Page 1 of 1

Dynamic solver solution techniques?

Posted: 16 Dec 2012, 19:09
by Fem_Guy
Dear All,

I wish to know what is the normal procedure or technique in FEM world for solving dynamic problems which is K-lamda*M=F , as you might guess those are stiffness, mass and external force matrices , repectively, non of the matrices are time dependant. So how those Linear equation is solved in scientific world and which method uses the Elmer?

Your comments will be appreciated,

Regards,

Re: Dynamic solver solution techniques?

Posted: 17 Dec 2012, 12:16
by carstenp
Hi,

The basic FEM approach is:

1. Discretise the domain of interest with a number of simple geometric elements (tetraeder, bricks, prisms)
2. Approximate the field quantities (pressure, temperature, velocity ...) with a set of basis functions with compact support
3. Find some form of weak formulation by applying a set of test function => Basically your set of differential equations is only fulfilled on average and not pointwise
4. Assemble your discrete operators
5.1. If your system is linear, use a standard algorithm to solve it (see Elmer Solver manual for details).
5.2. If your system is nonlinear a fix point interation is performed => linearise the system, compute a new solution, linearise at the new solution, ...

I am not sure if this answers your question.

carsten

Re: Dynamic solver solution techniques?

Posted: 17 Dec 2012, 14:59
by Fem_Guy
Thanks casrtenp,

The thing you depicted is basically summarization of the FEM procedure we follow on for generic problem solving. Actually each item you specified is worth a tomes of books in itself. But I mainly concentrate on method we do for the solution of eigenproblems. To be more specific if I've K*X + C*X'+ M*X'' = F equation in hand how to transform it into eigen value problem and how to find the solution of it ?

Regards,