Fluent and Gambit CFD

himanshu121

Hi
i want to know the difference b/w a good mesh and a bad mesh in gambit, pls give me suggestions and if possible a good link which gives a lot of tutorial on gambit and fluent,

I am attaching a msh file which i created using gambit can anyone pls suggest what is wrong with the above mesh

All the help is appreciated and thanks in adv.

Regards
Himanshu

himanshu121

i donn know how to upload such a large file of .msh format

himanshu121

u can find the the document here
http://home.iitk.ac.in/~himan/

Clausius2

There are no good meshes and bad meshes. The only thing a mesh must fulfill is the well posed algebraic transformation: it's jacobian must be well defined (no physical point should be described by two computational points at the same location).

What makes a mesh a good mesh is the kind of computation you want to calculate. Before working with fluent o gambit, you should have in your mind how your physical realisation is going to result, where are going to be the largest gradients and what are the critical localizations where the propagation of numerical errors is more likely to occur. A lot of manuals are on line (there are manuals attached to the cd of each program). Or go to the library and check out the Fluent manual.

Astronuc

In general, 3D FEM has particlar stability criterion regardless of software.

Here is a nice summary for CFD from Flow Science (Flow 3D)
http://www.flow3d.com/Cfd-101/criteria.htm

More generally -
http://www.flow3d.com/Cfd-101/cfd101.htm

http://www.fluent.com/software/gambit/index.htm

http://www.fluent.com/software/fluent/index.htm
http://www.fluent.com/software/fluent/grids.htm
http://www.fluent.com/software/fluent/dynamic.htm

Certainly mesh generation is important. I am not particularly familiar with GAMBIT, but I expect there is a default node/element density.

In certainly problems, the meshing would need adjustment, particular where there are large thermal gradients, otherwise the elements require appropriate shape/form functions. Computational efficiency is another issue that must be considered.

One can register for FLUENT support as a professional or academic/student -
http://www.fluent.com/support/fuscreg.htm

One may also try the following CFD forum -
http://www.cfd-online.com/Forum/

saadsami

hi
i wanted to do a 3D meshing of a wind turbine propeller in gambit.then i wanted to do aero acoustic analysis on it .could any one help me out. thanks

gmwsy

HI,I am gonna to do a similar project too. Thus I am very pleased to keep in touch with you if possible. Please do not hesitate to contact me.Thank you very much.
My email:gmwsy@sjp.buaa.edu.cn

minger

I don't mean to crap on your day, but I wouldn't trust those type of codes very much for aeroacoustics. We use CFX, and they claim at best 2nd order, and I've heard that the actual code runs at first order. This in layman's terms means that as you add more grid points, the solution doesn't converge quickly towards the actual solution (which as we know we will never get).

If you really want to resolve the acoustic waves, which can be magnitudes of orders down from the mean flow, you're going to need a LOT of grid. Also make sure you do whatever you can get to double precision for your solver, otherwise don't waste your time.

As with any third-party CFD solver, you'll get an answer, and it will probably seem right, but you have no idea what its really doing.

[/biased] :)

jaap de vries

A good grid is a grid that has no influence on the solution. When a grid is too fine you could get numerical round of errors and when to course one could get truncation errors.
Typically you want to watch aspect ration and skew angle, both can be checked in Gambit.
Also, the grid type should really depend on your problem. hexahedral grids are typically preferred in combustion modeling.

minger

The only time I've heard of there being a problem using too fine a mesh is if one is using a DRP scheme and just happens to start and end at the right point. Unless the grid size is approaching machine zero, IMHO you cannot ever make the mesh too small (making sure of course each line is at least d2 differentiable (assuming structured grid).

justin0741

Like they said before, when you apply small changes to a GOOD mesh, there will be little difference in the solution. The opposite for bad meshes.

I like to make a uniform mesh and use the grid adaptation tools in fluent to refine the mesh for me. Otherwise, you will be spending A LOT of time building grids.

tunzawcn

Hi , gmwsy@sjp.buaa.edu.cn I m student of beihang university. at the same time, i m studying gambit and fluent, too. i would like to know how to construct 3d nozzle in gambit. if you have experience with gambit please guide me how to construct that nozzle. my email is tunzawcn@gmail.com.

arman2_civil

Hi all,
I am working on my project using gambit for geometry and mesh generation,
in edge meshing I have no problem and in face meshing by tri-pave element I can mesh my faces successfuly but for volume meshing I can not mesh my volume,I dont know why but for every selections of my elements gambit provides an error.
only steirstep meshing is compatible and successfull.But this type of meshing creates additional faces and volumes that provides errors for boundary conditions and fluent mesh checking.is there any way to mesh my project successfully with other types of meshing?
please help me if you can.
with regards,
arman.
arman2h_civil@yahoo.com

vrox4eva

hi guys...am doing a project on turbine blades too...but there is a small problem in my mesh creation...there are no errors generated but my final pressure contour is not right... I am doing a 3d model by using a hex/tri mesh.

minger

You probably should have started a new thread, but why do you think that the problem lies in the mesh?

vrox4eva

because the only warning i get is during the mesh export....!!

minger

Warning for poorly shaped elements? How poor? Where in the grid? What's the solution look like there? There are MANY questions you need to figure out before you can try debugging a problem like this and truly identifying where the problem lies.

Maybe you're just running to large CFL.