| Advanced planning and scheduling (APS) has been implemented in major corporations for over twenty years. In comparison, multi-echelon inventory optimization systems (IOS) are still at an earlier point in their adoption cycle. The adoption of inventory optimization will create the opportunity to jointly solve the production and inventory problem, enabling planners to bring model-based results to a revised sales, inventory, and operations process.;Traditionally, APS and IOS are considered belong to difference planning layers and are solved independently. APS systems are every day use operational planning tool, and generally assume deterministic demand and assign a linear penalty cost for both demand and inventory shortfalls. While resource capacity is always taken into consideration in APS systems, managing uncertainty is beyond its capability. On the other hand, IOS is taken as part of the tactical planning and has long been developed to treat the demand uncertainty. But current development always sacrifices some system complexity, such as capacity constraint. Hierarchical decision making and monolithic model are the two approaches exist to solve the production-inventory problems. However, the hierarchical approach does not guarantee the solution of top-level aggregate model can generate a feasible disaggregation for the item-level problem. Most of current monolithic models still lack the capability to handle the safety stock issue in a general planning system.;This research formulates the integrated production-inventory problem faced in a semiconductor supply chain. The emphasis of our work is on finding practical solutions to support planners making production and inventory decisions as part of their operation planning process. Our work will start with iteratively solving two known optimization problems: a production planning optimization and a multi-echelon inventory optimization. While this iterative approach can admittedly produce a suboptimal solution, it represents a good start to understand the correlation of the inventory optimization problem and the production planning problem. Then we will work on an integrated model that optimizes inventory and production planning problems simultaneously.;While the problem context is one major semiconductor manufacturer, and semiconductor manufacturing does have domain-specific nuances, our resulting formulation is general enough to apply to any discrete-parts manufacturing operation. |
