Research on Pavement Temperature Prediction Based on Neural Network

Through data of pavement temperature, air temperature, humidity, wind speed and rainfall in the Dianjiang-Wanzhou section of G42 Shanghai-Chengdu Expressway from 2014 to 2016, it is found from the change law of the pavement temperature of this expressway section that the annual average values of pavement temperature of most monitoring stations gradually lower, but extent is not quite big; the lowest pavement temperature is between -3.2°C and 1.4°C, the highest pavement temperature is 57.9°C to 76.5°C; no matter which season, the pavement temperature and air temperature are characterized by rapid temperature rise after sunrise and significant temperature drop after sunset; the correlation between pavement temperature and air temperature is the best, the correlation coefficient reaches 0.986, and then its correlation with humidity, wind speed and rainfall decrease progressively, and they are -0.467, 0.189, 0.034, respectively. It can be found from the prediction results of pavement temperature in Chongqing Dianjiang-Wanzhou Expressway section based on neural network that the summer error is the largest, the spring is second, the autumn error is less, and the winter error is the smallest; from the angle of fitting degree of time series, the autumn fitting is optimal, the spring and winter is second, the summer effect is not good enough; the correlation coefficient between the model prediction sequence and the live sequence of the four seasons is greater than 0.89, and the overall effect is ideal. Through the research on pavement temperature, the cooperation between meteorological department and transportation departments can be strengthened, accident losses can be reduced, and the maximum benefit of road transportation can be brought into playr.