A Reactive Scheduling Approach Based on Domain-Knowledge

Abstract

Real world industrial environments frequently face unexpected events that generallydisrupt in-progress production schedules. This contribution presents advances in thedevelopment of a support framework to address the repair-based reactive scheduling of industrial batch plants. When facing an unforeseen event, the framework is capable of capturing the current operational plan and its status. Based on this information, a rescheduling problem specification is developed. Tasks to be rescheduled are identified and, for them, the set of the most suitable rescheduling action types (e.g. shift, reassign, etc.) is specified. For a given specification, many solutions to the problem could exist. Then, the second step of this approach relies on the generation of a constraint programming (CP) model to address the rescheduling problem just specified. To create such model each rescheduling action type is automatically transformed into different constraints. In addition, a search strategy based on domain knowledge can also be developed. Finally, the solution of the CP model and its associated search strategy will render the repaired schedule in which the repair action types that were suggested will be instantiated. A case study of a multiproduct multistage batch plant is presented, where an event of unit failure is considered.