[Trac #123] CBC.exe produces an "optimal" IP solution that violates a BigM constraint - a scaling problem?

[s-c-e]

Attachment: https://github.com/s-c-e/cbc-trac-migration-attachments/blob/master/trac-ticket-123.zip

An OpenSolver user has sent me a spreadsheet that has some large BigM values. Solving with these large values leads CBC.exe to return a solution reported as optimal that violates at least 1 constraint. Reducing the BigM by a factor of 100 produces a new, now feasible solution.

I have attached (1) Sue's Model - CBC gives wrong answer.txt My summary of the issue (2) model2.lp and modelsolution2.txt Original Input and solution showing the violated constraint (3) model.lp and modelsolution.txt Modified input (smaller BigM) and a new solution that does not violate the modified constraint

Please contact me if you need any more information.

Thanks, Andrew

[s-c-e]

I've executed both model.lp and model2.lp and I get the same solutions - with objective = 10.0. Log files are respectively linked here:
https://gist.github.com/s-c-e/cbc31a29eaff767ca8ad570c5abb9cdf
https://gist.github.com/s-c-e/b0f9baea014ca0ceb8026df42734a882

[s-c-e]

Because I got the same solution in both cases, both pre- and post- modification of the big M, I think @tkralphs this can be marked as closed/resolved. The version of standalone CBC I've used is shown below.

Welcome to the CBC MILP Solver 
Version: 2.8.12 
Build Date: Feb 22 2016