AI-Driven Predictive Maintenance for Industrial Robots in Automotive Manufacturing: A Case Study

Abstract

Manual assembly tasks are often labor-intensive and prone to errors. However, with the rise of artificial intelligence technologies, these tasks can be automated with the help of robots. Most automotive manufacturers are implementing the robotization of assembly tasks. However, the type of assembly processes in the automobile manufacturing industry is complex, meaning that maintenance of the robotic arms is exceptionally critical. Scheduled maintenance costs massive downtime of the robotic arm, which concerns both manufacturing throughput and financial losses. On-demand predictive maintenance could optimize repair plans to be performed only when necessary while maintaining a high uptime of the robotic arms.
Therefore, a new framework that learns functional generic product representations and transfers knowledge across different domains is proposed. Then, a case study on on-demand predictive maintenance for industrial robots in the automobile manufacturing industry is presented. The experimental results show that the proposed framework could work in an unseen assembly environment, and knowledge transfer increases predictive maintenance performance. One of the first fully developed robotic arms with torque sensors in the automobile manufacturing industry is used, which is allowed to be studied offline. In contrast to industrial settings, all configurations are controlled directly on the robotic operation script level, making it easy to construct different scenarios of different assembly processes.