me with this since I'm a bit lost when interpreting this results. Thanks!

July 20, 2009, 11:31		#2
ak6g08 Member Akour Join Date: May 2009 Posts: 79 Rep Power: 17	hi, first thing you should do is have a uniform mesh, and start increasing the cell number by a factor each time (uniformly), do this until the solution stops changing by a lot (5%)...the quantity you should monitor depends on what type of flow it is really...but I suppose as a start you could just have a look at the value of say maximum velocity,temperature (if applicable), and maybe pressure...if it is an isotropic problem or homogenous along a certain plane then look at the mean velocity (along with other mean values) along that plane and see how they change according to cell number, it is probably better if you look at how a statistic changes rather than a single value. You should then refine in the areas of interest...boundary layers, areas where you expect things like recirculation, sharp changes in pressure/velocity etc...as far as aspect ratio is concerned...if it is a real issue, FLUENT will tell you when you do the grid check (the check will fail), always check your grid before running a simulation (look in the manual to see how to do this)...and as a rule of thumb the value you monitor (according to grid) should not vary by more than 5% from one grid choice to the next. I hope that helps __________________ akour

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
unstructured mesh grid independence for Fluent	Shane Schouten	FLUENT	0	October 11, 2006 17:50
Grid independence check	Mattis Voss	FLUENT	1	November 23, 2005 11:30
Grid independence check	Mattis Voss	FLUENT	0	November 23, 2005 08:31
Grid Independence problem for simple pipe geometry	chanchala	FLUENT	13	November 20, 2003 09:24
Combustion Convergence problems	Art Stretton	Phoenics	5	April 2, 2002 06:59

July 21, 2009, 05:50		#3
hbf New Member Join Date: Mar 2009 Posts: 5 Rep Power: 17	Thank you very much for the thorough answer. If I plot the resulting pressure drop, I always get a minimum more or less where I also get an aspect-ratio minumum. For coarser as well as for finer grids I get higher pressure drops. If I apply the 5% rule, I will get two possible grids, one coarser, one finer. This is the reason why I'm not so sure which one I should use.

July 21, 2009, 06:42		#4
Farhat New Member Farhat Rasool Khan Join Date: Jul 2009 Posts: 5 Rep Power: 17	If you have experimental data you can easily verify because such results will be grid free and what you want. Also for grid independence start from very coarse grid and use first order equations, i.e keep things simple.