An optimization method for TBM tunnel support parameters based on mechanical response prediction

Abstract

The present invention discloses an optimization method for TBM tunnel support parameters based on mechanical response prediction, which relates to the technical field of tunnel construction. The key points of its technical solution are as follows: This method generates a large number of mechanical response samples by establishing a finite element calculation model under different surrounding rock states and support parameter combinations, and at the same time collects on-site measured data for supplementation and correction. A surrounding rock-support mapping relationship model is established using a deep learning algorithm. According to the relationship between specific surrounding rock conditions and support responses, the optimal support parameter combination that meets the safety requirements is intelligently selected, thereby realizing the precise optimization of support design.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel construction, and more specifically, to an optimization method for TBM tunnel support parameters based on mechanical response prediction. Background Art

[0002] In the western region of China, tunnel construction faces complex engineering geological environments and multiple adverse factors, such as high ground temperature, high ground stress, high altitude, and high seismic intensity. These special conditions pose severe challenges to the design and construction of deep-buried tunnels. The most prominent problem is the complexity of the in-situ stress state and distribution, which makes it extremely easy for problems such as support failure and support waste to occur during tunnel construction, affecting the construction safety and economy of the tunnel. Therefore, how to optimize the support parameters to adapt to different surrounding rock conditions and ensure the safety and economy of tunnel construction has become a key problem to be solved urgently.

[0003] Currently, the support design of deep-buried tunnels mainly relies on means such as empirical methods, analogy methods, and numerical simulation methods. Although empirical methods and analogy methods can provide references to a certain extent, due to their dependence on the empirical data of previous projects, they cannot accurately reflect the complex and changeable surrounding rock conditions, easily leading to conservative or overly risky support designs. Numerical simulation methods can predict the mechanical responses during tunnel construction, but traditional numerical simulations often target single working conditions or simple surrounding rock conditions, lacking comprehensive research on the optimization of support parameters under various complex geological conditions. In addition, these methods usually take a long time, require a large amount of calculation and parameter adjustment processes, and are difficult to quickly adjust the support plan during construction, resulting in insufficient response speed and flexibility in on-site applications.

[0004] The limitations of traditional support design are mainly reflected in the following aspects: First, the support design fails to fully consider the complexity and diversity of different surrounding rocks, resulting in insufficient universality and reliability of the support plan; second, it is impossible to achieve rapid parameter optimization and adjustment, and it is difficult to adapt to the complex geological condition changes in actual construction; third, the existing methods often have a certain degree of conservative design, leading to waste of support materials and an increase in construction costs, and failing to achieve a balance between economy and safety. Summary of the Invention

[0005] The purpose of the present invention is to provide an optimization method for TBM tunnel support parameters based on mechanical response prediction. This method establishes a finite element calculation model under different combinations of surrounding rock states and support parameters, generates a large number of mechanical response samples, and at the same time collects on-site measured data for supplementation and correction. Using deep learning algorithms, a surrounding rock-support mapping relationship model is established. According to the relationship between specific surrounding rock conditions and support responses, the optimal support parameter combination that meets the safety requirements is intelligently selected, thereby achieving precise optimization of the support design.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions: An optimization method for TBM tunnel support parameters based on mechanical response prediction, the method comprising the following steps:

[0007] 1) Determine the mechanical characteristics of the surrounding rock: The mechanical characteristics of the surrounding rock include uniaxial compressive strength, Young's modulus, unit weight, cohesion, internal friction angle, Poisson's ratio, joint spacing, and burial depth;

[0008] 2) Form a support parameter library: Combine the steel arch frame models and spacings, wire mesh diameters and spacings, steel bar row diameters and spacings, shotcrete models and thicknesses, bolt models, lengths and spacings, lining thicknesses, and other various support parameters used at the construction site to form a support parameter library;

[0009] 3) Confirm the mechanical parameter values characterizing the support behavior of the TBM tunnel: Combine the mechanical characteristics of the surrounding rock in step 1) with the support parameter library in step 2), conduct numerical simulation tests, calculate based on the discrete element model, and select the mechanical parameter values of the steel arch frame axial force, steel arch frame bending moment, steel arch frame displacement, bolt axial force, lining bending moment, lining displacement, and other multiple parameters characterizing the support behavior of the TBM tunnel at 5 measuring points;

[0010] 4) Widely collect on-site actual mechanical characteristics of the surrounding rock and support parameter samples, and jointly form a big data sample library of the surrounding rock-support mechanical response with the mechanical parameter values of the support behavior monitored in step 3);

[0011] 5) Construct a surrounding rock-support mapping model framework, and normalize the mechanical characteristics of the surrounding rock and its mechanical response by formula (1):

[0012]

[0013] In the formula, max(x) and min(x) are the maximum and minimum values of the protective gear, and x′ is the value after normalization; at the same time, perform discrete data encoding operations on the support parameters;

[0014] Perform 1D causal convolution operation through formula (2):

[0015]

[0016] In the formula, X = [x1, x2,..., x T is the input sequence, Y = [y1, y2,..., y T is the output sequence, W = [w1, w2,..., w K is the convolution kernel weight, and d is the dilation factor;

[0017] Perform Dropout operation through formula (3):

[0018] Y dropout = Y cov ⊙ d(3)

[0019] In the formula, Y cov is the output result of causal convolution, d is a random mask matrix of the same size as Y cov , and the pruning operation is implemented through multiplication;

[0020] The non-linear transformation is realized by using the ReLU activation function through Equation (4):

[0021] Y ReLU = max(0, W·Y dropout + b)(4)

[0022] In the formula, W is the weight and b is the bias;

[0023] The data flow of the GRU layer is controlled by three gating mechanisms (reset gate, update gate, candidate activation). The calculation formula of the update gate is:

[0024] z t = σ(W z x t + U z y t-1 + b z )(5)

[0025] In the formula, W z and U z are weight matrices, and b z is the bias term;

[0026] The calculation formula of the reset gate is:

[0027] r t = σ(W r x t + U r y t-1 + b r )(6)

[0028] The calculation formula of the candidate activation is:

[0029]

[0030] The final output is:

[0031]

[0032] 6) Combine the surrounding rock - support mechanical response big data sample library in step 4) with the surrounding rock - support mapping model framework formed in step 5) for model training to form a trained surrounding rock - support mapping model;

[0033] 7) For each surrounding rock condition encountered in actual construction, different combinations of support parameters and their corresponding mechanical responses are matched, and the safety score, efficiency score, and economic score are calculated respectively. Different weights are selected for the three scores for different situations, and the WOA optimization algorithm is used to achieve the optimization process, and finally the parameter optimization and adjustment are formed.

[0034] The present invention is further configured such that: the WOA optimization is divided into three stages:

[0035] The motion trajectory equation in the initial stage is:

[0036]

[0037] In the formula: k is the number of iterations; Q k and respectively represent the vectors between the starting point and the search agent position and the best position; D is the vector between the current search point and the current best individual; A and C are coefficient vectors that control the movement mode of the search point; it is set that ρ1 and ρ2 are random numbers between 0 and 1; a is a convergence factor that linearly decreases from 2 to 0;

[0038] The motion trajectory equation in the local search stage is:

[0039]

[0040] In the formula: b is the logarithmic spiral constant, and l is a random number between [-1, 1];

[0041] The motion trajectory equation in the global search stage is:

[0042]

[0043] In the formula: represents a randomly selected position vector.

[0044] In summary, the present invention has the following beneficial effects:

[0045] 1. Strong pertinence: By analyzing the mechanical responses under different surrounding rock states, the present invention can customize the support scheme for specific geological conditions, greatly improving the rationality and pertinence of the support design, and avoiding the situation of support failure or support waste in traditional designs.

[0046] 2. High efficiency: The process of generating a large number of samples by numerical calculation can greatly shorten the time of support design. At the same time, through the pre-constructed response database, the optimal combination of support parameters can be quickly obtained, realizing the rapid iteration and optimization of support design, and greatly improving the response speed at the construction site.

[0047] 3. Good economy: By optimizing the combination of support parameters, the present invention can significantly reduce the usage amount of support materials, avoid over-design of support at the same time, reduce construction costs, and achieve the best balance between the economy and safety of support design.

[0048] 4. Intelligence and flexibility: The core of the present invention lies in the dynamic correlation between support parameters and surrounding rock conditions. The support plan can be adjusted in real time according to on-site monitoring data to ensure the safety and stability of tunnel construction. In addition, this method also has strong scalability and can be applied to different types of tunnel projects and various complex geological environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the measuring points of the support structure in Embodiment 1 of the present invention;

[0050] Figure 2 It is the surrounding rock-support mapping model framework in Embodiment 1 of the present invention;

[0051] Figure 3 It is the support parameter optimization framework in Embodiments 1 and 2 of the present invention;

[0052] Figure 4 It is the prediction effect of the mapping relationship model in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0053] The following further elaborates on the present invention Figures 1-4 in conjunction with the attached drawings.

[0054] Embodiment 1: A method for optimizing the support parameters of a TBM tunnel based on mechanical response prediction, as Figures 1-3 shown, includes the following steps:

[0055] 1) Determine multiple surrounding rock mechanical characteristics including uniaxial compressive strength, Young's modulus, unit weight, cohesion, internal friction angle, Poisson's ratio, joint spacing, buried depth, etc.

[0056] 2) Combine various support parameters such as the type and spacing of steel arch frames, the diameter and spacing of steel mesh, the diameter and spacing of steel bars, the type and thickness of shotcrete, the type, length and spacing of bolts, and the thickness of lining used at the construction site to form a support parameter library.

[0057] 3) Combine the surrounding rock mechanical characteristics in step 1) with the support parameter library in step 2), conduct numerical simulation tests, and calculate based on the discrete element model. Select Figure 1 the mechanical parameter values of multiple indicators of the TBM tunnel support behavior, such as the axial force of the steel arch frame, the bending moment of the steel arch frame, the displacement of the steel arch frame, the axial force of the bolt, the bending moment of the lining, and the displacement of the lining, shown in

[0058] 4) Widely collect the mechanical characteristics of the surrounding rock at the site, the support parameter samples, and the corresponding mechanical response parameter values monitored in step 3), and jointly form a big data sample library of the mechanical response of the surrounding rock - support with the mechanical response samples in step 3).

[0059] 5) Construct the framework of the surrounding rock - support mapping model as shown in Figure 2 , and normalize the mechanical characteristics of the surrounding rock and its mechanical response by Equation (1):

[0060]

[0061] In the formula, max(x) and min(x) are the maximum and minimum values of the protective gear, and x′ is the value after normalization. At the same time, perform discretized data encoding operations on the support parameters.

[0062] Perform 1 - D causal convolution operation through Equation (2):

[0063]

[0064] In the formula, X = [x1, x2,..., x T is the input sequence, Y = [y1, y2,..., y T is the output sequence, W = [w1, w2,..., w K is the convolution kernel weight, and d is the dilation factor.

[0065] Perform Dropout operation through Equation (3):

[0066] Y dropout = Y cov ⊙d (3)

[0067] In the formula, Y cov is the output result of the causal convolution, and d is a random mask matrix of the same size as Y cov , and the pruning operation is achieved through the product.

[0068] Perform non - linear transformation through Equation (4) using the ReLU activation function:

[0069] Y ReLU = max(0, W·Y dropout + b) (4)

[0070] In the formula, W is the weight and b is the bias.

[0071] Control the data flow of the GRU layer through three gating mechanisms (reset gate, update gate, candidate activation). The calculation formula of the update gate is:

[0072] z t = σ(W z x t+U z y t-1 +b z ) (5)

[0073] Wherein, W z and U z are weight matrices, and b z is the bias term.

[0074] The calculation formula for the reset gate is:

[0075] r t = σ(W r x t +U r y t-1 +b r ) (6)

[0076] The calculation formula for the candidate activation is:

[0077]

[0078] The final output is:

[0079]

[0080] 6) Combine the surrounding rock - support mechanical response big data sample library in step 4) with the surrounding rock - support mapping model framework formed in step 5) to conduct model training and form a trained surrounding rock - support mapping model.

[0081] 7) For each surrounding rock situation encountered in actual construction, with different combinations of support parameters and their corresponding mechanical responses, calculate the safety score, efficiency score, and economic score respectively. Select different weights for the three scores for different situations and use the WOA optimization algorithm to achieve the optimization process. Finally, form parameter optimization and adjustment. The overall implementation framework is as Figure 3 shown.

[0082] The WOA optimization is divided into three stages:

[0083] The movement trajectory equation in the initial stage is:

[0084]

[0085] Where: k is the number of iterations; Q k and respectively represent the vectors between the starting point and the search agent position and the best position; D is the vector between the current search point and the current best individual; A and C are coefficient vectors that control the movement mode of the search point; ρ1 and ρ2 are set as random numbers between 0 and 1; a is the convergence factor that linearly decreases from 2 to 0.

[0086] The trajectory equation in the local search stage is as follows:

[0087]

[0088] Where: b is the logarithmic spiral constant, and l is a random number between [-1, 1].

[0089] The trajectory equation in the global search stage is as follows:

[0090]

[0091] Where: represents the randomly selected position vector.

[0092] Example 2: Take a certain railway tunnel as an example to demonstrate the implementation path of the present invention

[0093] (1) Select Young's modulus, unit weight, cohesion, internal friction angle, and Poisson's ratio as the indexes characterizing the mechanical properties of surrounding rock, which are represented by 8 cases shown in Table 1.

[0094] Table 1 Schematic diagram of mechanical indexes of surrounding rock

[0095]

[0096]

[0097] (2) Select different buried depth conditions and on-site support parameters shown in Table 2, and carry out numerical calculations in combination with 8 surrounding rock conditions shown in Table 1. A total of 8 * 8 * 9 * 8 * 6 * 4 * 4 * 9 = 3,981,312 working conditions are calculated, and according to Figure 1 the 5 monitoring points shown, record the mechanical response values of the support structure such as the axial force of the steel arch, the bending moment of the steel arch, the displacement of the steel arch, the axial force of the bolt, the bending moment of the lining, and the displacement of the lining.

[0098] Table 2 Support parameter combinations

[0099]

[0100] (3) Based on the big data sample library of the mechanical response of the surrounding rock - support formed in step (2), carry out the training of the surrounding rock - support mapping model according to the Figure 2 surrounding rock - support mapping model framework shown, and establish its mapping relationship model, Figure 4 which is the prediction effect diagram of the corresponding mapping relationship model.

[0101] (4) Based on Figure 3The optimized support parameter framework shown above sets the safety score weight, efficiency score weight, and economic score weight to 0.5, 0.3, and 0.2 respectively, and sets up the optimization framework to finally form the TBM tunnel support parameter optimization method. When the buried depth is 1000m, the deformation model is 4350MPa, the unit weight is 25kN / m 3 , and the recommended support parameters for the surrounding rock with a cohesion of 1.809MPa, an internal friction angle of 28.8°, and a Poisson's ratio of 0.26 are a shotcrete thickness of 10cm, a bolt length of 3.5m, a bolt spacing of 1.2*1m, a bolt range of a semi-circle of 180°, a steel mesh in the arch part of 90°φ8, 20*20, and a steel arch frame of HW100 with a spacing of 1.8m.

[0102] This specific embodiment is only an interpretation of the present invention and does not limit the present invention. After reading this specification, those skilled in the art can make modifications to this embodiment without creative contributions as needed, but as long as it is within the scope of the claims of the present invention, it is protected by the patent law.

Claims

1. An optimization method for TBM tunnel support parameters based on mechanical response prediction, characterized in that: The method includes the following steps: 1) Determine the mechanical characteristics of the surrounding rock: The mechanical characteristics of the surrounding rock include uniaxial compressive strength, Young's modulus, unit weight, cohesion, internal friction angle, Poisson's ratio, joint spacing, and buried depth; 2) Form a support parameter library: Combine the steel arch type and spacing, steel mesh diameter and spacing, steel bar row diameter and spacing, shotcrete type and thickness, bolt type, length and spacing, lining thickness, and other various support parameters used at the construction site to form a support parameter library; 3) Confirm the mechanical parameter values characterizing the support behavior of the TBM tunnel: Combine the mechanical characteristics of the surrounding rock in step 1) with the support parameter library in step 2), conduct numerical simulation tests, calculate based on the discrete element model, and select the axial force of the steel arch, bending moment of the steel arch, displacement of the steel arch, axial force of the bolt, bending moment of the lining, displacement of the lining, and other mechanical parameter values characterizing the support behavior of the TBM tunnel at 5 measuring points; 4) Extensively collect the actual mechanical characteristics of the surrounding rock and support parameter samples on site, and jointly form a big data sample library of the mechanical response of the surrounding rock - support with the mechanical parameter values of the support behavior monitored in step 3); 5) Construct a surrounding rock - support mapping model framework, and normalize the mechanical characteristics of the surrounding rock and its mechanical response with Equation (1): In the formula, max(x) and min(x) are the maximum and minimum values of the protective gear, and x′ is the value after normalization; at the same time, perform a discretized data encoding operation on the support parameters; Perform a one - dimensional causal convolution operation through Equation (2): where X = [x1, x2, …, x T is the input sequence, Y = [y1, y2, …, y T is the output sequence, W = [w1, w2, ..., w K is the convolutional kernel weight, and d is the dilation factor; Perform a Dropout operation through Equation (3): Y dropout = Y cov ⊙d(3) where Y cov is the output result of the causal convolution, and d is a random mask matrix of the same size as Y cov The pruning operation is implemented by multiplication; Perform a non - linear transformation using the ReLU activation function through Equation (4): Y ReLU = max(0, W·Y dropout + b) (4) In the formula, W is the weight and b is the bias; Control the data flow of the GRU layer through three gating mechanisms (reset gate, update gate, candidate activation). The calculation formula for the update gate is: z t = σ(W z x t + U z y t-1 + b z ) (5) where, W z and U z are weight matrices, and b z is the bias term; The calculation formula for the reset gate is: r t = σ(W r x t + U r y t-1 + b r ) (6) The calculation formula for the candidate activation is: The final output is: 6) Combine the big data sample library of the mechanical response of the surrounding rock - support in step 4) with the surrounding rock - support mapping model framework formed in step 5), conduct model training, and form a trained surrounding rock - support mapping model; 7) For each surrounding rock condition encountered in actual construction, match different support parameter combinations and their corresponding mechanical responses, calculate the safety score, efficiency score, and economic score respectively. For different situations, select different weights for the three scores, and use the WOA optimization algorithm to achieve the optimization process, and finally form parameter optimization and adjustment.

2. The optimization method for TBM tunnel support parameters based on mechanical response prediction according to claim 1, wherein: The WOA optimization is divided into three stages: The motion trajectory equation in the initial stage is: where: k is the number of iterations; Q k and respectively represent the vectors between the starting point and the search agent position and the best position; D is the vector between the current search point and the current best individual; A and C are coefficient vectors that control the movement mode of the search point; it is assumed that ρ1 and ρ2 are random numbers between 0 and 1; a is a convergence factor that linearly decreases from 2 to 0; The motion trajectory equation in the local search stage is: In the formula: b is the logarithmic spiral constant, and l is a random number between [-1, 1]; The motion trajectory equation in the global search stage is: Wherein: represents a randomly selected position vector.

Images