| In a modern manufacturing system, it is imperative that production go on efficiently. A good scheduler must allocate resources to processes with minimum waste while fulfilling all constraints of the scheduling environment. The Job Shop Scheduling Problem (JSSP) is among the most popular scheduling problems. The Flexible Job Shop Scheduling Problem (FJSP) relaxes the restrictive machine assignment of JSSP, moving it closer to a real-world application. However, it is still far from a real-world manufacturing environment, in which disruptions such as machine failure must be taken into account.;The goal of this dissertation is to create a Genetic Algorithm (GA) approach to FJSP that can adapt to disruption to reflect more closely the real-world manufacturing environment. We hope that by using just-in-time machine assignment and adapting scheduling rules, we can achieve the robustness and flexibility we desire.;The adaptive representation (AdRep) was tested in both a static environment and a disruption-prone dynamic environment. In the static environment, benchmark problems and published results were compared with the result of our approach to test its utility there. Although not as scalable as some approaches usable only for static cases, our approach discovered all the best-so-far published results on a series of commonly used benchmark problems in a strong way---it consistently produced, in almost every run, all of the points on the Pareto front produced when FJSP is formulated as a multi-objective problem, whereas most of the other approaches maximized only a single one of the objectives. Then, in the dynamic model (i.e., in which machines break down), we compared our adaptive method to two benchmark algorithms: a right-shifting rescheduler and a prescheduler. A right-shifting rescheduler repairs schedules by delaying affected operations until the disruption is over. A prescheduler works on each disruption scenario separately, treating disruptions like prescheduled downtime. Experiments showed that our approach was able to adapt to disruptions in a manner that minimized lost time. |
