Combining Inertial Sensing and Predictive Modeling for Biomechanical Exposure Assessment in Specific Material Handling Work

Abstract

Occupationally-relevant low back disorders are an important health concern in the workplace. Measuring workers exposure to biomechanical risk factors such as non-neutral postures and high force exertions is an important step for managing and mitigating the risk of such cumulative disorders. Accurate and precise quantification of exposures to biomechanical risk factors in non-repetitive material handling present unique challenges, particularly in jobs where the magnitude of hand loads vary during the workday and/or are difficult to measure directly (e.g., warehousing, construction). Wireless wearable inertial sensors offer new opportunities to acquire continuous posture data in situ and has received considerable interest in ergonomics research. This research explores several emerging issues and opportunities afforded by body-worn inertial sensing technology for quantifying biomechanical exposures and low back disorder risk particularly in relation to non-repetitive Manual Material Handling (MMH) jobs.

The specific goal of this research was to develop, implement, and assess a framework that combines occupational biomechanics principles and empirical findings, wearable inertial sensing, and predictive modeling for quantifying biomechanical exposures associated with low back disorder risk in non-repetitive MMH work. The research specifically targeted manual load carriage, lifting and lowering, which represent common MMH tasks associated with an increased risk of low back disorders.

A series of four laboratory studies were conducted. The first study quantified the effects on manual load carriage mode and load magnitude on gait kinematics measured using body-worn inertial sensors. Load carriage was found to induce systematic alterations in gait patterns and pelvic-thoracic coordination. Leveraging this information, the second study developed and assessed a statistical prediction algorithm for classifying carrying mode and normalized hand-loads using specific inertial sensor-based kinematic features as predictors. The third study extended the statistical classification model to different task conditions in load carriage and lifting-lowering. The fourth study incorporated the developed algorithms into a framework implementation for quantifying biomechanical exposures in manual load carriage, lifting, and lowering by leveraging continuous inertial sensor-based kinematic information and predictive modeling. A quantitative evaluation of the implemented framework based on empirical data from simulated manual load carriage, lifting, and lowering indicated promising results for detecting load carriage tasks (>97.0% on average) and for classifying the carrying mode (>83.1% on average) and load level (>86.2% on average). Prediction accuracy in the lifting and lowering task for classifying height and mode were also high, though accuracy was comparatively less when classifying load level (>43.1% on average).

Overall, this research makes important contributions towards our theoretical understanding of adaptations in gait and posture resulting from transient loads and task conditions. It also demonstrates the potential for combining biomechanical information, wearable inertial sensing, and predictive modeling to quantify physical exposures associated with low back disorder risk in non-repetitive MMH work. Inertial sensing in ergonomics has largely been limited to characterizing the intensity, repetitions and duration of postures. The ability to extract information about work content from motion data is underutilized. Likewise, the application of predictive modeling techniques in ergonomics remains underexplored. The presented research substantially expands the applications and opportunities afforded by body-worn inertial sensing technology and predictive modeling techniques in ergonomics.