Entwicklung und praktische Anwendung eines hybriden Umlagerungsmodells in der Prototypenteilelogistik der Automobilindustrie

Abstract

In logistics, transshipment models can lead to a significant reduction in transport costs and to realization of time advantages. A large number of reactive and proactive stock transfer models have already been presented in literature. Hybrid transshipment models, which combine reactive and proactive transshipments, have been researched less in literature. The special characteristic of prototype parts logistics is a batch size = 1, which means that each part can only be stored in one warehouse and cannot be replenished from the upstream echelon. As the few existing models proactively balance the stock > 1 of a part across several warehouse locations, they cannot be adapted to the special characteristic of prototype parts logistics in the automotive industry. Due to short-term demands of various assembly sites for the assembly of prototype vehicles and the need for a fast delivery, transportations to assembly sites and between warehouses are planned on a daily basis. The objective of the present work is to develop a new transshipment model specifically for unique parts and to validate it in practice. The aim is to store unique items in the best possible warehouse so that the right part can be delivered at the right time, to the right place and at the right costs. In the newly developed transshipment model, demand probabilities are determined for each part and each assembly location, which are used to calculate approximated future shipping costs as an indicator of the expected best warehouse. By combining a weighted and lexicographic target function, in which, for example, the minimization of the approximated future shipping costs is included, remaining vehicle capacities are used in the hybrid approach to implement proactive transshipments. The influence of various characteristics in the logistics network on the hybrid transshipment approach is investigated in a numerical study. For the model validation, the transshipment model is adapted to a practical application in the prototype parts logistics of Volkswagen AG and a case study is carried out. A machine learning model is trained and applied to forecast demand probabilities in the case study. In addition, a rule-based heuristic algorithm is developed to implement the hybrid approach in practice. The decision rules are integrated into a decision support system that generates a recommendation for the logistics planners on which parts to transship. The case study does not only show that the hybrid approach saves transportation costs, but also that the delivery time for customers can be significantly reduced. The model is formulated in such a way that it can be adapted to different use cases with unique items and demand probabilities.