Ansys error help

[alphaden]

Hi all,

I'm doing a fluid structure interaction problem using Ansys 9.0. I'm getting a strange error, "The inversion subroutine was supplied with a singular matrix". what does that mean ??? i tried modifying almost everything. in the output window the last message is ... "applying boundary conditions" and immediately after that "The inversion subroutine was supplied with a singular matrix". what is wrong ????
any help would be grateful.
thanks a lot

Alphaden

 

[prost]

I am not an ANSYS user, however, I can take an educated guess as to what is going on, I think. I can only guess as the depth of your FE knowledge, so I apologize if you know all this. A finite element code like ANSYS constructs large matrices (in a structures problem, this is the stiffness matrix, K) which are then inverted: Ku=f, you know the 'f' vector, you use an inversion technique to find the vector 'u'. The stiffness matrix K, a square matrix, itself is singular, in the case of two dimensional problems (say plane strain or plane stress) that means there are at least 3 rows in the K matrix that are dependent on other rows. Since K is singular, it cannot be inverted without some constraints. In the case of 2D problems, you need at least 3 boundary constraints that in effect eliminate rows in the big K matrix and therefore allow it to be inverted and the solution 'u' to be found. My guess would then be from what you have written that you do not have enough boundary conditions to constrain the stiffness matrices fully to allow the solver to invert the matrices and find the solutions.

 

[alphaden]

thanks prost,

this is true. and when I do the analyses separately .. both of them run fine. but while performing FSI (combining both together) it fails. I have also considered structural disp. BCs for fluid for FSI which otherwise I would never consider if I was to do CFD all alone. but still failure is what I end up in !! but I'll try different approach for BCs. and let you know if it works.

thanks again,
alphaden

 

[cbrn]

Hi,
you could try to run a modal analysis on the struct part of your model alone, in order to see if there are rigid-body modes (i.e. unconstrained fundamental DOF of your system) and gather info about which they are, if any.

Regards