Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Uncertainty in deflection of mechanical structures
Posted Mar 28, 2013, 1:08 p.m. EDT 1 Reply
Please login with a confirmed email address before reporting spam
I designed a silicon structure which is supposed to deflect in the transverse direction (Y axis) under axial buckling (compressive) force (applied along X axis). Even after increasing the number of iterations from default 25 to 500, the solution of the problem never converged. However, when I applied the same force in the opposite direction (tensile), the solution converged. I used the 'Structural Mechanics' module and performed 'Stationary' study find out the deflection.
There is equal probability for the transverse deflection to occur in either of the two transverse directions (+Y or -Y) when compressive buckling force is applied along X axis. In case of applied tensile stress along X axis, however, there is no possibility of transverse deflection but only a little stretching displacement along X axis.
Is it the uncertainty in the direction of deflection that is preventing the solution from converging in the first case? It seems that COMSOL lacks the capability to handle uncertainty or randomness on its own. Is there any other method that can be used to simulate mechanical models of random nature?
There is equal probability for the transverse deflection to occur in either of the two transverse directions (+Y or -Y) when compressive buckling force is applied along X axis. In case of applied tensile stress along X axis, however, there is no possibility of transverse deflection but only a little stretching displacement along X axis.
Is it the uncertainty in the direction of deflection that is preventing the solution from converging in the first case? It seems that COMSOL lacks the capability to handle uncertainty or randomness on its own. Is there any other method that can be used to simulate mechanical models of random nature?
1 Reply Last Post Mar 28, 2013, 3:50 p.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
my experience, with buckling in structures, also MEMS devices, is that the little initial difference that helps to decide which of the two or more solution COMSOL selects is rather critical for convergence, it much metter to give a little puff with a small lateral load until the buckling is initiated than just to rely on meshing inhomogenitites.
But that is not the only reason why buckling models can give solver convergence issues: often it also depends on the load type: force or imposed displacements.
Normally I like to use forces on boundaries, as these are "free" in space and in no way overconstrains the model, while displacements can be tricky to apply without adding parasitic BC or loads (always check the coherence of the reaction forces).
However, when a solver case comes to a bifurcation, the numerical algorithm has no way to choose one or the other soluton and might start to jump from one to the other close to the crossing points. Therefore when you apply a force load and go above the critical load, you get a sudden large displacement and convergence failures, while solving for imposed displacements (and monitoring the reaction forces) avoids the run offs of the back looping stiffness or convergence points
--
Good luck
Ivar
my experience, with buckling in structures, also MEMS devices, is that the little initial difference that helps to decide which of the two or more solution COMSOL selects is rather critical for convergence, it much metter to give a little puff with a small lateral load until the buckling is initiated than just to rely on meshing inhomogenitites.
But that is not the only reason why buckling models can give solver convergence issues: often it also depends on the load type: force or imposed displacements.
Normally I like to use forces on boundaries, as these are "free" in space and in no way overconstrains the model, while displacements can be tricky to apply without adding parasitic BC or loads (always check the coherence of the reaction forces).
However, when a solver case comes to a bifurcation, the numerical algorithm has no way to choose one or the other soluton and might start to jump from one to the other close to the crossing points. Therefore when you apply a force load and go above the critical load, you get a sudden large displacement and convergence failures, while solving for imposed displacements (and monitoring the reaction forces) avoids the run offs of the back looping stiffness or convergence points
--
Good luck
Ivar