Abstract
Manual assembly processes are prevalent in manufacturing facilities despite the proliferation of automation and robotics. Humans are adaptable and our capability to make on-the-fly decisions about assembly quality is not yet replaced by artificial intelligence. Therefore, there is a need to ensure that humans are trained well to perform manual assembly tasks. To this end, the research presented in this paper investigates the use of tolerance bands around accelerometer data for the evaluation of manual assembly tasks.
Previous research has identified methods to assess consistency of manual assembly tasks performed. One such approach is based on accelerometer time series data obtained from wrist-worn inertial measurement units. Consistency is computed as a function of point-to-point matches between time series data from two assembly process cycles. This approach assesses minor task performance variations, which may have been a result of desired human adaptability, as inconsistent.
To account for human adaptability within the context of assembly process consistency, the use of tolerance bands on accelerometer data is investigated in this research. Two experiments were conducted where three assemblers assembled a total of three products multiple times. Wrist-worn inertial measurement unit data from these experiments were analyzed based on tolerance bands and inconsistencies are identified using a statistical process control tool - control charts.
It was found that the proposed method was able to identify inconsistencies in the assembly processes performed. A post-hoc video analysis of the inconsistent assembly cycles showed that there are opportunities to improve workstation layout, product design, and work instructions.
