Intelligent prediction method and system for settlement of TBM crossing pebble stratum

Abstract

The application provides an intelligent prediction method and system for settlement of TBM crossing pebble stratum, and belongs to the technical field of stratum settlement prediction. The method comprises the following steps: acquiring position information of a ground monitoring point, position information of an excavation surface and buried depth, and inputting the position information and the buried depth into a first deep neural network model to acquire tunnel vault displacement, arch waist convergence and ground settlement; a second deep neural network model takes position coordinates formed by the position information of the ground monitoring point, the position information of the excavation surface and the buried depth and the ground settlement as inputs, and outputs elastic modulus and Poisson's ratio of the ground; stress and strain components of the ground settlement are acquired, and a control equation and boundary conditions are constructed based on the stress and strain components, the elastic modulus and Poisson's ratio; prior geological information is acquired by using a control equation residual; the control equation, the boundary conditions and the prior geological information are taken as constraint conditions, and a settlement prediction value of TBM crossing pebble stratum is acquired based on the tunnel vault displacement, so that accurate prediction of the settlement of the pebble stratum is realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of stratum settlement prediction, and particularly relates to an intelligent prediction method and system for TBM crossing pebble stratum settlement. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] With the continuous acceleration of urbanization, the development and utilization of underground space has become an effective way to solve the pressure of urban traffic and infrastructure construction. Tunnel boring machine (TBM) as an important equipment for modern tunnel construction, is widely used in various tunnel projects due to its high construction efficiency and small disturbance to the ground. However, TBM is prone to cause stratum deformation and ground settlement when crossing complex geological conditions, especially pebble stratum, which brings potential risks to surrounding buildings and infrastructure. Therefore, how to accurately predict the settlement induced by TBM construction and ensure construction safety has become a key technical problem to be solved in the field of tunnel engineering.

[0004] Traditional settlement prediction methods are mainly based on empirical formula, elastoplastic analytical formula and numerical simulation. However, the physical knowledge model on which these methods are based is difficult to interact with the actual measured ground settlement, resulting in low efficiency in practical engineering application. Machine learning is an effective method for predicting TBM-induced ground settlement, but it relies on high-quality data mining mapping relationship, ignoring the physical knowledge of tunnel-induced stratum displacement. Most of the physical knowledge that can be mastered by the project is relatively limited, and it is difficult to obtain sufficient monitoring data to train the machine learning model during the construction process, especially in pebble stratum. The non-uniformity of the stratum, the uncertainty of the geological parameters and the complex stress distribution increase the difficulty of settlement prediction. SUMMARY

[0005] To overcome the shortcomings of the prior art, the present application provides an intelligent prediction method and system for TBM crossing pebble stratum settlement, which fully combines the respective advantages of physical knowledge and data-driven, and proposes a deep learning framework that integrates physical constraints and prior geological information for intelligent prediction of ground settlement when TBM crosses pebble stratum. It provides more intelligent prediction and decision support for TBM construction in complex geological conditions, and has a wide application prospect.

[0006] To achieve the above purpose, one or more embodiments of the present application provide the following technical solutions:

[0007] The first aspect of the present application provides an intelligent prediction method for TBM crossing pebble stratum settlement;

[0008] An intelligent prediction method for TBM crossing pebble stratum settlement, comprising:

[0009] obtaining a surface monitoring point position, an excavation face position, and a buried depth;

[0010] inputting the surface monitoring point position, the excavation face position, and the buried depth into a trained settlement prediction knowledge-data dual driving model, and outputting a surface settlement value when the TBM passes through the gravel stratum;

[0011] The trained settlement prediction knowledge-data dual driving model comprises a first deep neural network model and a second deep neural network model connected in series with each other. The first deep neural network model takes the surface monitoring point position, the excavation face position, and the buried depth as input, and outputs a tunnel crown displacement, a haunch convergence, and a surface settlement. The second deep neural network model takes the surface monitoring point position, the excavation face position, and the buried depth and the surface settlement as input, and outputs an elastic modulus and a Poisson's ratio of the surface.

[0012] Based on the basic principles of elastic mechanics, stress and strain components of the surface settlement are solved. Control equations and boundary conditions are constructed based on the stress and strain components, the elastic modulus, and the Poisson's ratio. Prior geological information constraints are applied based on control equation residual terms.

[0013] A dual driving model loss function is constructed to include data driving terms, control equation residual terms, and boundary conditions. The intelligent algorithm is used to optimize the combination of hyperparameters to obtain the final settlement prediction value.

[0014] As a further technical solution, the tunnel crown displacement includes a uniform radial displacement caused by stratum loss during the tunnel excavation process and an elliptical deformation of the tunnel contour caused by anisotropic initial stress.

[0015] As a further technical solution, the control equation comprises:

[0016] The control equation satisfied by the stress, strain, and displacement components of the surface is as follows:

[0017]

[0018]

[0019]

[0020] In the formula, x is surface monitoring point position information, z is a buried depth; is a partial derivative of a stress tensor with respect to a coordinate; is a Kronecker function; is a volumetric strain tensor; is a strain tensor; isx - z Elastic modulus in plane; 、 Partial derivative of displacement tensor to coordinate;

[0021] Control equation between crown displacement and convergence of haunch, as follows:

[0022]

[0023] In the formula, y is position of excavation face; is crown displacement; is convergence of haunch; r is radius of tunnel excavation;

[0024] Control equation between crown displacement and ground settlement, as follows:

[0025]

[0026] In the formula, is ground settlement; is uniform radial displacement caused by stratum loss in tunnel excavation process; is elliptical deformation of tunnel contour caused by anisotropic initial stress; υ is Poisson's ratio; h is distance from tunnel axis to ground.

[0027] As a further technical solution, the ground settlement is:

[0028]

[0029] In the formula, θ is included angle between line connecting boundary point around hole and center of circle and horizontal direction, 、 is crown displacement at different positions of excavation face.

[0030] As a further technical solution, the boundary condition is boundary condition satisfied by ground stress and displacement component, and the boundary condition is:

[0031]

[0032]

[0033] In the formula, is normal stress in direction; x is normal stress in direction; is normal stress in direction; z is normal stress in direction; is normal stress in direction; x ​Direction displacement.

[0034] As a further technical solution, the process of applying a priori geological information constraints based on the residual term of the control equation is as follows:

[0035] The original value of the second deep neural network model output Apply exponential function constraints to ensure E Always positive:

[0036]

[0037] The original value of the second deep neural network model output Apply the scaled Sigmoid function constraint:

[0038]

[0039] After the constraint E and υ Substitute it into the control equation to calculate the residual term of the control equation.

[0040] As a further technical solution, the trained settlement prediction knowledge data dual-driven model also includes a loss function, which is:

[0041]

[0042] Where, is the total number of sample points containing position coordinates and actual displacement values; is the actual value of surface settlement; is the predicted value of surface subsidence; is the total number of sample points in the study area that do not require actual displacement values; is the total number of sample points on the boundary that do not require actual displacement values; ~ is the control equation; λ is the weight coefficient of the physical knowledge driving item.

[0043] A second aspect of the present invention provides an intelligent prediction system for settlement of a TBM passing through a pebble stratum.

[0044] An intelligent prediction system for TBM settlement during traversal of pebble strata, comprising:

[0045] The settlement information acquisition module is configured to: obtain the location of the surface monitoring point, the location of the excavation surface and the burial depth;

[0046] The data processing module is configured to: input the ground surface monitoring point position, the excavation face position and the buried depth into a first deep neural network model to obtain a tunnel vault displacement, a hance convergence and a ground surface settlement; and input the ground surface monitoring point position, the excavation face position and the buried depth and the ground surface settlement into a second deep neural network model to output an elastic modulus and a Poisson's ratio of the ground surface;

[0047] The control equation obtaining module is configured to: solve stress and strain components of the ground surface settlement based on basic principles of elastic mechanics, construct a control equation and a boundary condition based on the stress and strain components, the elastic modulus and the Poisson's ratio, and apply a prior geological information constraint based on a control equation residual term.

[0048] The settlement prediction value obtaining module is configured to: construct a double-driving model loss function containing a data driving term, a control equation residual term and a boundary condition, optimize a hyperparameter combination by using an intelligent algorithm to minimize the loss function, and obtain a final settlement prediction value.

[0049] The third aspect of the present application provides a computer readable storage medium having a program stored thereon, the program being executed by a processor to implement the steps of the intelligent prediction method for TBM tunneling settlement in pebble stratum.

[0050] The fourth aspect of the present application provides an electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, the processor executing the program to implement the steps of the intelligent prediction method for TBM tunneling settlement in pebble stratum.

[0051] The above one or more technical solutions have the following beneficial effects:

[0052] The present application couples the physical and mechanical mechanism of TBM construction induced ground surface settlement to the deep learning framework, uses the control equation, the boundary condition and the prior geological information as the constraint condition, effectively makes up for the problem of ignoring the physical law in the data driven method, reduces the dependence on a large amount of measured data, and improves the prediction accuracy and the generalization ability. The method provides more intelligent prediction and decision support for TBM construction under complex geological conditions, and has a wide application prospect.

[0053] The advantages of the additional aspects of the present application will be partially given in the following description, partially become obvious from the following description, or be learned by the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings, which form a part of the specification, are included to provide a further understanding of the application and are incorporated herein by reference. The embodiments illustrated in the drawings are shown for the purposes of exemplification only and are not intended to limit the present application.

[0055] Figure 1 The method flowchart of the first embodiment.

[0056] Figure 2 The architecture diagram of the settlement prediction knowledge data double-driven model of the first embodiment.

[0057] Figure 3 The system structure diagram of the second embodiment. DETAILED DESCRIPTION

[0058] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0059] It should be noted that the terms used herein are only intended to describe specific embodiments and are not intended to limit the exemplary embodiments according to the present application.

[0060] The embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0061] Embodiment one

[0062] The present embodiment discloses an intelligent prediction method for settlement of TBM crossing pebble stratum;

[0063] As shown in Figure 1 and Figure 2 , an intelligent prediction method for settlement of TBM crossing pebble stratum, comprising:

[0064] Step S1, obtaining ground monitoring point position information, excavation face position information and buried depth;

[0065] In the present embodiment, the buried depth of the TBM construction tunnel is less than 5-7 times the tunnel excavation radius. According to the construction information, the ground monitoring point position information x , excavation face position information y and buried depth z。

[0066] Step S2, inputting the ground monitoring point position information x , excavation face position information y and buried depth z to the trained settlement prediction knowledge data double-driven model, and outputting the ground settlement value when the TBM crosses the pebble stratum;

[0067] The trained settlement prediction knowledge data double-driven model includes two parallel first and second deep neural network models; the first deep neural network model takes the ground monitoring point position informationx , excavation face position y and buried depth z As input, output tunnel vault displacement, arch waist convergence and ground settlement . Among them, the tunnel vault displacement is composed of two parts, including uniform radial displacement caused by stratum loss during tunnel excavation and tunnel contour elliptical deformation caused by anisotropic initial stress . Tunnel hole radial convergence u ( θ ) can be expressed by vault displacement:

[0068]

[0069] In the formula, θ is the angle between the connecting line of the hole boundary point and the center and the horizontal direction, , is the vault displacement at different excavation face positions .

[0070] According to the classical V-B analytical solution of ground displacement induced by semi-infinite space tunnel, when the tunnel vault displacement is determined, the ground settlement can be determined by the following formula:

[0071]

[0072] In the formula, θ is the angle between the connecting line of the hole boundary point and the center and the horizontal direction, , is the vault displacement at different excavation face positions .

[0073] The second depth neural network model takes the surface monitoring point position x , excavation face position y , buried depth z and the ground settlement as input, and outputs the surface elastic modulus and Poisson's ratio.

[0074] Step S3, based on the basic principles of elasticity, the stress and strain components of the ground settlement are solved, and the control equation and boundary conditions are constructed based on the stress and strain components, elastic modulus and Poisson's ratio; the prior geological information constraint is applied based on the control equation residual term; the double driving model loss function containing data driving term, control equation residual term and boundary condition is constructed, and the intelligent algorithm is used to optimize the combination of hyperparameters (learning rate, hidden layer number and knowledge driving term weight coefficient, etc.), and the ground settlement value when the loss function is minimum is the final settlement prediction value. Specifically:

[0075] Step S31, based on the basic principles of elastic mechanics, solving the stress and strain components of the surface settlement, and constructing the control equations and boundary conditions based on the stress and strain components, elastic modulus and Poisson's ratio;

[0076] The surface settlement vector can be further solved for its strain component according to the basic principles of elastic mechanics : 、 、 , For rock and soil x The normal strain in the direction, For rock and soil z The normal strain in the direction, For rock and soil x - z In-plane shear strain; On the surface x Directional displacement; On the surface z Direction displacement.

[0077] The stress and strain components satisfy the governing equations ~ and boundary conditions; governing equations ~ The elastic modulus involved E and Poisson's ratio υ Output by the second deep neural network model and used as the control equation ~ 、 The control equation is ~ As shown in the following formula:

[0078]

[0079]

[0080]

[0081] Where, Find the partial derivative of the stress tensor with respect to the coordinates; is the Kronecker function; is the volume strain tensor; is the strain tensor; for x - z In-plane elastic modulus; 、 Find the partial derivative of the displacement tensor with respect to the coordinates;

[0082] The boundary conditions are as follows:

[0083]

[0084]

[0085] wherein, is the normal stress in the direction; x is the normal stress in the direction; is the displacement in the direction; z is the displacement in the direction; is the displacement in the direction on the ground surface; x is the displacement in the direction;

[0086] The vault displacement and the haunch convergence satisfy the control equation as shown in the following formula:

[0087]

[0088] wherein, is the vault displacement; is the haunch convergence; r is the tunnel excavation radius.

[0089] The vault displacement and the ground surface settlement satisfy the control equation as shown in the following formula:

[0090]

[0091] wherein, is the ground surface settlement; is the elliptical deformation of the tunnel profile caused by the anisotropic initial stress; υ is the Poisson's ratio; h is the distance from the tunnel axis to the ground.

[0092] In step S32, the prior geological information constraint is applied based on the control equation residual term, wherein in the output layer of the second deep neural network model, the elastic modulus E and the Poisson's ratio υ The prior geological information constraint is applied in the following way:

[0093] (1) The original value output by the second deep neural network model is subjected to an exponential function constraint to ensure that E is always positive:

[0094]

[0095] (2) The original value output by the second deep neural network model is subjected to a scaled Sigmoid function constraint:

[0096]

[0097] The constrained E and υ are substituted into the control equation , The control equation residual term is calculated.

[0098] Step S33, a dual-drive model loss function containing the data-driven term, the control equation residual term and the boundary condition is constructed, wherein the loss function is as follows:

[0099]

[0100] In the formula, is the total number of sample points containing position coordinates and actual displacement values; is the actual value of ground subsidence; is the predicted value of ground subsidence; is the total number of sample points in the study area that do not require actual displacement values; is the total number of sample points on the boundary that do not require actual displacement values; ~ is the control equation; λ is the weight coefficient of the physics knowledge-driven term.

[0101] In the present embodiment, the dual-drive model loss function includes but is not limited to the form of addition of the data-driven term, the control equation residual term and the boundary condition: It can also be constructed in the form of multiplication: wherein , and respectively represent the bias coefficients of the data-driven term, the control equation residual term and the boundary condition, which are very small positive numbers to ensure that the loss function will not be numerically unstable due to excessive dependence on a certain term. Similarly, it can also be regarded as a hyperparameter of the dual-drive model to be solved by an intelligent algorithm to obtain an optimal value to make the loss value of the model lower.

[0102] Step S33, the loss function value is calculated, and the intelligent algorithm is used to optimize the combination of hyperparameters (such as learning rate, number of hidden layers and weight coefficient of knowledge-driven term, etc.) with the goal of minimizing the loss function. The ground subsidence value at the minimum loss function is the final subsidence prediction value.

[0103] Specifically, for samples whose actual displacement values in the research area and on the boundary are unknown, only the control equation residual term and the boundary condition residual term are calculated respectively, for samples whose actual displacement values in the research area and on the boundary are known, in addition to calculating the above residual terms, the mean square error (MSE) between the predicted displacement value and the actual displacement value also needs to be calculated, and the above mean square error, the control equation residual term and the boundary condition residual term are substituted into the double driving model loss function to calculate the loss value of the first iteration, wherein the weight coefficient of the initial physical knowledge driving term is set to 0.1.

[0104] Through multiple iterations of training, the total loss value is gradually minimized, and the ground subsidence at the time when the total loss value is minimized is the final output value.

[0105] Embodiment Two

[0106] The embodiment discloses an intelligent prediction system for TBM subsidence in pebble stratum;

[0107] As shown in Figure 3 An intelligent prediction system for TBM subsidence in pebble stratum, comprising:

[0108] An intelligent prediction system for TBM subsidence in pebble stratum, comprising:

[0109] A subsidence information acquisition module configured to acquire a ground monitoring point position, an excavation face position and a buried depth;

[0110] A data processing module configured to input the ground monitoring point position, the excavation face position and the buried depth into a first deep neural network model to acquire a tunnel crown displacement, a hogging convergence and a ground subsidence, and input the ground monitoring point position, the excavation face position and the buried depth and the ground subsidence into a second deep neural network model to output an elastic modulus and a Poisson's ratio of the ground;

[0111] A control equation acquisition module configured to solve stress and strain components of the ground subsidence based on basic principles of elastic mechanics, construct a control equation and boundary conditions based on the stress and strain components, the elastic modulus and the Poisson's ratio, and apply a prior geological information constraint based on a control equation residual term;

[0112] A subsidence prediction value acquisition module configured to construct a double driving model loss function containing a data driving term, a control equation residual term and a boundary condition, optimize a hyperparameter combination by using an intelligent algorithm to obtain a final subsidence prediction value, with the loss function being minimized as a target.

[0113] Embodiment Three

[0114] The purpose of the embodiment is to provide a computer readable storage medium.

[0115] A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method of intelligent prediction of settlement of TBM crossing through cobble strata as claimed in embodiment 1.

[0116] Embodiment four

[0117] An object of the present embodiment is to provide an electronic device.

[0118] An electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, the processor implementing the steps of the method of intelligent prediction of settlement of TBM crossing through cobble strata as claimed in embodiment 1 when executing the program.

[0119] The steps and methods involved in the devices of embodiments two, three and four above correspond to embodiment one, and the detailed description can be found in the relevant description section of embodiment one. The term "computer readable storage medium" should be understood to include a single medium or multiple media of one or more instruction sets; it should also be understood to include any medium that is capable of storing, encoding or carrying instruction sets for execution by a processor and causing the processor to perform any of the methods of the present invention.

[0120] It should be apparent to one skilled in the art that the modules or steps of the present invention described above can be implemented using a general purpose computer device, alternatively, they can be implemented using program codes executable by a computing device, so that they can be stored in a storage device for execution by the computing device, or they can be made into individual integrated circuit modules, or a plurality of modules or steps can be made into a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0121] The above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, but is not a limitation on the scope of protection of the present invention, and those skilled in the art should understand that various modifications or variations made by those skilled in the art on the basis of the technical solutions of the present invention without creative labor are still within the scope of protection of the present invention.

Claims

1. A method for intelligent prediction of settlement of TBM crossing boulder strata, characterized by, The method comprises the following steps: obtaining the position of a ground monitoring point, the position of an excavation face and the depth of burial; inputting the position of the ground monitoring point, the position of the excavation face and the depth of burial into a trained settlement prediction knowledge-data double driving model to output the ground settlement value when a TBM passes through a gravel stratum; wherein the trained settlement prediction knowledge-data double driving model comprises a first deep neural network model and a second deep neural network model connected in series; the first deep neural network model takes the position of the ground monitoring point, the position of the excavation face and the depth of burial as input and outputs the tunnel crown displacement, the tunnel haunch convergence and the ground settlement; the second deep neural network model takes the position of the ground monitoring point, the position of the excavation face, the depth of burial and the ground settlement as input and outputs the ground elastic modulus and the Poisson's ratio; based on the basic principles of elastic mechanics, the stress and strain components of the ground settlement are solved, and the control equation and the boundary condition are constructed based on the stress, the strain components, the elastic modulus and the Poisson's ratio; the prior geological information constraint is applied based on the control equation residual term; a double driving model loss function containing the data driving term, the control equation residual term and the boundary condition is constructed, and the intelligent algorithm is used to optimize the hyperparameter combination to obtain the final settlement prediction value.

2. A method of intelligent prediction of settlement of TBM crossing boulder strata as claimed in claim 1 wherein, The tunnel crown displacement comprises the uniform radial displacement caused by the stratum loss in the tunnel excavation process and the tunnel contour ovalization deformation caused by the anisotropic initial stress.

3. A method of intelligent prediction of settlement of TBM crossing boulder strata as claimed in claim 1 wherein, The control equation comprises: the control equation satisfied by the stress, the strain and the displacement components of the ground, as follows: wherein, x is surface monitoring point position information, z is buried depth; is a partial derivative of a stress tensor with respect to a coordinate; is a Kronecker function; is a volumetric strain tensor; is a strain tensor; is x - z is an in-plane elastic modulus; , is a partial derivative of a displacement tensor with respect to a coordinate; the control equation between the tunnel crown displacement and the tunnel haunch convergence, as follows: wherein y is the location of the face; is the crown displacement; is the haunch convergence; r is the tunnel excavation radius; the control equation between the tunnel crown displacement and the ground settlement, as follows: wherein is the ground settlement; is the uniform radial displacement caused by strata loss during tunnel excavation; is the tunnel profile ovalization deformation caused by anisotropic initial stress; The ground settlement is: is the Poisson's ratio; h is the distance from the tunnel axis to the ground.

4. A method of intelligent prediction of settlement of TBM crossing boulder strata as claimed in claim 3 wherein, The boundary condition is the boundary condition satisfied by the stress and the displacement components of the ground, specifically, the boundary condition is: In the formula, The process of applying the prior geological information constraint based on the control equation residual term is: is the angle between the line connecting the boundary point of the hole and the center of the circle and the horizontal direction, , is the vault displacement at different excavation face positions .

5. A method of intelligent prediction of settlement of TBM crossing boulder strata as claimed in claim 1 wherein, The trained settlement prediction knowledge-data double driving model further comprises a loss function, as follows: wherein is the normal stress in the direction x is the normal stress in the direction is the normal stress in the direction z is the normal stress in the direction is the displacement in the direction x is the displacement in the direction 6. A method of intelligent prediction of settlement of TBM crossing boulder strata as claimed in claim 1 wherein, The method comprises the following steps: raw values output by the second deep neural network model applying an exponential function constraint, ensuring E always positive values: raw values output by the second deep neural network model applying a scaled sigmoid function constraint: The constrained E and a settlement information acquisition module configured to obtain the position of a ground monitoring point, the position of an excavation face and the depth of burial; Substitute into the control equation , Calculate the control equation residual term.

7. A method of intelligent prediction of settlement of TBM crossing boulder strata as claimed in claim 1 wherein, a data processing module configured to input the position of the ground monitoring point, the position of the excavation face and the depth of burial into a first deep neural network model to obtain the tunnel crown displacement, the tunnel haunch convergence and the ground settlement; and input the position of the ground monitoring point, the position of the excavation face, the depth of burial and the ground settlement into a second deep neural network model to output the ground elastic modulus and the Poisson's ratio; wherein is the total number of sample points comprising position coordinates and actual displacement values; is the actual value of the ground subsidence; is the predicted value of the ground subsidence; is the total number of sample points within the study area for which no actual displacement values are required; is the total number of sample points on the boundary for which no actual displacement values are required; is the control equation;​ a control equation acquisition module configured to solve the stress and strain components of the ground settlement based on the basic principles of elastic mechanics, and construct the control equation and the boundary condition based on the stress, the strain components, the elastic modulus and the Poisson's ratio; and apply the prior geological information constraint based on the control equation residual term; is the weight coefficient for the physical knowledge driven item.

8. An intelligent prediction system for settlement of TBM crossing through cobble strata characterized by: a settlement prediction value acquisition module configured to construct a double driving model loss function containing the data driving term, the control equation residual term and the boundary condition, and use an intelligent algorithm to optimize the hyperparameter combination to obtain the final settlement prediction value. ​ ​ ​ ​ 9. A computer-readable storage medium having stored thereon a program, characterized in that, The program, when executed by a processor, implements the steps of the intelligent prediction method of settlement of a TBM crossing a gravel stratum as claimed in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a program stored on the memory and executable on the processor, characterized in that, The processor, when executing the program, implements the steps of the intelligent prediction method of settlement of a TBM crossing a gravel stratum as claimed in any one of claims 1-7.

Images