Shield tunnel geotechnical parameter inversion and tunneling parameter optimization method based on approximate model

Abstract

The application provides a shield tunnel geotechnical parameter inversion and tunneling parameter optimization method based on an approximate model. First, the geotechnical parameters are inverted by using the ground surface displacement of the tunneling section, then the ground surface displacement of the subsequent excavation is predicted by using the inverted geotechnical parameter results, and the construction tunneling parameters are optimized in time. Specifically, the corresponding relationship between the geotechnical parameters theta and the ground surface displacement y is established based on the approximate model method: y=Appro(theta); the target optimization function minF(theta) is established according to the obtained ground surface displacement monitoring data, and the particle swarm optimization algorithm is used to solve the optimal geotechnical parameter solution theta best ; the corresponding relationship between the shield construction parameters alpha and the ground surface displacement y is established based on the approximate model method: y=Appro(alpha); the target optimization function minF(alpha) is established according to the maximum displacement value u lim allowed by the specification, the particle swarm optimization algorithm is used to solve the target function, and the recommended value alpha best of the shield construction parameters is obtained. The application can reduce the calculation cost caused by frequent calling of the numerical model during parameter optimization, and improve the calculation efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent technology for shield tunnel engineering, and in particular to a method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels based on an approximate model. Background Technology

[0002] With the development of computing technology, numerical simulation, especially the finite element method, has become an important method for analyzing the mechanical behavior of shield tunnels. Its basic steps can be divided into: (1) Engineering site information analysis, including borehole sampling, in-situ site experiments, and laboratory experiments. (2) Establishing and solving the corresponding numerical model according to the engineering scale and problem requirements. This can be achieved through existing commercial finite element software such as Abaqus, Midas, and Plaxis. (3) Comparing the calculation results of the numerical simulation with the monitoring values ​​on site, evaluating the reliability of the model calculation results, and adjusting the subsequent construction plan.

[0003] However, due to the inherent uncertainty of soil and rock media, the values ​​of soil and rock parameters have a significant impact on the numerical calculation results, and this impact is difficult to eliminate through experimental means.

[0004] To improve the reliability of numerical results, inverting geotechnical parameters based on field measured data is an effective method to improve the calculation results of numerical models. The essence of geotechnical parameter inversion calculation is to continuously adjust the geotechnical input parameters so that the calculation results of the numerical model are as close as possible to the field monitoring data. This process requires repeated calls to the numerical calculation model. However, considering that typical 3D numerical models for tunnel boring machine (TBM) construction typically contain a large number of finite element meshes and have multiple construction and calculation steps, this incurs significant computational costs. Furthermore, TBM construction may encounter situations where it passes under existing sensitive structures. Numerical simulation can provide suggested values ​​for TBM construction parameters during such passages, aiming to reduce surface settlement. These problems can be categorized as parameter optimization problems. Solving this optimization problem still requires frequent use of TBM construction numerical models, and the high computational cost remains unavoidable.

[0005] Therefore, the existing methods for inverting geotechnical parameters and optimizing tunneling parameters based on the finite element method for shield tunnels suffer from low computational efficiency and high computational cost. Summary of the Invention

[0006] This invention provides a method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels based on an approximate model to address the aforementioned technical problems. First, geotechnical parameters are inverted using the surface displacement of the already excavated section. Then, the inverted geotechnical parameter results are used to predict the surface displacement of subsequent excavation and optimize the construction and tunneling parameters in a timely manner. This solves the problems of low computational efficiency and high computational cost inherent in finite element method-based methods for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels.

[0007] This invention provides a method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels based on an approximate model, comprising the following steps:

[0008] Step S30: Based on the approximate model method, establish the correspondence between the soil and rock parameters θ and the surface displacement y: y = Appro(θ);

[0009] Step S40: Based on the acquired surface displacement monitoring data, establish the objective optimization function minF(θ), and use the particle swarm optimization algorithm to solve for the optimal soil and rock parameter solution θ. best ;

[0010] Step S50: Based on the approximate model method, establish the correspondence between the shield tunneling construction parameter α and the ground surface displacement y: y = Appro(α);

[0011] Step S60, according to the maximum allowable displacement value u required by the specification. lim A target optimization function minF(α) is established, and the particle swarm optimization algorithm is used to solve the target function to obtain the suggested values ​​α for the shield tunneling construction parameters. best .

[0012] In one implementation, step S30 specifically includes the following steps:

[0013] Step S301: Determine the design variables θ = [θ1, θ2, ..., θ n ] T And the upper and lower limits of the corresponding variables (θ) L ,θ U Based on the hypercubic Latin sampling method, the input dataset X = [θ] is generated. (1) ,θ (2) ,…,θ (m) ];

[0014] The input dataset can be represented by an n×m dimensional matrix:

[0015]

[0016] Step S302: Based on the input dataset X, establish and run m sets of finite element models, and extract the observed nodal displacements of the m sets of finite element models. The input-output dataset {X,Y} is divided into a training set (TrainingSet) and a test set (TestSet) according to a preset ratio.

[0017] Step S303: Based on the training set TrainingSet generated in step A2, establish the approximate model method set ApproSet = {Appro1, Appro2, Appro3};

[0018] Among them, Appro1 is implemented based on the Kriging interpolation method, Appro2 is implemented based on the support vector regression method, and Appro3 is implemented based on the radial basis function method.

[0019] Step S304: Based on the test set TestSet generated in step S302 and the approximate model method set ApproSet established in step S303, the optimal approximate model y = Appro(θ) is obtained by combining the root mean square error RMSE, the multiple correlation coefficient MSE, and the mean absolute error AEE.

[0020] In one implementation, in step S302, the input-output dataset {X,Y} is divided into a training set (TrainingSet) and a test set (TestSet) in an 8:2 ratio.

[0021] In one implementation, step S40 specifically includes the following steps:

[0022] Step S401: Obtain the monitored displacement u = [u1, u2, ..., u] of the ground surface at different construction stages. l Construct the objective function:

[0023] minF(θ)=‖Appro(θ)-u‖

[0024] stθ∈(θ L ,θ U );

[0025] Step S402: Solve the objective function θ using the particle swarm optimization algorithm. best =minF(θ), to obtain the inversion result θ of the soil and rock parameters. best .

[0026] In one implementation, step S50 specifically includes the following steps:

[0027] Step S501: Determine the shield tunneling construction parameters to be optimized: α = [α1, α2, ..., α n ] T And the corresponding engineering experience value range (α) for construction parameters. L ,α U );

[0028] Step S502: Based on the hypercubic Latin sampling method, generate the input dataset X = [α] (1) ,α (2) ,…,α (m) ];

[0029] Step S503: Based on the input dataset X, establish and run m sets of finite element models, and extract the observed nodal displacements of the m finite element models. The input-output dataset {X,Y} is divided into a training set (TrainingSet) and a test set (TestSet) according to a preset ratio.

[0030] Step S504: Based on the training set generated in step S303, establish an approximate model method set ApproSet = {Appro1, Appro2, Appro3};

[0031] Among them, Appro1 is implemented based on the Kriging interpolation method, Appro2 is implemented based on the support vector regression method, and Appro3 is implemented based on the radial basis function method.

[0032] Step S505: Based on the test set TestSet generated in step S503 and the approximate model method set ApproSet established in step S504, the optimal approximate model y = Appro(α) is obtained.

[0033] In one implementation, step S60 specifically includes the following steps: Step S601, obtaining the displacement limit value u of the ground surface at different construction stages. lim =[u lim1 ,u lim2 ,…,u liml Construct the objective function:

[0034] minF(α)=‖Appro(α)-u lim ||;

[0035] Where, stα∈(α L ,α U );

[0036] Step S602: Solve the objective function α using the particle swarm optimization algorithm. best =minF(α), to obtain the optimized result α of the shield tunneling construction parameters. best .

[0037] In one embodiment, before step S30, there is a step S20 to determine the soil and rock parameters to be inverted, wherein the soil and rock parameters to be inverted include at least one of elastic modulus E, Poisson's ratio μ, and unit weight γ of the soil layer.

[0038] In one implementation, before step S20, step S10 is included: analysis of engineering background and site information.

[0039] In one implementation, step S10 specifically includes: obtaining the value range of parameters reflecting the mechanical properties of each stratum and selecting a suitable shield tunneling section for inversion; wherein the suitable shield tunneling section for inversion has the conditions for setting up observation points.

[0040] In one embodiment, in step S503, the input-output dataset {X,Y} is divided into the training set TrainingSet and the test set TestSet in an 8:2 ratio.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0042] 1. The inverted geotechnical parameters can improve the accuracy of the numerical model. Therefore, applying the inverted geotechnical parameters to optimize construction parameters can yield more reliable results. Furthermore, the approximate model method is also used to establish the relationship between shield tunneling parameters and surface displacement. On the one hand, this reduces the computational cost caused by frequent calls to the numerical model during parameter optimization, improving computational efficiency. On the other hand, it can also quantitatively provide suggested values ​​for construction parameters.

[0043] 2. An input-output relationship between soil and rock parameters and surface displacement was established based on an approximate model, which significantly improved the computational efficiency of the inversion and reduced the computational cost of the inversion. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the method for inverting ground parameters and optimizing tunneling parameters in shield tunnels based on an approximate model, as described in this embodiment of the invention.

[0045] Figure 2 This is a flowchart of the construction method based on an approximation model in an embodiment of the present invention.

[0046] Figure 3 This is a flowchart illustrating the establishment and verification of an approximate model in an embodiment of the present invention.

[0047] Figure 4 This is a schematic diagram of a model in an embodiment of the present invention. Detailed Implementation

[0048] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] like Figure 1 As shown, this invention provides a method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels based on an approximate model, comprising the following steps:

[0050] Step S30: Based on the approximate model method, establish the correspondence between the soil and rock parameters θ and the surface displacement y: y = Appro(θ);

[0051] Step S40: Based on the acquired surface displacement monitoring data, establish the objective optimization function minF(θ), and use the particle swarm optimization algorithm to solve for the optimal soil and rock parameter solution θ. best ;

[0052] Step S50: Based on the approximate model method, establish the correspondence between the shield tunneling construction parameter α and the ground surface displacement y: y = Appro(α);

[0053] Step S60, according to the maximum allowable displacement value u required by the specification. lim A target optimization function minF(α) is established, and the particle swarm optimization algorithm is used to solve the target function to obtain the suggested values ​​α for the shield tunneling construction parameters. best .

[0054] Specifically, in one embodiment, step S30 specifically includes as follows: Figure 3 The relevant steps are shown below:

[0055] Step S301: Determine the design variables θ = [θ1, θ2, ..., θ n ] T And the upper and lower limits of the corresponding variables (θ) L ,θ U Based on the hypercubic Latin sampling method, the input dataset X = [θ] is generated. (1) ,θ (2) ,…,θ (m) ];

[0056] The input dataset can be represented by an n×m dimensional matrix:

[0057]

[0058] Step S302: Based on the input dataset X, establish and run m sets of finite element models, and extract the observed nodal displacements of the m sets of finite element models. The input-output dataset {X,Y} is divided into a training set (TrainingSet) and a test set (TestSet) according to a preset ratio.

[0059] Step S303: Based on the training set TrainingSet generated in step A2, establish the approximate model method set ApproSet = {Appro1, Appro2, Appro3};

[0060] Among them, Appro1 is implemented based on the Kriging interpolation method, Appro2 is implemented based on the support vector regression method, and Appro3 is implemented based on the radial basis function method.

[0061] Step S304: Based on the test set TestSet generated in step S302 and the approximate model method set ApproSet established in step S303, the optimal approximate model y=Appro(θ) is obtained by combining the root mean square error RMSE, the multiple correlation coefficient MSE, and the mean absolute error AEE.

[0062] Specifically, in one embodiment, in step S302, the input-output dataset {X,Y} is divided into a training set (TrainingSet) and a test set (TestSet) in an 8:2 ratio.

[0063] Specifically, in one embodiment, step S40 includes the following steps:

[0064] Step S401: Obtain the monitored displacement u = [u1, u2, ..., u] of the ground surface at different construction stages. l Construct the objective function:

[0065] minF(θ)=‖Appro(θ)-u‖

[0066] stθ∈(θ L ,θ U );

[0067] Step S402: Solve the objective function θ using the particle swarm optimization algorithm. best =minF(θ), to obtain the inversion result θ of the soil and rock parameters. best .

[0068] Specifically, in one embodiment, step S50 specifically includes as follows: Figure 3 The relevant steps are shown below:

[0069] Step S501: Determine the shield tunneling construction parameters to be optimized: α = [α1, α2, ..., α n ] T And the corresponding engineering experience value range (α) for construction parameters. L ,α U );

[0070] Step S502: Based on the hypercubic Latin sampling method, generate the input dataset X = [α] (1) ,α (2) ,…,α (m) ];

[0071] Step S503: Based on the input dataset X, establish and run m sets of finite element models, and extract the observed nodal displacements of the m finite element models. The input-output dataset {X,Y} is divided into a training set (TrainingSet) and a test set (TestSet) according to a preset ratio.

[0072] Step S504: Based on the training set generated in step S303, establish an approximate model method set ApproSet = {Appro1, Appro2, Appro3};

[0073] Among them, Appro1 is implemented based on the Kriging interpolation method, Appro2 is implemented based on the support vector regression method, and Appro3 is implemented based on the radial basis function method.

[0074] Step S505: Based on the test set TestSet generated in step S503 and the approximate model method set ApproSet established in step S504, the optimal approximate model y = Appro(α) is obtained.

[0075] Specifically, in one embodiment, step S60 includes the following steps: Step S601, obtaining the displacement limit value u of the ground surface at different construction stages. lim =[u lim1 ,u lim2 ,…,u liml Construct the objective function:

[0076] minF(α)=‖Appro(α)-u lim ||;

[0077] Where, stα∈(α L ,α U );

[0078] Step S602: Solve the objective function α using the particle swarm optimization algorithm. best =minF(α), to obtain the optimized result α of the shield tunneling construction parameters. best .

[0079] Specifically, in one embodiment, before step S30, there is a step S20 to determine the soil and rock parameters to be inverted, wherein the soil and rock parameters to be inverted include at least one of elastic modulus E, Poisson's ratio μ, and unit weight γ of the soil layer.

[0080] Specifically, such as Figure 1 As shown, in one embodiment, before step S20, step S10 is also included: engineering background and site information analysis.

[0081] Specifically, such as Figure 1 As shown, in one embodiment, step S10 specifically includes: obtaining the value range of parameters reflecting the mechanical properties of each stratum and selecting a suitable shield tunneling section for inversion; wherein the suitable shield tunneling section for inversion has the conditions for setting up observation points.

[0082] Specifically, in one embodiment, in step S503, the input-output dataset {X,Y} is divided into the training set TrainingSet and the test set TestSet in an 8:2 ratio.

[0083] The following is combined with Figure 2 , Figure 3 , Figure 4 This paper elaborates on the more specific optimization methods for the inversion of geotechnical parameters and tunneling parameters of shield tunnels based on an approximate model.

[0084] Step 1: Engineering Background and Site Information Analysis. This includes the type, distribution, and engineering characteristics of the soil and rock layers within the construction impact area. It also involves obtaining the value ranges of parameters reflecting the mechanical properties of each stratum and selecting a suitable shield tunneling section for inversion. This inversion section should have the conditions for setting up observation points. Figure 3 A schematic diagram of a type of shield tunneling under surface structures is provided. In this engineering example, it is assumed that the strata shown in the diagram are homogeneous and isotropic. Two sets of displacement observation points are set on the surface, denoted as point A and point B, respectively.

[0085] Step 2: Determine the soil and rock parameters to be inverted. During numerical calculations, the soil and rock input parameters are determined by the soil and rock constitutive model. Taking the widely used Mohr-Coulomb constitutive model as an example, four parameters are generally required: elastic modulus E, Poisson's ratio μ, cohesion c, and internal friction angle ψ. Furthermore, the unit weight γ of the soil layer also significantly affects the model's calculation results. It should be noted that in this example, to avoid singularity in the solution, the elastic modulus E (MPa), Poisson's ratio μ (-), and the unit weight γ of the soil layer (kN / m3) are chosen as the soil and rock parameters to be inverted, while the cohesion c and internal friction angle ψ are set to a fixed value, i.e., θ = [θ...]. E ,θ μ ,θ γ The surface displacement at observation point A will be used as the source of monitoring data. Assuming that the displacement of observation point A during the five construction steps is represented by the vector u... A =[u A1 ,u A2 ,…,u A5 ]express.

[0086] Step 3: Establish the displacement of observation point A in the numerical model With soil and rock parameters θ=[θ E ,θ μ ,θ γ The functional relationship between them. This step will be completed in the following three steps:

[0087] Step 3.1: Determine the design variable θ = [θ E ,θ μ ,θ γ ] and the range of design variables (θ) L ,θ U In this example, θ L = [60, 0.1, 15], θ U=[180,0.3,25]. Based on the hypercubic Latin sampling method (LHS), 60 sets of input samples were obtained, X = [θ 1 ,θ 2 ,…,θ (60) Then, based on the finite element numerical model, the formation response under 60 sets of input sample conditions was calculated, and the displacement at point A was extracted. The 60 displacements were represented as matrices:

[0088]

[0089] At this point, a dataset for training and testing the approximate model has been established. ,Will The dataset is divided into a training set (TrainSet) and a test set (TestSet) in an 8:2 ratio. Therefore, the TrainingSet has 48 input-input samples, and the TestSet has 12 input-output samples.

[0090] Step 3.2: Establish the approximate model method set ApproSet = {Appro1, Appro2, Appro3}. Appro1 is implemented based on the Kriging interpolation method, Appro2 on the support vector regression method, and Appro3 on the radial basis function method. Here, we will briefly introduce the implementation of the Appro3 radial basis function method as an example.

[0091] The radial basis function (RBF) method can approximate complex original functions through a simple linear superposition of basis functions. The principle of using RBF for regression fitting can be expressed as:

[0092]

[0093] Among them, c i These are the weight coefficients to be determined, and the function a(·) is the basis function. m is the number of training samples; in this example, m = 48. The weight coefficient c... i This can be achieved by solving the following system of linear equations:

[0094]

[0095] Where any element a in matrix A ij =a(‖θ) (j) -θ (i) The accuracy and local characteristics of radial basis functions will vary depending on the basis functions chosen. Common basis functions include Gaussian functions, multinomial functions, inverse quadratic functions, inverse polyquadratic functions, and multiple harmonic splines.

[0096] Kriging interpolation and support vector regression methods can be implemented using similar steps; their detailed working principles can be found in relevant literature.

[0097] Step 3.3: Evaluate the performance of the approximate model method set ApproSet = {Appro1, Appro2, Appro3} on the test set. Root Mean Square Error (RMSE), Mean Square Error (MSE), and Average Absolute Error (AEE) are used as accuracy evaluation criteria. Their calculation methods are as follows:

[0098]

[0099]

[0100]

[0101] Each approximation model method yields three accuracy evaluation criteria, which are defined in this invention as follows:

[0102] e RMSE = [χ1,χ2,χ3];

[0103]

[0104] e AEE =[ξ1,ξ2,ξ3];

[0105] In the formula, χ i ,η i ,ξ i Let e ​​represent the root mean square error, multiple correlation coefficient, and mean absolute error of the i-th approximation model method on the test set, respectively. Let the accuracy vector e be... RMSE , e AEE Normalization, in e RMSE For example, normalization:

[0106]

[0107]

[0108]

[0109] The normalized precision vectors are represented as e′ RMSE , 、e′ AEE .

[0110] Furthermore, an average accuracy evaluation index e is introduced. AVE Its definition is:

[0111]

[0112] Vector e AVE The approximate model corresponding to the largest element in the model is the best approximate model, denoted as y = APPROBest(θ).

[0113] Step 4: Utilize the surface displacement monitoring data obtained in Step 2. A =[u A1 ,u A2 ,…,u A5 And based on the best approximation model y = ApproBest(θ) established in step 3, the following objective function is established:

[0114]

[0115] Where, stθ∈[θ L ,θ U ];

[0116] The solution is obtained using the particle swarm optimization algorithm, and the result is θ. best =[97,0.27,19.5]. That is, the inversion results of the soil and rock parameters are: elastic modulus E = 97 MPa, Poisson's ratio μ = 0.27, and unit weight of the soil layer γ = 19.5 kN / m³. 3 Taking the i7-8700 processor as an example, the runtime of a single finite element model is approximately 10 minutes. The computational cost of this invention is mainly reflected in the establishment of 60 sets of input-output training samples, thus taking approximately 10 hours in total. If the original finite element model is used for inversion calculation, the finite element model needs to be called thousands of times in the particle swarm optimization algorithm, resulting in excessively long computation time.

[0117] Step 5: Determine the construction parameters to be optimized. During shield tunneling, the support force and grouting pressure at the excavation face have a significant impact on surface displacement. Therefore, this example mainly optimizes the support force (kPa) and grouting pressure (kPa) at the excavation face, which are represented by variables α1 and α2, respectively. Thus, the construction parameters to be optimized can be represented as α = [α1, α2].

[0118] Step 6: Establish an approximate model y = Appro(α) between observation point B and construction parameters. Its implementation process is similar to Step 3, with the main difference being in the design parameters. Following the details of Step 3, it is first necessary to determine the design variable α = [α1, α2] and its range of values ​​(α...). L ,α U According to existing research, the supporting force α1 and grouting pressure α2 at the excavation face should satisfy the following relationship:

[0119]

[0120]

[0121] Where, γ i h i Let k1 and k2 represent the unit weight and thickness of the i-th soil layer, respectively. k1 and k2 are the earth pressure coefficients. In this example, the range of the design variable α is set to α. L = [100, 300] kPa, α U = [120, 360] kPa. The sampling sample is set to 40 and divided into training and test sets in an 8:2 ratio. The best approximate model y = ApproBest(α) for observation point B and construction parameters is obtained according to the design scheme provided in step 3.

[0122] Step 7: Based on the settlement requirements when the tunnel boring machine (TBM) passes under different buildings, set the allowable displacement values ​​for the buildings at each stage. In this example, the surface settlement caused by the TBM passage is divided into 5 stages, therefore u limB =[u limB1 ,u limB2 ,u limB3 ,u limB4 ,u limB5 Then, the objective function is constructed by combining the approximate model y = ApproBest(α) from step 6:

[0123]

[0124] Where, stα∈[α L ,α U ]

[0125] The solution is obtained using the particle swarm optimization algorithm, and the result is α. best = [135, 240] kPa. That is, the optimized construction parameters are an excavation face support pressure of 135 kPa and a grouting pressure of 240 kPa. Taking an i7-8700 processor as an example, the runtime of a single finite element model is approximately 10 minutes. The computational time cost of this invention is mainly reflected in the establishment of 40 sets of input-output training samples, thus taking approximately 7 hours in total. If the original finite element model is used for inversion calculation, in the particle swarm optimization algorithm, the finite element model needs to be called thousands of times, resulting in excessively long computation time.

[0126] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for inverting soil and rock parameters and optimizing tunneling parameters for shield tunnels based on an approximate model, characterized in that, Includes the following steps: Step S30: Establish geotechnical parameters based on the approximate model method. With surface displacement Correspondence formula: ; Step S40: Based on the acquired surface displacement monitoring data, establish the objective optimization function. The optimal geotechnical parameters were solved using the particle swarm optimization algorithm. ; Step S50: Establish shield tunneling construction parameters based on the approximate model method. With surface displacement Correspondence formula: ; Step S60, according to the maximum allowable displacement value required by the specification. Establish the objective optimization function The objective function is solved using the particle swarm optimization algorithm to obtain suggested values ​​α for the shield tunneling construction parameters. best ; Step S30 specifically includes the following steps: Step S301, Determine design variables and the upper and lower limits of the corresponding variables. ; The input dataset is generated based on the hypercubic Latin sampling method. ; The input dataset can be one 3D matrix representation: ； Step S302, based on the input dataset , establish and run Create a finite element model and extract... Observation nodal displacements of the finite element model and input-output datasets The dataset is divided into a training set (TrainingSet) and a test set (TestSet) according to a preset ratio. Step S303: Based on the training set TrainingSet generated in step A2, establish an approximate model method set. ; Among them, Appro1 is implemented based on the Kriging interpolation method, Appro2 is implemented based on the support vector regression method, and Appro3 is implemented based on the radial basis function method. Step S304: Based on the test set TestSet generated in step S302 and the approximate model method set ApproSet established in step S303, the optimal approximate model is obtained by combining the root mean square error RMSE, the multiple correlation coefficient MSE, and the mean absolute error AEE. ; Step S50 specifically includes the following steps: Step S501: Determine the shield tunneling construction parameters to be optimized. And the corresponding engineering experience value range for construction parameters. ; Step S502: Generate the input dataset based on the hypercubic Latin sampling method. ; Step S503, based on the input dataset , establish and run Create a finite element model and extract... Observation nodal displacements of the finite element model Input-output dataset The dataset is divided into a training set (TrainingSet) and a test set (TestSet) according to a preset ratio. Step S504: Based on the training set generated in step S303, establish an approximate model method set. ; Among them, Appro1 is implemented based on the Kriging interpolation method, Appro2 is implemented based on the support vector regression method, and Appro3 is implemented based on the radial basis function method. Step S505: Based on the test set TestSet generated in step S503 and the approximate model method set ApproSet established in step S504, the optimal approximate model is derived. .

2. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 1, characterized in that, In step S302, the input-output dataset is... The dataset is divided into the TrainingSet and the TestSet in an 8:2 ratio.

3. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 1, characterized in that, Step S40 specifically includes the following steps: Step S401: Obtain the monitored displacement of the ground surface at different construction stages. Construct the objective function: ； ； Step S402: Solve the objective function using the particle swarm optimization algorithm. The inversion results of the geotechnical parameters were obtained. .

4. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 1, characterized in that, Step S60 specifically includes the following steps: Step S601, obtaining the displacement limit values ​​of the ground surface at different construction stages. Construct the objective function: ； in, ; Step S602: Solve the objective function using the particle swarm optimization algorithm. The optimized results of the shield tunneling construction parameters were obtained. .

5. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 1, characterized in that, Before step S30, step S20 is included, which determines the soil and rock parameters to be inverted, wherein the soil and rock parameters to be inverted include the elastic modulus. Poisson's ratio The unit weight of the soil layer At least one of them.

6. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 5, characterized in that, Before step S20, step S10 is also included: analysis of engineering background and site information.

7. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 6, characterized in that, Step S10 specifically includes: obtaining the value range of parameters reflecting the mechanical properties of each stratum and selecting a suitable shield tunneling section for inversion; wherein the suitable shield tunneling section for inversion has the conditions for setting up observation points.

8. The method for inverting geotechnical parameters and optimizing tunneling parameters for shield tunnels according to claim 3, characterized in that, In step S503, the input-output dataset is... The dataset is divided into the TrainingSet and the TestSet in an 8:2 ratio.

Images