Using augmented reality for industrial quality assurance: a shop floor user study

Quality control procedures are essential in many industrial production pipelines. These repetitive and precise tasks are frequently complex, including several steps that must be performed correctly by different operators. To facilitate these, quality control tests are often documented with static media like video recordings, photos, or diagrams. However, the need for the operator to divide attention between the visual instructions and the task, and the lack of feedback lead to slow processes, with potential for improvement. By using augmented reality (AR), operators can focus on the task at hand while receiving visual feedback where it is needed. Nevertheless, existing prototypes are still at early stages, being tested only in laboratory conditions, far from mimicking real scenarios. The major contributions of this work are twofold: first, we present an AR-based quality control system capable of generating virtual content to guide operators by overlaying information in a video stream while performing real-time validation. The system evaluates the current status of the procedure to ensure the automatic progression to the next phase. Second, an evaluation was conducted in an industrial shop floor during 1 week, with seven operators to verify if the system was robust and understand possible efficiency gains when compared to the alternative, i.e., video instructions. Results showed that AR had a significant impact in procedures' execution time (reduction of 36%), while reducing the risk of human errors, which means AR technologies may represent a profitable and sustainable solution when applied to real-world industrial scenarios, in the long run.

Automated summary

Quality control procedures are crucial in industrial production, and the use of AR technology can significantly improve efficiency by providing real-time visual feedback to operators. Current AR prototypes are still in early stages and have only been tested in laboratory settings, lacking real-world applicability.