Accuracy of Postures Predicted Using a Digital Human Model During Four Manual Exertion Tasks, and Implications for Ergonomic Assessments

Virtual ergonomic analyses can be used to guide significant engineering decisions and, therefore, must provide an accurate representation of real-world scenarios. To increase realism and reduce simulation effort, the Jack™ software includes the ability to predict postures that individuals would adopt when performing diverse tasks. In four simulation manual exertion tasks, we compared joint angles and the outputs of ergonomic analyses that were based on the postures produced by humans and those predicted the Jack™ software. Not all postures were well predicted by the software. This work shows the sensitivity of ergonomic tools to posture accuracy and highlights the importance of professional judgement when simulating human behaviors to ensure good decisions when evaluating work-task designs. Additionally, this work may help to refine the prediction algorithms within tools like Jack™. TECHNICAL ABSTRACTBackground: To help control work-related injuries, numerous companies have elected to use computer-aided digital human simulation tools to identify and resolve ergonomic issues early in the design of the products and processes. To reduce the burden of manually posturing a digital, the Jack™ Human Simulation Software has incorporated posture prediction capabilities. Purpose: The purpose of this study was to compare the results obtained from Jack™ (version 8.0.1), in terms of joint angles and ergonomic analyses, using postures obtained from lab-based manual tasks to those predicted by the software. Methods: Twenty-seven participants completed four hand-exertion tasks. Surface marker data were transferred into the Jack™ software, which were used to reconstruct whole-body postures. Using the same foot restriction zone, hand location(s) and force efforts, the whole-body postures were also predicted using the software. Joint angles and outputs of ergonomic analyses (lumbar spine forces and joint strength capabilities) generated using the two methods were compared. Results: Significant differences between “real” and predicted postures were found for some joint angles, lumbar spine forces, and joint strength capabilities. Specifically, in the four tasks at least 5 of the 10 joints examined had significantly different angles between the real and predicted postures, with the largest difference being 40.4°. Among 15 correlations assessed, only one strong positive relationship was found between the real and predicted postures. Conclusions: An important difference in posturing strategy was noted between the real and predicted results. Our work illustrates the importance of accurate manikin postures when ergonomically evaluating work-related tasks within a digital environment, and emphasizes the need for professional judgement when simulating human behaviors even if the software has the capability to predict such behaviors. While posture prediction capabilities in Jack™ are a good way to rapidly, and repeatably create a posture, much more work is needed to understand complex movements and motion sequences, to improve existing prediction algorithms and more accurately represent human behaviors within a digital environment.