Hi,

How is it possible to use CFX for store separation simulations?

I've tried all I can think of and CFX never comes close to even solving a simulation.

Firstly, there is no remeshing capability. Only a very limited mesh motion/deformation feature.

I've written a 6DoF model but had to greatly reduce it so as to put it in Expressions, as it wont read my MatLab 6DoF script. And I've never even heard of Perl before using CFX, so don't know how to write in it.

ANSYS tell me it's been done in the past but are of no help acutally trying a simulation. Any example they send is so simplifed it cannot even be classed as an attempt at store separation.

This was one of the main reasons for getting CFX as we were told by ANSYS that it could do this type of work.

Have we wasted money on the wrong CFD package?

Thanks.

Hi,

What is a store separation?

CFX can do 6DOF stuff but only over a limited range with the default solver, as you mention. I know the development guys have a more general capability and I have heard it is available - you will have to ask your support rep about it.

If your local support rep is no good then ask for a contact in ANSYS head office directly.

Glenn Horrocks

Yes you waste money. You should have bought FLUENT instead CFX. FLUENT have remesh module and 6DOF this give a opportunity to model you problem.

Glenn,

Store separation is the release of store(s); fuel tank, weapon etc from an aircraft.

There's loads of papers on the subject on the internet using Fluent and CFD-fastran. We were told by ANSYS that CFX will do the job but I cannot see how with it's current meshing ability.

The main problem is the meshing during the simulation as a store moves away from the aircraft. Of course as mesh deformation is used then the quality of the elements can become worse very quickly.

The 6DoF is taken care of with a user written subroutine. The other main problems are run time and validation/verification of results.

My issue with CFX for these studies is the meshing. I can use a technique that moves a store and the mesh remains good for quite a distance but it only works if theres no other body near the store - such as the aircraft at the first time point, which renders the method useless.

I recently saw simulations of this and indeed I think Fluent can do it and CFX can't. It requires mesh morphing (or something like that). Maybe Ansys told you they could do it, but mixed up on which of their CFD packages is capable of doing it.

Bart

PS. CFX still rules!!! (I don't do store separations ;-) )

ANSYS is right. CFX can be used for this type of calculation.

You have to use the moving/deforming mesh option. It is possible to do some store separation calculations with this option, but not all.

However, as you say, remeshing is not contained within the flow solver (like FLUENT say), so you can only go so far.

Obviously there is a communication problem between you and ANSYS. Perhaps you were not specific enough with your vendor.

Perl is a common scripting language.

Someone else mentioned using FLUENT. ANSYS also owns FLUENT so perhaps you can swap your licenses.

HekLer,

Have you done this type of work in CFX before?

If so can you tell me about the geometry (i.e. store to aircraft displacement at start of run), mesh (i.e. mesh resolution) and setup (i.e. did you use 6DoF in Expressions or separate subroutine) used.

Thanks

I've done a missile leaving a silo before done in CFX.

You just pack a bunch of grid under the missile and let the grid expand as the missile travels upwards.

I used explicit motion with CEL. I suppose you could write a 6-DOF solver in user FORTRAN and hook that up to the motion boundary conditions instead. I have not done that.

HekLeR -

Was your silo simulation in CFX 2D, 2D axisymmetric, or 3D? I am trying to do something similar and am having trouble with the moving mesh.

My simulation is 2D Axisymmetric, and I find that the mesh near the symmetry pole gets distorted much quicker than the rest of the mesh. If I do just 2D (extrude a rectangular mesh instead of rotation around the center axis) it works fine. I am using a mesh diffusivity equation that is K * Total Mesh Displacement Y where K is a large constant ~ 1E9 and Y is the direction of travel. Did you see any similar behavior in your sim? Any other words of wisdom?

I have done this before in CFD++ with overset and it was much much easier.

Thanks,

-Alex