Knowledge-based and data-based machine learning in intelligent TBM construction

This paper presents a comprehensive review on the application of tunneling feature parameters FPI and TPI within the context of the Yinchuo Jiliao Project (YC) and Yinsong Water Supply Project (YS) . These parameters are extracted using knowledge-driven methods in TBM (Tunnel Boring Machine) intelligent construction, leveraging theoretical analysis, statistical testing, and clustering techniques. The research encompasses intelligent predictions for both rock classifications and tunneling parameters. The findings of this study indicate that utilizing FPI and TPI obtained through knowledge-based approaches can lead to reduced data dimensions and noise, consequently enhancing prediction efficiency. In terms of intelligent prediction of surrounding rock classifications, both knowledge-based and data-based methods have demonstrated commendable accuracy levels. However, concerning the prediction of tunneling parameters, knowledge-based approaches exhibit significantly higher prediction accuracy than the data-driven methods. Therefore, the study suggests that the independent utilization of FPI and TPI or in conjunction with data-based parameters can all be used for the machine learning predictions within the TBM field. © 2024 Chinese Society of Civil Engineering. All rights reserved.