Prev: FXLMS Demo
Next: guide error
From: Ruffuz Hemlig on 14 Nov 2007 04:08
I am interested in finding the natural frequencies for a
structur. For this, I have achieved a mass and stiffness
matrix from a model (both are quadratic and of the same
size). The problem arise when I try to use the command:

[phi,om2]=eig(K,M)

which results in "Use eigs for sparse eigenvalues and vectors."
The eigs command cannot be used since the generalized matrix B
is not symmetrical positive.

Does anyone have a solution to this problem?
From: Greg Heath on 14 Nov 2007 05:58
On Nov 14, 4:08 am, "Ruffuz Hemlig" <rhem...(a)hotmail.com> wrote:
> I am interested in finding the natural frequencies for a
> structur. For this, I have achieved a mass and stiffness
> matrix from a model (both are quadratic and of the same
> size).

Terminology: ...both are SQUARE and of ...

Is either nonsymmetric?
Is either not positive definite?

If so, why would a stable structural model not have both M
and K symmetric and positive definite?

> The problem arise when I try to use the command:
>
> [phi,om2]=eig(K,M)
>
> which results in "Use eigs for sparse eigenvalues and vectors."
> The eigs command cannot be used since the generalized matrix B
> is not symmetrical positive.

The terms "generalized eigenvalues" and "generalized
eigenvectors" are well defined. The term "generalized matrix"
is not.

However, I assume you mean the dummy variable B in

[V, D] = eig(A,B)

> Does anyone have a solution to this problem?

Verify that having M not positive definite makes physical sense.
If it does but K is positive definite, try

[phi,om2]=eigs(M,K)

Hope this helps.

Greg

 | 
Pages: 1
Prev: FXLMS Demo
Next: guide error