[BahaZero]

Hi,

I'm trying to figure out if it's possible to create a structured hex mesh without using the blocking strategy in ICEM (it's so messy and tedious!). I can create a structured hex mesh in Workbench Mesh using the sweep function and export it as a .prj to read into ICEM.

My question is: Is there anything special that ICEM does to the output file that requires it to be blocked within ICEM for the CFD solver to know that it's a structured grid?

The reason why I still need ICEM is to extrude the inflation layers since the WB mesher cannot do that reliably yet. Does anyone know if this grid will stay structured?

Thanks in advance.

[macfly]

If the geometry have plane surfaces:

- create a blocking that is cut at every surfaces

- create parts with the blocks where you want to have internal boundaries (see top right image if I'm not clear)

- mesh the blocking

- convert to unstruct mesh

- there you go, you have a 'structured' mesh without the hassle of edge associations

[diamondx]

multizone is your only option. you need a strong and perfect geometry for that.
when initializing a blocking, don't select 3d block, go for surface blocking, give it the surface you want to block and see how it worked out for you

[Far]

Originally Posted by macfly

If the geometry have plane surfaces:

- create a blocking that is cut at every surfaces

- create parts with the blocks where you want to have internal boundaries (see top right image if I'm not clear)

- mesh the blocking

- convert to unstruct mesh

- there you go, you have a 'structured' mesh without the hassle of edge associations

But it is not the practical way to deal with complex problems in ICEM

[macfly]

Far, look at my right-angled geometry and the corresponding blocking, I'm pretty glad that I don't have to perform the edge associations!

[diamondx]

If the geometry have plane surfaces:

- create a blocking that is cut at every surfaces

- create parts with the blocks where you want to have internal boundaries (see top right image if I'm not clear)

- mesh the blocking

- convert to unstruct mesh

- there you go, you have a 'structured' mesh without the hassle of edge associations

Still i call it blocking...

[Far]

Originally Posted by diamondx

Still i call it blocking...

ya right.

[macfly]

I'm presuming that the 'messy and tedious' that BahaZero was referring to is mainly the edge associations (from my own experience). So I suggested a method without edge associations that he may not know of, and may constitute a solution to his 'messy and tedious' situation.

my own experience with blocking:
time for geometry cutting = 10 % of the blocking process
time for edge associations = 90 % of the blocking process

When I discovered the 'blocking without association' technique, it changed my world for right-angled geometries.

[Far]

If you use some smart buttons like : Auto Associate, group curves + tangential and associate the major edges in beginning and you can save alot of time.

In my experience either your blocking is correct or totally wrong depending on how you imagined it. So my time is spent mostly in getting the right blocking. Once it is done then time is spent in making the vertex on right place and off-course association (but time can be reduced as discussed above).

[BrolY]

My solution to avoid spending too much time on edge association is the following

1) Do NOT group all the curves to one part. I've seen a lot of project on this forum which have one part for curves, one part for points etc.. From my experience, it's not a good idea.
2) Use the index blocking to show only the edges you want to associate, and use the option "all visible" to select all the edges you need in one click. Show only the parts you need and associate edges to curves.
3) I have never associated all the edges of the blocking, only the most important ones.

At the end, it takes you only 10% of your time to associate, so you keep 80% to think to a better strategy

[Far]

) Do NOT group all the curves to one part. I've seen a lot of project on this forum which have one part for curves, one part for points etc.. From my experience, it's not a good idea.

I follow this strategy for 3d cases. You should have separate elements in boundary conditions otherwise you will get mixed boundaries in BOCO specification panel. For example you should have shells in inlet part and should not have volume, point or line elements. (Although it is not required by the Fluent, other codes may require it)

[BrolY]

It means you need a proper geometry before starting blocking
If you put the inlet curves/surfaces in your outlet part, it should results in weird results at the end

[Far]

Originally Posted by BrolY

It means you need a proper geometry before starting blocking
If you put the inlet curves/surfaces in your outlet part, it should results in weird results at the end

hmmm...........

[Far]

Originally Posted by macfly

Far, look at my right-angled geometry and the corresponding blocking, I'm pretty glad that I don't have to perform the edge associations!

Did you output that file to Fluent? Were you able to apply the boundary conditions? what will happen if these were circles instead of square....

[macfly]

Did you output that file to Fluent? Were you able to apply the boundary conditions? what will happen if these were circles instead of square....

Yes, it works in Fluent, check mesh is fine and simulations are running.

I've specified many times in this thread that it works for right-angled geometries. Or can be extended to geometries with plane surfaces only.

I've experimented with the 'smart' buttons like Auto Associate before, and their 'smartness' didn't impress me at all. There is always a flaw, if not many. I also work a lot with the blocking index and it doesn't improve the auto associations. Actually, I find ICEM pretty dumb when it can't associate a block edge to a geometry edge that is at the exact same coordinates. All those comments are based on my experience with the geometry I've showed in post #5. I'm working with Ansys 14.0, maybe the auto association algorithms have been improved in 14.5?...

[Far]

wrong thread. see next post

[Far]

Hmmm. You are sure that the blocks you cut are of same size as cubes you have? I mean you are may be compromising on geometry details. For example if you have cube of dimension 3*3*3 and you are may modeling it as 3.1*2.9*2.8 or 3*3*2.2.

[macfly]

Ducts, furnaces, room with simplified furniture, simplified buildings, etc.

I cut the blocks at geometry points (Split Method => Prescribed point), the points that are at the end of geometry curves, so blocks are cut exactly over geometry.

[BrolY]

Originally Posted by macfly

Yes, it works in Fluent, check mesh is fine and simulations are running.

I've specified many times in this thread that it works for right-angled geometries. Or can be extended to geometries with plane surfaces only.

I've experimented with the 'smart' buttons like Auto Associate before, and their 'smartness' didn't impress me at all. There is always a flaw, if not many. I also work a lot with the blocking index and it doesn't improve the auto associations. Actually, I find ICEM pretty dumb when it can't associate a block edge to a geometry edge that is at the exact same coordinates. All those comments are based on my experience with the geometry I've showed in post #5. I'm working with Ansys 14.0, maybe the auto association algorithms have been improved in 14.5?...

About the index blocking and auto association, I have never said that the first one would improve the second one. Only that it's really fast to associate edges when using index blocking. Considering the fact you don't have to associate all the edges, I have never spent more than 10% of my time on it.

[Far]

Ok......