Research on Prediction of Excavation Parameters for Deep Buried Tunnel Boring Machine Based on Convolutional Neural Network-Long Short-Term Memory Model

Hard rock tunnel boring machines (TBMs) are increasingly widely used in tunnel construction today; however, TBMs are deeply buried underground and have a low perception of the underground surrounding rock conditions and excavation parameters. In order to ensure the safety of TBM digging, this paper describes the research carried out relating to the accurate prediction of TBM digging parameters and the precise prediction of tunnel surrounding rock grades. Based on the on-site excavation parameters and geological data of a certain water diversion project in Xinjiang, the thrust, torque, rotational speed, net excavation speed, construction speed, and excavation specific energy of the stable section of TBM excavation are selected as the input parameters for the model. A convolutional neural network optimized–long short_term time series prediction model (CNN-LSTM model) is established to predict the excavation parameters under various levels of surrounding rock conditions. The research results indicate that the CNN-LSTM model has a high prediction accuracy, with most data having a relative prediction error rate (Er) within 10%, root mean square error (RMSE) within 5%, mean absolute percentage error (MAPE) within 10%, and goodness of fit (R2) above 0.9. The model can assist in parameter setting, engineering planning and disposal of high-risk holes in the TBM digging process, and improve the safety level of TBM digging.

Cette oeuvre a été publiée sous la license Creative Commons Attribution 4.0 (CC-BY 4.0). Il est autorisé de partager et adapter l'oeuvre tant que l'auteur est crédité et la license est indiquée (avec le lien ci-dessus). Vous devez aussi indiquer si des changements on été fait vis-à-vis de l'original.