Robust and reactive scheduling for multi-objective flexible job shop problems with unpredicted arrival of new jobs

Abstract

Scheduling in production systems concludes the proper coordination of activities in order to increase productivity and reduce operational costs. In dynamic manufacturing environments, scheduling solutions based on the classical objectives such as makespan will not be sufficient. In fact, because of random disruptions that may occur in the system, additional criteria that have capability to counter such disruptions should be considered. To maintain system performance effective, rescheduling is often used to counteract the effects of random disruptions. In practical production environments, the scheduling process starts with determining an initial schedule. Then, when a disruption arises, the initial schedule should be revised in order to keep its feasibility and performance quality. The type of scheduling that is actually carried out in shops is known as real schedule. As it is clear, real schedule can be different from the initial schedule. This difference depends on the level of failure and disruption and also the changes of the setting. There are two policies to achieve a high level of system performance for the real schedule after occurring of any disruption. These strategies are entitled reactive scheduling and proactive scheduling. The “reactive approach” does not consider the uncertainty when an initial schedule is determined. However, when a random event occurs, it modifies the initial schedule and performs the necessary reaction to obtain better result. This reaction can be in the form of modification and improvement of the initial schedule or the formulation of a totally-new schedule. On the other hand, the “proactive approach” considers the stochastic and unexpected events to create the initial schedule. In this approach, in addition to classical criteria such as makespan and tardiness, performance measures such as robustness and stability is also considered to establish a schedule. Optimization of stability is concerned with the deviation of the modified schedule relative to the initial schedule. Optimization of robustness is concerned with the different in terms of objective function (performance criteria) between initial and modified schedules. An integrated proactive–reactive approach can also be considered to generate better and practical results. In this thesis, a two-step proactive–reactive method is presented foflexible job shop scheduling to achieve a more stable and robust solution. In the first step, it is attempted to generate an initially robust schedule by using robust optimization approach. The initial robust schedule handles the uncertain processing times. In the second step, when a random disruption occurs (which is the arrival of an unpredicted new job), an appropriate reaction is adopted to determine the best modified schedule