Answers are grounded in this paper and cite their sources.
Accurately forecasting tunnel boring machine (TBM) cutterhead torque is crucial for optimizing tunneling parameters and warning jamming disasters. However, the current research pays little attention to the robustness of the model when there is a sign of jamming and the generalization performance of the model in other similar projects. This research presents a novel hybrid multiscale TCN-CBAM-LSTM (HMTCL) network-based method for TBM cutterhead torque forecast. Firstly, a set of data preprocessing methods are proposed, including parameter selection, data cleaning, and data normalization. Sequentially, a novel HMTCL network is designed for cutterhead torque forecast, which can accurately capture long-term dependencies and local changes related to tasks. Lastly, the proposed model’s generalization ability in another tunnel and robustness in the presence of jamming symptoms are verified. The results show that the HMTCL model is superior to other comparative models with a R 2 of 0.747 in another tunnel, which is 4.20%, 4.60%, 4.90%, 3.10%, 12.00%, 5.30%, and 3.30% higher than the LSTM, Bi-LSTM, GRU, Bi-GRU, temporal convolutional network (TCN), WaveNet, and Transformer models, respectively. Moreover, when there are signs of jamming, the performance of each model decreases to varying degrees, but the HMTCL model still performs the best, and its R 2 is 11.00%, 38.69%, 70.60%, 83.78%, 179.54%, 23.06%, and 5.16% higher than these comparative models, respectively. Therefore, the proposed method has good generalization performance and robustness, and has important academic and engineering application values.
No takes yet. Share an insight, caveat, or question.
Yu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: