Abstract
The primary objectives of this study were: (1) Development of an artificial neural network (ANNs)-based physiological model that could accurately predict the energy cost associated with manual lifting/lowering-activities of short duration, and (2) Development of an artificial neural network (ANNs)-based physiological model that could accurately predict the energy cost associated with manual lifting activities performed for long duration. Both models apply to a wide range of workers and task characteristics.For each, the following steps were implemented: (1) Design of the artificial neural network models, (2) Evaluation of the network's prediction ability to different input combinations, (3) Evaluation of the network's prediction ability to different network variants, (4) Assessment of the network's validity to novel situations, (5) Selection of the ANN model with the best predictive ability, (6) Development of prediction models using Multiple Linear Regression (MLR) techniques, (7) Comparison of the the ANN models' predictive ability to the MLR models' predictive ability.Two short-duration ANNs were developed to predict energy cost in response to different short-duration manual-lifting/lowering tasks: a floor-height model and a table-height model. The data utilized in developing these models were adopted from Asfour (1980), and the models were constructed using Multilayer Feedforward networks (MLFF).Similarly, four ANN models were developed to predict the energy cost to different long-duration manual-lifting tasks: a floor-height model, a table-height model, a shoulder-height model, and an all-heights model. The objective was to develop models that could be used to predict an individual's oxygen consumption at five-minute intervals up to the endurance-time limit.The ANN-based physiological models demonstrated predictive ability superior to that of the corresponding MLR physiological models. This study provides evidence that ANN could be a valuable tool in alleviating many current modeling problems in the Ergonomics field. ANNs provide a useful alternative to the contemporary modeling techniques used to predict human responses to physiological stresses associated with different activities. The developed ANN-based physiological models could be utilized in risk assessment and management. (Abstract shortened by UMI.)