Manual material handling (MMH) tasks are associated with the development of work-related musculoskeletal disorders (WMSDs). Minimizing the frequency and intensity of handling objects is an ideal solution, yet MMH remains an integral part of many industry sectors, including manufacturing, construction, warehousing, and distribution. Physical exposure assessment can help identify high-risk tasks, guide the development and evaluation of ergonomic interventions, and contribute to understanding exposure-risk relationships. Physical exposure can be evaluated using self-assessment, observational methods, and direct measurements. Nevertheless, implementing these methods in situ can be challenging, time consuming, expensive, and infeasible or inaccurate in many cases. Thus, there is a critical need to improve physical exposure assessments to protect workers and save costs.
This dissertation assessed the accuracy of a markerless motion capture system (MMC) to quantify physical exposures during MMH tasks using three studies. Specifically, the first study investigated the performance of an MMC system, together with machine learning algorithms, for classifying diverse MMH tasks during a simulated complex job. In the second study, the feasibility of predicting dynamic hand forces was determined, using alternative measures, such as kinematics from MMC and/or in-sole pressure systems, coupled with a machine learning algorithm. Finally, in the third study, we systematically evaluated MMC for assessing biomechanical demands, by comparing outputs from a full-body musculoskeletal model driven by kinematic and kinetics from gold standard input and estimates derived from the MMC and in-sole pressure measurement system.
Overall, the findings of these studies demonstrated the potential of using MMC to classify several common occupational tasks and to estimate the associated biomechanical demands for a given worker (automatically and with minimal physical contact). Additionally, the methods developed here can help stakeholders rapidly assess an individual worker's exposure to physical demands during diverse tasks. / Doctor of Philosophy / Manual material handling (MMH) tasks expose workers to known risk factors for work-related musculoskeletal disorders (WMSDs) such as back and shoulder pain. Accurately quantifying workplace exposures to these risk factors is an essential aspect of identifying high-risk working conditions and for developing/evaluating workplace interventions to reduce WMSD risks. Current physical exposure assessment tools are labor-intensive, offer crude measures, and have limited application due to costs or feasibility. Using markerless motion capture (MMC) systems in the workplace could enable full or partial automation for the collection of critical measures such as the tasks a worker performs, the hand forces involved, and their biomechanical demands. New approaches are needed, though, since such automation is challenging due to variations in the type of input data required for different physical exposure assessments. In this dissertation, our goal was to assess the accuracy of MMC as a tool to quantify physical exposures during MMH tasks. In support of our goal, three studies were completed.
In the first study, we investigated the accuracy of using data from MMC together with machine learning algorithms to classify diverse MMH tasks, and distinguish among different task conditions. Our results emphasized that classification performance was satisfactory, though it differed between feature sets, MMH tasks, and between males and females. The second study explored combining MMC and IPM data with machine learning algorithms to predict hand forces during MMH tasks. Our results were encouraging overall, but predictions were less accurate in pushing and pulling tasks. In the third study, we evaluated an approach for estimating biomechanical demands on data obtained from MMC and in-sole pressure measurement systems. We compared estimates from a musculoskeletal model driven by kinematics from a whole-body inertial measurement unit and kinetics from direct measures of hand loads, and kinematics from MMC. Our findings support using MMC and kinetics from predicted hand forces as input for estimating biomechanical demands.
Overall, findings from these studies show that MMC can automatically classify common occupational tasks, predict dynamic hand forces, and estimate biomechanical demands with minimal physical contact. This new approach could allow stakeholders to assess worker's exposure and the efficiency of ergonomic interventions.
| Identifer | oai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/118403 |
| Date | 12 March 2024 |
| Creators | Ojelade, Aanuoluwapo Ezekiel |
| Contributors | Industrial and Systems Engineering, Nussbaum, Maury A., Kim, Sun Wook, Dickerson, Deborah Elspeth, Paige, Frederick Eugene, Tsui, Kwok |
| Publisher | Virginia Tech |
| Source Sets | Virginia Tech Theses and Dissertation |
| Language | English |
| Detected Language | English |
| Type | Dissertation |
| Format | ETD, application/pdf, application/vnd.openxmlformats-officedocument.wordprocessingml.document |
| Rights | In Copyright, http://rightsstatements.org/vocab/InC/1.0/ |
Page generated in 0.0026 seconds