Data-driven predictions of shield attitudes using Bayesian machine learning

Wang, Lai; Pan, Qiujing; Wang, Shuying

doi:10.1016/j.compgeo.2023.106002

Cited by 8 publications

(2 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Currently, the prediction model of shield tunnelling parameters is based on the fixed dataset, and the fixed model is applied to predict the subsequent construction [17,27,37,40]. The trained model is not updated in realtime to consider the expansion of the dataset, which obviously wastes a lot of high-quality data.…”

Section: Introductionmentioning

confidence: 99%

An intelligent construction model in whole process for shield tunnelling

Lu,

Liu,

Kong

et al. 2024

Preprint

View full text Add to dashboard Cite

Predicting shield tunnelling parameters in the whole construction process is of great importance, which can effectively control ground stability and improve tunnelling efficiency. A novel deep learning method is developed considering transfer learning, incremental learning and Bi-LSTM fusing with available data of the next ring to be excavated (ADNRE) to predict shield tunnelling parameters in the whole process. Before construction, transfer learning uses data from similar projects to determine initial network parameters, then solve the insufficient data in the prophase of the project. As the shield machine begins to excavate, incremental learning is used to continuously accept new data and adjust model parameters in real time during the whole process. A feature fusion module in Bi-LSTM is proposed to integrate ADNRE and data of the adjacent excavated rings. The proposed Bi-LSTM method can consider the mutation of stratum conditions during tunnelling. The applicability of the proposed method is explored by predicting the shield cutter head torque of a tunnel project in Qingdao, China. The influence of fine-tuning epochs and project similarity on model performance is further discussed. Overall, the proposed method can provide reasonable whole process prediction for shield tunnelling parameters, which improves construction safety and efficiency.

show abstract

Section: Introductionmentioning

confidence: 99%

An intelligent construction model in whole process for shield tunnelling

Lu,

Liu,

Kong

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The existing control methods for the shield tunnel misalignment (STM) problem exhibit hysteresis, because the misalignment growth cannot be quickly stopped when the operator alters the operating parameters [5]. Failure to timely adjust the tunneling attitude can lead to serious engineering issues, such as segment misalignment [6]. Moreover, there exists a strong nonlinear relationship between shield machine operating parameters and shield attitude, making it challenging for inexperienced shield operators to effectively identify the cause of deviation from numerous parameters at an early stage.…”

Section: Introductionmentioning

confidence: 99%

Prediction of shield tunneling attitude: a hybrid deep learning approach considering feature temporal attention

Zeng,

Chen,

Zhang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Accurate prediction of shield attitude deviation is essential to ensure safe and efficient shield tunneling. However, previous studies have predominantly emphasized temporal correlation, which has limitations in engineering guidance and prediction accuracy. This research proposes a hybrid deep learning approach considering feature temporal attention (FTA-N-GRU) for shield attitude prediction. To elucidate the contributions of each parameter, the Integrated Gradients algorithm is leveraged for parameter sensitivity analysis. The results from the Bangladesh Karnaphuli River Tunnel Project indicate that: the proposed model outperforms other models in prediction accuracy. The integration of feature attention can adaptively allocate attention weights to input parameters, facilitating inexperienced operators in discerning crucial parameter variations and decision-making. By incorporating temporal attention, the model effectively explores the connection among different output time steps, improving overall prediction accuracy and reliability. Consequently, operators are empowered with timely information to proactively adjust operations before deviations occur, underscoring the significance of this approach in promoting safe and efficient shield tunneling practices.

show abstract