Pruning Redundancy in Answer Set Optimization Applied to Preventive Maintenance Scheduling

Abstract

Multi-component machines deployed, e.g., in paper and steel industries, have complex physical and functional dependencies between their components. This profoundly affects how they are maintained and motivates the use of logic-based optimization methods for scheduling preventive maintenance actions. Recently, an abstraction of maintenance costs, called miscoverage, has been proposed as an objective function for the preventive maintenance scheduling (PMS) of such machines. Since the minimization of miscoverage has turned out to be a computationally demanding task, the current paper studies ways to improve its efficiency. Given different answer set optimization encodings of the PMS problem, we motivate constraints that prune away some sub-optimal and otherwise redundant or invalid schedules from the search space. Our experimental results show that these constraints may enable up to ten-fold speed-ups in scheduling times, thus pushing the frontier of practically solvable PMS problem instances to longer timelines and larger machines.