Abstract
This paper presents a fused metric for the assessment of physical workload that can improve fatigue detection using a statistical visualization approach. The goal for considering this combined metric is to concisely reduce the number of variables acquired from multiple sensors. The sensor system gathers data from a heart rate monitor and accelerometers placed at different locations on the body including trunk, wrist, hip and ankle. Two common manufacturing tasks of manual material handling and small parts assembly were tested. Statistical process control was used to monitor the metrics for the workload state of the human body. A cumulative sum (CUSUM) statistical analysis was applied to each of the single metrics and the combined metric of heart rate reserve and acceleration (HRR*ACC). The sensor data were transformed to linear profiles by using the CUSUM plot, which can be monitored by profile monitoring techniques. A significant variation between the lifting replications was observed for the combined metric in comparison to the single metrics, which is an important factor in selecting a fused metric. The results show that the proposed approach can improve the ability to detect different states (i.e., fatigue vs. non-fatigued) in the human body.