Research On The Driving Load Of Spiral Tunnels

Spiral tunnels has the advantage of spiral spreading to effectively overcome the natural height difference,but at the same time,it has the characteristics of poor visibility,small radius,continuous longitudinal slope,monotonous and closed internal driving environment,which is the bottleneck point of highway traffic safety.In the closed-loop system of human-vehicle-road,the driver is the most critical link,and the excessive driving load will easily make the driver make mistakes in perceiving the environment,making judgments and handling the vehicle,while the accumulation of driving load is also one of the main causes of driving fatigue,which affects the traffic safety of spiral tunnels.In this thesis,we investigate the driving load of drivers in spiral tunnels conditions,and further propose a driving load warning algorithm by establishing a driving load quantification model,classifying the driving load level,and combining it with driving fatigue detection,in order to explore the inner mechanism of driving load changes in spiral tunnels and improve traffic safety.The specific research contents are as follows:(1)Driving load quantification model construction.A real vehicle driving experiment was conducted in the Yaxi spiral tunnels in Sichuan Province to collect vehicle motion data,driver eye movement data and physiological data.By selecting the driver’s eye movement,ECG signal and vehicle data indicators and determining the driving load characterization indicators based on factor analysis and polynomial curve fitting,the driving load quantification model of the spiral tunnels was constructed and the spatial distribution characteristics of the driving load in the spiral tunnels were studied.(2)Comparative study of driving load in spiral tunnels.According to the driving load quantification results,the driving load distribution characteristics of the two spiral tunnels are studied separately,and the differences between the two spiral tunnels are analyzed by horizontal comparison,and the difference points are analyzed.(3)Driving load level classification prediction research.In this thesis,unsupervised machine learning algorithm K-means is selected for driving load level classification,and the driving load is classified into high,medium-high,medium-low and low load levels.By comparing machine learning algorithms such as the plain Bayesian algorithm,decision tree and random forest algorithm,the driving load characterization index and driving load class are trained to further predict the driving load class,and the random forest algorithm has high accuracy and can be applied to driving load class prediction.(4)Driving load warning algorithm.Based on Open CV + Dlib model,this thesis extracts the driver’s eye,mouth and head feature points for driver fatigue recognition,and combines with driving load level prediction algorithm to realize real-time monitoring of driver load status and timely warning.The research results can provide reference for spiral tunnel construction and provide theoretical basis for improving the traffic safety level of spiral tunnels.