MPC Research Reports |
Title: | Assessment of Safe Work Indicators in Transportation Construction Using Personal Monitoring Systems |
Authors: | Caroline Clevenger, Moatassem Abdallah, Kevin Rens, and Mahdi Ghafoori |
University: | University of Colorado Denver |
Publication Date: | Mar 2024 |
Report #: | MPC-24-520 |
Project #: | MPC-649 |
TRID #: | 01920232 |
Keywords: | cardiovascular system, forecasting, heart rate, machine learning, occupational safety, road construction workers |
Construction projects require long hours where workers are subjected to intensive tasks such as hard manual labor, heavy lifting, and constrained working postures. Among the physiological metrics, heart rate (HR) is reported to be a good indicator of physical stress and workload. HR forecasting models have been used in various areas including cardiopathy research, heart attack warning indicator, and early physical fatigue detection. However, there are no reported studies on HR modeling and forecasting in the construction field. Modeling and forecasting the HR of construction workers using construction field data is of paramount importance due to the direct relationship between activity level and HR. The objective of this study is to (1) analyze the effect of physiological factors such as breathing rate, acceleration of torso movements, torso posture, and impulse load on the HR of construction workers; and (2) model and forecast one-minute-ahead HR for construction workers based on their physical activity using deep learning algorithms. To this end, physiological metrics of five bridge maintenance workers performing several construction activities were collected. According to the Pearson correlation and entropy based mutual information analysis, time-lagged variables, including acceleration of torso movements, torso posture, and impulse load, have a significant effect on HR data. The results of deep learning models indicate that long short-term memory network (LSTM), bidirectional LSTM (BiLSTM), gated recurrent unit (GRU), and bidirectional GRU (BiGRU) have similar predictive performance. However, LSTM had the best overall performance in HR prediction with mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) of 5.4%, 7.34%, and 5.77%, respectively. These models have the potential to facilitate the mitigation of cardiovascular strain and enable proactive prevention of accidents in the construction industry.
Clevenger, Caroline, Moatassem Abdallah, Kevin Rens, and Mahdi Ghafoori. Assessment of Safe Work Indicators in Transportation Construction Using Personal Monitoring Systems, MPC-24-520. North Dakota State University - Upper Great Plains Transportation Institute, Fargo: Mountain-Plains Consortium, 2024.