Research on prediction model of asphalt pavement subsidence development based on back propagation neural network

In order to improve the detection efficiency of expressway asphalt pavement, the development of asphalt pavement subsidence is predicted based on BP neural network. Relying on the pavement subsidence of an asphalt pavement as the research object, with time, trafficvolume, filling height and degree of foundation consolidation compression as input layer of BP neural network model of four neurons, once every half a month is carried out on the road surface subsidence measurement. Using before 24, 28, 32, 36 group data as the training sample, four BP neural network prediction model is set up respectively, and the pavement subsidence volume of 37-40 groups was predicted respectively. The obtained values were compared with the predicted results and measured values of the conic method. The results show that the prediction accuracy of the neural network model is improved with the increase of training data. Based on the consideration of engineering efficiency and prediction accuracy, 32 groups of data are selected as the best sample number. The maximum error between the predicted output value and the measured value of the network model trained by 32 sets of data is only 3.08%, and the maximum error between the predicted value and the measured value of the quadratic curve method is 9.25%. The accuracy of the BP neural network model in predicting the asphalt pavement subsidence is significantly better than that of the quadratic curve method. This proves the feasibility and effectiveness of the BP neural network model, and it provides the data basis for the research on the development trend of asphalt pavement subsidence.