problem coming during prism creation in icemcfd

ashish agrahari · August 5, 2010, 02:22

hey frnds...this is the first time i m writing thread..i ve done unstructured mesh in icemcfd for full aircraft and after that i tried to create 20 prism layers but it ended with one problem showing "null dir in move_v_to_levelset"...it wrote upto 8 prism layers but after that its showing the above mentioned prob...can anybody help me regarding this......as soon as possible...waiting eagerly for ur replies....

thank you

ashish agrahari

PSYMN · August 5, 2010, 09:32

u don't know such an easy answer? - Karma

On CFD Online, you help others with questions that are easy for you and people will help you with questions that are easy for them...

Think about it because Karma 's a bitch.

In this case, prism is complaining that it can't figure out the direction to grow in. I usually see this in a situation where a node is located between two overlapping plates... Imagine a situation where two steel bars with square cross sections cross. They touch and form a square contact patch. At the corners of the contact patch, the normal directions are not clear (there are two possibilities and neither is right... You probably have a configuration like that in your model. You can fix it by either adjusting your geometry so this doesn't happen (add a fillet or something between the bars) or just move the corner node a little off the corner (this may result in fairly poor quality, but it will work).

I can help more if you can be more specific about the results of the error, etc. What problem is it showing?

ashish agrahari · August 6, 2010, 01:04

hey thanks for replying...i got u but adjusting nodes of full aircraft is very difficult and it will take lot of time ...can u suggest any other way..if it is there...is there any problem in prism settings given at the starting..?

ashish agrahari · August 6, 2010, 01:08

i got 10 prism layers but getting surface orientation problem...while trying for 20 layers it showing "null dir in move_v_to_levelset"...it is d same prob showing again ...my mesh(unstructured _quality is very low...is there anything to do with that...but no errors are coming ...

ashish agrahari · August 6, 2010, 01:43

i got 10 prism layers but getting surface orientation problem...while trying for 20 layers it showing "null dir in move_v_to_levelset"...it is d same prob showing again ...my mesh(unstructured _quality is very low...is there anything to do with that...but no errors are coming ...

PSYMN · August 6, 2010, 08:39

If you got 10 layers, then it wasn't the geometry issue I was thinking of.

But yes, you should smooth as much as possible to get the surface mesh quality as high as possible before running prism. I also usually run a few iteration with Laplace since it helps smooth out any volume transitions. Once prism is grown, it restricts the smoother (to prevent twisting prisms) and it becomes very difficult to smooth, so smooth with tetra while it is easy.

If you have inside curvature (prisms growing in toward a tighter space) you may need to reduce the layer height, etc just to fit them in...

I would need to see a screen shot for a better guess of what you could do to fix it, can you run a cut plane thru the worst area and post the image?

Seattle_Hokie · August 11, 2010, 10:31

I have run into these errors before. No fun at all. Try a few things.

1) If there are any surfaces that are parallel, or nearly parallel then your 20 prism layers may run into each other. In these regions, you would want to select less than 20 layers, or limit the height

2) Instead of extruding 20 layers, try to extrude 1 layer that is the full height of the intended 20 layers. You can then split this large prism into 20 smaller prisms. This usually ends up yielding a better grid in the end anyway

3) Sometimes when fillets or corners have insufficient grid, the initial tet mesh can overlap. Although tedious, these errors must be corrected. You have to go looking for them, destroy the cells effected and patch it back together manually.

4) Along with 3, you could start out with a defeatured aircraft with less detail, this should help 1 2 and 3.

5) There is an option to write out the mesh after each level. If you do this, and load the mesh with 10 layers, you may see upon close inspection where the mesh will fail if an additional layer is added.
5) If none of this works, try, try again...

PSYMN · August 11, 2010, 15:51

A lot of ICEM CFD users like that 1 layer and split method that Seattle_Hokie suggested. Personally, I prefer to grow at least 3 layers and then split... So for 20 layers, instead of one layer split into 20, I might do 3 layers, each split into 7 and then redistributed, or the other way round, 7 layers split into 3. It is a compromise that gives a better result (in my opinion).

The prism tool has lots of smarts in it that help prisms turn corners, etc. You lose all these benefits when you grow a single layer. If you are not interested in those benefits, then I suggest you just turn off the layer smoothing, etc. and run your larger number of layers. It is very quick without all that fancy stuff.

August 5, 2010, 02:22	problem coming during prism creation in icemcfd	#1
ashish agrahari New Member ashish Join Date: May 2010 Posts: 6 Rep Power: 15	hey frnds...this is the first time i m writing thread..i ve done unstructured mesh in icemcfd for full aircraft and after that i tried to create 20 prism layers but it ended with one problem showing "null dir in move_v_to_levelset"...it wrote upto 8 prism layers but after that its showing the above mentioned prob...can anybody help me regarding this......as soon as possible...waiting eagerly for ur replies.... thank you ashish agrahari

August 5, 2010, 09:32	Trouble calculating Prism direction...	#2
PSYMN Senior Member Simon Pereira Join Date: Mar 2009 Location: Ann Arbor, MI Posts: 2,663 Blog Entries: 1 Rep Power: 47	u don't know such an easy answer? - Karma On CFD Online, you help others with questions that are easy for you and people will help you with questions that are easy for them... Think about it because Karma 's a bitch. In this case, prism is complaining that it can't figure out the direction to grow in. I usually see this in a situation where a node is located between two overlapping plates... Imagine a situation where two steel bars with square cross sections cross. They touch and form a square contact patch. At the corners of the contact patch, the normal directions are not clear (there are two possibilities and neither is right... You probably have a configuration like that in your model. You can fix it by either adjusting your geometry so this doesn't happen (add a fillet or something between the bars) or just move the corner node a little off the corner (this may result in fairly poor quality, but it will work). I can help more if you can be more specific about the results of the error, etc. What problem is it showing?

August 6, 2010, 08:39	Hmmm...	#6
PSYMN Senior Member Simon Pereira Join Date: Mar 2009 Location: Ann Arbor, MI Posts: 2,663 Blog Entries: 1 Rep Power: 47	If you got 10 layers, then it wasn't the geometry issue I was thinking of. But yes, you should smooth as much as possible to get the surface mesh quality as high as possible before running prism. I also usually run a few iteration with Laplace since it helps smooth out any volume transitions. Once prism is grown, it restricts the smoother (to prevent twisting prisms) and it becomes very difficult to smooth, so smooth with tetra while it is easy. If you have inside curvature (prisms growing in toward a tighter space) you may need to reduce the layer height, etc just to fit them in... I would need to see a screen shot for a better guess of what you could do to fix it, can you run a cut plane thru the worst area and post the image?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
a problem of icemcfd in linux	anne	Main CFD Forum	0	March 12, 2008 04:55
How to mesh axisymmetric 2D problem using prism...	sameer	CFX	2	March 7, 2006 13:14
need help on Prism Mesh Generation	Pete	CFX	0	December 6, 2004 02:11
extremely simple problem... can you solve it properly?	Mikhail	Main CFD Forum	40	September 9, 1999 09:11
Is this problem well posed?	Thomas P. Abraham	Main CFD Forum	5	September 8, 1999 14:52

August 6, 2010, 01:04		#3
ashish agrahari New Member ashish Join Date: May 2010 Posts: 6 Rep Power: 15	hey thanks for replying...i got u but adjusting nodes of full aircraft is very difficult and it will take lot of time ...can u suggest any other way..if it is there...is there any problem in prism settings given at the starting..?

August 6, 2010, 01:08		#4
ashish agrahari New Member ashish Join Date: May 2010 Posts: 6 Rep Power: 15	i got 10 prism layers but getting surface orientation problem...while trying for 20 layers it showing "null dir in move_v_to_levelset"...it is d same prob showing again ...my mesh(unstructured _quality is very low...is there anything to do with that...but no errors are coming ...

August 6, 2010, 01:43		#5
ashish agrahari New Member ashish Join Date: May 2010 Posts: 6 Rep Power: 15	i got 10 prism layers but getting surface orientation problem...while trying for 20 layers it showing "null dir in move_v_to_levelset"...it is d same prob showing again ...my mesh(unstructured _quality is very low...is there anything to do with that...but no errors are coming ...

August 11, 2010, 10:31		#7
Seattle_Hokie New Member Scott Join Date: Aug 2010 Posts: 2 Rep Power: 0	I have run into these errors before. No fun at all. Try a few things. 1) If there are any surfaces that are parallel, or nearly parallel then your 20 prism layers may run into each other. In these regions, you would want to select less than 20 layers, or limit the height 2) Instead of extruding 20 layers, try to extrude 1 layer that is the full height of the intended 20 layers. You can then split this large prism into 20 smaller prisms. This usually ends up yielding a better grid in the end anyway 3) Sometimes when fillets or corners have insufficient grid, the initial tet mesh can overlap. Although tedious, these errors must be corrected. You have to go looking for them, destroy the cells effected and patch it back together manually. 4) Along with 3, you could start out with a defeatured aircraft with less detail, this should help 1 2 and 3. 5) There is an option to write out the mesh after each level. If you do this, and load the mesh with 10 layers, you may see upon close inspection where the mesh will fail if an additional layer is added. 5) If none of this works, try, try again...

August 11, 2010, 15:51		#8
PSYMN Senior Member Simon Pereira Join Date: Mar 2009 Location: Ann Arbor, MI Posts: 2,663 Blog Entries: 1 Rep Power: 47	A lot of ICEM CFD users like that 1 layer and split method that Seattle_Hokie suggested. Personally, I prefer to grow at least 3 layers and then split... So for 20 layers, instead of one layer split into 20, I might do 3 layers, each split into 7 and then redistributed, or the other way round, 7 layers split into 3. It is a compromise that gives a better result (in my opinion). The prism tool has lots of smarts in it that help prisms turn corners, etc. You lose all these benefits when you grow a single layer. If you are not interested in those benefits, then I suggest you just turn off the layer smoothing, etc. and run your larger number of layers. It is very quick without all that fancy stuff.