Rolling Partial Rescheduling Driven by Disruptions on Single-machine Based on Genetic Algorithm

This paper discusses large-scale single-machine rescheduling problems with efficiency and stability as bi-criterion, where more than one disruption arises during the execution of an initial schedule. Partial rescheduling (PR), which involves only partial unfinished schedules, is adopted in response to each disruption and forms a PR sub-problem. The remaining unfinished schedule is just right-shifted or not following the solution of PR sub-problem. During the process of schedule execution, a rolling PR strategy is driven by disruption events. Each global rescheduling consisting of two segments of local rescheduling revises the original schedule into a new schedule, which is exactly the next original schedule. Two types of local objective functions are designed for PR sub-problems locating in the process or the terminal of original schedules respectively, where the global information of bi-criterion problems is reflected to an extent. The analytical results demonstrate that each local PR objective is consistent to the global one. For PR sub-problems with such a particular criteria, a genetic algorithm is used to solve it. Extensive computational experiments were performed. Computational results show that the rolling PR can greatly improve schedule stability with a little sacrifice in schedule efficiency and consistently outperforms the rolling right-shift rescheduling. The rolling PR strategy is effective to address large-scale rescheduling problems with more disruptions