Dual-engine prediction method and system for ground deformation caused by shield tunneling

Abstract

The present invention relates to the field of shield tunneling and stratum deformation prediction, and provides a dual-engine prediction method and system for stratum deformation caused by shield tunneling. The method comprises: S1, determining an analytical solution model; S2, obtaining the stratum loss rate of a geological section based on the geological parameters of the geological section, the field-measured deformation value of the geological section, and the analytical solution model; S3, repeating S2 to obtain the stratum loss rate of multiple geological sections; S4, selecting and training a neural network; S5, using the geological parameters of the section to be predicted as input, substituting them into the neural network, and outputting the stratum loss rate of the section to be predicted; S6, substituting them into the analytical solution model to obtain the predicted stratum deformation value of the section to be predicted. The present invention has a clear theoretical basis, is simple to apply, requires a small number of samples, and can be easily applied in engineering under the premise of meeting the prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the field of shield tunneling and stratum deformation prediction, and in particular to a dual-engine prediction method and system for stratum deformation caused by shield tunneling. Background Art

[0002] Ground deformation caused by tunnel excavation is a critical concern during tunnel construction. While shield tunneling can minimize ground disturbance caused by excavation, shield-induced ground deformation is still unavoidable. The ground loss rate is a crucial parameter in the analysis of ground deformation caused by tunnel excavation. This parameter is defined as the ratio of the difference between the tunnel excavation volume and the actual tunnel space occupied to the tunnel excavation volume. The ground loss rate is an input in many ground deformation analysis methods for tunnel excavation, and is generally determined based on engineering experience and carries a certain degree of uncertainty.

[0003] In recent years, with advances in computer technology, artificial neural networks have been widely used across various industries. Numerous researchers have used various neural networks to conduct extensive research and analysis on ground deformation caused by tunneling, achieving remarkable results. However, existing neural network methods for analyzing ground deformation caused by shield tunneling generally use shield tunneling parameters as the input layer and ground deformation as the output layer. While this approach lacks a rigorous theoretical basis, the powerful nonlinear fitting capabilities of neural networks allow for reasonable fitting results when the sample size is large enough. However, these methods suffer from shortcomings such as unclear theoretical frameworks, large sample sizes, lengthy network training times, and insufficiently reliable results. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a dual-engine prediction method and system for stratum deformation caused by shield tunneling. It has a clear theoretical basis, is simple to apply, and requires a small amount of samples. It can be convenient for engineering applications while meeting the prediction accuracy.

[0005] The present invention adopts the following technical solutions:

[0006] In one aspect, the present invention provides a dual-engine prediction method for ground deformation caused by shield tunneling, comprising:

[0007] S1. Determine an analytical solution model for stratum deformation caused by shield tunneling; the analytical solution model inputs are geological parameters of the geological section and the stratum loss rate, and outputs an analytical solution for stratum deformation of the geological section;

[0008] S2. Obtaining a formation loss rate of a geological section based on geological parameters of the section, the field-measured deformation value of the section, and the analytical solution model of step S1;

[0009] S3, repeating step S2 to obtain formation loss rates of multiple geological sections;

[0010] S4. Select a neural network, use the geological parameters of multiple geological sections as sample inputs of the neural network, use the formation loss rates of multiple geological sections obtained in step S3 as sample expected outputs of the neural network, and train the neural network to obtain a trained neural network;

[0011] S5. Select a section to be predicted, use the geological parameters of the section to be predicted as input, substitute them into the trained neural network obtained in step S4, and output a predicted value of the formation loss rate of the section to be predicted;

[0012] S6. Substitute the geological parameters of the section to be predicted and the predicted value of the formation loss rate of the section to be predicted obtained in step S5 into the analytical solution model to obtain the predicted value of the formation deformation of the section to be predicted.

[0013] Any possible implementation method as described above further provides an implementation method, wherein the method also includes: S0, constructing a tunnel geological section database, the database including geological parameters of multiple geological sections of the tunnel, and on-site measured deformation values ​​corresponding to the geological sections; the geological parameters of the geological sections are read from the database.

[0014] In any of the possible implementations described above, a further implementation is provided, in step S1, the analytical solution model of the formation deformation is expressed as:

[0015] u(x,y,z)=f(x,y,z,r,h,E s ,v s ,u w ,k0,η,σ' v ,q,E q ,v q )

[0016] Where: u(x, y, z) is the ground deformation at the coordinate (x, y, z) in a geological section; x, y and z are the coordinate variables in the shield tunneling direction, horizontal direction and vertical direction respectively; r is the tunnel radius; h is the tunnel depth; E s and v s are the soil compression modulus and Poisson's ratio at the tunnel center, respectively; σ' v is the effective vertical earth pressure at the tunnel center; u w is the water pressure at the center of the tunnel; k0 is the static earth pressure coefficient at the center of the tunnel; q is the grouting volume per unit length; E q and v q are the elastic modulus and Poisson's ratio of the grouting body respectively; η is the formation loss rate.

[0017] For any of the possible implementations described above, a further implementation is provided, in which the specific method of step S2 is:

[0018] S21, select geological section X, read geological parameters of the geological section X, and measured deformation values ​​u at different positions i of the geological section X i , and record the coordinate positions of different positions i;

[0019] S22, given the initial formation loss rate η0 and the formation loss rate incremental step Δη, and taking η0 as the initial formation loss rate trial value, and giving the accuracy ξ;

[0020] S23, using the geological parameters of the geological section X in step S21 and the initial formation loss rate η0 in step S22 as trial values, substitute them into the analytical solution model to obtain the analytical solution of the formation deformation of the geological section X under the trial value of the formation loss rate, and obtain the deformation analytical value u at the same position as the measured formation deformation i ';

[0021] S24, calculate |u i -u i '|, determine whether it is less than the precision ξ;

[0022] S25. If the accuracy requirement is met, the current calculated value of the formation loss rate is the true value of the formation loss rate corresponding to the measured deformation value at the position i on the geological section X. If the accuracy requirement is not met, the formation loss rate is increased by Δη as a new calculated value of the formation loss rate, and steps S23 and S24 are repeated until the accuracy requirement is met. The formation loss rate η corresponding to the measured deformation value at different positions i on the geological section X can be obtained. i ;

[0023] S26, the formation loss rate η corresponding to the measured deformation value at different positions i of the geological section X i The formation loss rate vector η X , calculate η based on the weight vector k of the measured position X k, and obtain the formation loss rate of geological section X.

[0024] Any possible implementation as described above further provides an implementation, in which, in step S4, the output layer of the neural network is one node, and the number of input layers corresponds to the sample input vector; the sample input vector is composed of multiple geological parameters of the geological section.

[0025] As for any possible implementation described above, a further implementation is provided, wherein the neural network is a BP neural network or a RF neural network, and other neural networks may also be used.

[0026] Any possible implementation as described above, further provides an implementation, in step S1, the geological parameters include multiple ones of tunnel radius, tunnel burial depth, compression modulus and Poisson's ratio of soil, effective vertical earth pressure, water pressure, earth pressure coefficient, grouting volume per unit length, grouting elastic modulus and Poisson's ratio, lining elastic modulus and Poisson's ratio, and lining thickness.

[0027] Any of the possible implementations described above further provides an implementation, in which in step S0, the method for constructing a tunnel geological section database is:

[0028] S01. Create a soil layer dictionary, where the key is the soil layer number and the corresponding value is a list of soil layer parameters. The parameters in the list are: soil layer density, deformation modulus, soil pressure coefficient, and Poisson's ratio.

[0029] S02. Divide the geological section, read the soil layer numbers and corresponding thicknesses of the section on the geological map, and create three corresponding lists: one is the soil layer number list, i.e., the distribution of soil layers from the surface downward; the second is the soil layer thickness list, i.e., the thickness of each soil layer corresponding to the elements in the soil layer number list; and the third is a list of other parameters such as tunnel burial depth and groundwater depth;

[0030] S03. Establish a geological condition dictionary. Each section has three key-value pairs. The first key is the mileage number + soil layer number, and the corresponding value is a list of soil layer numbers; the second key is the mileage number + soil layer thickness, and the corresponding value is a list of soil layer thicknesses; the third key is the mileage number + tunnel depth and groundwater, and the corresponding value is a list of tunnel depth, groundwater depth, and other parameters.

[0031] Any possible implementation as described above further provides an implementation, in which, as the shield advances, the original predicted section is used as a sample to enter the training of the neural network as a continuous correction to the neural network.

[0032] On the other hand, the present invention also provides a dual-engine prediction system for ground deformation caused by shield tunneling, which is used to implement the above method and includes:

[0033] The analytical solution module inputs the geological parameters of the geological section and the stratum loss rate, and outputs the analytical solution of the stratum deformation of the geological section based on the analytical solution model of the stratum deformation caused by tunneling;

[0034] a formation loss rate iterative determination module, which iteratively obtains the formation loss rate of a geological section based on geological parameters of the geological section, the field measured deformation value of the geological section, and the analytical solution model;

[0035] A neural network training module selects a neural network, uses geological parameters of multiple geological sections as sample inputs of the neural network, uses the formation loss rates of multiple geological sections obtained by the formation loss rate iterative determination module as sample expected outputs of the neural network, trains the neural network, and obtains a trained neural network;

[0036] The stratum deformation prediction module takes the geological parameters of the section to be predicted as input, substitutes them into the trained neural network, and outputs the stratum loss rate of the section to be predicted; then the geological parameters of the section to be predicted and the stratum loss rate of the section to be predicted are substituted into the analytical solution model to obtain the stratum deformation prediction value of the section to be predicted.

[0037] Any possible implementation as described above further provides an implementation, wherein the system also includes a tunnel geological section database, the database including geological parameters of multiple geological sections of the tunnel and field measured deformation values ​​corresponding to the geological sections.

[0038] The beneficial effects of the present invention are:

[0039] The present invention takes into account the empirical nature of the formation loss rate in the analysis of formation deformation caused by shield excavation, adopts a neural network to predict the formation loss rate, and combines it with the analytical solution of formation deformation caused by shield tunneling to achieve a dual-engine prediction of formation deformation. This prediction approach gives full play to the advantages of analytical methods and neural networks. Compared with other prediction methods, it has the advantages of clear theory, short prediction time, small sample requirements, and high credibility of results. Furthermore, this method provides a prediction approach for formation deformation caused by underground excavation, and can also be used to predict formation deformation caused by mechanical excavation of vertical shafts, foundation pits and other underground engineering projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Shown is a flowchart of a dual-engine prediction method for ground deformation caused by shield tunneling according to an embodiment of the present invention.

[0041] Figure 2 Shown is a structural diagram of the tunnel geological section database in the embodiment.

[0042] Figure 3 The figure shows a comparison curve of the measured deformation value of the ground surface just above the center of the tunnel and the dual-engine prediction result when the shield engineering (a shield engineering of Shanghai Metro) in the embodiment is applied in a typical shield engineering in soft soil strata.

[0043] Figure 4 The figure shows a comparison curve of the measured deformation of the ground surface directly above the center of the tunnel, the numerical simulation and the dual-engine prediction results when the shield project (a shield project of the Guanghua Intercity Railway) is applied in a typical shield project in hard rock formations in the embodiment. DETAILED DESCRIPTION

[0044] The following will describe in detail specific embodiments of the present invention with reference to the accompanying drawings. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated, and they can be combined with each other to achieve better technical effects.

[0045] like Figure 1 As shown, an embodiment of the present invention provides a dual-engine prediction method for ground deformation caused by shield tunneling, comprising:

[0046] S1. Determine an analytical solution model for stratum deformation caused by tunneling; the analytical solution model inputs are geological parameters of the geological section and the stratum loss rate, and outputs an analytical solution for stratum deformation of the geological section;

[0047] S2. Obtaining a formation loss rate of a geological section based on geological parameters of the section, the field-measured deformation value of the section, and the analytical solution model of step S1;

[0048] S3, repeating step S2 to obtain formation loss rates of multiple geological sections;

[0049] S4. Select a neural network, use the geological parameters of multiple geological sections as sample inputs of the neural network, use the formation loss rates of multiple geological sections obtained in step S3 as sample expected outputs of the neural network, and train the neural network to obtain a trained neural network;

[0050] S5. Select a section to be predicted, use the geological parameters of the section to be predicted as input, substitute them into the trained neural network obtained in step S4, and output a predicted value of the formation loss rate of the section to be predicted;

[0051] S6. Substitute the geological parameters of the section to be predicted and the predicted value of the formation loss rate of the section to be predicted obtained in step S5 into the analytical solution model to obtain the predicted value of the formation deformation of the section to be predicted.

[0052] In a specific embodiment, the method also includes: S0, constructing a tunnel geological section database, the database including geological parameters of multiple geological sections of the tunnel, and field measured deformation values ​​corresponding to the geological sections; the geological parameters of the geological sections are read from the database.

[0053] In a specific embodiment, the analytical solution model of formation deformation can adopt any one of the existing technologies, such as:

[0054] Reference [1] Zhao Jiale, Li Binwei, Fu Yonggui, Liu Yang. Parameter determination of analytical method for ground deformation caused by tunnel excavation [J]. Railway Construction, 2023, 63(2): 99-102. and reference [2] Zhao Jiale. Three-dimensional analysis and prediction of ground deformation caused by shield tunneling [D]. University of Science and Technology Beijing, 2024, provide a specific analytical solution model.

[0055] In a specific embodiment, in step S1, the analytical solution model of the formation deformation is expressed as:

[0056] u(x,y,z)=f(x,y,z,r,h,E s ,v s ,u w ,k0,η,σ' v ,q,E q ,v q )

[0057] Where: u(x, y, z) is the ground deformation at the coordinate (x, y, z) in a geological section; x, y and z are the coordinate variables in the shield tunneling direction, horizontal direction and vertical direction respectively; r is the tunnel radius; h is the tunnel depth; E s and v s are the soil compression modulus and Poisson's ratio at the tunnel center, respectively; σ' v is the effective vertical earth pressure at the tunnel center; u w is the water pressure at the center of the tunnel; k0 is the static earth pressure coefficient at the center of the tunnel; q is the grouting volume per unit length; E q and v q are the elastic modulus and Poisson's ratio of the grouting body respectively; η is the formation loss rate.

[0058] In a specific embodiment, the specific method of step S2 is:

[0059] S21, select geological section X, read geological parameters of the geological section X, and measured deformation values ​​u at different positions i of the geological section X i , and record the coordinate positions of different positions i;

[0060] S22, given the initial formation loss rate η0 and the formation loss rate incremental step Δη, and taking η0 as the initial formation loss rate trial value, and giving the accuracy ξ;

[0061] S23, using the geological parameters of the geological section X in step S21 and the initial formation loss rate η0 in step S22 as trial values, substitute them into the analytical solution model to obtain the analytical solution of the formation deformation of the geological section X under the trial value of the formation loss rate, and obtain the deformation analytical value u at the same position as the measured formation deformation i ';

[0062] S24, calculate |u i -u i '|, determine whether it is less than the precision ξ;

[0063] S25. If the accuracy requirement is met, the current calculated value of the formation loss rate is the true value of the formation loss rate corresponding to the measured deformation value at the position i on the geological section X. If the accuracy requirement is not met, the formation loss rate is increased by Δη as a new calculated value of the formation loss rate, and steps S23 and S24 are repeated until the accuracy requirement is met. The formation loss rate η corresponding to the measured deformation value at different positions i on the geological section X can be obtained. i ;

[0064] S26, the formation loss rate η corresponding to the measured deformation value at different positions i of the geological section X i The formation loss rate vector η X , calculate η based on the weight vector k of the measured position X k, and obtain the formation loss rate of geological section X.

[0065] In a specific embodiment, in step S4, the output layer of the neural network is a node, and the number of input layers corresponds to the sample input vector; the sample input vector is composed of multiple geological parameters of the geological section.

[0066] Generally, the sample input vector should include: tunnel radius, tunnel burial depth, soil compression modulus and Poisson's ratio, effective vertical earth pressure, water pressure, earth pressure coefficient, grouting volume per unit length, grouting elastic modulus and Poisson's ratio, lining elastic modulus and Poisson's ratio, lining thickness, etc.

[0067] However, in practice, the sample input vector can be adjusted based on actual conditions. For example, when this method is applied to a specific project, the tunnel radius, lining elastic modulus and Poisson's ratio, and lining thickness remain constant and can be omitted from the input. Because the geological conditions and construction parameters of different shield tunnels vary significantly, training samples from different projects should not be mixed when the input vector contains few parameters and the training sample size is small.

[0068] It is important to note that the geological parameters in the sample input vector do not need to completely correspond to the geological parameters required by the analytical solution model. The selection of geological parameters in the input vector should be determined through experiments or experience based on specific circumstances.

[0069] In a specific embodiment, in step S1, the geological parameters include multiple ones of tunnel radius, tunnel burial depth, compression modulus and Poisson's ratio of soil, effective vertical earth pressure, water pressure, earth pressure coefficient, grouting volume per unit length, grouting elastic modulus and Poisson's ratio, lining elastic modulus and Poisson's ratio, and lining thickness.

[0070] In a specific embodiment, in step S0, the method for constructing the tunnel geological section database is:

[0071] S01. Create a soil layer dictionary, where the key is the soil layer number and the corresponding value is a list of soil layer parameters. The parameters in the list are: soil layer density, deformation modulus, soil pressure coefficient, and Poisson's ratio.

[0072] S02. Divide the geological section, read the soil layer numbers and corresponding thicknesses of the section on the geological map, and create three corresponding lists: one is the soil layer number list, i.e., the distribution of soil layers from the surface downward; the second is the soil layer thickness list, i.e., the thickness of each soil layer corresponding to the elements in the soil layer number list; and the third is a list of other parameters such as tunnel burial depth and groundwater depth;

[0073] S03. Establish a geological condition dictionary. Each section has three key-value pairs. The first key is the mileage number + soil layer number, and the corresponding value is a list of soil layer numbers; the second key is the mileage number + soil layer thickness, and the corresponding value is a list of soil layer thicknesses; the third key is the mileage number + tunnel depth and groundwater, and the corresponding value is a list of tunnel depth, groundwater depth, and other parameters.

[0074] In a specific embodiment, a simplified tunnel geological section database is established using a python list corresponding method. The structure of the tunnel geological section database is as follows: Figure 2 shown.

[0075] Using this method to build a geological section database can greatly simplify the parameter calculation process and is the basis for programming and automation of tedious steps such as neural network sample acquisition and model correction.

[0076] In a specific embodiment, as the shield machine advances, the original predicted section is used as a sample and input into the training of the neural network as a continuous correction to the neural network.

[0077] An embodiment of the present invention provides a dual-engine prediction system for ground deformation caused by shield tunneling, which is used to implement the above method. The system includes:

[0078] The analytical solution module inputs the geological parameters of the geological section and the stratum loss rate, and outputs the analytical solution of the stratum deformation of the geological section based on the analytical solution model of the stratum deformation caused by tunneling;

[0079] a formation loss rate iterative determination module, which iteratively obtains the formation loss rate of a geological section based on geological parameters of the geological section, the field measured deformation value of the geological section, and the analytical solution model;

[0080] A neural network training module selects a neural network, uses geological parameters of multiple geological sections as sample inputs of the neural network, uses the formation loss rates of multiple geological sections obtained by the formation loss rate iterative determination module as sample expected outputs of the neural network, trains the neural network, and obtains a trained neural network;

[0081] The stratum deformation prediction module takes the geological parameters of the section to be predicted as input, substitutes them into the trained neural network, and outputs the stratum loss rate of the section to be predicted; then the geological parameters of the section to be predicted and the stratum loss rate of the section to be predicted are substituted into the analytical solution model to obtain the stratum deformation prediction value of the section to be predicted.

[0082] In a specific embodiment, the system further includes a tunnel geological section database, which includes geological parameters of multiple geological sections of the tunnel and field measured deformation values ​​corresponding to the geological sections.

[0083] Compared with existing methods, this method fully utilizes the advantages and disadvantages of neural networks and analytical theories, and has the advantages of clear theoretical basis, short prediction time, simple application, and small sample requirements. When applied to a typical shield project in soft soil (a shield project of Shanghai Metro), the comparison curve between the measured surface deformation value directly above the tunnel center and the dual-engine prediction result is shown in the figure below. Figure 3 As shown in the figure, when applied to a typical shield project in hard rock formations (a shield project of the Guanghua Intercity Railway), the measured deformation of the ground surface just above the tunnel center, the numerical simulation and the dual-engine prediction results are compared. Figure 4 The application results of a shield tunneling project in Guanghua Intercity Railway show that when the number of samples is only 12 groups, a simple single hidden layer BP neural network can achieve Figure 4 The prediction effect shown takes about 3.0 seconds; conventional neural network prediction methods often require thousands of sets of samples. The numerical simulation method takes about 6 hours for the solution process alone, which verifies that this method has good engineering applicability.

[0084] The present invention is applicable to the prediction of ground deformation caused by shield excavation, and can also be used for the prediction of ground deformation caused by mechanical excavation of vertical shafts, foundation pits and other underground projects.

[0085] Although several embodiments of the present invention have been described herein, those skilled in the art will appreciate that modifications may be made to the embodiments herein without departing from the spirit of the present invention. The above embodiments are merely exemplary and should not be used as limitations on the scope of the present invention.

Claims

1. A dual-engine prediction method for ground deformation caused by shield tunneling, characterized in that: The method comprises: S1. Determine an analytical solution model for stratum deformation caused by shield tunneling; the analytical solution model inputs are geological parameters of the geological section and the stratum loss rate, and outputs an analytical solution for stratum deformation of the geological section; S2. Obtaining a formation loss rate of a geological section based on geological parameters of the section, the field-measured deformation value of the section, and the analytical solution model of step S1; S3, repeating step S2 to obtain formation loss rates of multiple geological sections; S4. Select a neural network, use the geological parameters of multiple geological sections as sample inputs of the neural network, use the formation loss rates of multiple geological sections obtained in step S3 as sample expected outputs of the neural network, and train the neural network to obtain a trained neural network; S5. Select a section to be predicted, use the geological parameters of the section to be predicted as input, substitute them into the trained neural network obtained in step S4, and output the formation loss rate of the section to be predicted; S6. Substitute the geological parameters of the section to be predicted and the formation loss rate of the section to be predicted obtained in step S5 into the analytical solution model to obtain a predicted value of the formation deformation of the section to be predicted.

2. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 1, characterized in that: The method also includes: S0, constructing a tunnel geological section database, the database including geological parameters of multiple geological sections of the tunnel and field measured deformation values ​​corresponding to the geological sections; the geological parameters of the geological sections are read from the database.

3. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 1, characterized in that: In step S1, the analytical solution model of the formation deformation is expressed as: u(x,y,z)=f(x,y,z,r,h,E s ,v s ,u w ,k0,η,σ' v ,q,E q ,v q ) Where: u(x, y, z) is the ground deformation at the coordinate (x, y, z) in a geological section; x, y and z are the coordinate variables in the shield tunneling direction, horizontal direction and vertical direction respectively; r is the tunnel radius; h is the tunnel depth; E s and v s are the soil compression modulus and Poisson's ratio at the tunnel center, respectively; σ' v is the effective vertical earth pressure at the tunnel center; u w is the water pressure at the center of the tunnel; k0 is the static earth pressure coefficient at the center of the tunnel; q is the grouting volume per unit length; E q and v q are the elastic modulus and Poisson's ratio of the grouting body respectively; η is the formation loss rate.

4. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 1, characterized in that: The specific method of step S2 is: S21, select geological section X, read geological parameters of the geological section X, and measured deformation values ​​u at different positions i of the geological section X i , and record the coordinate positions of different positions i; S22, given the initial formation loss rate η0 and the formation loss rate incremental step Δη, and taking η0 as the initial formation loss rate trial value, and giving the accuracy ξ; S23, using the geological parameters of the geological section X in step S21 and the initial formation loss rate η0 in step S22 as trial values, substitute them into the analytical solution model to obtain the analytical solution of the formation deformation of the geological section X under the trial value of the formation loss rate, and obtain the deformation analytical value u at the same position as the measured formation deformation i '; S24, calculate |u i -u i '|, determine whether it is less than the precision ξ; S25. If the accuracy requirement is met, the current calculated value of the formation loss rate is the true value of the formation loss rate corresponding to the measured deformation value at the position i on the geological section X. If the accuracy requirement is not met, the formation loss rate is increased by Δη as a new calculated value of the formation loss rate, and steps S23 and S24 are repeated until the accuracy requirement is met. The formation loss rate η corresponding to the measured deformation value at different positions i on the geological section X can be obtained. i ; S26, the formation loss rate η corresponding to the measured deformation value at different positions i of the geological section X i The formation loss rate vector η X , calculate η based on the weight vector k of the measured position X k, and obtain the formation loss rate of geological section X.

5. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 1, characterized in that: In step S4, the output layer of the neural network is a node, and the number of input layers corresponds to the sample input vector; the sample input vector is composed of multiple geological parameters of the geological section.

6. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 1, characterized in that: In step S1, the geological parameters include multiple ones of tunnel radius, tunnel burial depth, soil compression modulus and Poisson's ratio, effective vertical earth pressure, water pressure, earth pressure coefficient, grouting volume per unit length, grouting elastic modulus and Poisson's ratio, lining elastic modulus and Poisson's ratio, and lining thickness.

7. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 2, characterized in that: In step S0, the method for constructing the tunnel geological section database is: S01. Create a soil layer dictionary, where the key is the soil layer number and the corresponding value is a list of soil layer parameters. The parameters in the list are: soil layer density, deformation modulus, soil pressure coefficient, and Poisson's ratio. S02. Divide the geological section, read the soil layer numbers and corresponding thicknesses of the section on the geological map, and create three corresponding lists: one is the soil layer number list, i.e., the distribution of soil layers from the surface downward; the second is the soil layer thickness list, i.e., the thickness of each soil layer corresponding to the elements in the soil layer number list; and the third is a list of other parameters such as tunnel burial depth and groundwater depth; S03. Establish a geological condition dictionary. Each section has three key-value pairs. The first key is the mileage number + soil layer number, and the corresponding value is a list of soil layer numbers; the second key is the mileage number + soil layer thickness, and the corresponding value is a list of soil layer thicknesses; the third key is the mileage number + tunnel depth and groundwater, and the corresponding value is a list of tunnel depth, groundwater depth, and other parameters.

8. The dual-engine prediction method for ground deformation caused by shield tunneling according to claim 1, characterized in that: As the shield machine advances, the original predicted section is used as a sample to enter the training of the neural network, and the neural network is continuously corrected.

9. A dual-engine prediction system for ground deformation caused by shield tunneling, characterized in that: The system is used to implement the method according to any one of claims 1 to 8, and the system includes: The analytical solution module inputs the geological parameters of the geological section and the stratum loss rate, and outputs the analytical solution of the stratum deformation of the geological section based on the analytical solution model of the stratum deformation caused by tunneling; a formation loss rate iterative determination module, which iteratively obtains the formation loss rate of a geological section based on geological parameters of the geological section, the field measured deformation value of the geological section, and the analytical solution model; A neural network training module selects a neural network, uses geological parameters of multiple geological sections as sample inputs of the neural network, uses the formation loss rates of multiple geological sections obtained by the formation loss rate iterative determination module as sample expected outputs of the neural network, trains the neural network, and obtains a trained neural network; The stratum deformation prediction module takes the geological parameters of the section to be predicted as input, substitutes them into the trained neural network, and outputs the stratum loss rate of the section to be predicted; then the geological parameters of the section to be predicted and the stratum loss rate of the section to be predicted are substituted into the analytical solution model to obtain the stratum deformation prediction value of the section to be predicted.

10. The dual-engine prediction system for ground deformation caused by shield tunneling according to claim 9, characterized in that: The system further comprises a tunnel geological section database, wherein the database comprises geological parameters of a plurality of geological sections of the tunnel and field measured deformation values ​​corresponding to the geological sections.

Images