Based On Neural Network Model For The Transport Of Atmospheric Environmental Quality Evaluation

In this paper, on basis of Artificial Neural Network(ANN), the traffic atmospheric environmental quality has been assessed in a comprehensive way and raod traffic noise has been predicted within the range of 200m along the expressway. Besides, the structural properties of network model is also discussed here m. By using LVQ network model, the concentrations of pollutants like CO, COx and TSP and noise levels at the monitoring locations along the road have been also assessed in an integrative way. The results obtained were compared with those by principal component analysis method and they were proved to be quite the same. Still, by using BP network model with LM calculation method, and based on the traffic flow, average vehicle speed, height differences at sensitive locations and the distance between the noise reception location and the road shoulder at the monitoring spots, the traffic equivalent acoustic levels have been predicted. The results turned out to be basically the same. This shows that the devised nerve network model has favorable nonlinear mapping and generalization. This paper has also dealt with the elements affecting ANN function like Hidden layer, neuron and sample number. The ralation between neuron number and errors has been considered. It has been found that the increase of neuron number can reduce errors and increase accuracy. However, it makes the network more difficult and will prolong training time. By increasing sample, generalization can be 駈proved, but too much sample can increase the scale of the network, thus leading to over-fitting. The sample and hidden layer number will need to be researched further.