Prev: bug with MATLAB example avifile
Next: Pseudo Eigen-Decomp?
From: Hugo on 3 Jun 2010 17:43
hi all

I´m doing a Finite element model updating program on matlab. i´m using the optimization toolbox.Basically, I´ve created and input file in Ansys programming language that return the deformation in the finite element mesh nodes of my model. the matlab optimizes the geometric parameters of the model and returns the results to ansys, where the calculations of the new deformation using the new values for the geometric parameters are done.

the program runs without errrors or warnings. the problem is that the fval of fmincon function returns always same value in every iteration. Can you tell me what might be wrong?

regards
hugo silva
From: us on 3 Jun 2010 17:55
"Hugo " <hugolopessilva(a)gmail.com> wrote in message <hu97l9$h1b$1(a)fred.mathworks.com>...
> hi all
>
> I´m doing a Finite element model updating program on matlab. i´m using the optimization toolbox.Basically, I´ve created and input file in Ansys programming language that return the deformation in the finite element mesh nodes of my model. the matlab optimizes the geometric parameters of the model and returns the results to ansys, where the calculations of the new deformation using the new values for the geometric parameters are done.
>
> the program runs without errrors or warnings. the problem is that the fval of fmincon function returns always same value in every iteration. Can you tell me what might be wrong?
>
> regards
> hugo silva

how can CSSMers possibly be of any help with this kind of information...

us
From: Matt J on 3 Jun 2010 18:04
"Hugo " <hugolopessilva(a)gmail.com> wrote in message <hu97l9$h1b$1(a)fred.mathworks.com>...

> the program runs without errrors or warnings. the problem is that the fval of fmincon function returns always same value in every iteration. Can you tell me what might be wrong?
==============


A few guesses (which are all that are possible right now, without seeing more of your code and output).

1. The input parameter guesses are already optimal, or they lie at a stationary point of the cost function.

2. You've coded your cost function incorrectly, e.g., so that it always returns a constant value.

3. You're supplying a user-defined gradient which has been incorrectly coded, e.g. such that it is always zero or always points uphill.

4. The fvals are not really the same. You're just displaying them to limited precision.
 | 
Pages: 1
Prev: bug with MATLAB example avifile
Next: Pseudo Eigen-Decomp?