![]() | Only 14 pages are availabe for public view |
Abstract In the dynamic environment of production, the task of managing and controlling production systems becomes more difficult as a result of internal disruptions like machine breakdowns, and external disruptions (such as urgent jobs). Once the initial production schedule is disrupted, a corrective action should be taken to minimize the negative effects of disruption on the overall performance. Some of the practical strategies to handle disruptions are: working for overtime, outsourcing some of the production tasks and production rescheduling. Rescheduling refers to the process of updating an existing production schedule in response to disruptions or other changes to maintain schedule performance. Production rescheduling studies considered the rescheduling problem from different perspectives. Some studies are focused on the nature of the scheduling environment which is static or dynamic scheduling. Some are concerned with the strategy of considering the uncertainty in the production system. Some are concerned with the policy that should to be followed to respond to disruptions whenever occur. Some are concerned with the method by which the disrupted schedule would be updated. Some are dedicated to find a solution for a specific machine environment problem under a specific set of assumptions. Some are concerned with finding appropriate performance measures for the rescheduling problem. Others are concerned with the optimization tool that is used to find an optimal or near-optimal solution. Most of rescheduling studies consider static set of jobs to be rescheduled upon disruption occurrence. Two alternative policies are mostly followed to respond to disruptions either upon occurrence, event-driven policy, or periodic policy to be performed at equally spaced time intervals. The method of updating a disrupted schedule by generating a completely new schedule has been proven to be superior to other methods by computational results. The disruption of urgent job arrival has been considered in most of rescheduling studies. Despite its importance, the single machine environment was rarely considered in the rescheduling studies. The objective of this study is to answer the following questions: Is rescheduling needed when disruptions occur? If yes, when to reschedule? and how often? This study followed a reactive strategy to handle unexpected disruptions in a single machine environment such that the unexpected disruptions are not accounted for while generating the initial schedule. Instead, the schedule is revised when unexpected disruptions occur according to the followed rescheduling policy. The performance of different rescheduling policies is compared in terms of the total tardiness and total cost at different disruption parameters. The disruption parameters considered in this study are: disruption duration, time of disruption occurrence and the number of disruptions. In this study, a genetic algorithm is used to generate the initial schedule and the updated schedules. Since production rescheduling may be affected by many factors and our concern is to estimate the effect of these factors on the rescheduling decision, as well as the effect of possible interactions between these factors. Using Minitab® statistical software, a general factorial experiment, in which more than two factors are considered and each factor has different number of levels, is designed to study the performance of different rescheduling policies. In each complete trial of the experiment, all possible combinations of all levels of the factors are investigated. The experiment includes the following factors: 1) rescheduling policy. 2) time of disruption occurrence. 3) Number of disruptions during the planning period. 4) Disruption duration. 5) jobs’ processing time represented by different number of jobs to be scheduled in the same planning horizon. and 6) the optimization objective (minimum total tardiness or minimum total cost). The results are first analyzed using ANOVA to identify parameters with statistically significant variation. Then the results are further analyzed and discussed using different graphical methods. The computational results showed that the timing of the occurrence of disruption as related to scheduling horizon has a major effect on determining the best rescheduling policy. Event-driven policy is found to be superior to other policies for short infrequent disruptions. On the other hand, the periodic rescheduling policy is found to be superior to other policies for long and/or frequent disruptions. |