Discussion Forum

Hi

nothing fundamentally wrong there, at least not from what I see,

but I suppose you can do a little better with the subtilities.

You are asking for a transtient time solving sequence, this means you are adding in the rho*d2_u_solid / dt^2 terms (check the equations) but in this model your steel cylinder will certainly not be deformed significantly by the liquid flowing by, so it's internal dynamic behaviour is of no true interest, particularly as you have fixed it solid on the lower part (such that you can measure force and torque on this fixed part, I suppose ;)

Hence you can greatly simplify your model if you set the main FSI physics part to "Quasi static - Structural transient behaviour"

Then you should do a little better with your "turn-on" BCs and initial conditions to help the solver start. First of all you ramp up smoothly your inletvelocity, but you have added a 2m/s constant starting speed, while the default initial conditions is "0" m/s, hence you are sending a shock wave down your model, either the +2 is too much, or you should use an initial velocity of 2 m/s for your fluid (not the easiest as you have a truew profile with no-sip v=0m/s along the boundaries, not easy to set up as initial conditions with your special geoemtry.)

In certain cases you could also add a small pressure drop along your tube flow, this is easy for long tubes, you take Poiseuille as starting point, but les evident for your geometry so hopefully it's OK with the default all "0" initial conditions.

LAst point you have selelcted user mesh, but only thets, no boundary mesh along the "no-slip" walls, this is not the best approach, try to let COMSOL propose a default mesh (set it back to physics controlled mesh settings) and look closely at the mesh

--
Good luck
Ivar

Hi,

Thank you for your explanation.

As you mentioned, I want to see how much torque and force exert to the cylinder( this is originally complicated model like ship. I simplified it for less computing).
I change structural transient behavior to "Quasi-static" and modified initial velocity starting from 0 using step function so that it looks no shock wave existing.
Also I used physics-controlled mesh with coarse element size.
But I still have problem to simulate this model, error is still same too. Nonlinear solver did not converge..

It was OK using either laminar and turbulent to simulate stationary study with constant velocity. But what should be added for time-dependent?. Regardless how the velocity changes, I would like to see time dependant analysis.

Thanks you.