In the construction of a soft rock tunnel,it is critical to accurately estimate the pre-stressed anchor support parameters for surrounding rock reinforcement;otherwise,engineering disasters may occur.This paper presen...In the construction of a soft rock tunnel,it is critical to accurately estimate the pre-stressed anchor support parameters for surrounding rock reinforcement;otherwise,engineering disasters may occur.This paper presents a support parameter selection method that aims to allow deformation as a control objective,which was applied to the tunnel located in Muzailing Highway,Min County,Dingxi City,Gansu Province,China.Through theoretical analysis,we have identified five factors that influence pre-stressing anchorages.The selection of mechanical parameters for the rock mass was carried out using an inverse analysis method.Compared with the measured data,the maximum displacement error of the numerical simulation results was only 0.07 m.The length of anchor cable,circumferential spacing of anchor cable,longitudinal spacing,and pre-stress index are adopted as the input parameters for the support vector machine neural network model based on particle swarm optimization(PSO-LSSVM).Besides,the vault subsidence and the maximum deformation of surrounding rock are considered as output values(performance indices).The goodness of fit between the predicted values and the simulated values exceeds 0.9.Finally,all support parameters within the acceptable deformation range are calculated.The optimal support variables are derived by considering the construction cost and duration.The field application results show that it is feasible to construct the sample database utilizing the numerical simulation approach by taking the displacement as the control target and using the neural network to specify the appropriate support parameters.展开更多
The robust parameter design method is a traditional approach to robust experimental design that seeks to obtain the optimal combination of factors/levels. To overcome some of the defects of the inflatable wing paramet...The robust parameter design method is a traditional approach to robust experimental design that seeks to obtain the optimal combination of factors/levels. To overcome some of the defects of the inflatable wing parameter design method, this paper proposes an optimization design scheme based on orthogonal testing and support vector machines （SVMs）. Orthogonal testing design is used to estimate the appropriate initial value and variation domain of each variable to decrease the number of iterations and improve the identification accuracy and efficiency. Orthogonal tests consisting of three factors and three levels are designed to analyze the parameters of pressure, uniform applied load and the number of chambers that affect the bending response of inflatable wings. An SVM intelligent model is established and limited orthogonal test swatches are studied. Thus, the precise relationships between each parameter and product quality features, as well the signal-to-noise ratio （SNR）, can be obtained. This can guide general technological design optimization.展开更多
基金supported by the Open Fund of State Key Laboratory of High speed Railway Track Technology(2022YJ127-1)National Natural Science Foundation of China(52104125,41941018)+1 种基金the Natural Science Basic Research Plan in Shaanxi Province of China(2022JQ-304)the Young Elite Scientists Sponsorship Program by CAST(No.2021QNRC001)。
文摘In the construction of a soft rock tunnel,it is critical to accurately estimate the pre-stressed anchor support parameters for surrounding rock reinforcement;otherwise,engineering disasters may occur.This paper presents a support parameter selection method that aims to allow deformation as a control objective,which was applied to the tunnel located in Muzailing Highway,Min County,Dingxi City,Gansu Province,China.Through theoretical analysis,we have identified five factors that influence pre-stressing anchorages.The selection of mechanical parameters for the rock mass was carried out using an inverse analysis method.Compared with the measured data,the maximum displacement error of the numerical simulation results was only 0.07 m.The length of anchor cable,circumferential spacing of anchor cable,longitudinal spacing,and pre-stress index are adopted as the input parameters for the support vector machine neural network model based on particle swarm optimization(PSO-LSSVM).Besides,the vault subsidence and the maximum deformation of surrounding rock are considered as output values(performance indices).The goodness of fit between the predicted values and the simulated values exceeds 0.9.Finally,all support parameters within the acceptable deformation range are calculated.The optimal support variables are derived by considering the construction cost and duration.The field application results show that it is feasible to construct the sample database utilizing the numerical simulation approach by taking the displacement as the control target and using the neural network to specify the appropriate support parameters.
文摘The robust parameter design method is a traditional approach to robust experimental design that seeks to obtain the optimal combination of factors/levels. To overcome some of the defects of the inflatable wing parameter design method, this paper proposes an optimization design scheme based on orthogonal testing and support vector machines （SVMs）. Orthogonal testing design is used to estimate the appropriate initial value and variation domain of each variable to decrease the number of iterations and improve the identification accuracy and efficiency. Orthogonal tests consisting of three factors and three levels are designed to analyze the parameters of pressure, uniform applied load and the number of chambers that affect the bending response of inflatable wings. An SVM intelligent model is established and limited orthogonal test swatches are studied. Thus, the precise relationships between each parameter and product quality features, as well the signal-to-noise ratio （SNR）, can be obtained. This can guide general technological design optimization.
