Prediction of Dust Emissions in Highway Subgrade-Filling Construction Based on Deep Neural Network

Abstract

Dust pollution can harm the urban environment and the health of citizens. Each stage in highway construction generates unorganized dust emissions to varying degrees, which complicates their quantification. To precisely forecast dust emissions during the construction of highway subgrades and reduce the associated pollution risks, this study introduces a predictive model based on a deep neural network (DNN) for dust emissions during highway subgrade-filling operations. Dust concentration is treated as a nonlinear multivariate problem, with predictive indicators encompassing particulate matter 2.5 (PM2.5), particulate matter 10 (PM10), ground surface temperature, wind speed, air temperature, surface pressure, and relative humidity. Using a DNN model, this study forecasts the concentrations of PM2.5 and PM10 at highway construction sites. Based on a highway project in Hebei Province, this study predicts dust-emission concentrations via field monitoring conducted using self-developed equipment. The model’s predictions exhibit a small mean-absolute-percentage error and root-mean-square error compared with the actual values, and the model’s accuracy significantly surpasses that of conventional regression models. Accurate forecasting can facilitate the timely control of dust concentrations at construction sites, thus facilitating more environmentally friendly and efficient construction. Doi: 10.28991/HIJ-2024-05-02-03 Full Text: PDF