Demographic specific musculoskeletal models of factory worker performance, fatigue, and injury

We have developed a system that facilitates performance analysis of workers engaged in factory floor processes for commercial airliner production. The initial function of this system is to obtain demographically specific musculoskeletal models that reflect worker populations of interest. To this end we have developed an algorithm that adaptively tunes muscle parameters in generic musculoskeletal models using Hill-type active state muscles to generate performance data consistent with data from ergonomic subject (e.g. strength) studies of worker demographic groups. This allows flexible tuning of generic musculoskeletal models to human subject data acquired from both isometric strength tests as well as exercise dynamometers. Each dynamic exercise is replicated in a physics-based musculoskeletal simulation environment. We employ a novel algorithm to track, in simulation, the motion trajectory followed by the subject's limb(s) during the exercise, while the forces recorded on the limb(s) during the exercise are applied in the simulation. Maximum isometric forces in the muscles are then adjusted based on whether the current maximum isometric muscle forces specified are sufficient for the limb(s) to track the motion within some prescribed bounds on the tracking error. If the tracking error in the motion is too large the muscles are strengthened in the simulation until the threshold is reached at which the tracking error is just within the prescribed bounds. The tuned models derived from demographic data are used in conjunction with process inputs, like worker motions and external forces involved in a production process, to derive biomechanical performance output. This output is in the form of the muscle activation states and other biomechanical variables associated with the muscles-tendon units involved in the movement. A muscle fatigue model from existing literature been been adopted to estimate loss of muscle strength with repetitions of motions, as well as muscle recovery during worker rest periods. This provides a quantitative means for forecasting worker performance based on process inputs and offers ergonomists more detailed biomechanical tools to analyze and ultimately optimize production processes for specific worker demographic groups. Given additional data related to injury thresholds (e.g. tendon failure thresholds) the system has the potential to forecast injury likelihood, and to test protocol modifications for their ability to reduce worker injury. While simulated results, even at the level of detailed musculoskeletal models, cannot be expected to reliably predict all injury mechanisms in individual workers, our system provides valuable analytical tools that augment the limited means currently employed by ergonomists to estimate injury risk in production processes. Some preliminary results are reported for shoulder injury.