Real-time inversion method for surrounding rock mechanical parameters of tunnel boring machine (TBM) construction roadway

Abstract

The invention relates to the technical field of real-time inversion of mechanical parameters of surrounding rocks of construction roadways, in particular to a real-time inversion method of mechanical parameters of surrounding rocks of TBM construction roadways. According to the technical scheme, the method comprises the following steps that a finite element geologic model is built, and the finite element model reflecting geologic features is built according to the actual geologic condition of a TBM construction roadway; initial model physical property parameters are determined, and initial surrounding rock mechanical parameters including elastic modulus and cohesion are set; constraint is added, forward excavation calculation is conducted, and constraint conditions are added to the finite element model. According to the method, the finite element geologic model is established, the error function is reasonably defined, the effective parameter updating mechanism is adopted, the convergence condition is clear, and the scientific and accurate method is provided for TBM construction roadway surrounding rock mechanical parameter inversion. The geological condition can be accurately simulated, the inversion process is optimized, parameters continuously approach true values, the reliability of an inversion result is guaranteed, and powerful support is provided for safety and stability of roadway construction.

Description

Technical Field

[0001] The present invention relates to the technical field of real-time inversion of mechanical parameters of surrounding rock in construction roadways, and particularly to a method for real-time inversion of mechanical parameters of surrounding rock in TBM construction roadways. Background Art

[0002] During the process of constructing a roadway by a TBM (Tunnel Boring Machine), accurately understanding the mechanical parameters of the surrounding rock is crucial for ensuring construction safety, optimizing the construction plan, and improving the project quality. Traditional methods for determining the mechanical parameters of the surrounding rock often rely on empirical estimates or limited on-site tests, which have significant uncertainties and limitations.

[0003] With the development of numerical simulation technology, the finite element method has been widely applied in the engineering field. However, in the inversion of mechanical parameters of surrounding rock in TBM construction roadways, the simple finite element method faces some challenges. For example, numerical instability may occur during the inversion process, resulting in inaccurate parameter estimation; at the same time, the overfitting problem may also cause the inversion results to perform poorly on new data.

[0004] Adjusting parameters by the momentum method is a commonly used optimization method, but in practical applications, it may also affect the inversion effect due to improper parameter adjustment. Therefore, new technical means need to be introduced to solve these problems and improve the accuracy and reliability of the inversion of mechanical parameters of surrounding rock in TBM construction roadways.

[0005] Therefore, this application proposes a method for real-time inversion of mechanical parameters of surrounding rock in TBM construction roadways. Summary of the Invention

[0006] The purpose of the present invention is to propose a method for real-time inversion of mechanical parameters of surrounding rock in TBM construction roadways to solve the problem of poor inversion effect of mechanical parameters of surrounding rock in TBM construction roadways in the background art.

[0007] The technical solution of the present invention: A method for real-time inversion of mechanical parameters of surrounding rock in TBM construction roadways includes the following method steps:

[0008] S1. Establish a finite element geological model, and construct a finite element model reflecting geological characteristics according to the actual geological conditions of the TBM construction roadway;

[0009] S2. Determine the initial model physical property parameters, and set the initial mechanical parameters of the surrounding rock, including elastic modulus and cohesion;

[0010] S3. Add constraints and perform forward excavation calculation. Add constraint conditions to the finite element model, and the constraint conditions include boundary conditions and initial stress conditions. Perform forward excavation calculation to simulate the mechanical response of the surrounding rock of the roadway during the TBM construction process;

[0011] S4. Calculate the error function, define the target value and the error function. Assume that the known target vector y is of dimension n×1, and it is necessary to infer the unknown target vector y of dimension m×1 based on the target vector y. Establish a forward model h that maps s to the approximate target value h(s), that is, in the forward model h, when the unknown parameter s is input, the output target value h(s) is an approximation of the target vector y; define the difference μ = y - h(s), where μ represents the difference between the calculated value of the forward model and the target value, and the difference includes model error, measurement error, and theoretical error; select the mean square error function as the error function.

[0012] That is

[0013] where n is the number of target values. During the iteration process, the smaller the error function error, the closer the inferred parameter is to the true value.

[0014] S5. Adjust the parameters using the momentum method, which specifically includes the following steps:

[0015] S501. Calculate the gradient matrix: During the process of updating from s k to s k+1 , the calculation method of the gradient matrix for the k-th iteration is as follows. Run m groups of three-dimensional finite element forward models. Among them, for the i-th (1 ≤ i ≤ m) group of forward models, the output parameter is based on the unknown vector in the (k - 1)-th iteration, and a small variable is added to the i-th term of the vector while the other parameters are the same, forming the input vector of the i-th group of forward models. Run m groups of forward models to obtain a vector composed of m error functions. According to the definition of partial derivatives, the gradient of the error function is:

[0016] S502. Calculate the velocity matrix: According to the momentum method, introduce a velocity matrix v of dimension w×1, with an initial value v 0 = 0. Define the velocity matrix for the k-th iteration as v k-1 . From the (k - 1)-th iteration to the k-th iteration, the update process of the velocity matrix v is as follows: v k = α * v k-1 - λ * G k-1 , where α is the attenuation coefficient in the momentum method, generally taking a value between 0.5 and 0.9, and λ is the step size parameter, whose value is determined according to the order of magnitude of the partial derivative of the error function and the inversion parameter.

[0017] S503. Update the parameter matrix: After calculating the velocity matrix v k , update the unknown vector. After the update, obtain the unknown vector s k for the k-th iteration;

[0018] S6. Determine the convergence condition. Manually set the upper error limit ε. The condition for the inversion to converge is that the error function calculated based on the current forward modeling input parameters is less than the set upper error limit, that is, error < ε. It is considered that within a certain error range, the input parameters are consistent with the true parameters. At this time, the input parameters are an approximation of the true parameters and are defined as the optimal parameters under the given conditions.

[0019] S7. Output the inversion result. After the convergence condition is met, output the final inversion result of the mechanical parameters of the surrounding rock of the TBM construction roadway.

[0020] Optionally, when constructing the finite element model, the actual geological structure and rock properties of the TBM construction roadway are considered, and the physical and mechanical properties of the geological body are automatically identified and extracted, including but not limited to rock type, joint distribution, and stress state.

[0021] Optionally, the initial mechanical parameters of the surrounding rock are determined based on past similar engineering experience and on-site preliminary exploration results, and are determined through statistical analysis or machine learning methods.

[0022] Optionally, in the forward excavation calculation, by considering the actual excavation steps during the TBM construction process, including cutting, crushing, and loading and unloading, the mechanical response of the surrounding rock of the roadway is simulated.

[0023] Optionally, in the mean square error function of the error function, the target value is provided by on-site measured data or historical construction data.

[0024] Optionally, in the parameter adjustment of the momentum method, an adaptive adjustment strategy is adopted, and the attenuation coefficient α and the step size parameter λ are dynamically adjusted according to the error change during the iteration process.

[0025] Optionally, the setting of the upper error limit ε comprehensively considers the engineering accuracy requirements and the feasibility of the inversion calculation.

[0026] Compared with the prior art, the present application includes at least one of the following beneficial technical effects:

[0027] By establishing a finite element geological model, the actual geological situation can be more accurately simulated, providing a reliable basis for subsequent inversion. The clear inversion analysis process includes steps such as setting the physical property parameters of the initial model and forward excavation calculation, making the entire inversion process proceed in an orderly manner, and improving the reliability and accuracy of the inversion.

[0028] By defining the target value and the error function, the difference between the calculated value of the forward model and the target value is clearly expressed, which helps to accurately evaluate the deviation between the inversion result and the actual situation. Selecting the mean square error function as the error function, during the iteration process, the smaller the error function means that the inversion parameters are closer to the true values, providing a clear direction for optimization.

[0029] The iterative update of parameters is achieved through a conversion layer. By inputting the inversion parameters of the previous iteration and outputting the updated inversion parameters of the current iteration, it ensures that the parameters are continuously adjusted towards the optimal direction. When calculating the gradient matrix, multiple groups of three-dimensional finite element forward models are run, and the gradient matrix is calculated according to the definition of partial derivatives, improving the accuracy of parameter update. A velocity matrix is introduced, and the momentum method is used to update the velocity matrix. Then, the unknown vector is updated according to the velocity matrix, making the parameter update more stable and efficient.

[0030] By setting an artificial error upper limit, when the error function calculated based on the current forward input parameters is less than the set error upper limit, it is considered that the inversion process converges, and the input parameters are approximations of the true parameters, providing a guarantee for the reliability of the inversion results.

[0031] Through establishing a finite element geological model, reasonably defining an error function, adopting an effective parameter update mechanism, and clarifying the convergence conditions, the present invention provides a scientific and accurate method for the inversion of mechanical parameters of surrounding rocks in TBM construction roadways. It can accurately simulate the geological conditions, optimize the inversion process, make the parameters continuously approach the true values, ensure the reliability of the inversion results, and provide strong support for the safety and stability of roadway construction. Brief Description of the Drawings

[0032] Figure 1 It is the flowchart of the inversion analysis in the embodiment;

[0033] Figure 2 It is the diagram of the inversion iteration process in the embodiment. Specific Embodiments

[0034] The following further describes the technical solutions of the present invention in conjunction with the drawings and specific embodiments.

[0035] Embodiment: This embodiment provides a real-time inversion method for mechanical parameters of surrounding rocks in TBM construction roadways. The inversion process of mechanical parameters of surrounding rocks includes: establishing a finite element geological model, initial model physical property parameters, adding constraints and performing forward excavation calculations, calculating the error function, adjusting parameters by the momentum method until the convergence conditions are met, and outputting the inversion results. The flowchart of the inversion analysis is as Figure 1 shown.

[0036] Inversion Principle:

[0037] (1) Define the target value and the error function

[0038] Assume that the known n×1-dimensional target vector y is given, and it is necessary to infer the m×1-dimensional unknown target vector y based on y. A forward model h that maps from s to y needs to be established. That is, in the forward model h, when the unknown parameter s is input, the output target value h(s) is an approximation of y. Define the difference between h(s) and y as the deviation μ, as shown in formula (1).

[0039] y = g(s) + μ (1)

[0040] In the above formula represents the parameter vector to be inverted, and m is the number of inversion parameters. In geotechnical engineering, they are generally the physical properties of material S such as elastic modulus and cohesion. The target value y=(y 1 , y 2 ,…y n )(n is the number of target values) is the displacement of the observation point in geotechnical engineering. μ represents the difference between the calculated value of the forward model and the target value, usually including model error, measurement error, and theoretical error.

[0041] Define the error function. In the optimization process, it is required that the error function is continuously differentiable. Therefore, the mean square error function is selected as the error function, as shown in formula (2). In the iterative process, the smaller the error function error, the closer the inverted parameters are to the true values.

[0042]

[0043] (2) Parameter update process

[0044] Define: is the initial value of the inversion parameter vector before iteration, is the inversion parameter vector at the k-th iteration step. The inversion parameters start from and after a certain number of iterations, they are updated to the optimal inversion parameters In each iteration, the parameter update is achieved by using a transformation layer: in a transformation layer, the input is the inversion parameter of the previous iteration, and the output is the updated inversion parameter of this iteration. In the transformation layer, operations such as calculating the gradient matrix, calculating the momentum matrix, and updating the inversion parameters are performed. Through the alternating connection of the inversion parameters and the transformation layer, the continuous iterative update of the parameters is realized. The inversion iteration process is as shown in Figure 2 .

[0045] 1) Calculate the gradient matrix

[0046] In the process of updating from to , the calculation method of the gradient matrix for the k-th iteration is as follows: Run m groups of three-dimensional finite element forward models respectively. Among them, the output parameter of the i-th (1≤i≤m) group of forward models is That is, on the basis of the unknown vector in the (k - 1)-th iteration, the i-th item of the vector is increased by a small variable and the remaining parameters are the same, forming the input vector of the i-th group of forward models. Run m groups of forward models to obtain a vector composed of m error functions. According to the definition of partial derivatives, the gradient of the error function is:

[0047]

[0048] where error k-1 is the error function of the (k - 1)-th iteration, is the error function calculated from the i-th (1 ≤ i ≤ m) group of forward models during the (k - 1)-th iteration run.

[0049] 2) Calculate the velocity matrix

[0050] According to the momentum method, introduce an m×1 dimensional velocity matrix V, with the initial value V 0 = [0.,..0,..0]'. Define the velocity matrix of the k-th iteration as V k . From the (k - 1)-th iteration to the k-th iteration, the update process of the velocity matrix V is as follows:

[0051] V k = α * V k-1 - λ * G k-1 (4)

[0052] In the above formula, α is the attenuation coefficient in the momentum method, generally taking values between 0.5 - 0.9. λ is the step size parameter, and its value is determined according to the order of magnitude of the partial derivative of the error function and the inversion parameters.

[0053] 3) Update the parameter matrix

[0054] After calculating the velocity matrix V k , update the unknown vector.

[0055]

[0056] After the update, obtain the unknown vector of the k-th iteration Complete the parameter update from the (k - 1)-th step to the k-th step, and at the same time input into the forward model to obtain the error error k of the k-th iteration step, and make the following judgment:

[0057] error < ε (6)

[0058] ε is the upper error limit set artificially. The condition for inversion convergence is that the error function calculated according to the current forward input parameters is less than the set upper error limit. It is considered that within a certain error range, the input parameters are consistent with the true parameters. At this time, the input parameters are an approximation of the true parameters, defined as the optimal parameters under the given conditions. Further, in the squared error function of the error function, the target value is provided by on-site measured data or historical construction data. The setting of the upper error limit ε comprehensively considers the engineering accuracy requirements and the feasibility of the inversion calculation.

[0059] Repeat the above formulas (3) to (6) until the inversion process converges.

[0060] It should be noted that when constructing the finite element model, the geological structure and rock properties of the TBM construction roadway are considered in actual situations, and the physical and mechanical properties of the geological body are automatically identified and extracted, including but not limited to rock types, joint distributions, and stress states.

[0061] Among them, the initial mechanical parameters of the surrounding rock are determined according to past similar engineering experiences and the results of preliminary on-site surveys, and are determined through statistical analysis or machine learning methods.

[0062] In this embodiment, in the forward excavation calculation, by considering the actual excavation steps during the TBM construction process, including cutting, crushing, and loading and unloading, the mechanical response of the roadway surrounding rock is simulated.

[0063] In this embodiment, in the momentum method parameter adjustment, an adaptive adjustment strategy is adopted, and the attenuation coefficient α and the step size parameter λ are dynamically adjusted according to the error changes during the iteration process.

[0064] In this application, by establishing a finite element geological model, the actual geological situation can be more accurately simulated, providing a reliable basis for subsequent inversion. The clear inversion analysis process includes steps such as setting the physical property parameters of the initial model and forward excavation calculation, making the entire inversion process orderly, and improving the reliability and accuracy of the inversion. By defining the target value and the error function, the difference between the calculated value of the forward model and the target value is clearly expressed, which helps to accurately evaluate the deviation between the inversion result and the actual situation. The square difference function is selected as the error function. During the iteration process, the smaller the error function, the closer the inversion parameter is to the true value, providing a clear direction for optimization.

[0065] It should be noted that the iterative update of the parameters is achieved through the conversion layer. The inversion parameters of the previous iteration are input, and the updated inversion parameters of this iteration are output, ensuring that the parameters are continuously adjusted in the optimal direction. When calculating the gradient matrix, multiple groups of three-dimensional finite element forward models are run, and the gradient matrix is calculated according to the definition of partial derivatives, improving the accuracy of parameter update. The velocity matrix is introduced, and the momentum method is used to update the velocity matrix, and then the unknown vector is updated according to the velocity matrix, making the parameter update more stable and efficient. By setting an artificial error upper limit, when the error function calculated according to the current forward input parameters is less than the set error upper limit, the inversion process is considered to converge, and the input parameters are approximations of the true parameters, providing a guarantee for the reliability of the inversion result.

[0066] The above specific embodiments are merely several alternative embodiments of the present invention. Based on the technical solution of the present invention and the relevant inspirations of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel, characterized in that: The method comprises the following steps: S1. Establish a finite element geological model. According to the actual geological conditions of the TBM construction tunnel, a finite element model reflecting the geological characteristics is constructed; S2. Determine the initial model physical parameters and set the initial surrounding rock mechanical parameters, including elastic modulus and cohesion; S3, adding constraints and performing forward excavation calculation, adding constraints to the finite element model, the constraints include boundary conditions and initial stress conditions, performing forward excavation calculation, and simulating the mechanical response of the tunnel surrounding rock during TBM construction; S4. Calculate the error function, define the target value and the error function. Assuming that the target vector y of n×1 dimensions is known, it is necessary to infer the unknown target vector y of m×1 dimensions based on the target vector y. Establish a forward model h that maps from s to the approximate target value h(s). That is, in the forward model h, input the unknown parameter s, and output the target value h(s) as the approximate value of the target vector y. Define the difference μ=yh(s), where μ represents the difference between the forward model calculated value and the target value. The difference includes model error, measurement error, and theoretical error. Select the square error function as the error function. Right now Among them, n is the number of target values. During the iteration process, the smaller the error function error is, the closer the inverted parameter is to the true value; S5, momentum method parameter adjustment, specifically including the following steps: S501, find the gradient matrix: k Update to s k+1 In the process, the calculation method of the gradient matrix of the kth iteration is as follows: m groups of three-dimensional finite element forward models are run respectively, where the output parameter of the i-th (1≤i≤m) group of forward models is based on the unknown vector of the k-1th iteration, and the i-th term of the vector adds a small variable The remaining parameters are the same, forming the input vector of the i-th group of positive models. Running m groups of positive models, we get a vector of m error functions. According to the definition of partial derivatives, the gradient of the error function is: S502, calculate the velocity matrix: according to the momentum method, introduce a w×1-dimensional velocity matrix v, with an initial value v0=0, and define the velocity matrix of the kth iteration as v k-1 , from k-1 iterations to k iterations, the update process of the velocity matrix v is as follows: k =α*v k-1 -λ*G k-1 , where α is the attenuation coefficient in the momentum method, which is generally 0.5-0.9, and λ is the step size parameter, which is determined by the order of magnitude of the partial derivative of the error function and the inversion parameter; S503, update parameter matrix: calculate and obtain velocity matrix v k After that, the unknown vector is updated. After the update, the unknown vector s of the kth iteration is obtained k ; S6. Determine the convergence condition and artificially set the error upper limit ε. The condition for inversion convergence is that the error function calculated based on the current forward input parameters is less than the set error upper limit, that is, error<ε. It is considered that within a certain error range, the input parameters are consistent with the true parameters. At this time, the input parameters are approximate to the true parameters and are defined as the optimal parameters under given conditions. S7. Output the inversion results. When the convergence conditions are met, output the final inversion results of the mechanical parameters of the surrounding rock of the TBM construction tunnel.

2. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel according to claim 1, characterized in that: The finite element model is constructed by taking into account the actual geological structure and rock characteristics of the TBM construction tunnel, and automatically identifying and extracting the physical and mechanical properties of the geological body, including but not limited to rock type, joint distribution, and stress state.

3. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel according to claim 1, characterized in that: The initial surrounding rock mechanical parameters are determined based on previous similar engineering experience and preliminary on-site survey results, and are determined through statistical analysis or machine learning methods.

4. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel according to claim 1, characterized in that: In the forward excavation calculation, the mechanical response of the tunnel surrounding rock is simulated by considering the actual excavation steps in the TBM construction process, including cutting, crushing, and loading.

5. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel according to claim 1, characterized in that: In the square difference function of the error function, the target value is provided by field measured data or historical construction data.

6. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel according to claim 1, characterized in that: In the momentum method parameter adjustment, an adaptive adjustment strategy is adopted to dynamically adjust the attenuation coefficient α and the step size parameter λ according to the error changes in the iterative process.

7. A real-time inversion method for mechanical parameters of surrounding rock in a TBM construction tunnel according to claim 1, characterized in that: The setting of the upper limit of the error ε comprehensively considers the engineering accuracy requirements and the feasibility of the inversion calculation.

Images