Scheduling Steel Production using Mixed-Integer Programming and Constraint Programming

We present an overview of a system we have developed for generating short-term production schedules for a large steel manufacturer. We have developed a decomposition based approach which uses mixed-integer programming to generate a plan for production on the downstream casting processes, and constraint programming to generate a schedule upstream of these casting processes, taking into account detailed scheduling constraints and preferences. One drawback to this approach is that downstream production plans can be generated which cannot be scheduled on the upstream processes, since downstream planning does not take into account bottlenecks that may occur when scheduling upstream. We investigate an integration mechanism, which uses constraint programming to build local estimations of upstream capacity utilization, which are used in the integer programming formulation for downstream production planning. We present results illustrating the effectiveness of this approach.