[Help-glpk] building feasibility cuts from Phase 1 solution

[Sebastien . deMentendeHorne]

Hi,

I must build optimality/feasibility cuts in a Benders algorithm.
However, when the LP is infeasible, the phase I of the simplex algorithm
fails and I can't find, in the result that GLPK gives me, the useful
information for building the cut.

The (simplified) problem is:
        v(nprev) = max .138078 * q + z
                        -1 <= q <= 1
                        10 <= n <= 10
                        0 <= z <= 0
                        nprev = n - q
and as example, I took nprev = 8.1 which makes the primal infeasible.
The cut I would like to build is 
                        nprev >= 9

The result after calling lpx_simplex(lp) are showned below

Problem:    vnprev
Rows:       1
Columns:    3
Non-zeros:  2
Status:     INFEASIBLE (FINAL)
Objective:  0.1380784877 (MAXimum)

   No.   Row name   St   Activity     Lower bound   Upper bound    Marginal
------ ------------ -- ------------- ------------- -------------
-------------
     1 nprevious    B              9           8.1             =

   No. Column name  St   Activity     Lower bound   Upper bound    Marginal
------ ------------ -- ------------- ------------- -------------
-------------
     1 q            NU             1            -1             1
0.138078
     2 nnext        NL            10            10            10         <
eps
     3 z            NF             0             0             0
1

End of output

I have the impression that I need marginal cost for the rows and not for the
columns.
Has anyone faces a similar problem ?
Any solution to retrieve a cut on the r.h.s. of the LP ?

TX


Sebastien de Menten               |    ELECTRABEL
Tel:  ++32 10 48 51 76            |    Risk Asset and Liability Management,
Fax:  ++32 10 48 51 09            |    Place de l'universite, 16, 4eme etage
Gsm:  ++32 478 789 444            |    B-1348 Louvain-la-Neuve, BELGIUM