Noise generated at construction sites can cause serious health problems, such as noise-induced hearing loss (NIHL) among workers. However, the impacts on workers have not been sufficiently addressed, particularly regarding equipment-specific noise characteristics and the effect of the distance between sources and workers. Hence, this study proposed a framework for assessing occupational noise exposure using construction equipment monitoring. A convolutional neural network–based sound classification model was developed to identify equipment-specific noise with 94.59% accuracy. The proposed model also achieved an average inference latency of 0.3965 ms, outperforming visual geometry group (VGG)16 (93.99% accuracy; 1.0977 ms latency) and residual neural network (ResNet)50 (62.40% accuracy; 1.4780 ms latency), demonstrating both high reliability and real-time applicability for on-site deployment. Based on the estimated sound pressure levels, workers’ health impacts were evaluated using disability-adjusted life years associated with NIHL. The results revealed that health impacts increased as the distance between equipment and workers decreased, with the pile driver showing the greatest effect, causing an annual health damage cost of about USD 2,419.50. This framework quantitatively assesses equipment-specific NIHL, providing practical insights for improving worker health protection and preventing occupational accidents.
Oh et al. (Fri,) studied this question.