[RectangleGuillotine] first_stage_orientation and "empty cut"

[bbeaulant]

Hello Florian,

Let us consider this instance with the problem type RectangleGuillotine :
instance_parameters.csv
instance_bins.csv
instance_items.csv

As you can see in parameters, the first_stage_orientation parameter is set to vertical.

I know that it doesn't necessary directly influence the orientation of items. But this instance generate a solution where some bin contain the first vertical cut that cut nothing.

SolutionBin (id = 4)

Isn't it possible in this case to prefer item rotation and use first level of cut vertically ?

[fontanf]

Hi,

The rational behind the first cut orientation and the number of stages is that some factories really have stages with the glass pieces moving through different machines. In this context, it's not an issue if some stage has 0 cut to perform.

Forbidding empty first stage would make the solver miss the optimal solution in some cases. So I don't consider this acceptable.

"Prefering" a solution over another requires formally defining the criterion to compare two solutions. I'm not very enthusiast about appending a criterion related to the number of cuts in the first stage to the objective

Is there a real issue with solutions where the first stage has no cut?

[bbeaulant]

Is there a real issue with solutions where the first stage has no cut?

Yes, the total length of cut is wrong. In the timber industry, the "cutting price" is often linked to the total length of the cut.

---

The particular thing here, is that if I isolate those 3 parts with the same parameters, they will be placed vertically, but in a larger run, not necessary.

[fontanf]

Yes, the total length of cut is wrong. In the timber industry, the "cutting price" is often linked to the total length of the cut.

So the issue is not the first stage to have no cut, but the objective of the solver that doesn't match the real objective

[bbeaulant]

So I don't understand the real goal of this first cut orientation parameter if it can be associated with an abstract cut even if a real solution is possible.

[fontanf]

If the number of stages is limited, then imposing the first stage orientation restrict the set of feasible solutions.

If the number of stages is unlimited, the first-stage orientation has no reason to be.

[bbeaulant]

If the number of stages is limited, then imposing the first stage orientation restrict the set of feasible solutions.

Is this a problem or an expected behavior ?

-- In the case above, the number of stages was limited to 2.

Is there a real issue with solutions where the first stage has no cut?

One other real issue is the number of stage. The more stage involved, the longer the cutting process will take, and therefore the higher the cost. But I understand that global cost is not necessary a goal of the BinPacking objective.
So, reducing the number of stage is not a goal of the BinPacking objective ?

[fontanf]

Is this a problem or an expected behavior ?

It's an expected behavior

One other real issue is the number of stage. The more stage involved, the longer the cutting process will take, and therefore the higher the cost. But I understand that global cost is not necessary a goal of the BinPacking objective.
So, reducing the number of stage is not a goal of the BinPacking objective ?

The goal of the BinPackingWithLeftovers is to minimize the number of bins used and then maximize the value of the leftover. Minimizing the cutting time or the cutting cost are different objectives, which are not implemented in the solver