Earth Pressure Balanced Shield Data Mining And Driving Parameter Prediction

Shield machine is a kind of heavy equipment specially used for tunnel construction.It has the advantages of fast excavation speed and high efficiency,and has gradually become the mainstream method of tunnel excavation.During the excavation process of the shield machine,it is necessary to move forward according to the designed tunnel axis,but during the specific construction period,it is inevitably affected by factors such as the driver’s operation level and the geological environment,which will cause the shield machine to deviate from the actual trajectory and cause deviation.If it is not corrected in time,it will affect the quality of the project and even cause the failure of the entire project.Relying on the driver’s subjective experience to correct deviations,which is related to their emotional state and personal skills,is sometimes unreliable,so it is necessary to develop intelligent assisted driving technology.In the past ten years,this technology has just emerged,and it is still immature,mainly because the technical solution is not established,and the result is in the experimental stage,so it is very meaningful to further study and analyze the intelligent attitude control of shield machine.In this paper,data mining and fusion model construction are carried out on the actual construction data of a subway line in a city,and the pressure of the four groups of cylinders that control the driving attitude of the shield is grouped and predicted.According to the needs of modeling,the line segmentation,geological classification,periodic processing and other operations are carried out on the line data to prepare for modeling.In this paper,the parameters that have a great influence on the attitude of the shield are selected first,and then new features with practical significance are designed through the mechanism analysis of the parameters.The importance of each feature is obtained by using the random forest model,which proves that the new feature has a high degree of importance and can participate in the modeling.In this paper,under different geological conditions,the cylinder pressure is divided into different groups,and three single models of random forest,XGBoost and LightGBM are established.The recommended number of groups is given through the analysis of the different number of groups under each model.Under the relatively good grouping,the accuracy rate of each model is almost above 0.90,but the attitude control problem is related to the quality and even the success or failure of the entire project,so the accuracy of the model prediction is extremely high.The Stacking fusion model is further used to improve the accuracy.The results show that the accuracy of the fusion model is improved compared with the single model,whether it is soft soil or rock geology,which proves that the model fusion has a certain effect on improving the performance of the model.The fusion model is tested in an actual construction line,and the test results further prove the effectiveness of the fusion model in improving the performance of the model.