[fawidmer]

Dear all

I'm currently trying to simulate a mixing device which operates conducts a complex movement (small displacement of the COG, at the same time rotating around a moving axis of rotation with variable angular velocity).
The mixing system uses an "oloid". To give you an idea, have a look at the following video:
https://www.youtube.com/watch?v=pPbSIwmIQYk

I want to simulate only the mixing body, not the mechanics.
I have tried to do it with an immersed solid. Now I noticed that you can only define a constant "Specified Angular Velocity" for the domain movement. Even if you use expressions as inputs, it only takes the constant value of the first time step as constant angular velocity vector for the rest of the simulation.

Does anyone of you know an approach how I could still simulate this kind of movement?

Thank you very much in advance!

fawidmer

[fawidmer]

First of all I want to bump this topic in case anyone of you has ideas about how to help me I would much appreciate it!

I managed to model the Oloid geometry including its dynamics in a "Rigid Dynamics" Block in Workbench (see attached files).

Now maybe some of you know how to use the results from this Rigid Dynamics block in CFX? I tried some stuff but didn't succeed up until now ..

Help is much appreciated!

[monkey1]

Hi!
I was just thinking about the following. Divide your Domain into 2 parts. The fluid domain, and a spherical domain including the oliod (without the mechanical connections that will induce the movement, so just the "mixer"/oloid body). Then you should be able in Pre to import both domain parts and define the spherical as moving/rotating domain. There, I think, it should be possible to define rotation of the oliod spherical domain in every direction desired, as to represent the movement shown in your video.
I'm not sure at all if this might be a practical solution, but the idea just crossed my mind.
Especially since immersed solids are very limited in their application (mostly only fluid flow without any complex behaviours, especially no heat transport)

[fawidmer]

Dear monkey1
Thank you for your reply!
I have thought about the possibility of the Oloid inside a sperhical domain as well. This would probably faciliate meshing.

The problem is, that the movement stays the same whether the Oloid is inside a spherical domain or whether it is an immersed solid. The movement consists of a superposition of translational movement of the COG and a rotation with moving vector of rotation.

So I still don't know how to implement this movement ..

[monkey1]

Hmm ok...if I get it right your problem is that you have to combine two different movements into one to represent correctly the behaviour of the oloid right?
Could it be a possible solution to combine movement of the oloid AND movement of the surrounding domain? Especially if you design the surrounding asymmetrically it could simulate kind of a translation ....
or better would be to find out a function (certainly with combination of different trigonometrical functions) that describes the overall movement of the oloid body around an (excenter) axis....???

[monkey1]

Just found the following
http://www.cfd-online.com/Forums/cfx/67997-how-set-domain-translation-rotation-cfx.html
even if it does not offer a direct solution...maybe it can help

[fawidmer]

Thank you for your effort, monkey1!

Hmm ok...if I get it right your problem is that you have to combine two different movements into one to represent correctly the behaviour of the oloid right?

-> yes! You could explain it like this: The COG of the Oloid is doing the first movement. The second movement is described by a 3D changing angular velocity vector. I have the equations of this movement.

Could it be a possible solution to combine movement of the oloid AND movement of the surrounding domain? Especially if you design the surrounding asymmetrically it could simulate kind of a translation ....
or better would be to find out a function (certainly with combination of different trigonometrical functions) that describes the overall movement of the oloid body around an (excenter) axis....???

I'm not sure if I'm understanding this correctly: Do you mean I could only rotate the Oloid (without translation). And then translate the whole rest of the domain with the inverse Oloid COG-movement? This could work, but will there be no problem with the inertia of the fluid when it is constantly moved? Also I can imagine this could be a pain in the a** to post-process .. Or am I understanding you wrong?

Thank you also for the link to the other topic. It seems this guy had a similar problem. The problem he had would be solvable using the immersed solid approach, because there you can set "General Motion" with separate "Origin Motion" and "Body Rotation". Because his rotation is pretty easy, it should be possible. But altough you can use (time-dependent) CEL expressions to define the angular velocity vector, CFX will only evaluate the expressions for the first time step and take it as a constant angular velocity vector for the rest of the simulation
So I'm not sure if the possibility with the mesh deformation proposed in this post would work:
http://www.cfd-online.com/Forums/cfx...tml#post228345
(I don't have any experience with deforming mesh .. if anyone has, please tell me if this could work or not!)

[monkey1]

Maybe one last hint:
https://www.youtube.com/watch?v=hF8e4iKDvXI
It's a Video of an Ansys CFX simulation of a Kanu Paddle moving in the Water.
The shape of the paddle is comparrably irregular to your oloid and the movement or path of the body of the same complexity.
It was Simulated by CFX Berlin (an official Ansys CFX reseller and research center in Germany), maybe you can contact them and ask if they might help even though you certainly are not part of their customers...

One last question...why do you say that for the rotational speed it will take only the first value of an equation...when i define a time dependent function, the function is reevaluated at each new time step. At least it was possible in a simulation I did some time ago to implement an if-statement varying the rotational speed from a value to 0 in order to stop the movement at some point, wich worked perfectly fine...

[fawidmer]

Thank you again, monkey1.

Thank you for the video link. Maybe I'll try to contact them in the next days (btw I'm Swiss, so the German homepage should be no problem for me .. )

One last question...why do you say that for the rotational speed it will take only the first value of an equation...when i define a time dependent function, the function is reevaluated at each new time step. At least it was possible in a simulation I did some time ago to implement an if-statement varying the rotational speed from a value to 0 in order to stop the movement at some point, wich worked perfectly fine...

You are right, it works under some conditions.
If you define your movement with the "Rotating"-Option, you can use time-dependant CEL expressions, but you can only use a constant axis (e.g. "Tow Points"). See attached image.
However, if you chose the option "Specified Angular Velocity", you can insert the 3 components of the Angular Velocity Vectors. But even though (time-dependant) CEL expressions are allowed, it will evaluate the expression at the first timestep and take it as constant for the rest of the simulation.

Currently I'm trying to implement the system as a coupled FSI-system. I'm trying to do it exactly like the tutorial "Oscillating Plate with Two-Way Fluid-Structure Interaction".
The mechanical part of the system is defined as rigid body, only the oloid is "flexible". When you define the Oloid as rigid body, you can't define a fluid-solid interface, right? (Why is that? Actually I don't want to simulate the deformation of the oloid ...)
In the fluid part of the simulation, I defined the fluid volume. I neglected the mechanical parts, only a cutout for the oloid is there. Can I do it like that or is it a problem that there are surfaces in "Mechanical", which are not there in CFX?
The "Mechanical" part alone can be solved (see short video https://dl.dropboxusercontent.com/u/...d_movement.avi).

When I try to solve the coupled system, I get the classical error messages about "highly distorted elements" in the ANSYS Out File respectively "Fatal error occurred when requesting Total Mesh Displacement for interface." in the Out File. Could this be because of the Mesh Movement (I used exactly the same definitions as in the tutorial mentioned above, even though the movement is completely different - is this the right approach?)? As I said - i have no experiences with moving mesh .....

Any help is much appreciated!

[monkey1]

Not an expert in FSI but the "rigid body" and "flexible" thing maybe due to the fact, that a rigid body is not subject to any deformation, therefore it does not need a fluid/solid Interface, as there is no interaction between solid and fluid (acting like a "Wall" boundary), whereas the "flexible" bodys do show a reaction and interaction with the fluid.

And for your element problems: the oscillating plate has a limited displacement when I remember well, therefore the elements of the mesh are subject to a "controllable" or "limited" deformation. When looking at your oloid and its movement and shape, I would expect that during its movement elements will have to "disappear" and others to be created to fill the gaps. With pure mesh deformation you will have to deform the mesh to a point where it can no longer be called a mesh.

Would have to google that but: Wasn't there also an option in CFX WB with remeshing the case during the solving process? Maybe this could help you....

[fawidmer]

Dear reader(s), dear monkey1

@ monkey1: Thank you again for helping me.

I managed to implement the movement of the oloid including the movement of the mesh. In case there are other people stumbling across this topic having a similar problem, I thought I'd share my solution.
Reading through the posts of this thread I suppose that monkey1 proposed this solution already in his first post, but at that point I didn't understand exactely what he meant yet:

I was just thinking about the following. Divide your Domain into 2 parts. The fluid domain, and a spherical domain including the oliod (without the mechanical connections that will induce the movement, so just the "mixer"/oloid body). Then you should be able in Pre to import both domain parts and define the spherical as moving/rotating domain. There, I think, it should be possible to define rotation of the oliod spherical domain in every direction desired, as to represent the movement shown in your video.
I'm not sure at all if this might be a practical solution, but the idea just crossed my mind.

I used the mesh displacement option "Specified Displacement" under the "Boundary Details" Tab in the settings of the boundaries.

What I did was the following:

• In the geometry, i made a spherical domain around the COG of the Oloid, which contains the whole Oloid body (the Oloid is subtracted (boolean) from the sphere. The sphere is subtracted as well from the rest of the tank.
• Defined the interface between the sphere and the rest of the tank. Defined a boundary for the oloid surface.
• "Specified Displacement" for the outer interface of the sphere is set to the motion of the cog only
• "Specified Displacement" for the inner interface and for the Oloid boundary is exactely the same (rotation and translation), derived from the equations of motion.

Thanks again!
CU

[monkey1]

Yes I think this corresponds more or less to my first suggestion.

Thank you a lot for sharing the solution. This is very helpful and a nice from you!