Abstract The beam-spring model calculation formula is concise and easy to use in numerical calculation, it can be used to effectively evaluate the jointing effect of lining segments, and it has been widely used for design of lining structure of shield tunnel. As an important parameter that represents the joint performance, the joint stiffness (particularly the flexural stiffness) selected will directly determine the calculation accuracy of the beam-spring model and it is very important for internal forces analysis of lining structure of shield tunnel. With the internal forces (bending moment and axial force) as the input features and the flexural stiffness of joint as the output, this paper offers a deep neural network model that can indicate the nonlinear feature of flexural stiffness of joint. Based on the joint fullscale test data and through forward propagation and back propagation algorithms training, the nonlinear mapping function between joint flexural stiffness and internal forces that has analytical form and that is universally applicable has been obtained. This overcomes the limitedness of fitting results of traditional method and solves the problem that it is difficult to obtain the expression for joint flexural stiffness in the entire internal forces space. This comes as a new approach for nonlinear prediction and uniform continuous expression of joint flexural stiffness based on joint full-scale test data. The comparison between the joint flexural stiffness value predicted by the nonlinear model based on deep neural network and the full-scale test data and the application of the method in calculation and analysis of full-ring lining structure have proven the correctness and effectiveness of this method
Abstract：
The beam-spring model calculation formula is concise and easy to use in numerical calculation, it can be used to effectively evaluate the jointing effect of lining segments, and it has been widely used for design of lining structure of shield tunnel. As an important parameter that represents the joint performance, the joint stiffness (particularly the flexural stiffness) selected will directly determine the calculation accuracy of the beam-spring model and it is very important for internal forces analysis of lining structure of shield tunnel. With the internal forces (bending moment and axial force) as the input features and the flexural stiffness of joint as the output, this paper offers a deep neural network model that can indicate the nonlinear feature of flexural stiffness of joint. Based on the joint fullscale test data and through forward propagation and back propagation algorithms training, the nonlinear mapping function between joint flexural stiffness and internal forces that has analytical form and that is universally applicable has been obtained. This overcomes the limitedness of fitting results of traditional method and solves the problem that it is difficult to obtain the expression for joint flexural stiffness in the entire internal forces space. This comes as a new approach for nonlinear prediction and uniform continuous expression of joint flexural stiffness based on joint full-scale test data. The comparison between the joint flexural stiffness value predicted by the nonlinear model based on deep neural network and the full-scale test data and the application of the method in calculation and analysis of full-ring lining structure have proven the correctness and effectiveness of this method
YAN Pengfei CAI Yongchang ZHOU Long
.Nonlinear Model for Segment Joint Stiffness Based on Deep Neural Network and Its Application[J] MODERN TUNNELLING TECHNOLOGY, 2023,V60(3): 24-33
2023, Vol. 60 