Study On The Performance Prediction Of Asphalt Pavement Based On Big Data Technology

The performances of asphalt pavement include many aspects, such as riding quality, pavement surface condition, skidding resistance, pavement structure strength and so on. Each asphalt pavement performance index is affected by various internal factors and external factors, which have cross effects among themselves. Too many performance indexes as predictors make against proper and efficient forecast through big data technology. According to surveyed data, approximately 80% of asphalt pavement maintenance is caused by the existence of rutting depth, of which harm is of the highest level compared with other diseases damaging the asphalt pavement. Besides, rutting has an adverse impact on the comfort and safety of traveling, even initiates or worsen the diseases. For these reasons, rut depth is an very important index to the asphalt pavement performance, and how to predict the difference of rut depth effectively has great significance for the predition of asphalt pavement performance. This thesis presents a method of rutting difference prediction of asphalt pavement, based on big data techniques and viscoelasticity and analysis of LTPP data, in order to predict the change of asphalt pavement porfermance and lay a foundation for the survey of the change of asphalt pavement performance. The data used in this thesis is from LTPP with big data characteristics of complex source, large volume and high potential value and it is suitable to use big data technology to achieve integration, sharing and cross multiplex to form intellectual resources and knowledge service platform.First of all, asphalt mixture is a typical heterogeneous viscoelastic composites. This thesis analyses the causes of rutting formation and influence factors of rutting difference based on the viscoelasticity. Then air temperature is chosen as the representative of the climate module, cumulative equivalent axle loading times as the representative of the traffic module and air void as representative of the materials characteristics module.Next, rutting and influence factors data are extracted from LTPP database and reconstructed as four new databases by distributed storage system SQL Server. Jmp is used to analyse the distribution of air temperature, traffic loads and air void in space and time on the basis of big data techniques. Through this means, one year is divided into four quarters and four corresponding models are established to predict the rutting difference of each quarter respectively.Furthermore, the three influence factors are analyzed by factor analysis method using SPSS and the results report that the correlation coefficient among three arguments is very low and every factor is independent from each other. Sensitivity analyses of them are conducted through partial correlation method, acording to which that there is no significant linear relationship between air temperature and rutting difference as well as between air void and rutting difference, however, there is a relatively significant linear relationship between cumulative equivalent axle loading times and rutting difference. It is suggested that traffic loading has the greatest impact on the rutting difference followed by average value of highest air temperature and air void of top course has a relatively low impact on that.At last, four rutting difference predicting models are established considering temperature effect, traffic effect and asphalt mixture characteristics effect for four different quarters, in reference to the mathematical expression of viscoelastic mode. The extracted and pre-treated LTPP data are used to work out the parameters of the four models by fitting based on Levenberg-Marqurdt algorithms. The fitting results indicate that each model for four quarters has a good predictive ability with good predition accuracy.This thesis develops around key concept'let data itself says', based on many kinds of big data technology analysis, to predict differences of rutting depth happening in corresponding season section using LTPP data. Actually, rutting depth is expected to be in a slower development through adjusting influencing factors. According to relative criterions, the difference of rutting is used to analysize the change of other indexes of asphalt pavement performance, and effective maintenance measures are suggested to take to ensure that the road works functionally and provides sufficient safety.