Research On Asphalt Pavement Performance Prediction Of Shanxi Province

The pavement performance prediction model is an important part of the pavement management system,and maintenance decision making,planning and optimization all rely heavily on pavement performance model.The prediction of pavement performance is helpful for evaluating the condition of the expressway network,making decision of maintenance and optimizing the funds on maintenance.The main purpose of this paper is to establish a prediction model of asphalt pavement performance for Shanxi expressway network.First,the inventory data of the expressway network in Shanxi and the historical data of pavement performance,traffic volume,and etc.were collected.Then,the segmentation of the expressway network was done.The misising performance data were imputed and the irregular performance data were calibrated.Next,the sections were grouped by four indices including traffic volume,pavement type,climate zone,and last maintenance type.Pavement type was divided into pavement,bridge and tunnel.The last maintenance type was divided into rountine maintenance,preventive maintenance and rehabilitation maintenance.The K-means clustering algorithm was used to divide climate zone into two categories and traffic volume into three levels.The whole network was divided into 54 groups.The Markov model was used to predict the overall performance of the network.The particle swarm algorithm was used to solve the Markov transition matrixs.The calculated Markov transition matrixs were used to predict the the overall network future performance without maintenance projects.Based on the history of the pavement performance of Shanxi expressways and the past researches,a pavement performance formula form for Shanxi was determined.Then,the regression was done by least squares estimation and the formula parameters of each group were obtained.What’s more,the t-test was performed.Finally,the prediction accuracy of the model was evaluated using three evaluation indicators,including MAE,RMSE and correlation coefficient.For PCI,the MAE of the overall road network is 3.69,the RMSE is 5.21,and the correlation coefficient is 0.87.For RQI,the MAE of the overall road network is 0.82,RMSE is1.19,and correlation coefficient is 0.86.Futhermore,the performance model was applied in the process of maintenance decesicion and the maintenance plan was given.Finally,a new deep learning model LSTM model was used to predict the PCI and RQI.In this study,two LSTM networks were developed to predict three pavement performance indicators,including PCI and RQI.The LSTM Networks were designed to predict performance indicators PI in the flowing year PI(t+1),given the PIs in the preceding three years PI(t),PI(t-1),PI(t-2)with other variables that affect the process of pavement performance deterioration.The test results of the LSTM model showed that the prediction accuracy of the LSTM model was better than the traditional ANN model and the regression model,indicating that the LSTM model can be efficiently applied to predict network level pavement performance.