The semiconductor industry as one of the largest and fastest growing industries in the world needs to continuously reduce production costs to provide affordable products. Factory operations are likely to be major drivers to realize the necessary cost reductions in wafer fabrication facilities (waferfabs). For example operational scheduling systems powered by optimization techniques promise to replace dispatching systems as state-of-the-art control systems in the near future. Especially the capability of optimization makes scheduling systems superior to dispatching systems. Many authors share the opinion that exact optimization methods do not seem to be the method of choice in real-world scheduling systems. Instead metaheuristics and local search (LS) methods in particular are often used to solve scheduling problems with a variety of complicating constraints, since these algorithms can obtain good quality solutions within a reasonable time. The core of this work revolves around the implementation of a scheduling framework developed to deploy it as an operational batch scheduling system in the diffusion and oxidation area of a waferfab. The implemented framework is essentially a simulation-based scheduling system powered by Variable Neighborhood Search (VNS). This thesis provides the underlying theoretical background and reports valuable practical experiences from implementing a scheduling system in a real-world industrial environment. The focus of the theoretical part lies on two topics. First an extensive literature review about batch scheduling in wafer fabrication stands as one of the main pillars for this work. Secondly a detailed introduction to the batch scheduling topic with a detailed analysis of the complexity results of the most common batch scheduling problems completes the theoretical background of this work. The focus of the practical work lies between the poles of academia and industry. The implemented framework is a balancing act between academia and industry since it basically comprises two systems: the experimental system and the prototype. On one hand the experimental system offers the capability to properly investigate academic questions in the area of metaheuristic batch scheduling. The experiments show that even slight changes in the experimental setup can result in considerable changes of the output. The question is raised whether the problem instance's characteristics or the scheduling method settings have greater influence on the improvements. It is further shown by experiments how to ideally parametrize a VNS scheme optimizing schedules. On the other hand the framework's prototype is purposefully designed and developed for the needs in industry. The intention of the framework's design is to provide a functioning prototype that is suitable to run as a real-time scheduling system on the operational level. Designing, implementing and testing a scheduling system is a demanding task, but deploying it in a waferfab that relies on dispatching to that date is even more challenging. This thesis describes the top-level scheduling system with all its modeling, simulation and optimization functionalities and the underlying data level connected to the waferfab's manufacturing execution system (MES).    «

The semiconductor industry as one of the largest and fastest growing industries in the world needs to continuously reduce production costs to provide affordable products. Factory operations are likely to be major drivers to realize the necessary cost reductions in wafer fabrication facilities (waferfabs). For example operational scheduling systems powered by optimization techniques promise to replace dispatching systems as state-of-the-art control systems in the near future. Especially the capab...    »