Predictive Modeling of Tunnel Boring Machine Performance

Predictive modelling of tunnel boring machine performance integrates geological, mechanical and operational data to forecast key metrics such as penetration rate, advance rate and cutterhead torque. Early approaches relied on empirical correlations and statistical regression to link rock mass properties—such as uniaxial compressive strength and rock quality designation—with machine outputs. Advances in computational power and data acquisition have ushered in hybrid frameworks that combine optimisation algorithms, fuzzy inference systems and machine learning techniques. Contemporary models exploit deep learning architectures, including recurrent neural networks, to capture temporal dependencies in tunnelling operations and account for the heterogeneity of geological strata. By accurately predicting machine behaviour in real time, these models support cost estimation, schedule planning and risk mitigation for complex underground projects worldwide. Practical applications range from automated pressure regulation in earth pressure balance shields to anticipatory alignment corrections, enhancing safety and productivity in urban metro and intercity tunnelling schemes.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Predictive Modeling of Tunnel Boring Machine Performance publication trend

The graph below shows the total number of articles in predictive modeling of tunnel boring machine performance across all publications each year (not limited to Nature Index journals).

Technical terms

Rate of penetration (ROP): Speed at which the TBM advances through the ground, typically expressed in millimetres per minute or metres per day.

Advance rate: Increment of tunnel length achieved by the machine in a given time interval, encompassing cutterhead rotation and thrust.

Earth pressure balance (EPB): Shield tunnelling method that maintains balanced face pressure by controlling the properties of excavated soil and support slurry.

Gated recurrent unit (GRU): Recurrent neural network cell designed to capture temporal patterns in sequential data with fewer parameters than long short-term memory units.

Light gradient boosting machine (LightGBM): Efficient tree-based gradient boosting framework for classification and regression tasks, noted for fast training and high accuracy.

SHAP (Shapley additive explanations): Model-agnostic technique that quantifies the contribution of each feature to a prediction based on cooperative game theory.

Fuzzy C-means clustering: Unsupervised algorithm that assigns data points to clusters with varying degrees of membership, accommodating uncertainty in group definitions.

References

1. Real-Time Dynamic Earth-Pressure Regulation Model for Shield Tunneling by Integrating GRU Deep Learning Method With GA Optimization. IEEE Access (2020).
2. Prediction of Axis Attitude Deviation and Deviation Correction Method Based on Data Driven During Shield Tunneling. IEEE Access (2019).
3. Regression Models and Fuzzy Logic Prediction of TBM Penetration Rate. Open Engineering (2017).
4. Identification of geological characteristics from construction parameters during shield tunnelling. Acta Geotechnica (2022).
5. Prediction of rock mass class ahead of TBM excavation face by ML and DL algorithms with Bayesian TPE optimization and SHAP feature analysis. Acta Geotechnica (2023).
6. Real-time analysis and prediction of shield cutterhead torque using optimized gated recurrent unit neural network. Journal of Rock Mechanics and Geotechnical Engineering (2022).