NLP infeasibility problem on a MINLP

Archive of Gamsworld Google Group
Post Reply
Archiver
User
User
Posts: 7876
Joined: 7 years ago

NLP infeasibility problem on a MINLP

Post by Archiver »



Hi my name is Daniel, and I'm new using GAMS. I had to learn the
program because I'm developing a model for synthetise Heat Exchangers
Networks as a master thesis.
Well, I'm writting in this forum because I've a problem with DICOPT.
The model developed is parametrized with a variable called DTmin, and
what I do is make an optimization Loop with a pool of values for this
particular variable. In some cases there is an optimal MINLP solution,
but in other cases DICOPT reports the following sentence: "Too many
infeasible NLP subproblems consider the use of infeasder option"
when I analize the Log file it shows something that I think is a clue,
but I can't understand.

when DICOPT is running shows:

** Feasible Solution. Value of objetive = 45800 (or something like
that), this value is closer to the optimal solution reached for som
values of DTmin.

After this, the program search the solution during 3 iterations (in
phase =3) and then says:

** Loss of feasibility while tightening tolerances - Return to phase
0

the solver iterate once, and find the same old value (45800) and the
problem repeats until MAXCYCICLES gets to the maximum value.

¿What can I do for the convergence of my problem?
¿Is a good posibility change the infeasder option?, and if is ¿How can
I do this?

thanks and best regards!!

PS: I'm using as a solvers
CPLEX: LP
CONOPT: NLP
DICOPT: MINLP

--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---


Archiver
User
User
Posts: 7876
Joined: 7 years ago

Re: NLP infeasibility problem on a MINLP

Post by Archiver »



Daniel,

DICOPT calls an NLP and a MIP solver. The log you see is a mixure of
the NLP solver, MIP solver and DICOPT log. The message "** Feasible
Solution. Value of objetive = 45800 " looks like a CONOPT message and
does not give too much info about the solution of your MINLP problem.

You can try to work the INFEASDER option, but you can also try other
MINLP solvers like, AlphaECP, BARON, LindoGlobal, or SBB. I am always
interested in MINLP problems (I manage the MINLPLib at www.gamsworld.eu),
so you can also send a copy of the problem to me (mbussieck@gams.com).

Regards,
Michael Bussieck, GAMSWorld Coordinator

On Jul 21, 1:32 pm, Daniel wrote:
> > Hi my name is Daniel, and I'm new using GAMS. I had to learn the
> > program because I'm developing a model for synthetise Heat Exchangers
> > Networks as a master thesis.
> > Well, I'm writting in this forum because I've a problem with DICOPT.
> > The model developed is parametrized with a variable called DTmin, and
> > what I do is make an optimization Loop with a pool of values for this
> > particular variable. In some cases there is an optimal MINLP solution,
> > but in other cases DICOPT reports the following sentence: "Too many
> > infeasible NLP subproblems consider the use of infeasder option"
> > when I analize the Log file it shows something that I think is a clue,
> > but I can't understand.
> >
> > when DICOPT is running shows:
> >
> > ** Feasible Solution. Value of objetive = 45800 (or something like
> > that), this value is closer to the optimal solution reached for som
> > values of DTmin.
> >
> > After this, the program search the solution during 3 iterations (in
> > phase =3) and then says:
> >
> > ** Loss of feasibility while tightening tolerances - Return to phase
> > 0
> >
> > the solver iterate once, and find the same old value (45800) and the
> > problem repeats until MAXCYCICLES gets to the maximum value.
> >
> > ¿What can I do for the convergence of my problem?
> > ¿Is a good posibility change the infeasder option?, and if is ¿How can
> > I do this?
> >
> > thanks and best regards!!
> >
> > PS: I'm using as a solvers
> > CPLEX: LP
> > CONOPT: NLP
> > DICOPT: MINLP
--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---


Post Reply