Our client, given the variability of its order book and the modularity of its production lines, had to deal with more than 2 million daily scheduling possibilities.
Despite the tools in place, the solution was necessarily suboptimal in the long term. We were therefore involved in designing and developing a scheduling optimization engine based on numerous criteria (scrap, added value, production costs, deadlines, etc.)
As the factory is made up of many lines in parallel which can interface within each other, we opted for a purely financial modelling of the cost functions, and an agent-based optimization model to consider all the constraints and local earnings.
Each item, product, transfer, storage or product status was modelled as a source of costs or income in order to obtain a homogeneous space of constraints consistent with the field. This made it possible to produce a global solution which was then deployed in all the client’s factories.
in annual savings, mainly thanks to the reduction of rejects and chain reintroductions
by economy of means
On group-wide scheduling