Modelling and Solving Master Planning Problems in Semiconductor Manufacturing
This thesis deals with mid-term production planning problems, i.e., master planning, that arise in semiconductor manufacturing. Given the specifics of semiconductor manufacturing networks, the development of enterprise-wide planning approaches that are computationally tractable and address the uncertainties typically encountered in this industry remains particularly challenging. The purpose of production planning is to allocate limited resources to competing demands over time with respect to often conflicting economic objectives. Depending on the nature of the considered problems, production planning models may require the usage of integer-valued variables. Such models may be difficult to solve in a reasonable amount of time for large-scale instances as usually encountered in enterprise-wide planning environments. Therefore, efficient optimization approaches have to be used to reduce the computational effort while achieving optimal, or near optimal, problem solutions. The performance of the designed optimization algorithms is assessed, at first, using single problem instances. However, given the uncertainty that is typical for the semiconductor industry, there is a need for incorporating different sources of environment- and system-related disruptions into the evaluation of the planning approaches. For this purpose, a simulation model appears to adequately mimic the stochastic behavior of a semiconductor manufacturing network. In real-world situations, the planning activities occur on a regular basis; it allows for replanning the production plan by taking the current state of the input parameters into account. The investigation of the obtained rolling plans provides further insights into the performance of the planning approach used, which is usually not achievable when considering single problem instances. Production planning is complicated by the interaction between lead time and resource utilization. It is known from queueing theory that the cycle time increases nonlinearly with the utilization of the resources. However, the utilization is a result of the release schedule used. This leads to circularity in production planning. On the one hand, the planning approach determines the release schedule based on a prescribed lead time. On the other hand, the cycle time depends on the release schedule. The models presented previously in this thesis assume a fixed product lead time as an exogenous parameter of the planning approach. Although this assumption makes sense for highly aggregated strategic planning problems, it is not desirable for mid-term production planning decisions. Among other approaches, the iterations between a planning approach that determines the releases of production quantities based on a prescribed lead time and a simulation model that uses these production quantities to calculate cycle time estimates seem to adequately tackle the circularity in production planning.
Nutzung und Vervielfältigung:
Alle Rechte vorbehalten