Experiment Study on Rock Mass Classification Based on RCM-Equipped Sensors and Diversified Ensemble-Learning Model

Area of Science:

• Geotechnical Engineering
• Machine Learning Applications
• Tunneling Technology

Background:

• Geological condition monitoring is crucial for safe and efficient tunnel construction using Tunnel Boring Machines (TBMs).
• Real-time rock mass classification aids in adapting TBM operations to varying geological strata.
• Existing methods may lack the accuracy and stability required for dynamic underground environments.

Purpose of the Study:

• To develop and validate a real-time rock mass classification system using TBM-equipped sensors.
• To evaluate the performance of a stacking ensemble-learning model for geological identification.
• To enhance the safety and efficiency of tunnel construction through improved geological monitoring.

Main Methods:

• Full-scale rotary cutting experiments were conducted using TBM disc cutters.
• Thrust, torque, and vibration sensors were integrated with a rotary cutting machine (RCM).
• A stacking ensemble-learning model utilized statistical features, including a novel vibration spectrogram-based local amplification feature, for training.

Main Results:

• The stacking ensemble-learning model demonstrated superior accuracy and stability compared to individual models.
• The model effectively integrated thrust, torque, and vibration data for rock mass classification.
• The proposed method shows significant potential for real-time geological identification during tunneling.

Conclusions:

• The developed stacking ensemble-learning model offers a robust solution for real-time rock mass classification.
• TBM-equipped sensor data, particularly vibration features, can be effectively leveraged for geological monitoring.
• This approach holds promise for advancing the safety and efficiency of tunnel boring operations.