Study On Influencing Factors Of Pavement Icing And Construction Of Prediction Model

Road surface icing seriously affects road operation safety.Monitoring road surface icing status and predicting icing time can effectively guide road surface icing countermeasures,reduce accident frequency,study road icing influence factors,and construct road icing time prediction model to study road surface.One of the important aspects of icing monitoring.At present,the research on most road icing monitoring is limited by the development and application of icing sensors,which is difficult to meet the measurement range and accuracy.The prediction models are mostly constructed by nonlinear regression,and the accuracy of the model needs to be improved.In recent years,the icing sensor for roads has been gradually developed and applied.The network learning method is more suitable for the training and prediction of small sample data.Both of them provide theoretical and technical support.Therefore,based on the research and development and performance analysis of icing sensors,combined with the road weather station for outdoor and indoor simulated icing tests,the influencing factors and influence laws of icing time are analyzed,and the neural network prediction model is constructed based on the collected data samples.The main research contents and results are as follows:In the first place,comparing different types of icing sensors,a piezoelectric icing sensor is selected as a road sensor for the road environment.The sensor is frequencycalibrated under different conditions(dry,accumulated water,icing)to obtain the corresponding relationship between different states and frequencies on the sensor.The accuracy,long-term and anti-interference of the sensor are analyzed separately,which proves that the piezoelectric icing sensor has better working performance.Studying the layout method and data transmission method of piezoelectric icing sensor,all meet the needs of road icing monitoring.Secondly,the outdoor icing test was carried out,and the icing sensor installed outside was used to collect the road surface state data,and the meteorological data collected by the small weather station was used to analyze the influence of various factors on the icing time.According to the correlation between various influencing factors and icing time,the pavement temperature,wind speed and icing thickness were selected as the influencing factors of indoor quantitative test.According to the orthogonal experiment,the importance of different factors and the significance of the influencing factors were compared.Comparing the results of indoor and outdoor test analysis,the unity of indoor and outdoor test methods is proved.At last,the characteristics and shortcomings of the existing prediction models are analyzed,and the model suitable for road icing prediction is selected according to the characteristics of the model.The multivariate polynomial prediction model and BP neural network prediction model were constructed respectively.The quantitative test data was used as the data sample to construct the icing time prediction model.The running results of the prediction model under different parameters were compared and analyzed,and the optimal parameters of the model construction were selected.The two simulation models are validated by using outdoor simulation test data,which proves that the neural network prediction model is superior to the multivariate polynomial prediction model in accuracy and dispersion.