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.
Running FSI examples, "undefined value found"
Posted Mar 7, 2013, 7:13 a.m. EST MEMS & Nanotechnology, Fluid & Heat, MEMS & Piezoelectric Devices, Computational Fluid Dynamics (CFD), Structural Mechanics 4 Replies
Please login with a confirmed email address before reporting spam
i am trying to run an fluid-structure example with use of the MEMS module, byt anytime i get the error "undefined value found..." for the variable mod1.v_solid.
So i even downloaded the 2d fsi example from the comsol page and changed there only the geometry (i adapted the boundary condition to this new geometry but i took the same type of bc) I again get the above listed error.
I checked the bc several times....
I wonder, because the undefined values for u_solid apper in such a (x,y) coordinate, which is not involved in the domain for linear elastic model. Is that a problem? Or what i am doing wrong? I spent a lot of time with this.
I have another question. It is poissible to split the structural equation (i.e. the linear elastic model) to several step (something like operator splitting)?
I am trying to overcome this using Laplace equations instead of solid mechanics and solve them together with laminar flow and deformed geometry (ALE) . I take segregated solver, hereby i need inputs from one segregated step to different one. By this approach i have a problem with the FSI-interface condition ( i chose "moving wall") u_wall=u_laplace, where u_laplace is a solution from different segregated solver. The segregated solver for laminar flow does not know the variable u_laplace . Solving both problems in one segregated steps, or using coupled solver helps, but this is not, what i want to. How can i store the solution from one segregated solver and use it (for example as boundary, or initial condition) in the next timestep in another segregated solver?
Thanks for your help!
Please login with a confirmed email address before reporting spam
Somebody is familiar with this ?
Please login with a confirmed email address before reporting spam
certainly ! what does nu=0.5 mea ? check your equations how E and nu is entered into the stiffness matrix. You can find more about this on the Forum if you do a search. nu=0.495 is OK, but at some stage consider to turn on the "pressure" node dependent variable for"nearly incompressible" materials (see the doc too ;)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
I am not very familiar with the models for solid mechanics, so I had no idea about Poisson ratio. In my (simpflicated) structural model , which i solve with hyperbolic equations nu=0.5 s allowed.
Anyway, the fsi example with nu=0.495 works, but the convergence is still poor.
What about the second question about operator splitting and segregated solvers?
Thanks a lot!
Ann
Please login with a confirmed email address before reporting spam
check if your material can be considered as nearly incompressible, in which case you should add the tick in the physics setting (but only for the appropriate material, if required add a new linear material domain node, then it solves also for "p" and often its more stable.
If you have enough RAM and your problem is "linear" you might get around quicker by selecting a Direct solver method
--
Good luck
Ivar