Construction Fugitive Dust Quantification Modeling Based On BP Neural Network

Construction fugitive dust is the main source of atmospheric particulate pollutants, so it deserves to research and control it. Quantization dust emissions are an important means to assess effect of fugitive dust control measures. By monitoring the construction site near the boundary coordinates of the same plane within 3.9～18.9m under different weather conditions, different height of the dust concentration, summed up the construction dust generated by the main factors. The main factors affecting the construction fugitive dust are: dust moisture, air temperature, air humidity and construction strength. The results show that construction dust pollution in the foundation excavation, foundation construction, and backfilling the foundation stage are the most serious, followed by the main construction phase, the dust pollution produced by construction phase surface decoration is negligible, construction equipment Installation phase produces almost no dust pollution.Using BP neural network to simulate the process of construction fugitive dust, and get quantitative model under certain constraints. After BP neural network being trained, optimized, take 140 hidden layer neurons and the maximum training period of 1000 weeks; select the Levenberg-Marquardt optimization algorithm as hidden layer activation function to determine a three layer shape network model for 4-140-1 finally. The degree of input and output data related to 0.93786, the network performance of the final output after training is 5.94 x 10'28, results are satisfactory. The results show that the model can predict the dust pollution concentration trend in different weather conditions well, and provide the scientific basis for construction site dust migration, proliferation and pollution control. The method has wider application prospect.