.png){kind=icon}
.png){kind=icon}
Return
Multi-output prediction for TBM operation parameters based on stacking ensemble algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.tust.2024.105960.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Thrust and torque are two key operation parameters for tunnel boring machine (TBM), a rapid and accurate simultaneous prediction or determination of them is essential for TBM construction. In this study, based on the real dataset collected from the Chaoerhe River to Liaohe River project, a stacking algorithm framework is proposed for establishing a multi-output stacked prediction model for TBM thrust and torque. Such model is constructed by integrating the basic models including Random Forest (RF), XGBoost and CatBoost, Support Vector Regression (SVR), and Multilayer Perceptron Neural Network (MLPNN). Both Knowledge Driven and Data Driven approaches were employed to select feature parameters, and the importance of these parameters was analyzed in combination with the RF, XGBoost, and CatBoost algorithms. Meanwhile, the OPTUNA optimizer and Nondominated Sorting Genetic Algorithms III (NSGA-III) were used for hyperparameters optimization of models. The results indicate that the multi-output stacked prediction model outperforms the any of single basic models, with corresponding aCC, aRMSE and aRRMSE are 0.9524, 376.30 and 1.3260, respectively. Moreover, the method of multi-output prediction has about 5.8 times efficiency higher than the single-output prediction. The proposed framework provides an efficient, accurate, and promising method for the simultaneous prediction of multiple operation parameters of TBM.
Keywords:
Hyperparameter optimization
Multi -output prediction
Operation parameters
Stacking ensemble
Tunnel boring machine
Journal
IF:
7.4
Papers:
6.8K
Citations:
3.5W