Learning Forum

[Alberto]

Good morning,

I have a question about the six-dof solver inertia tensor. I have a non symmetric object and if i compute its inertia tensor using python libraries the output will be a 3x3 matrix whit non zero values in each entries (as expected). When I load the same cad on Spaceclaim and I compute mass properties it will give me only the principal axis values for the inertia. So the matrix will be diagonal. When I set the full matrix for the inertia tensor in the 6-dof panel, the software says "6dof angular solver has not converged". Also with implicit mesh update. Why when I insert the diagonal matrix with values in the principal axis frame the solver do not display that message ( so I suppose it converges ) ? In the user guide there's no requirement about being in the principal axis frame.

many thanks

[Federico]

Hello Alberto, 

can you show the inputs you have for both cases?

[Alberto]

Hello Federico,

sure

I_diag = [ 2.5e-10, 0, 0; 0, 4.5e-11, 0; 0, 0, 2.9e-10 ]

I_full = [ 2.5e-10,  2.8e-11, -3.1e-12; 2.8e-11, 5.1e-11, 1.4e-11; -3.1e-12,  1.4e-11,  2.9e-10 ]

 

[Federico]

Are you using double precision?

It might be that the determinant of your 6DOF solution matrix might be too small. Perhaps you can try rescaling your units

[Federico]

Please try my suggestions and share your findings here!

[Alberto]

Yes, i'm using it. My object is really thin and small, so this is why the tensor looks in that way. Rescaling the object maybe can help, but also the dynamics will change so i will look at different system in a certain sense. It will rotate fast, like 200 rad/s and falls at 3m/s. Is there any other way to stabilize the dynamics? Furthrmore i can't figure out why using implicit mesh update things get worse

[Federico]

Sorry I did say rescale, but I really only meant to change the units, perhaps to mm instead of m.

[Federico]

What is your time step size? You need to set a time scale that doesn't excessively rotate your object relative to the surrounding within a time step. A good rule of thumb is to start with 0.5-1deg of rotation per time step. Start with this and see if this helps

[Alberto]

My timestep goes from 1e-3 up to 5e-4, usually when it need to be updated the software says floating point exception or negative volume for the dynamic mesh. 

[Alberto]

Ok so i will try,  for what concern the determinant, its the same for both matrices (3e-30) .  

[Alberto]

Changing the units from SI to cgs solves the problem. Thank you

[Federico]

Awesome! Happy that this was resolved!