Minor bug in cellQualityC

lr103476 · February 13, 2009, 07:15

Dear all,

I observed for 2D cases that the nonOrthogonality() and faceNonOrthogonality() functions in cellQuality can cause a 'floating points exceptions'. This occurs in the calculation of cosDDotS = Foam::acos((d & s)/(mag(d)*magS + VSMALL)), apparently VSMALL is too small and indeed, when I change this in SMALL, everything works fine again.

Regards, Frank

henry · February 13, 2009, 07:45

The problem is that for very small geometries you will find that SMALL is too big. I guess we will need to make VSMALL a bit bigger, could you find out how much bigger VSMALL has to be for your case?

Thanks

H

lr103476 · February 13, 2009, 08:12

I already did that. in my case VSMALL need to be 1e-16, so almost SMALL.

The case is not very small, just a moving 2D block on a 2D domain which is 25 squared. Overall domain bounding box (-12.5 -12.5 -0.1) (12.5 12.5 0.1).

Frank

mattijs · February 13, 2009, 08:24

Is the problem in acos getting a number >1 or in the division?

Could you try something like

Foam::acos(min(1.0, (d & s)/(mag(d)*magS + VSMALL)))

gschaider · February 13, 2009, 08:28

Just my two cents: the point in that VSMALL constant is to avoid division by zero, and get zero if (d&s) is zero, is it? How about replacing it with max(VSMALL,SMALL*max(mag(d),mag(s))) or something similar that would correlate it to the local sizes. But I havn't looked at the full source so I may be talking nonsense

Bernhard

lr103476 · February 13, 2009, 08:28

Thanks Mattijs, this solved the problem.

Frank

February 13, 2009, 07:15	Dear all, I observed for 2	#1
lr103476 Senior Member Frank Bos Join Date: Mar 2009 Location: The Netherlands Posts: 340 Rep Power: 18	Dear all, I observed for 2D cases that the nonOrthogonality() and faceNonOrthogonality() functions in cellQuality can cause a 'floating points exceptions'. This occurs in the calculation of cosDDotS = Foam::acos((d & s)/(mag(d)*magS + VSMALL)), apparently VSMALL is too small and indeed, when I change this in SMALL, everything works fine again. Regards, Frank __________________ Frank Bos

February 13, 2009, 07:45	The problem is that for very s	#2
henry Senior Member Join Date: Mar 2009 Posts: 854 Rep Power: 22	The problem is that for very small geometries you will find that SMALL is too big. I guess we will need to make VSMALL a bit bigger, could you find out how much bigger VSMALL has to be for your case? Thanks H

February 13, 2009, 08:12	I already did that. in my case	#3
lr103476 Senior Member Frank Bos Join Date: Mar 2009 Location: The Netherlands Posts: 340 Rep Power: 18	I already did that. in my case VSMALL need to be 1e-16, so almost SMALL. The case is not very small, just a moving 2D block on a 2D domain which is 25 squared. Overall domain bounding box (-12.5 -12.5 -0.1) (12.5 12.5 0.1). Frank __________________ Frank Bos

February 13, 2009, 08:24	Is the problem in acos getting	#4
mattijs Senior Member Mattijs Janssens Join Date: Mar 2009 Posts: 1,419 Rep Power: 26	Is the problem in acos getting a number >1 or in the division? Could you try something like Foam::acos(min(1.0, (d & s)/(mag(d)*magS + VSMALL)))

February 13, 2009, 08:28	Just my two cents: the point i	#5
gschaider Assistant Moderator Bernhard Gschaider Join Date: Mar 2009 Posts: 4,225 Rep Power: 51	Just my two cents: the point in that VSMALL constant is to avoid division by zero, and get zero if (d&s) is zero, is it? How about replacing it with max(VSMALL,SMALL*max(mag(d),mag(s))) or something similar that would correlate it to the local sizes. But I havn't looked at the full source so I may be talking nonsense Bernhard __________________ Note: I don't use "Friend"-feature on this forum out of principle. Ah. And by the way: I'm not on Facebook either. So don't be offended if I don't accept your invitation/friend request

February 13, 2009, 08:28	Thanks Mattijs, this solved th	#6
lr103476 Senior Member Frank Bos Join Date: Mar 2009 Location: The Netherlands Posts: 340 Rep Power: 18	Thanks Mattijs, this solved the problem. Frank __________________ Frank Bos

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Minor buoyancy problem in Phoen2007	patti	Phoenics	3	December 1, 2007 19:43
Minor bug in CrankNicholsonDdtScheme	henry	OpenFOAM Bugs	5	May 22, 2007 12:52
Minor compilation problem with Gcc42	schmidt_d	OpenFOAM Bugs	1	May 9, 2007 14:11
Minor loss coefficients	Nick	Main CFD Forum	3	October 7, 2004 12:28
minor loss k values	clifford bradford	Main CFD Forum	1	August 2, 2000 17:58