Dynamic Mesh Problem.
I'm modelling interaction between two aerofoils, one upstream and one downstream.
It's not a 2D case, but it's a thin slice with slip planes at zmax and zmin. The downstream direction is +x.
The upstream aerofoil is fixed; the downstream aerofoil moves in the y direction, with a fixed velocity and no rotation. I'll be doing a transient simulation with the downstream foil moving through the wake of the upstream one. It's quite a lot like a stator-rotor interaction scenario.
For various reasons, I'm not using a slip plane between the two foils - I need to use a dynamic mesh, but am having trouble setting it up (I've followed the documentation, but am still not clear).
I have a profile file describing the downstream foil motion:
((linear_movement 2 point) (time 0 10) (y-velocity 12.344 12.344) )
And I have the following zones in my mesh: up_foil (upstream foil surface) dn_foil (downstream foil surface) ymin (there's a periodic condition on the ymax plane) xmin (inlet plane) xmax (outlet plane) zmin (no-shear-stress wall) zmax (no-shear-stress wall) fluid interior
In the dynamic mesh zone definition, I have set up the dn_foil to move as a rigid body, and have set up_foil to be stationary.
Clearly, the volume mesh will need to be adapted, along with mesh at the zmin and zmax planes (I've allowed for remeshing and spring smoothing in the Dynamic Mesh > Parameters dialogue).
In Define > Dynamic Mesh > Zones I've tried setting zmin, zmax, fluid and interior zones to 'deforming' (they can't be rigid-body as the upstream foil has to be stationary). Having tried a number of different things, when I attempt to preview the mesh motion, it recalculates, but never alters the mesh.
Has anyone got any clues about what I might be doing wrong? I'm uneasy that my mesh is unstructured (i.e. a mixture of hex and tets) - would that affect things? Perhaps I need to put different boundary conditions on (currently all walls are not moving, but the ANSYS tutorial says that this is overriden by the dynamic mesh setup).
Thanks in advance for any help you can offer,
Re: Dynamic Mesh Problem.
I made this error before.
You should not define zmin, zmax and the interior as deforming. In fact, you do not need to do anything about them. i.e. do not add them in the zone list.
as i understand, what you have defined is a deformation of your computational domain for example if your plane would deform into a cylinder.
I do not know exactly what will happen if your moving foil is connected through the boundaries zmin and zmax because i never tried that.
Hello Tom Clarke,
I know this is an old post and you have probably solved this by now. But I am having similar problems to you. Similar in a sense that my airfoil is connected on the zmin and zmax boundaries. So far i have only defined my airfoil as a rigid body with a pitching motion at the quarter chord. What I find is that after approx 1000 time steps, negative cell volumes always appears on a cell just adjacent to the zmin/zmax boundaries. When I inspect the boundary mesh after the preview, I can see the tri surface mesh beside the airfoil trailing edge are mostly collasped at the pressure side and stretched at the suction side. This led me to think that the dynamic mesh algorithm is only remeshing the volume cells and not the surface cells on the boundaries zmin and zmax, which led to high skewness and negative volume cells. I am currently trying local face remeshing to see if those boundary face mesh can be remeshed together with the volume cells. If you have exoerience with this problem, I hope you can confirm/reject my reasoning and tell me what is actually wrong?
Thanks a bunch
Same Problem - Any hints?
Hi Darren and Tom,
I have a similar problem and would therefore love to know if and how u have solved this.
I'm trying to simulate the motion of a 3d-alveole which is regularly expanding and shrinking during breathing. When I try to simulate the mesh motion I also get negative volume after several steps. Unfortunately it didn't help a lot to change the smoothing and remeshing parameters (maximal improvement was 5 iterations plus) nor the time step or expansion factor in the UDF.
Just like in Darren's it also looks as if fluent is only remeshing the volume.
Pleaz let me know if you found any solution. I would appreciate any hints how to solve this problem.
Thanks a lot in advance!!
Such problems can be easily managed using Radial Basis Functions and mesh morphing. RBF Morph is an add on that implements this functionality within Fluent even for a transient analysis. I have recently developed a test (that will be published very soon) to account for the structural deformation of a wing (using a real life mesh of the aircraft with 14 millions of cells).
I'm currently set-up the transient flapping and twisting of helicopter rotor blades...
ups, posted double... so deleting this one
thanks for ur reply. i actually came across ur rbf mesh in another thread of this forum and also checked out ur web-site. it sounds really interesting and i'm curious for the tests. unfortunately i'm not in a position to buy it. apart from that i still can't believe that a powerful solver like fluent is not able to perform a mesh motion as described... but didn't find a solution anyway... :(
do u maybe happen to know more about the capabilities and possibilities within fluent it-self?
is it somehow possible to expand and extract an organically shaped 3d-mesh without producing negative volumes just by means of a udf? is there anything else i could try?
btw, i forgot to mention last time: the motion is defined by the most and the least expanded state of the alveole, for which i hav real data.
thanks a lot for any hints or comments. m
Fluent is really a very powerful solver and is open to be expanded using UDF. RBF Morph was born after years of experience in UDF programming. And basically it interacts with fluent that allows to use a user defined movement for MDM.
If you have the nodal position for the two configurations you can use an UDF to move surface nodes with a blending criteria (linear or non linear), i.e. (linear):
n (lambda)= n1 + lambda * (n2-n1)
where n is the generic position of the node, n1 is the position at configuration1 n2 at 2; lambda is in the range 0-1 (0 gives n1, 1 gives n2).
n, n1, n2 are vectors.
If you are able to store such positions for the surface mesh you can update it and use fluent smoothing and remeshing to move the volume.
the problem is that the meshes for the two geometries differ a lot in shape and number of nodes. I wrote a program in matlab to delete additional nodes but the resulting geometry looks crap. Another try was to project the nodes of one geometry onto the surface of the other and only take those for the mesh. Unfortunately many elements of the projected mesh become too distorted during the projection, resulting in negative volume during fluent's calculations. So far I didn't find a possibilty to automatically mesh one of the geometries with a specified number of nodes (the same number as the other one) in order to use the method u also suggested.
Do you maybe have an idea how I could solve that problem?
Thanks a lot in advance and also for all help u offered so far.
It seems a challenging application.
I did some successful test on two state of the same geometry (with different mesh exported as STL) and using FEM solution (and of course in this case the two mesh are conformal, but coarser with respect to the CFD one). Such shape modifier are handled by RBF Morph changing also the volume mesh (RBF smoothing has better performance if compared to the spring model used within fluent).
If you have a relevant difference between the two shapes I'm not sure you can face it using RBF Morph. I'm open to make a test on your model if you can send me the original mesh and the surface target exported as an STL file.
|All times are GMT -4. The time now is 17:05.|