In [1], the following scenario is described:

- A MIP model is solved, delivering a
**feasible**solution. - All integer variables are fixed to their levels.
- The resulting LP model is solved in order to produce duals. This model turns out to be
**infeasible**.

**elastic**.

**elastic model**, we replace some of the hard constraints with soft constraints. This often makes economic sense: throwing money at a problem can almost always make it go away. Here are some examples:

- If running out of storage capacity, rent expensive extra capacity.
- If we cannot deliver orders on time, allow being late at a cost.
- If we have a capacity shortage in production capacity or personnel, allow expensive outsourcing or hiring of temps or pay for overtime.

*infeasible, no solution*". Or my favorite message: "

*Infeasible or unbounded.*" (Solvers do not always minimize the time the user has to spend on deciphering error messages). It always produces a solution that can be inspected. This solution may show the use of expensive extra resources. This emergency usage is minimized by the solver (so it makes sense). An end-user can see that we exhausted our normal resources and had to use these emergency resources to make a schedule or plan.

#### References

- Avoid infeasibility in “fixed MIP problem” in CPLEX, https://or.stackexchange.com/questions/6048/avoid-infeasibility-in-fixed-mip-problem-in-cplex