Machine Learning Based Optimization of EPB-TBM Control Parameters

Konstantinos N. Sioutas^*, Andreas Benardos
School of Mining and Metallurgical Engineering, National Technical University of Athens, Athens, Greece
* Corresponding Author: Konstantinos N. Sioutas. Email: email
(This article belongs to the Special Issue: Advances in Artificial Intelligence for Geotechnical Engineering)

Computer Modeling in Engineering & Sciences https://doi.org/10.32604/cmes.2026.087755

Received 22 June 2026; Accepted 25 August 2026; Published online 10 September 2026

Download PDF

Abstract

Tunnel operations performed with Tunnel Boring Machines (TBMs) require the selection of feasible setpoints under changing ground and site constraints. In this study, an inverse Machine Learning (ML) setup was constructed to infer operating parameters from geotechnical context and target performance thresholds for Penetration Rate (PR) and the utilization (UTIL) of the machine. Inputs included geological and geotechnical descriptors, relative depth, a categorical geological profile, harmonic time features, and the two targets (PR and UTIL). Outputs comprised Cutterhead Rotation and Torque, Total Thrust, Screw-conveyor Rotation and Working Pressure, Excavating Rate, and Earth Pressure on the face of the tunnel. A Feed-Forward Artificial Neural network (OPT_ANN) trained with Adam, batch normalization, dropout, and early stopping achieved test Root Mean Squared Error (RMSE) ≈ 0.75 and Mean Absolute Error (MAE) ≈ 0.53 aggregated across outputs. Indicative maps using thresholds yielded coherent operating ranges that reflect coupled controls such as torque, thrust and screw-conveyor pressure, throughput. The approach converts prediction into decision support by summarizing feasible scenarios for field operations.