Domaine: Manufacturing ⊕
Factsheet:
The DeepScheduling project aims to optimize steel annealing, quenching, and hot rolling operations to reduce energy consumption (and thus CO2 emissions).
Given a set of orders (i.e., steel slabs to be processed, each requiring a specific treatment), the project seeks to determine the optimal sequence for processing each steel slab at each stage. A significant energy expense is the need to adjust the annealing furnace temperatures for each steel slab, as the temperature is prescribed by the specific treatment it must undergo. Similarly, the duration and final temperature of the cooling process are also dictated by the required treatment.
Creating a schedule is thus a complex constrained optimization problem. To achieve this, the project will utilize a range of optimization technologies, including local search (oscar.cbls) and a variant called "actor-based optimization," deep learning, and optimization based on MDDs currently being studied at CETIC (DDOlib project).
