by Klara Pejić, Leonie Schmidt, Christian Ennes, Konstantin von Haugwitz, Jan Sender
Abstract
This paper deals with the application of AI algorithms to improve the automatic positioning of screws in a jigless assembly line for lower seat structures, a process common in the automotive industry where positioning errors are common due to system and mechanical inconsistencies. The primary objective of the study is to evaluate the effectiveness of two AI approaches, Soft Actor-Critic (SAC) and Evolutionary Strategy (ES), in autonomously correcting these errors and improving the assembly process. The study uses a simulation-based methodology where the assembly process is simplified and modelled in a controlled environment. The SAC algorithm is used for real-time continuous control, while ES is used for global optimization. Both algorithms are tested with different parameter settings to evaluate their performance. The results show that SAC approximates solutions but lacks consistency, while ES provides more accurate results but is less efficient for re-al-time use. The study makes a contribution by exploring AI-driven solutions for process automation, filling a gap in the understanding of how these technologies can improve industrial applications. The findings highlight the potential for AI to reduce manual teachings and interventions in assembly processes, with practical implications for manufacturing industries seeking to improve production efficiency and reduce errors.
Video presentation
Presenting author
| Name: Affiliation: Email: | Klara Pejić Fraunhofer Institute for Large Structures in Production Engineering IGP, Germany klara.pejic@igp.fraunhofer.de | |
