Prev: bar location
Next: Very Slow GUI in 64-bit Ubuntu linux
From: Nikita on 4 Jul 2010 18:22
Hello Everyone,

I need to minimize a function that computes the cost of a pipeline system based on the diameters of pipes.

cost=cost(D1,D2,D3,...,Dn)

Every Diameter, D, can take one of these values and no other ones:

0.5
0.75
1
1.25
1.5
2
2.5
3
4

What type of optimization should I use? Discrete or constrained continuous? Which matlab function should I use?

I tried fminsearch. It gave me some negative values for the diameter.

Thanks!
From: John D'Errico on 4 Jul 2010 18:31
"Nikita " <sharni(a)vt.edu> wrote in message <i0r1id$qng$1(a)fred.mathworks.com>...
> Hello Everyone,
>
> I need to minimize a function that computes the cost of a pipeline system based on the diameters of pipes.
>
> cost=cost(D1,D2,D3,...,Dn)
>
> Every Diameter, D, can take one of these values and no other ones:
>
> 0.5
> 0.75
> 1
> 1.25
> 1.5
> 2
> 2.5
> 3
> 4
>
> What type of optimization should I use? Discrete or constrained continuous? Which matlab function should I use?
>
> I tried fminsearch. It gave me some negative values for the diameter.
>
> Thanks!

Well, since fminsearch is neither discrete nor does it allow
constraints, nor does it work well for many parameters, it
would seem a poor choice. But you have found that out
already.

You have essentially a nonlinear integer programming
problem.

John
From: Marcus M. Edvall on 5 Jul 2010 17:47
Hi Nikita,

If you are lucky you can model the problem with tomSym in TOMLAB and
automatically get 1st and 2nd order derivatives. Then KNITRO and
minlpBB are suitable solvers to test.

For small scale cases you can try glcDiredct.

Best wishes, Marcus
http://tomopt.com/
http://tomdyn.com/
 | 
Pages: 1
Prev: bar location
Next: Very Slow GUI in 64-bit Ubuntu linux