An artificial neural network model for predicting fatigue of construction workers in humid environments

Climate change is experienced in many countries located in tropical/subtropical regions with generally hot/humid condition. Heat illness, particularly heat stroke, has caused a substantial increase in morbidity and mortality during heat waves. Thus, the high incidence of heat stroke is a pressing concern in the construction industry. Construction workers, being exposed to such unpleasant working environment, are at a higher risk of heat stress while undertaking physically demanding tasks. This paper aims to establish a model for predicting fatigue of construction workers in hot weather. During the period of summer months in 2010 and 2011, we conducted 39 field measurements on six construction sites in Hong Kong and collected a series of meteorological, personal, and work-related parameters. A total of 550 synchronized datasets were measured to establish the model. Artificial neural networks (ANNs), a type of artificial intelligence technology which implements more complex data-analysis features into existing applications, was applied to forecast the fatigue of construction workers. Performance measures including mean absolute percentage error (MAPE), R2, and root-mean-square deviation (RMSE) confirm that the established model is a good fitting with high accuracy. The ANN-based model presents a reliable and scientific forecast physical condition of workers which may enhance the occupational health and safety (OHS) in the construction industry.