AI-Based Production System Configuration in a Learning Factory

Institut

Institut für Werkzeugmaschinen und Betriebswissenschaften (TUM-ED)

Typ

Bachelorarbeit / Semesterarbeit / Masterarbeit /

Inhalt

 

Beschreibung

Initial Situation

The introduction of different products into existing production systems requires identifying feasible combinations of available manufacturing resources. In flexible production environments, multiple workstations may be capable of performing the same assembly operations, resulting in a large number of possible production system configurations. Conventional configuration approaches rely on combinatorial optimization algorithms, which can become computationally expensive as problem complexity increases. Recent advances in artificial intelligence and machine learning offer opportunities to accelerate configuration synthesis by learning patterns from previously solved planning problems. However, their applicability and generalization capabilities in production system configuration remain largely unexplored.

Objective

The objective of this work is to investigate the potential of machine learning for the automated synthesis of feasible production system configurations using a learning factory as an application case. Initially, the available products, assembly operations, workstation capabilities, and relevant constraints will be analyzed and represented in a structured configuration model. Based on this model, suitable machine learning approaches for predicting promising operation-to-resource assignments will be investigated. A selected approach will be prototypically implemented and trained using generated configuration instances. Predicted assignments will be checked against explicit feasibility constraints and, where necessary, refined using conventional optimization methods. The proposed approach will be evaluated against a conventional constraint-based configuration algorithm using different products and problem instances of increasing complexity. The evaluation will focus on computational efficiency, feasibility of generated configurations, scalability, and generalization to previously unseen configuration problems.

Voraussetzungen

Interest in artificial intelligence, machine learning, and production planning. Good Python skills are required. Experience with machine learning frameworks such as PyTorch or scikit-learn is advantageous. Independent and scientifically structured working skills are expected.

Möglicher Beginn

sofort

Kontakt

Jannik Stöver
Raum: 1325
jannik.stoeveriwb.tum.de