'mandrel' simulation project by tbarnes


#1

I created a new simulation project called 'mandrel':

mandrel


More of my public projects can be found here.


#2

@cjquijano, @ahmedhussain18, @BenLewis @rszoeke

Hi guys,
Thanks for all your help with getting me to understand the coordinates and axis orientation question from last week. I tried modeling this simulation after @rszoeke similar model, I believe i have set up the simulation in terms of BCs and initial conditions the same however my simulation has timed out after an out when his only too 4-6 mins. I realize I am using a more refined mesh with nearly ~2x as many nodes, would this cause the simulation to take over an hour tom complete ? Or is there something else that is causing the excessive computation time. Sorry for the trivial question, i am not yet very adept at deciphering the simulation logs

Again thanks for all your support!


#3

Hi @tbarnes,
usually when doubling the number of nodes, the simulation time scales with about a factor of 4 to 6. In you case I guess the long simulation time is die to a combination of remote displacement constraint together with MultFront solver. There is a hint in the documentation regarding the use of remote constraints with a high number of connected nodes. In this case the MUMPS solver should be preferred. With this the simulation time should be only a few mins.

Additionally to save time you should reduce the amount of connected nodes for the remote displacement constraint. I noticed you are assigning the complete bottom surface to the remote displacement. Wouldn’t be the bolt holes be enough, as the deformation behavior is set to deformable and thus does not add any stiffness to the model?
You need also to constrain the remote point in all 6 directions, otherwise you will have a rigid body motion and thus a singular system.

I noticed you are using a dynamic analysis. Is there any reason you need a dynamic analysis? For this simple static loading (a constant centrifugal force is considered a static load, as force direction and amplitude stay constant) a linear static analysis should be preferred.

Best,
Richard


#4

@rszoeke

  Thanks for the help. I have taken your tip and switched to the MUMPS solver, and switched the analysis type to  static advanced. Could you please spend some time explaining in more detail the "stiffening free base fixation" BC you have in your model. 
  As I understand we are defining the allowable motion of the model in relation to a remote point, in this case the origin (center of rotation).  so the the Dx,Dy,Dz  translation Bcs make sense to me we fix the motion along the axis of rotation while allowing the model to "slide" or translate on the plane formed by the two unconstrained axis.
  Where I get lost is on the rotational properties. I copied your set-up in that I prescribed a 0 rotation  along the axis of rotation, this confused me a bit, isn't this the axis we want to allow rotation about ? as is defined in the load properties. Also what is gained my unconstraining the the rotation of the two axis orthogonal to the models assumed axis of rotation?

I hope this question is clear I am having a difficult time explaining my confusion.


#5

Hi @tbarnes,

You are applying a centrifugal force to your mandrel but you are not actually rotating the part. The centrifugal force is applied in the radial direction to the model. This is why you do not need to have the rotation of the remote force unconstrained.

You have already changed from a Dynamic Analysis to a Static Analysis - Advanced which is good. If you contain all six dof of the remote point, it should solve fairly quickly.

Good Luck,
Christopher

EDIT: I just constrained the remote displacement point in all six directions and ran your model. It completed in in under five minutes.
If you let us know the purpose of the model we can help you fine tune it a little more.