TBM jamming risk prediction method based on fuzzy theory and Bi-LSTM

Abstract

Tunnel Boring Machine (TBM) is an indispensable equipment in modern tunneling projects, but real-time safety risk assessment during its operation is still a major challenge. In this study, a method for real-time safety risk assessment of TBM interference accidents is proposed. The method combines the hierarchical analysis method (AHP) and fuzzy comprehensive evaluation (FCE) for safety risk assessment before construction. When the pre-assessment result meets the standard, the construction stage can be entered; if it does not meet the standard, corresponding measures will be taken and re-assessed until the safety score meets the requirement. During the construction process, the study collected data from the normal boring section and utilized the Bi-LSTM model to predict the thrust and torque of the TBM in real time. Geological parameters were processed by solo thermal coding and label coding techniques, and the coded features were feature vector spliced with operational parameters from the Bi-LSTM layer in a fully connected layer. A 5-fold cross-validation method was used for model optimization. The study further evaluated the residual rate of the model and correlated it with the safety pre-assessment score through vector analysis. The results of the study showed that the pre-construction safety assessment classified the safety level as IV, which meets the criteria for tunnel boring. During the actual tunneling process, the safety level gradually decreased to level III due to changes in geological conditions. In addition, it was found that the fluctuating trend of the safety scores became more and more obvious as the TBM approached the blockage condition. Therefore, the rationality of the proposed method is verified. This study provides new perspectives and methods for the dynamic safety assessment of TBMs.