Study On The Torque Prediction Of EPB Cutter-head During The Process Of Tunneling

The construction of urban rail transit plays an important role in improving urban traffic congestion and promoting urban economic development.At the meanwhile,Shield driven method plays an important role in urban rail transit construction because of its high quality of security and effectiveness.The abrasion of cutter in sand stratum during the process of shield tunneling has an obviously influence on practical engineering,such as total cost and period of construction,therefore,it is important to control the variation of cutter abrasion.Due to the lack of efficient monitoring methods,the relationship between cutterhead torque and related tunneling parameters become the evaluation criterion of cutter abrasion.By analyzing the variation of cutterhead torque in practical engineering,it could describe the status of cutter in sand stratum during the shield tunneling.Based on the monitoring values of engineering parameters of shield tunneling interval of line 6 and line 9 of shenyang metro,mathematical statistics,theoretical calculation and artificial intelligence was adopted to analyze the variation of cutterhead torque in this paper.At first,the impact factors of cutterhead torque were divided into two categories:geological conditions and tunneling parameters.The relationship between impact factors and cutterhead torque were obtained by analyzing related monitoring results.It is shown that the tunneling velocity has the most significant influence on cutterhead torque,the influence of earth chamber pressure could be ignored.The relevant data of impact factors of cutterhead torque were optimized by multicollinearity diagnosis and logarithmic transformation,after that,the theoretical solution of cutterhead torque was derived by multiple linear regression.The results of theoretical analysis showed that the torque of cutter head was affected by the cross pollination of various influencing factors.and there was an obvious nonlinear influence relationship.Two types of artificial intelligence models—neural network analysis models and support vector regression machine models were adopted to predict the value of cutterhead torque,meanwhile,the accuracy’and applicability of predictions were evaluated.By comparing the results of cutterhead torque prediction models which contains 14 impact factors,it is shown that the predictions of support vector regression machine models possess a higher accuracy,and its applicability are verified in practical engineering.