A novel intelligent displacement prediction model of karst tunnels.
Sci Rep
; 12(1): 16983, 2022 Oct 10.
Article
en En
| MEDLINE
| ID: mdl-36216860
Karst is a common engineering environment in the process of tunnel construction, which poses a serious threat to the construction and operation, and the theory on calculating the settlement without the assumption of semi-infinite half-space is lack. Meanwhile, due to the limitation of test conditions or field measurement, the settlement of high-speed railway tunnel in Karst region is difficult to control and predict effectively. In this study, a novel intelligent displacement prediction model, following the machine learning (ML) incorporated with the finite difference method, is developed to evaluate the settlement of the tunnel floor. A back propagation neural network (BPNN) algorithm and a random forest (RF) algorithm are used herein, while the Bayesian regularization is applied to improve the BPNN and the Bayesian optimization is adopted for tuning the hyperparameters of RF. The newly proposed model is employed to predict the settlement of Changqingpo tunnel floor, located in the southeast of Yunnan Guizhou Plateau, China. Numerical simulations have been performed on the Changqingpo tunnel in terms of variety of karst size, and locations. Validations of the numerical simulations have been validated by the field data. A data set of 456 samples based on the numerical results is constructed to evaluate the accuracy of models' predictions. The correlation coefficients of the optimum BPNN and BR model in testing set are 0.987 and 0.925, respectively, indicating that the proposed BPNN model has more great potential to predict the settlement of tunnels located in karst areas. The case study of Changqingpo tunnel in karst region has demonstrated capability of the intelligent displacement prediction model to well predict the settlement of tunnel floor in Karst region.
Texto completo:
1
Colección:
01-internacional
Base de datos:
MEDLINE
Tipo de estudio:
Prognostic_studies
/
Risk_factors_studies
Idioma:
En
Revista:
Sci Rep
Año:
2022
Tipo del documento:
Article
País de afiliación:
China
Pais de publicación:
Reino Unido