A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Dynamic multiclass prediction of tunnel squeezing intensity with stacking model and Markov process
Graphical abstract Display Omitted
Highlights 162 tunnel squeezing samples from public literature and actual engineering data were collected. The quality of the dataset was effectively improved due to use of various data preprocessing measures. A three-classification model with excellent performance was trained by using stacking ensemble learning. The dynamic prediction of squeezing intensity matching the excavation is purposed by combining the Markovian geologic model.
Abstract Tunnel squeezing is a deformation behavior influenced by various nonlinear factors. This not only increases the construction time and the budget but also threatens the safety of construction personnel. In this study, a novel approach to improve the accuracy of predicting squeezing intensity during the design and excavation stage is proposed. To achieve this goal, a comprehensive dataset comprising 162 tunnel squeezing samples was built from those in the published literature and from engineering practices, and a dynamic multiclass prediction method was developed that combines a stacking ensemble model and a Markovian geological model. Through data preprocessing and the implementation of the stacking ensemble algorithm, a highly accurate multiclass prediction model was trained. In the next step, appropriate predictors for the Markovian geological model, which enables dynamic prediction, were identified. This method is first applied during the tunnel design stage to obtain initial predictions of the squeezing intensity. Subsequently, the model is dynamically updated during the excavation stage by using the Markovian geological model. The results of our study demonstrate the effectiveness of the stacking ensemble model in achieving a higher level of accuracy in multiclass prediction, with an impressive accuracy rate of 90.77%. Furthermore, the dynamic prediction approach employed during the excavation stage effectively addresses the limitations caused by insufficient authenticity of the design and the survey data. The findings of this study provide reliable references for tunnel excavation methods and support measures, particularly in terms of managing squeezing intensity.
Dynamic multiclass prediction of tunnel squeezing intensity with stacking model and Markov process
Graphical abstract Display Omitted
Highlights 162 tunnel squeezing samples from public literature and actual engineering data were collected. The quality of the dataset was effectively improved due to use of various data preprocessing measures. A three-classification model with excellent performance was trained by using stacking ensemble learning. The dynamic prediction of squeezing intensity matching the excavation is purposed by combining the Markovian geologic model.
Abstract Tunnel squeezing is a deformation behavior influenced by various nonlinear factors. This not only increases the construction time and the budget but also threatens the safety of construction personnel. In this study, a novel approach to improve the accuracy of predicting squeezing intensity during the design and excavation stage is proposed. To achieve this goal, a comprehensive dataset comprising 162 tunnel squeezing samples was built from those in the published literature and from engineering practices, and a dynamic multiclass prediction method was developed that combines a stacking ensemble model and a Markovian geological model. Through data preprocessing and the implementation of the stacking ensemble algorithm, a highly accurate multiclass prediction model was trained. In the next step, appropriate predictors for the Markovian geological model, which enables dynamic prediction, were identified. This method is first applied during the tunnel design stage to obtain initial predictions of the squeezing intensity. Subsequently, the model is dynamically updated during the excavation stage by using the Markovian geological model. The results of our study demonstrate the effectiveness of the stacking ensemble model in achieving a higher level of accuracy in multiclass prediction, with an impressive accuracy rate of 90.77%. Furthermore, the dynamic prediction approach employed during the excavation stage effectively addresses the limitations caused by insufficient authenticity of the design and the survey data. The findings of this study provide reliable references for tunnel excavation methods and support measures, particularly in terms of managing squeezing intensity.
Dynamic multiclass prediction of tunnel squeezing intensity with stacking model and Markov process
Liang, Ming (author) / Peng, Hao (author) / Xie, Weiwei (author) / Yu, Bo (author) / Han, Yu (author) / Zhu, Menglong (author) / Song, Guanxian (author) / Huang, Nenghao (author)
2024-01-31
Article (Journal)
Electronic Resource
English
Dynamic and Probabilistic Multi-class Prediction of Tunnel Squeezing Intensity
| Online Contents | 2020
Prediction of tunnel deformation in squeezing grounds
| British Library Online Contents | 2013
Prediction of tunnel deformation in squeezing grounds
| Elsevier | 2013
Prediction of tunnel deformation in squeezing grounds
| Online Contents | 2013