TBM tunneling parameters prediction based on Locally Linear Embedding and Support Vector Regression

Tunnel boring machine (TBM) tunneling parameter prediction was conducted based on the Yinsong project in Jilin. A TBM tunneling data segmentation method was proposed to extract features from rising phase and stable phase. Thrust, cutter head speed, advance rate, torque, cutter head speed setting, advance rate setting, penetration rate, field penetration index (FPI) and torque penetration index (TPI) in the first 30 s of rising phase were extracted as the input of the prediction models. The locally linear embedding (LLE) was used to reduce the dimensions of the characteristic data of rising phase. A prediction model for TBM construction control parameters (propulsion speed, cutter head speed) and load parameters (total propulsion force, cutter head torque) was established based on the support vector regression (SVR). The impact on the prediction effect of whether to combine the FPI and TPI indexes of the previous tunneling cycle was analyzed and compared. Results show that favorable prediction effects for propulsion speed, cutter head speed, total propulsion force and cutter head torque were obtained based on the proposed model. The mean absolute percentage errors on the test set were all below 15%. The proposed method can provide guidance for TBM site operation due to the high prediction accuracy. © 2021, Zhejiang University Press. All right reserved.