help-glpk
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-glpk] Bilinear problem


From: Jeffrey Kantor
Subject: Re: [Help-glpk] Bilinear problem
Date: Mon, 6 Apr 2015 10:07:55 -0400

Keep in mind the method outlined in the AIMMS book is a only an approximation for case of two continuous variables.

Bilinear constraints frequently come up in chemical process applications. For example, decision variables for a particular process stream might be stream flow rate and concentration of a species, and the constraint written on flowrate of that species.  These situations lead to non-convex solution spaces that can be solved using global optimization techniques.

One way to treat these situations in a linear programming context is to replace the constraint with a set of four linear constraints that provide a convex 'outer approximation'.  The outer approximation itself has been known for a long time, and is sometimes called 'McCormick Relaxations.' 

The approximation can be improved by subdividing the solution space, then using a branch-and-bound search.  Repeating this process will eventually give a global solution. The scheme is described in http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.208.4249 by Ruiz and Grossmann.  

On Sun, Apr 5, 2015 at 7:22 PM, Antonio Carlos Moretti <address@hidden> wrote:
If the bilinearity involves

(1) two binary variables or
(2) a binary variable and a continuous variable or
(3) two continuous variable

then is it possible to write as an integer linear programming.
See ''AIMMS - OPTIMIZATION MODELING' pages 83,84 and 85. (you can get the
pdf through internet)

Best regards,
Antonio

>>> Is it possible to solve a problem with bilinear constraints using
>>> GLPK?
>>>
>>
>> No, because bilinear constraints are non-linear while glpk allows only
>> linear ones.
>>
>
> Can we do something about it? From my experience with commercial systems -
> solving NLP problems involves successive linearisation until convergence,
> although it could lead to a local optimum.
>
> I?d like to participate.
>
> --
> Ruslan Gazizov
> _______________________________________________
> Help-glpk mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/help-glpk
>




_______________________________________________
Help-glpk mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/help-glpk


reply via email to

[Prev in Thread] Current Thread [Next in Thread]