Research On Tunneling Parameters Of Earth Pressure Balance Shield Based On Data Mining

In recent years,with the rapid development of China’s economy and society,the pressure on urban traffic was increasing,and the subway has become an important measure to alleviate ground traffic congestion.The shield method was the main safety mean of subway tunnel construction.However,in the past tunnel construction,the control of shield tunneling parameters mostly depends on the past experience and the operator’s on-site operation.Therefore,it has become an urgent task to find the law between the shield driving data,to clarify the relationship between the various parameters,and to provide scientific guidance for analogous case of shield construction in the future.Based on the shield data of Jinan Metro R2 Line,research on shield data mining was carried out.Data mining technologies of data preprocessing,regression analysis and machine learning modeling were used in this paper.The main research contents and results were as follows:(1)The engineering geology of Jinan Metro Line R2 was analyzed.According to the engineering geological conditions,the section between Lishan North Road Station and East Second Ring Road Station of Jinan Metro R2 Line was divided into three stratums of short columnar strong weathered diorite,medium weathered diorite.Geological characteristics and its excavation characteristics of each stratum were analyzed.(2)The useless data generated during non working hours of the shield machine were eliminated,according to the normal working state of the shield machine.Shield driving data preprocessing were carried out by means of box diagram in Python.And then the processed data were analyzed by statistical method.Research shows that the trends of total thrust,cutter head power,and cutter head torque were basically accordant,and they were very sensitive to geological conditions.The shield advance speed was basically consistent with penetration rate,and the sensitivity to geological conditions is equally high.(3)In condition of the above three stratums,correlation analysis was carried out each other,among the driving parameters of cutter head speed,cutter head torque,cutter head power,propulsion pressure,advance rate,total thrust and penetration rate.Their correlation equations were achieved.Then taking the propulsion speed as the dependent variable,the other six parameters as independent variable,multiple linear regression analysis was carried out according to the above three stratums.Penetration rate was the most important factor which influencing advance rate and cutter head speed was the secondly factor.The equation was tested by error,and the model error was within 10%.It had a better predictive effect was in line with the actual engineering.(4)Machine learning of shield performance was carried out based on tunneling big data in Python Programming Environment.Three prediction models of support vector machine,random forest and BP neural network were selected to predict the shield tunneling parameters of the above three stratums.The optimum prediction model for each stratum was the BP neural network model in accordance with error comparison.Then,geological parameters were added to predict the propulsion speed by BP neural network modeling.Comparing with the actual engineering value,the average error was 6.67%,and good prediction accuracy was obtained.