Anti-fragmentation in aversion dynamics scheduling

Aversion dynamics (AD) scheduling is a general approach used to create special job sequences that reduce (mitigate) the potential negative impacts associated with events such as machine upgrades or repairs. Upgrades and repairs may go as planned, but they can also be unstable processes that require one or more shifts until performance re-stabilizes to a steady-state level. To consider these phenomena, special AD sequences are constructed using hybrid scheduling heuristics that combine an underlying heuristic for robust performance throughout the scheduling horizon and a specialized modifying heuristic for time periods around the anticipated event. Previous research into the behaviour of the AD heuristics examined the deterministic and stochastic characteristics of the problem and investigated the benefits associated with special job sequencing around known and anticipated events. This research explores the concept of protecting priority jobs around the anticipated event to ensure that the high-valued or important jobs do not become fragmented by the interruption. A fragmented job may require additional set-ups and the integrity of the entire job may be questionable if the process is stopped and then restarted after a repair or upgrade. This paper proposes a strategy that incorporates information regarding how far into the impact zone a job will run to determine if the job should be delayed until the risk has decreased. Sensitivity experiments wll be reported to provide insight into the tradeoffs that occur in the immediate time horizon – just before and after an event takes place. Experimental results indicate that fragmentation can be significantly reduced in many cases with only a small impact on the weighted tardiness objective function.