Operational production planning for multi-plant metal casting using mixed integer linear programming - A case study from the aluminum industry
MetadataShow full item record
Significant market shares in the global aluminum industry belong to producers located in low cost countries. Norsk Hydro ASA is a Norwegian firm in this industry, and operate as a Make-to-Order company. They deliver cast products from multiple plants to a wide range of customers. However, the quantities, products and delivery dates of demand is only partly known at any given day. An operational production planning model is developed, aimed at minimizing costs. The model allocates production to a series of plants, and accounts for demand being a mixture of confirmed orders as well as forecasts. All this is done while taking complications such as batch production and updated demand information into account. The model is a mixed integer linear program, and it is applied with a rolling horizon approach, where demand information is updated daily. The model successfully allocates demand in a way that is consistent with Norsk Hydro's business practices, although a trade-off between optimality gaps and nervousness is observed. The potential of Lagrangian relaxations is tested on certain model constraints, with the subgradient method being used to update Lagrangian multipliers. The approach is successful in providing reduced optimality gaps. The model is developed with the aluminum industry in mind, but can be applied to any metal industry utilizing casting processes.