I am currently using a genetic algorithm for optimising the production schedules of a factory that produces bespoke insulation panels.

The factory has a list of bespoke panels that need to be produced in the most optimised way.

The panels have:

  • a thickness (10 sizes)
  • type of material (20 types)
  • unique width and length
  • angle left/right (not all panels are 90 degrees)

The panels are produced on tables. There are max 2 panels per table (they most both fit on the table (combined with of the panels must be smaller than the width of the table).

It is best to combine panels with the same thickness and type of material. This saves a lot of manual work. If 2 panels with (almost) the same angle can be combined on a table, only one manual action needs to be done by a worker.

We are currently using a genetic algorithm to calculate the best combinations in the given list. The fitness is calculated by using the thickness, material type, angle (and sum of panel width < table width)

Are there other types of algorithms/AI that can be used and would be much less resource intensive? How could we approach these?

  • $\begingroup$ GA sounds not too bad, have you tuned stuff like mutation rate and population size? It could also be a fit for integer programming. $\endgroup$
    – maxy
    Commented Aug 1, 2023 at 11:09


You must log in to answer this question.