[siw]

Hi,

There's a tutorial in CFX for flow through a pipe with a butterfly valve, which uses an unstructured mesh. So for mesh generation practice I'm trying to make an ICEM Hexa mesh for this type of geometry. Can anyone suggest a blocking topology for around the open valve?

I can never work blocking topologies out, goodness knows how anyone can learn this stuff

Thanks.

[BrolY]

Could you post a picture of the geometry ?

[PSYMN]

I used to give a lot of training classes before I got into product management and I was starting to formulate a theory about a "hexa" gene. It just seemed like some people just naturally took to hexa blocking and others could never wrap their heads around it...

In this case, the pipe requires an Ogrid along its length (with faces at the corners...

The butterfly valve can be imagined as in a sphere (so you could move it to any angle...) so we would usually put the sphere in the geometry and do it as two blocking files, one to the outside of the sphere and one within the sphere and then we could rotate the inner blocking to any angle...

If you want a single volume and you know your valve is mostly open, then you could imagine it as a second pipe perpendicular to the first... This can be easily blocked by splitting upstream and downstream of the valve and then putting an OGrid in that block. You could do it with our without faces on two sides of the pipe (pretend there is a perpendicular pipe and the butterfly is just a slice of it...)

Then you slice that imaginary perpendicular pipe to cut out the butterfly valve and put it into a solid part (or just delete those blocks).

If you are modeling the pin for the butterly valve, then that is another Ogrid in that direction...

Best regards,

Simon

[PSYMN]

Oh yes, to be clear...

When doing the Ogrid for the valve or the pin, you will need the ogrid to start around the object to capture boundary flow. However, for the valve, which has mesh on its side also, you will need the central Hgrid to be well within the valve where it is not associated to any curves. Otherwise you will have 180 degree elements at the corners of the valve. The same would apply to the pin if you are modeling the solid for FSI...

So, make the Ogrid start outside the valve and extend to well inside the valve, then split it for the edge of the valve and associate that...

Get it?

If not, start down the path and send pics. It is easier to steer a moving car.

[Far]

How to draw the sphere ...

[PSYMN]

There are options under create surfaces to create primitives such as a sphere or cube... Usually, I create two points first (to be the "poles" of my sphere) and then select them during the sphere creation process...

Put the sphere in a new part, something like "INTERFACE"...

[Far]

Originally Posted by PSYMN

There are options under create surfaces to create primitives such as a sphere or cube... Usually, I create two points first (to be the "poles" of my sphere) and then select them during the sphere creation process...

Put the sphere in a new part, something like "INTERFACE"...

Previously I understood that you are talking in terms of the blocking.

Creating the sphere in the circular pipe would create the single point contact with the pipe? Circular hinge will be the part of the that sphere or not? Where this sphere should be created?


Here I have made the blocking for butterfly vavle geometry from the CFX tutorial.

https://dl.dropbox.com/u/68746918/project5tin.zip

[Far]

Originally Posted by PSYMN

There are options under create surfaces to create primitives such as a sphere or cube... Usually, I create two points first (to be the "poles" of my sphere) and then select them during the sphere creation process...

Put the sphere in a new part, something like "INTERFACE"...

Should I make the two fluids : fluid1 and fluid2 and then make the shared wall?


Here is the blocking after making the o-grid around the valve and hinge, but quality is no good.

What is main idea/trick behind blocking the such cases?

https://dl.dropbox.com/u/68746918/Bu...e_project6.rar

[Far]

I read about the FE modeller that is very efficient to get the good quality geometry from the mesh file. Used following steps:

1. Import mesh file (.def in my case)
2. Skin the mesh using the option "curves"
3. Create the initial goemetry
4. Convert to "Paralsolid"

But got the problem in parasolid, where geometry is distorted (last pic). Any hint please...

[snpradeep]

I think starting with quarter geometry is a good idea..Once you have made blocking for quarter section properly, delete unwanted blocks and generate a quality mesh..Later copy the blocking for full model and reassociate and merge nodes if required..This can be quite effective..But only problem with this is it will lead to lot of splits..

[snpradeep]

Originally Posted by BrolY

Could you post a picture of the geometry ?

Dear Mr. Boles,

Please find the eclosed pic. of the geometry..
I am sorry for the wrong post..

Regards,
Pradeep

[PSYMN]

When I suggested a sphere, I was assuming that you would have a non-conformal interface...

I was assuming you would add a sphere around the moving valve, call that INTERFACE. The sphere should not go to the wall of the tube, but should cut across the bar between the valve and the wall. Then you block the model in two parts.

One part is just the pipe assembly and the outside of the sphere... (blocking is just an ogrid in the pipe, delete the central block.)

The other part is the valve inside the sphere. This starts with an Ogrid to capture the sphere, then split to capture the valve. You will probably need to drill an ogrid tube thru the sphere for the bar in the valve, and the valve its self is just a couple splits...

Then output the mesh from both of these models and either bring them together in ICEM CFD, adjust the position of the valve and output to the solver, or load them into the solver separately. Set up the two zones and the interface. The one zone can rotate about an axis controlled by a udf or something like that ...

Get what I mean?

[Far]

Got it...

Can we get the good quality hexa mesh with this valve setting for full geometry without inserting interface i.e.. full conformal mesh.

[PSYMN]

Yes, you can do it as one piece, but you end up needing 2 or 3 different blockings to cover different ranges in the valve motion...

You may have one blocking where the valve is a split along the duct direction. As the valve closed, This would get deflected upward until the quality was considered too poor... THen you would switch to another blocking, perhaps one with a vertical split for the valve, or perhaps some intermediate and more advanced blocking would be needed first...

You would then select the blocking based on the angle of the valve.

I have seen it done in WB where the geometry system is connected to 3 mesh systems (each of which loads an ICEM CFD replay file corresponding to a specific topology) and then those all connect back to a solver. The solver selects which mesh to use based on the valve angle.

Best regards,

Simon