tracking infeasibility in CONOPT

[Vijay Hanagandi]

Hi All:

I am trying to hunt down infeasibility in my problem (NLP) which is pretty small in size.
The message I get in the CONOPT.sta is pretty cryptic (see text below).
Any way to interpret this? I have browsed through the conopt.pdf manual you get at
http://www.gams.com/dd/docs/solvers/conopt.pdf but has no reference as to how to interpret this.

FYI... I've turned on infeasibility listing at every solve, set Display infeasibility Analysis to 'On', Set Nonlinear Pre-Solve to 'On', and CONOPT logical switch for scaling to be 'On'

Any help will be appreciated.
Thanks,
-Vijay

----------------------------------------------------------------
    C O N O P T 3   version 3.14G
    Copyright (C)   ARKI Consulting and Development A/S
                    Bagsvaerdvej 246 A
                    DK-2880 Bagsvaerd, Denmark
 
 Using default options.
 
 
    The model has 16 variables and 19 constraints
    with 64 Jacobian elements, 24 of which are nonlinear.
    The Hessian of the Lagrangian has 3 elements on the diagonal,
    6 elements below the diagonal, and 8 nonlinear variables.
 
 ** Infeasible solution. There are no superbasic variables.
 
 
 CONOPT time Total                            0.004 seconds
   of which: Function evaluations             0.000 =  0.0%
             1st Derivative evaluations       0.000 =  0.0%


--
_____________
Vijay Hanagandi
Applied Optimization, LLC.
17 Kershaw Ct.,
Bridgewater, NJ 08807

Phones:
(908) 393 1316 (Land)
(812) 205 8551 (Cell)

www.applied-optimization.com
Browse to http://www.applied-optimization.com/TransportDemo/ for a cool optimization demo built using ASP

[Chuck Teeter]

Hi Vijay,

The message seems to indicate that your problem really is infeasible.  The fact that there are no superbasic variables (ie, all variables are at bounds) leads me to believe that there are insufficient degrees of freedom.  Since your model is small, it should be fairly simple/quick to eliminate equations until feasibility is obtained.  If you want, I could take a look at it.

Poor scaling could also be contributing to infeasibility.  Have you tried looking at the generated matrix with the Math Program Inspector?  It will identify areas of the model that are improperly scaled by color code.  The scaling option in CONOPT may help somewhat, but CONOPT relies primarily on the model builder to generate/submit a well-scaled model.  Use the Unit attribute for variables and constraints as needed for scaling changes.  Note that in addition to units of measure, you can also enter scalar values.

Another suggestion is to ensure that all of your nonlinear variables are properly initialized to non-zero values prior to the solve.  Ideally, you'd like to have most of your constraints initialized in such a way as to be initially feasible as well.

[Vijay Hanagandi]

thanks Chuck, for the pointers. I was able to ID 2 constraints that are responsible for the infeasibility. scaling maybe a potential issue. I will also try the math program inspector.
-vijay

[Peter.Ni...@aimms.com]

Vijay,

Another way that can help you to determine if the model is truly
infeasible is the nonlinear
presolve. This will try to reduce the bounds of the variables. If
during the process of bound
reduction, the presolve concludes that a lower bound should be larger
than the upper
bound, it will return with an infeasibility status. In that case, you
are sure that the model
is infeasible, regardless of start solution and scaling.

The nonlinear presolve can be activated by setting the option
"Nonlinear Presolve" to "on",
and to get more information on the infeasibility set the option
"Display Infeasibility Analysis"
to "on" as well.

Peter Nieuwesteeg
Senior AIMMS specialist
Paragon Decision Technology

On Jun 6, 10:37 am, Vijay Hanagandi <vi...@applied-optimization.com>
wrote:

> thanks Chuck, for the pointers. I was able to ID 2 constraints that are
> responsible for the infeasibility. scaling maybe a potential issue. I will
> also try the math program inspector.
> -vijay
>

> >>http://www.gams.com/dd/docs/solvers/conopt.pdfbut has no reference as to

> >> Browse tohttp://www.applied-optimization.com/TransportDemo/for a cool

> >> optimization demo built using ASP
>
> --
> _____________
> Vijay Hanagandi
> Applied Optimization, LLC.
> 17 Kershaw Ct.,
> Bridgewater, NJ 08807
>
> Phones:
> (908) 393 1316 (Land)
> (812) 205 8551 (Cell)
>
> www.applied-optimization.com

> Browse tohttp://www.applied-optimization.com/TransportDemo/for a cool