Production control in a semiconductor production facility is a very complex and timeconsuming task. Different demands regarding facility performance parameters are defined by customer and facility management. These requirements are usually opponents, and an efficient strategy is not simple to define. In semiconductor manufacturing, the available production control systems often use priorities to define the importance of each production lot. The production lots are ranked according to the defined priorities. This process is called dispatching. The priority allocation is carried out by special algorithms. In literature, a huge variety of different strategies and rules is available. For the semiconductor foundry business, there is a need for a very flexible and adaptable policy taking the facility state and the defined requirements into account. At our case the production processes are characterized by a low-volume high-mix product portfolio. This portfolio causes additional stability problems and performance lags. The unstable characteristic increases the influence of reasonable production control logic. This thesis offers a very flexible and adaptable production control policy. This policy is based on a detailed facility model with real-life production data. The data is extracted from a real high-mix low-volume semiconductor facility. The dispatching strategy combines several dispatching rules. Different requirements like line balance, throughput optimization and on-time delivery targets can be taken into account. An automated detailed facility model calculates a semi-optimal combination of the different dispatching rules under a defined objective function. The objective function includes different demands from the management and the customer. The optimization is realized by a genetic heuristic for a fast and efficient finding of a close-to-optimal solution. The strategy is evaluated with real-life production data. The analysis with the detailed facility model of this fab shows an average improvement of 5% to 8% for several facility performance parameters like cycle time per mask layer. Finally the approach is realized and applied at a typical high-mix low-volume semiconductor facility. The system realization bases on a JAVA implementation. This implementation includes common state-of-the-art technologies such as web services. The system replaces the older production control solution. Besides the dispatching algorithm, the production policy includes the possibility to skip several metrology operations under defined boundary conditions. In a real-life production process, not all metrology operations are necessary for each lot. The thesis evaluates the influence of the sampling mechanism to the production process. The solution is included into the system implementation as a framework to assign different sampling rules to different metrology operations. Evaluations show greater improvements at bottleneck situations. After the productive introduction and usage of both systems, the practical results are evaluated. The staff survey offers good acceptance and response to the system. Furthermore positive effects on the performance measures are visible. The implemented system became part of the daily tools of a real semiconductor facility.
«Production control in a semiconductor production facility is a very complex and timeconsuming task. Different demands regarding facility performance parameters are defined by customer and facility management. These requirements are usually opponents, and an efficient strategy is not simple to define. In semiconductor manufacturing, the available production control systems often use priorities to define the importance of each production lot. The production lots are ranked according to the defined...
»