Abstract A circular tunnel is prone to elliptical deformation under the impact of seismic shear wave. Main meth? ods for evaluating the deformation are analytical solution and numerical simulation. So, in this paper, a new method based on optimized neural network is proposed to accurately predict the elliptical deformation through the constructed algorithm model. In the first part of this paper, the back propagation neural network (BPNN) optimized by mind evolutionary algorithm (MEA) is adopted to determine the elliptical deformation ΔD of the circular tunnel lining.Then, a sample database containing 370 data sets is collected from the existing literatures and numerical analysis examples. The numerical analysis is in line with the assumption of the existing analytical solutions, and the data collected from the literatures include the measured results on the site. Since not considered in most analytical solutions, the interface strength Rinter and buried depth h are introduced as additional input parameters. The values of three statistical performance indexes R2, MAPE and RMSE indicate that the improved BPNN has good generalization performance. In the second part, the mean impact value (MIV) algorithm is used to analyze the impact of parameters in the trained network. The calculation results reflect the correlation between input parameters and output ones, and the prediction results are in good agreement with that of analytical solutions and numerical analysis.
Abstract：
A circular tunnel is prone to elliptical deformation under the impact of seismic shear wave. Main meth? ods for evaluating the deformation are analytical solution and numerical simulation. So, in this paper, a new method based on optimized neural network is proposed to accurately predict the elliptical deformation through the constructed algorithm model. In the first part of this paper, the back propagation neural network (BPNN) optimized by mind evolutionary algorithm (MEA) is adopted to determine the elliptical deformation ΔD of the circular tunnel lining.Then, a sample database containing 370 data sets is collected from the existing literatures and numerical analysis examples. The numerical analysis is in line with the assumption of the existing analytical solutions, and the data collected from the literatures include the measured results on the site. Since not considered in most analytical solutions, the interface strength Rinter and buried depth h are introduced as additional input parameters. The values of three statistical performance indexes R2, MAPE and RMSE indicate that the improved BPNN has good generalization performance. In the second part, the mean impact value (MIV) algorithm is used to analyze the impact of parameters in the trained network. The calculation results reflect the correlation between input parameters and output ones, and the prediction results are in good agreement with that of analytical solutions and numerical analysis.
HUANG Diwen HUO Hongbin CHEN Dong
.Analysis on Shear Deformation in a Circular Tunnel Based on Optimized Neural Network[J] MODERN TUNNELLING TECHNOLOGY, 2021,V58(6): 102-110
2021, Vol. 58 