CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [GAMBIT] Savonius 2D "S" Profile (

Rich1979 April 13, 2010 12:51

Savonius 2D "S" Profile
Hi all,
i am starting a study about the characterization of a Savonius wind generator.
The first thing i have to do is the creating the two 2d profiles with gambit.They are semi-circumference with an overlap.I'm now searching some acceptable values for the geometric dimensions,and then i'll start with gambit.
This is what i should report on gambit :

I'd like to know how to begin : which approch have i to use?have i to use some particular solution?
I have read all the gambit tutorials but i am not sure how to proceed...
Have i simply to draw 2 semi circunferences?This just for the geometry,meshing will follow later...
Thanks very much for reading,i need help :).

Edit:if you don't see the image this is the direct link

-mAx- April 14, 2010 14:01

1 Attachment(s)
Just create the 2 semi circles. Adjust the position of one circle with Move option.
Then create the outter disk (surface)
Then use split surface tools: split the outter disk (surface) with each semi-circle (edge)
Attachment 2959
Regarding mesh part, for sure you will have problem at the intersection between outter circle and semi-circles, as they are tangential.
It can be easy resolved with some splits

Rich1979 April 15, 2010 04:12

Thx for answering MaX!
Sorry to say but my teacher yesterday decided to change the geometry of the savonius,attacching the 2 profiles (so the overlap now is 0).
And he also told me to consider a thickness of 5mm ,considering a radius for each profile of 1,20 m.
So i have to :
1)Draw the geometry
2)Setting up the integration domain
3)Mesh the domain

For the geometry i have to draw something like this:

but i am having troubles,because i am not confidential with autocad and by gambit i'm trying to find a way to make it,by the help of the grid tool,even if it is difficult.
Considering the integration domain,in 2d i was thinking about a circle of 10 meters of diameter (hoping the fluid trail will not be too long) and finally for the mesh the teacher told me to use an un-structured mesh,but i don't know how to proceed.
Yes..i am very very noob..but i hope to improve my knowlowdge about cfd.
Thx again for your time and help!

Edit: I Succeeded in making geometry,just using create vertex and edges.Now geometry is ready and i have to proceed in making points 2) and 3).

-mAx- April 16, 2010 03:16

the 5mm thickness in comparison with radius 1,2m will give you some trouble with your mesh (but not severe)
I would work with 0 thickness, and consider your spline as a baffle.
you will have to pay attention with the BC, and enforce gambit setting the S-spline as wall (else it will consider as interior).
*Your S is constructed? but are they connected together? check if you don't have 2 superposed vertex at their intersection.
*Regarding the integration domain, create a disk (surface) with the right radius, and then split it with your spline.
Now the spline belongs to the surface and you can start meshing
But before meshing you have to make you clear with the location for your inlet and outlet.
Then decompose your domain into subdomains which can be mappable (for unstructured mesh)

Rich1979 April 19, 2010 02:17

Hi Max,
thx again for your advices :)
What do you mean by "spline" and "baffle"?Sorry but i am not english and i haven't found a traduction term which fits :(
About the inlet and outlet my purpose is to find,by the numerical simulation through gambit and fluent,the expression of the lift,drag and momentum coefficient in function of the incident angle "α".
So how do you is better to set inlet and outlet?
Thx as ever!

-mAx- April 19, 2010 02:35

the spline is just your 2 semi-circles.
Baffle is a thin wall (0-thickness)
Regarding your inlet and outlet, create for instance a straight pipe and rotate it with your given incidence.

Rich1979 April 19, 2010 03:31

Ok,this is what i've done about the geometry,even if i have to work with the 5mm thickness (if i worked with 0 thickness i would not know how to consider the effect of the thickness).
Now,following your advices,i have to split my domain with the spline.How do i do?i think it's not the same if there is thickness or not?Thx

-mAx- April 19, 2010 03:40

ok I understand now why you want to work with the thickness.
So forget what I said about baffle etc...
If you already built the curves (with the thickness), then you can do followings:
*create a surface by picking all the curves which describes your body of interest.
*Then create your disk (integration's domain)
*Split the disk with the surface body (choose Face)
Enable shadded modus and check if the interior of your body is empty (black)

Rich1979 April 19, 2010 05:27

Here it is my result :
Now i have to :
1)set the inlet and outlet
2)Decompose the domain into subdomain which could be meshed

About the inlet-outlet i think i will create,as you said,an straight pipe which i'll rotate with different incident angles (i think to use 15 od interval to divide the 360 domain).The dimensions of the pipe could be 10m x 10cm ?
About the meshing i have to understand first which mesh scheme fits better my needings,then decompose the domain in subdomain which could be meshed and finally meshing.
I hope to find soon a solution,thx as ever for your support.

-mAx- April 19, 2010 07:09

1 Attachment(s)
ok now delete the face representing your solid domain (ie: face1)
PS: can you use the attachment fonction on this forum for posting pcitures, etc...instead of megaupload, becasue this one opens "weird" pop-up windows --> I am at the office :rolleyes:
Attachment 3018

Rich1979 April 20, 2010 10:53

Hi Max,
I've finally finished the geometry,with some changes about the integration domain,that now is a circle of 20 meters diameter,since there are long trails.
I've splitted the domain with the "S" face and meshed the edges of the "S" and of the domain.
I have totally 6 edges on ths "S":2 semicircunferences (upper and lower) on each side(left and right),and 2 straight edge (thickness) at the the extremes of the face.
For the domain i've used a successive ratio with ratio 1 and an interval size of 125,6.
For the 4 semicircunferences i've used a successive ratio (double sided to have more cells nears the extremes of the face and lesser near the centre) with ratio 1.05 and interval count of 187(for the larger) and 189 (for the smaller).
I've choosen the ratio according to the number and the distribution of the nodeds i've seen on the screen.
For the 2 straight edges interval count 1,because i want my smaller cell will be of 5 mm (the measure of the thickness).
Assuming i've done right (but as always i'm not sure),i've now to mesh the face,and now there is a big problem since i don't know which elements,and type i have to use to obtain the best mesh (either structured or un-structured).I'am looking for some books or sites that could explain me the differences between the various mesh types used by gambit (map,submap,pave,tri primitive).Have you some ideas?
Thx as ever!

-mAx- April 21, 2010 01:24

You won't be able to mesh you domain in hexa since you don't decompose it into subdomains, which will be mappable (take for instance a disk which is not mappable. But if you decompose it with a square in the middle and splits from the 4 square corners, then you can mesh the domain with hexa)
The question is: can you afford spending time to mesh it with hexa?
Else a combination of Boundary Layer and Size function should be acceptable.
PS: If you use Size Function, you don't need to mesh edges --> Gambit will mesh them automatically according to SF parameters

Rich1979 April 23, 2010 03:29

1 Attachment(s)
Hi Max,
i've spent some time reading about the possible mesh types available (map,submap,pave,tri-primitive,wedge primitive) and now i know how they work and when i can use them and which restrictions they have (thx to gambit modeling guide).
Now it's time to choose how to proceed.I'd like to try first to make a structured mesh with hexa,and if it is too difficult or it takes too long,i'll pass to an un-structured mesh.
You suggested to me to decompose the domain with a square in the center:does it mean i have to draw a square inside the circle and then split the domain with the square ad mesh the four faces outside the square with a type mesh (for example quad-tri map since the are trielements at the end of each face) and the square (which contains the spline) with another mesh type (for example submap since there is the spline) ?Before meshing the faces anyway i have to mesh the edges (if i don't use your 2nd method by boundary layer and size function).i think i'll have to respect the edge mesh intervals restrictions for the mesh type i'm using,right?Thx as ever,have a nice day!

-mAx- April 23, 2010 04:17

the story of the square inside the circle was just an example of surface decomposition. It has nothing to deal with your model.
Once you did the surface decomposition (with spits) I would try to mesh all the surfaces with hexa (without meshing any edge) The grading of edge will be given through the Surface Mesh element size (gambit will automatically adapt edge mesh)

Rich1979 April 23, 2010 04:27

Oh...i didn't understand it was just an example :/
Could you explain me better what does "decompose with spits" mean?Sorry for my poor tech english.
When you say "mesh with hexa",do you mean meshing with a quad pave?
Thx again,sorry to be boring.

-mAx- April 26, 2010 01:15

I meant splits... ;)
Check Tutorial 3 form Gambit, it can help you understanding Splits (3. MODELING A THREE-PIPE INTERSECTION (3-D)
With hexa, I meant quad-map. As you are in 2d, you can mesh your domain on the fly with quad-pave without split

Rich1979 April 26, 2010 02:23

Ok!I've completed the mesh,using both tri-pave and quad pave and the result it's not so bad :) As soon i can i'll post the gambit file to make you see how it looks.
Anyway with pave i created an un-structured mesh,if i have time i'll try to make a structured mesh to make a fine work.
Thx as evere for your support Max,have a nice day!

All times are GMT -4. The time now is 03:12.