CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ANSYS Meshing] Concept of conformal and non-conformal interface (http://www.cfd-online.com/Forums/ansys-meshing/105880-concept-conformal-non-conformal-interface.html)

Anna Tian August 11, 2012 11:42

Concept of conformal and non-conformal interface
 
Hi

Could you tell me what's the concept of conformal and non-conformal interface? I tried to search for it online, but I didn't find any good answer. Thanks!

Far August 11, 2012 14:15

Both are interfaces and there is averaging of data on both sides before transferring to other side.

Conformal interface : An interface with equal no of nodes on both sides.

Non Conformal interface : An interface which does not have equal no of nodes on both sides

flotus1 August 13, 2012 02:49

Just a minor correction here:

A conformal interface is not defined by an equal number of nodes on both sides.
Instead, the meshes on both sides of the interface have to be conformal:D, which means that every node on one side of the interface can be matched with a node on the other side of the interface with a very low tolerance.
The advantage is that there is no (additional) interpolation necessary at a conformal interface. This makes the computation faster and more accurate.

Consequently, a non-conformal interface can be set at two faces with the same number of nodes.

sagarmore May 7, 2015 10:17

non conformal interface
 
Since data across non conformal interface is averaged and transferred is there significant loss in accuracy due to interpolation error?If error is large then which approach is better
1.domain divided in structured meshes joined with non conformal interface.or,
2.unstructured mesh in whole domain without need of interface
I am using multiblock method for horizontal axis wind turbine mesh which is separated in 3 multiblocks and merged

ghost82 May 7, 2015 11:52

Quote:

Originally Posted by flotus1 (Post 376743)
Just a minor correction here:

A conformal interface is not defined by an equal number of nodes on both sides.
Instead, the meshes on both sides of the interface have to be conformal:D, which means that every node on one side of the interface can be matched with a node on the other side of the interface with a very low tolerance.
The advantage is that there is no (additional) interpolation necessary at a conformal interface. This makes the computation faster and more accurate.

Consequently, a non-conformal interface can be set at two faces with the same number of nodes.

So, by your definition, if I have an interface which splits a moving from a static zone, even if I have the two adjacent surfaces with the same number of nodes and at the same distance, this is a non-conformal interface, as when sliding the nodes will not overlap.
So, a conformal interface can be obtained only for an interface which connects 2 static zones?

Thanks

Aohnia March 11, 2016 10:04

Sagarmore's question is very interesting, it's a pity that no one seemed to answer it. Did you manage to get an answer ?

sagarmore March 12, 2016 01:36

unstructured vs. non conformal mesh approach
 
No I couldn't get answer.I prefer non conformal mesh approach though.

scipy March 12, 2016 18:39

I'll take a stab at this...

Since hexa meshes are usually of far superior quality compared to tetra meshes (at least in cases where flow direction is aligned with direction of elements), I've always opted for an option to use hexa (conformal structured or non-conformal with interfaces) instead of tetra. My typical usage was to create a structured hexa wind tunnel and then create hexacore meshes for volumes containing vehicles (race cars). Now, as far as non-conformal interfaces go ANSYS said the following: interfaces work without any problems, in fact they're used in such complex cases involving heat transfer and enthalpy and other crap with turbomachinery examples (where interfaces are common place) that I've never worried about "lack of accuracy". However, user's guide and other documentation says that large jumps in element sizing on the interfaces can lead to more time needed for the interpolation and some errors.. so, to deal with this I've usually tried to match element sizes on both sides of the interface. Even though I could usually match this so that it looks as if it could be a completely conformal mesh from a far, I couldn't actually merge the meshes (fuse) because the number of nodes didn't really exactly match (hexacore had that 1 layer of tetras to transition from a floor prism layer to the hexa volume - which screwed up the node count on those faces).

All in all, any loss in accuracy over the interface was never a concern for me, especially when you take into consideration the huge leaps in accuracy obtained by using a hexa mesh for the whole wind tunnel. The wake effects were really well captured even meters behind the vehicle while tetra meshes looked all smudged and diffused even very close to the car (and let's not even talk about levels of refinement needed to get a good wake resolution with a tetra mesh).

My $0.02 = hex meshes + interfaces for the win. :p

Fabio88 April 27, 2016 05:52

Hi,

I am also dealing with interfaces for the first time.
My case is a wheel study.
I have created my wheel and outside it a cylindrical interface to allow the definition of a moving reference frame.
I was using now tet mesh, but do you think is better to use hexacore + hex outside it?

Thanks a lot,

Fabio

scipy April 28, 2016 04:59

If you could post a picture/screenshot things might be clearer. Anyway.. if your wheel is sitting on the ground and you want it to "rotate" and are conducting an unsteady simulation then it'd be enough to just separate the volume inside the rim (including the spider) so it can rotate freely - the rest of the wheel (tire & rim part) which don't really affect the rotation can just be given rotating wall boundary conditions without using the MRF & moving mesh.

As far as hex/hexacore goes.. yes. This case is no different, maybe just keep the inside volume of the rim & spider pure prism + tetra since flow in there is not really going to be aligned with the hex element direction, but the rest of the domain can be hexacore for sure. Same principle applies if you want the whole tunnel to be pure hexa (to avoid floor prisms & triangles), you just add more interfaces for the two rectangular domains (wheel domain and tunnel domain).

Fabio88 April 29, 2016 04:03

1 Attachment(s)
Hi Scipy,

thanks for replying.
I am actually evaluating two wheels, one disc wheel and one with spokes.

My first idea, as you can see in the picture, was to create a displacement between the wheel+interface and the floor, and using the MRF on the interface.
Later I will create another cylindrical interface to allow investigation of different yaw angles.

My first mesh attempt, was using tetra+prism. Now I started to watch some tutorial about hexcore to understand how to use them.

Your method, splitting the inner volume from rim/wheel, sounds also very interesting. I just need to figure out how to split in two parts the wheel: I do not know if it is better to do at CAD level (PTC Creo) or when I import it in design modeler. Do you have any suggestion?

Thanks a lot for the help.

Fabio

scipy April 29, 2016 06:39

Yeah.. you don't need all of this for a pure disk wheel. You just need a wall boundary condition of "rotating wall" and can run this in a simple steady state case. Also, the wheel needs to be touching/intersecting the ground :)

Once you go for a spoked wheel, you just need to separate the inner volume and have that rotate, the rest stays the same.


All times are GMT -4. The time now is 07:17.