| This research aims at the development of analytical models to predict the performance of dynamic manufacturing systems that operate under frequent rescheduling, where setups occur whenever a machine starts processing a different job type. In this type of production environment, jobs arrive for processing dynamically over time and machines can fail randomly.;Although rescheduling has been studied for many years, analytical modeling of its impact on system performance has been carried out very modestly, probably because of the difficulty in modeling the randomness intrinsic of manufacturing systems. Because there is still a lot of confusion caused by many different concepts, a framework has been developed to organize the most used rescheduling ideas in three categories, namely: rescheduling techniques, strategies, and approaches. With this framework, one can now better understand the differences and similarities among different topics in rescheduling.;Three types of production systems are studied: single machine, parallel machines, and flow shops. For these systems, several performance measures have been modeled. The primary performance measures modeled are average flow time, average rescheduling frequency, average setup frequency, average machine utilization, and average schedule execution time. The secondary measures are, average processing time percentage, average setup time percentage, average repair time percentage, and average idle time percentage.;Three rescheduling strategies were examined: periodic, event-driven based on the queue size, and hybrid. In the periodic strategy, rescheduling is performed at constant time intervals. In the event-driven rescheduling based on the queue size strategy, rescheduling occurs when the number of jobs arrived for processing reaches a certain limit. The hybrid rescheduling strategy uses a combination of periodic and event-driven principles, by considering machine failure and machine repair as rescheduling events.;Two types of scheduling algorithms have been considered: normal and emergency scheduling. Both approaches group jobs of the same type to save setups and use a first-in first-out rule to sequence the jobs. The main differences between them are the scheduling moments and the jobs considered for rescheduling.;To validate the analytical models developed for all types of manufacturing systems and for all rescheduling strategies, the performance results estimated by the analytical models are compared to experimental results obtained via simulation. For this, simulation programs for each type of manufacturing system considered were to be developed. Such simulation experiments confirmed the accuracy of the proposed analytical models.;This research also has presented possible applications for the proposed analytical models. Because rescheduling creates a tradeoff between setup and flow time, these applications have shown that doing rescheduling as often as possible is not the best policy, especially when similar jobs can be grouped to save setup times, which is the case for many industrial applications. |
