Biomechanical models are used in many situations to help understand the risks associated with performing different work tasks. A necessary input to most biomechanical models is the body posture of the worker. Measuring posture has proven to be a difficult and time-consuming process. The research reported in this thesis investigated if a posture can be predicted instead of measured.

The posture prediction model employs a whole-body sagittal plane representation of the worker with five links using inverse kinematic procedures to calculate the postures. The model chooses a posture by optimizing an objective function using a nonlinear programming search. Three separate models have been formulated to predict the postures of 16 subjects humans performing four static sagittal lifting tasks. Each model uses a different objective function or criterion defined relative to the torques on the human joints. These criteria are labeled as Total Torque, Percent Strength, and Balance. The influence of gender, hand position, and criteria on the prediction accuracy were investigated.

The results showed that there was less postural variability for higher hand positions compared to lower hand positions. For lower hand positions there were two distinct types of postures chosen by subjects which implies that there are two different types of criteria being used by subjects at these hand positions. Student t tests, which investigated the accuracy of the predictions, showed that all of the prediction errors were significantly greater than zero at α=0.05. A mixed factor, repeated measures ANOVA investigating the prediction error showed that the Total Torque criterion was more accurate than the two other criteria.