CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Simulation of Flow through Complex 3D Geometry (http://www.cfd-online.com/Forums/cfx/65969-simulation-flow-through-complex-3d-geometry.html)

 EmersonKB July 2, 2009 06:15

Simulation of Flow through Complex 3D Geometry

Hi,

As part of my final year engineering project, I am comparing the performance of a number of supercharger intake plates. Specifically I am looking at the intake plates of twin helical screw type compressors, similar to that shown below.

http://img194.imageshack.us/img194/1...arger13.th.jpg

http://img194.imageshack.us/img194/2...osuperc.th.jpg

Initially I developed a 3D model of the intake plate using Solidworks.

http://img199.imageshack.us/img199/8...3dmodel.th.jpg

http://img5.imageshack.us/img5/8205/...dmodel6.th.jpg

Now the geometry actually a number of webs which branch from the bearing housings out to the plate itself. These webs include quite complex geometry due to the profiling on the underside. This results in a face which not only joins the bearing housing tangentially, but also tapers down to zero width at the same time. This is best described with a picture,

http://img33.imageshack.us/img33/141...dmodel3.th.jpg

The plane was to (as a simplified flow model), model flow through a simple geometry matching the outline of the intake plate, with the plate removed from this geometry using a boolean cut opeartion. Again, this is made clear with a screenshot where the green geometry will be the flow region and the grey body is the plate which will be cut from this geometry.

http://img33.imageshack.us/img33/654...dmodel4.th.jpg

Resulting in the following geometry,

http://img6.imageshack.us/img6/1219/...dmodel5.th.jpg

This 3D model was exported for use in CFX via the Parasolid file format. Initially I had a lot of trouble getting CFX-Mesh to work with this geometry as whenever I selected 'Verify Geometry' or 'Generate Surface/Volume Mesh', the program would display 'Waiting for meshers to start...' for a long period of time and would not progress.

After extensive trial and error I decided the cause of this was poorly structured topology in the parasolid file. Using the program CADFix, I was able to restructure the parasolid files to a point where CFX would happily generate a mesh for the geometry.

After dealing with a few expected problem sliver faces and short edges, I was able to generate meshes with good resolution along the intake plate surface as well as inflation layers surrounding the plate and flow region walls to help resolve boundary layer flows.

http://img190.imageshack.us/img190/9...temesh1.th.jpg

http://img17.imageshack.us/img17/518...temesh2.th.jpg

http://img6.imageshack.us/img6/1161/...temesh3.th.jpg

However, the resulting volume mesh contains 800,000 elements. This seems very high and will result in long solution times. In addition to this I am not yet sure if my mesh is fine enough and may infact need (many) more elements.

I am after some guidance as to whether I am approaching this problem the right way.

Is there a more suitable method of modelling the flow then the boolean cut geometry I am currently using?

Is there a better way to deal with CFX freezing on 'Waiting for meshers to start' then using CADFix?

Is there a more suitable meshing method/application I should use?

I find there is a very limited amount of tutorials regarding the ANSYS Workbench/CFX workflow process and the tutorials that do exist deal almost entirely with 'perfect'/'ideal' geometry. Also there seems to be very little published literature where flow through complex 3D geometry has been simulated.

 ckleanth July 2, 2009 06:37

difficult project for final year engineering course. I'm surprised if this is not an msc course final year project at least.. anyway

to simplify get yerself access to v12 use the immersed solids beta function (although this will affect only the hydrodynamic equations) so turbulence modeling effects is no go at the moment. meshing is much easier as you only need to mesh the fluid domain and the solids... separately :p

get your project supervisor to agree with this and you will finish this on time. otherwise.... :(

 ckleanth July 2, 2009 06:38

however you could do this as a 2D analysis- ask your supervisor

 ghorrocks July 2, 2009 06:46

Hi,

First off - unless you have CFX V12 then you have no hope of doing this. The immersed body feature in V12 will make this relatively easy to do, but in any previous version of CFX this will be almost impossible. So make sure you have V12.

From your images it does not look like you are making full use of the CAD importing tools in ANSYS workbench. You should simplify the surfaces much more than you currently have to improve the mesh quality. (Virtual surfaces and the like) It will also improve meshing speed. Do the tutorials in V12 dealing with importing dirty CAD geometry.

You seem to have excessively fine mesh in some areas and excessively coarse mesh in others. You will need to improve this for analysis.

I would not go down the 2D path. A 2D model will not tell you anything of use in this analysis.

Regards,
Glenn

 EmersonKB July 2, 2009 07:00

I believe I can get access to ANSYS V12 at my university. I'll look into the immersed body feature some more.

Do you know of any published (or not) literature that deals with similar flow simulations over three-dimensional geometries such as this? I know companies do these types of simulations all the time, ex: flight and aerospace companies, however it seems there is very little publicly available research. Most papers consider only 2D geometries or simplify complex 3D geometry to 2D.

Thanks for your help, I'll see how I go with ANSYS v12...

 ckleanth July 2, 2009 08:17

just an alternative thought..

since you are in academia contact http://convergecfd.com/ and ask whether they would be interested to support you with their software regarding your project and all. they cant compete with the ansys multiphycis software capability but for the physics setup you would be using I think this is another good chance for you to complete this project.

 All times are GMT -4. The time now is 13:05.