I don’t know that because I tried to use it some yars ago in a MINLP problem, the best that you can do is listen the MP3 file and note if there is any restriction on the linking variables for the decomposition (note that you need to split the problem and then minimize certain gap)…

best regards

2010/3/29 HÃ¼seyin Demirkale

danieal,

is it possible to do langarian to NLP problems ?

2010/3/29 daniel navia

Hi Moein, if DICOPT doesn’t return integer values on binary variables is because the feasibility part of the solver couldn’t find a feasible solution, then the program returns the solution of the relaxed model (considering that the binary variables are continuous with bounds [0,1]. There are a lot of reasons for infeasibilities, but if your model is right, then you can try with several initial conditions, if this doesn’t work may be you should try to reformulate your problem using slack variables (s1 and s2) like:

original problem:

min J(x,y)

st.

h(x,y)=e=0

g(x,y)=g=0

with

continuous Variable: x

binary variable: y

Reformulated one:

min J(x,z)+sum(i,w1(i)*s1(i))+sum(i,w2(i)*s2(i))

s.t.:

h(x,z)=e=0

g(x,z)=g=0

aditional constraint–> y(i)=z(i)+s1(i)-s2(i)

with

continuous Variable: x

binary variable: y

continuous variable: z in [0,1]

positive continuous variable: s1 and s2

parameters: w1(i), w2(i) → weights

then if the problem is well formulated and scaled and there are not model missmatch, the solution of the reformulated problem must be the same of the original one, with the difference that in the reformulated one the binary variable only appears in one part of the model adding stability in the optimization, you can look in :

Szitkai, Z., Farkas, T., Lelkes, Z., Rev, E., Fonyo, Z., & Kravanja, Z. (2006). Fairly linear mixed integer nonlinear programming model for the synthesis of mass exchange networks. Industrial & Engineering Chemistry Research, 45, 236-244.

for the formulation.

If none of these steps works, maybe your problem is too big, then you might consider to use Lagrangean decomposition to split your original problem in different parts. If this is the case you can see this page:

http://egon.cheme.cmu.edu/ewocp/slides_seminars.html

and look for “Tutorial on Lagrangean Decomposition: Theory and Applications”, there are a MP3 file with a lecture of Professor Ignacio Grossmann and its corresponding PDF file about this technique.

I think that in

http://cepac.cheme.cmu.edu/pasi2008/slides/grossmann/index.html

you can also find some examples of relaxation and desompositions of MINLP problems

Hope this helps

Best regards!!

2010/3/29 Moein Parastegary

Hello every body

I have a problem with gams.

I have written a program solving a minlp problem. but any solver could not solve it. Just DICOPT solves it but its solution is not integer or binary.

What can I do?

Moein Parastegari

Daniel AndrÃ©s Navia LÃ³pez

Ingeniero Civil QuÃmico

Mg.Sc.Ciencias de la IngenierÃa, MenciÃ³n IngenierÃa QuÃmica

MÃ¡ster en InvestigaciÃ³n en IngenierÃa de Procesos y Sistemas

626752875

Daniel AndrÃ©s Navia LÃ³pez

Ingeniero Civil QuÃmico

Mg.Sc.Ciencias de la IngenierÃa, MenciÃ³n IngenierÃa QuÃmica

MÃ¡ster en InvestigaciÃ³n en IngenierÃa de Procesos y Sistemas

626752875

–

To post to this group, send email to gamsworld@googlegroups.com.

To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com.

For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en.