Additive Manufacturing machine scheduling for multi-material LPBF 3D printing : The Schaeffler Aerosint's case

Abstract

Additive Manufacturing (AM) techniques are gaining lots of importance in industry for various applications in recent years. Those techniques, still currently studied in literature, help create objects that have characteristics that are often unachievable with conventional industrial manufacturing techniques. In this context, Laser Powder Bed Fusion (LPBF) approach allows industries to create metallic 3D objects from metal fine powder particles. The module developed by Schaeffler Aerosint, called Selective Powder Deposition (SPD), expands the process capabilities by providing a way to create multi-material 3D printed objects. Although the technical capabilities are often studied, the business aspects related to the production of those objects are mainly not. In particular, the machine scheduling problem of Additive Manufacturing processes is not extensively tackled in scientific literature. This work aims at describing those business aspects in the context of a multi-material printing framework. The main objective is therefore to develop mathematical models and optimization techniques that help optimize an AM multi-material production pipeline. The optimization of the production is defined in this work as the minimization of the total production costs for a set of objects to print. This objective is first tackled by providing a strategic overview of Schaeffler Aerosint, both in terms of internal and market aspects. This then helps review the state-of-the-art in scientific literature for AM machine scheduling considerations. A rigorous literature review indeed subsequently allows defining the key model assumptions and mathematical choices that are used throughout the work. Building upon those assumptions, the determination of the optimization framework is then considered by defining a seven-category cost structure that models the multi-material LPBF process production costs. The different parameters of this cost structure are then estimated based on real production data. Leveraging those estimations, several Mixed-Integer Linear Programming (MILP) models are subsequently built, ranging from the most caricatural approach to complex models. Through the resolution of a realistic production scenario, the efficiency and accuracy of those models are assessed and studied. The work then delves into practical considerations of the optimization. It assesses the performance of the optimization algorithm on real production data and addresses several key business questions, among which are the pricing strategy, the practical implementation of the algorithm, and the best production configuration to minimize costs. By studying those aspects, this work demonstrates the importance and complexity of the machine scheduling problem and provides strategies to mitigate economic barriers of AM technologies.