VERY large displacement - move coord. system

[Broder1977]

Hello everybody,

im simulating the flight of a solar sail (highly nonlinear) with ANSYS and get VERY large displacements which cause the nonlinear transient Iteration to fail.
Thats why i want to move the coordinate system every n-th step.
So I have to restart the analysis using the displacements, velocities and acceleration of the previous step, with changing the displacement by the amount of the displacement of the coordinate system.
My Questions:
1) How to get the nodal velocities and accelerations in ANSYS?
2) How to apply the initial conditions? do i have to use a large mass method to apply the acceleration?

I can use Nastran if Ansys isn't capable of that.

Thx for suggestions
Bjoern

 

[Drej]

Bjoern

>1) How to get the nodal velocities and accelerations in ANSYS?

Use the variable viewer to differentiate the nodal displacements wrt time (Main Menu>TimeHist Postpro>Math Operations>Derivative), or use the DERIV function to do the same thing but manually.

>2) How to apply the initial conditions? do i have to use a large mass method to apply the acceleration?

Personally, yes, I would use the LMM.

However, see a recent XANSYS thread on this, as apparantly you can use the ACEL command to apply the accelerations wrt time to achieve the same result. You can draw your own conclusions from the XANSYS thread on this topic.

Cheers,

-- drej --

 

[pja]

I'm not sure doing what you want to do will be any better than the co-rotational formulation which ANSYS is using for the large displacement nonlinearity. Certainly using first order approximations for velocities and accelerations will introduce some errors.
Have you tried reducing your transient timestep?

 

[Broder1977]

Hi and thx for your replies,

to Drej:
I should have known about getting the velocities and acceleration by a derivation. Thx for the tip
About the LMM:
Am I right if I would connect every node of my solar sail with a Large mass and apply the equivalent Forces equal to nodal acceleration divided by mass?

to pja:
I don't get your idea? why do I use first order approximations?
Reducing the time step doesn't help to get convergence, I've tried that. I think the main problem is the large difference between the applied forces (air drag about 0.46E-08 N/mm^2 ) and the very large displacements (1.000.000 mm) which will cause the algorithm to fail due to numerical preciseness.

 

[Drej]

>Am I right if I would connect every node of my solar sail with a Large mass and apply the equivalent Forces equal to nodal acceleration divided by mass?

Not literally EVERY node. You'll need to connect only the applicable points of your model (the points that transfer the load for example). The equivalent acceleration is, like you say, obtained from f=ma. For a specific "a" you must know your mass (which you do) and an equivalent force (which you can calculate).

The best of luck to you,

-- drej --

 

[pja]

What I meant was if you calculate the velocities and accelerations using only a backwards/forward difference this will introduce some error since ANSYS is using a second order in time differencing scheme. I guess I'm not quite sure as to what you problems consists of. Is this deformation mainly due to rigid body motion
of the sail as in the whole sail is translating? Or is some portion of the sail fixed?

 

[Denial]

Broder1977,

I don't understand any of the detail of your problem; such detail as I think I do understand relates to a type of problem I have never had to investigate; and I have never used ANSYS so I do not know what sort of tools it provides. But I had a brief thought.

Perhaps a new perspective might lead to new, more tractible, approach. Perhaps you could use D'Alembert's Principle (sic) to turn your problem inside out. Convert it from a dynamics problem involving free-body accelerations (and consequent massive displacements) to a statics problem involving a stationary structure experiencing continuously varying inertial loadings.

Just a top-of-the-head transient thought.

 

[Broder1977]

Hi everybody,

yesterday a meeting with my mentor took place, and we considered the ridig body motion as the main problem. Because there are no displacements constraints set, the LMM would not be possible.
My next thougt was to avoid working in a moving coordinate system trough moving the global coordinate system every n-th step through a enforced displacement of all nodes with the time integration turned off.
This works so far, the problem now is that to continue with the transient solution with TIMINT,ON needs the acceleration and velocity information of the substeps before the moving of the coordinate system, but the initial conditions are taken by default from the last two substebs (help,TIMINT - notes).
Now this two substebs are static substebs:
1 - allpy moving with the D command
2 - remove constraints
Thats why no transient quantities remain .... arrrgh ;-)

@Denial:
I'm trying to avoid this, because than i have to know the path of the moving coordinate system in the center of the solar sail (SKetch), which i don't.

Sketch:
x x <- Sails between booms
x x
M <- Mass about 50kg
x x
x x

... no idea so far ...

Greetings Bjoern

 

[Drej]

Bjoern - how is your load applied? What loads are you applying wrt time?

 

[Broder1977]

The load is a pressure applied to surface elements in order to simulate a kind of wind- load. It is constant over time, but will change by the deformation of the sail (only the normal components are taken, not the tangential ones).

 

[brad]

Can you use inertia relief? This sounds like this would be an entirely appropriate assumption for your problem as described.
(I don't know whether or not ANSYS has inertia relief, but if it does this would be ideal)
Brad