Assessment of shear strength of fine-grained and coarse-grained soil using actual EPB-TBM operating data

• Published in: International journal of geo-engineering Vol. 15; no. 1; pp. 20 - 18
• Main Authors: Samadi, Hanan, Hassanpour, Jafar, Rostami, Jamal
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 27-09-2024; Springer Nature B.V; SpringerOpen
• Subjects: Accuracy; Algorithms; Boreholes; Boring machines; Civil Engineering; Coarse-grained soils; Computing machinery; Data logging; Drilling & boring machinery; Earth and Environmental Science; Earth pressure; Earth Sciences; Engineering geology; EPB-TBM; Foundations; Friction; Geoengineering; Geotechnical Engineering & Applied Earth Sciences; Hydraulics; Internal friction; Laboratory tests; Machine learning; Model testing; Neural networks; Original Research; Parameter estimation; Sensitivity analysis; Shear strength; Soft ground; Soil investigations; Soil strength; Soil strength parameters; Statistical analysis; Statistical methods; Subway tunnels; Subways; Support vector machines; Thrust; Tunnel construction; EPB-TBM; Soil strength parameters; Computing machinery; Sensitivity analysis; Machine learning
• ISSN: 2198-2783; 2092-9196; 2198-2783
• DOI: 10.1186/s40703-024-00220-6

The necessity of predicting geotechnical parameters in soft ground tunnelling is crucial for selecting the appropriate tunnel boring machine (TBM), evaluating the operating limit of earth pressure balance (EPB) machines’ parameters, and ensuring the safety and efficiency of TBMs during tunnel construction. In this research, various EPB operating parameters such as cutterhead torque, thrust force, chamber pressure, and cutterhead rotation speed (RPM) were utilized to estimate geotechnical parameters like friction angle (φ) and shear strength (τ) for engineering geological units ET1 to ET5 (fine-grained and coarse-grained soils) along the tunnels route, which serve as indicative units for the entire tunnels path. Statistical methods and computational techniques, namely support vector regression (SVR) and feed-forward neural network (FFNN), were trained using EPB operating parameters and geotechnical data from Tehran metro line 6—southern extension sector (TML-SE6) and the East–west section of line 7, Tehran metro project (TML-EW7). A comprehensive dataset comprising borehole logging results along the tunnel path was gathered, with 85% of the data randomly selected for training and the remaining 15% reserved for model testing. Various loss functions and statistical metrics were employed to evaluate the accuracy and precision of the method. The results of the proposed models demonstrate satisfactory and reliable accuracy of the approaches.
Highlights
Database collected, generated, and analyzed using daily operational records, TBM data logger, in-situ analysis, laboratory tests, and statistical investigations;
New formulas developed to predict geotechnical parameters of soil materials by incorporating various operating parameters of the Tunnel boring machine;
Investigation of predictive abilities of smart algorithms in estimating shear strength and internal friction angle;
Comprehensive evaluation conducted to compare and contrast empirical and machine learning outcomes;
Accuracy of models evaluated using multiple assessment metrics.