Multiple Criteria Scheduling Using Genetic Algorithms in a Shipyard Environment

The main aim of scheduling large manufacturing processes is to minimise the time taken to manufacture the product from start to finish, and maximise the utilisation of machinery/ personnel. However, as the number of tasks increases it becomes more difficult to find good schedules within a reasonable time. This paper explains the development of a Multiple Criteria Genetic Algorithm (MCGA) scheduler which evolves good schedules based on several criteria. The system is then demonstrated using a shipyard steel-shop problem. INTRODUCTION In most large scale engineering projects production scheduling plays an important role in making efficient use of resources and reducing manufacturing time. Research in scheduling has traditionally concentrated on using heuristics, ordered search or applying mathematical algorithms. However, as the size of the scheduling task increases so does the difficulty of finding good solutions within a reasonable time. It is common to split the task into a hierarchy (levels of scheduling), where the highest level considers the main events (milestones) and the lowest level is scheduling of workstations. This makes the project more manageable. The high level scheduling is performed first and is used to set targets for the next level down. If these targets cannot be met then re-scheduling occurs. If suitable schedules still cannot be found work may be sub-contracted or resources increased with associated cost penalties. For a major project three or four levels may be required. In the case of Made-To-Order (MTO) products such as offshore structures and specialised marine vehicles there are additional constraints to handle. This increases the complexity even further. It is important to ensure that the processes of design, manufacture and assembly are integrated to their fullest extent to ensure the realisation of appropriate benefits. It is also often the case that orders are governed by strict deadlines with