Lining structure proxy analysis method for road tunnel cross section

Abstract

The invention provides a lining structure proxy analysis method for a road tunnel cross section. The lining structure proxy analysis method comprises the following steps: step 1, constructing a plurality of groups of lining structure models for the cross section of a tunnel through a parametric modeling method; 2, calling a structure analysis kernel to sequentially calculate the constructed lining structure models; 3, constructing a multi-input neural network model; 4, training the multi-input neural network model through the input parameters and calculation results of each group of lining structure models; and step 5, storing the multi-input neural network model after training is finished. According to the method, multiple groups of lining structure models can be quickly generated, the dynamic modification of the later size and lining form is supported, the tedious process of traditional repeated modeling is avoided, and the subsequent calculation cost is remarkably reduced.

Description

Technical Field

[0001] This invention relates to the field of tunnel lining structure analysis technology, and in particular to a proxy analysis method for the lining structure of a highway tunnel cross section. Background Technology

[0002] Highway tunnels play a vital role in modern transportation. Their lining structures, as the main load-bearing and protective components, directly impact the tunnel's operational safety and service life. Analyzing highway tunnel lining structures provides a deep understanding of their stress conditions and deformation characteristics, offering a scientific basis for design optimization, construction quality control, and operation and maintenance. Therefore, conducting analysis of highway tunnel lining structures has significant engineering and social value.

[0003] In the past, when conducting cross-sectional lining structure analysis of highway tunnels, the engineering analogy method was often used. This method involves preliminary design based on similar engineering experience and determining support parameters through historical case analogies. It is fast and efficient and suitable for situations where geological conditions are clear and there are similar engineering references. However, it relies on existing engineering data and has poor adaptability to special geological or complex conditions.

[0004] With the development of computer technology, the conventional finite element analysis technique is widely used in current technologies. This technique analyzes the stress, deformation, and failure modes of linings based on numerical simulation, supporting complex boundary conditions and nonlinear material properties. However, it suffers from complex operation and cumbersome procedures. Conventional techniques typically require manual completion of multiple steps, including geometric modeling, mesh generation, attribute assignment, and boundary condition setting. This not only consumes significant time and human resources but also demands high levels of professional skill from the operators. Errors in these steps can easily lead to biased analysis results, affecting the accuracy of engineering decisions. Summary of the Invention

[0005] This invention addresses the aforementioned problems and aims to provide a proxy analysis method for the lining structure of highway tunnel cross sections. It constructs a lining structure model using parametric modeling, calculates the model using a structural analysis kernel, and rapidly learns the complex mapping relationship between input parameters and structural response through categorized inputs. After quickly training a multi-input neural network model, it can handle complex and variable lining structure proxy analysis situations without relying on a structural analysis kernel. Trained with a small number of finite element samples, it can replace high-cost finite element analysis calculations, providing efficiency, handling capability for complex and variable situations, and generalization ability for lining structure proxy analysis. It can learn the complex mapping relationship between input parameters and structural response in one go, significantly reducing subsequent computational costs.

[0006] Specifically, the first aspect of the present invention provides a proxy analysis method for the lining structure of a highway tunnel cross section, comprising the following steps: Step 1: Construct several sets of lining structure models for the tunnel cross-section using parametric modeling methods; Step 2: Call the structural analysis kernel to perform calculations on the constructed lining structure model in sequence; Step 3: Construct a multi-input neural network model, using the input parameters of the lining structure model as the model input and the corresponding calculation results of the structural analysis kernel as the model output; Step 4: Train the multi-input neural network model using the input parameters and calculation results of each lining structure model; Step 5: Store the multi-input neural network model after training is completed, and encapsulate it for use in proxy computation.

[0007] The cross-section of the tunnel includes different cases such as horseshoe shape, semi-circle, circle and rectangle. Parametric modeling methods include command flow-based parametric modeling methods in commonly used software such as AutodeskCAD, CATIA, Rhino or other software with parametric modeling functions.

[0008] When training a multi-input neural network model, at least 2000 training sets need to be prepared for the calculation results of each type of structure.

[0009] Techniques for storing trained multi-input neural network models include using the pickle toolkit in Python to store them in HDF5 format, or using the built-in storage capabilities of neural network frameworks such as TensorFlow, Keras, and PyTorch to encapsulate the models.

[0010] Furthermore, the lining structure model is a three-dimensional load-structure model based on shell elements, containing u×v non-repeating mesh nodes. Every three mesh nodes can form a triangular element patch, and each triangular element patch represents a shell element. The u×v non-repeating mesh nodes are automatically grouped using a pre-trained support vector machine model.

[0011] The nodes are divided into u groups of u×v, with each group containing v nodes evenly distributed along the outline of the tunnel cross section.

[0012] The node numbering of the lining structure model can be arranged and expressed in text format. The numbering method includes the following rules: ; in: It is a set of triangular unit facets; For a single triangular unit facet, i and j represent numbers u and v respectively, used to distinguish or retrieve specific triangular unit facets, and can represent the triangular unit facet in the i-th row and j-th column; For a single mesh node in the i-th row and j-th column, every three distinct mesh nodes can represent a triangular facet element; To retrieve the permutation function for each grid node; This represents the number of groups of the node in the u direction; This represents the number of groups of the node in the v direction; Let i be the i-th node in the u direction; Let j be the j-th node in the v direction; When automatically grouping u×v non-repeating mesh nodes using a pre-trained support vector machine model, the load arrangement nodes in the lining structure model will be selected according to three categories: vertical direction, horizontal direction, and lateral constraints (i.e., 90° angle range on both sides). The top load P1 and bottom load P2 of the lining are arranged vertically, the minimum horizontal load Q1 and maximum horizontal load Q2 are arranged horizontally, and the horizontal foundation stiffness S is arranged as a lateral constraint.

[0013] For node grouping in the vertical and horizontal directions, the pre-trained support vector machine model selects a linear kernel function, as shown in the following formula: ; in: A kernel function is used to evaluate a certain feature and to measure the performance of two samples. and Similarity under this feature; The sample is numbered i; Let j be the sample numbered j; For transpose; For laterally constrained node grouping, the pre-trained support vector machine model selects the radial basis kernel function, as shown in the following formula: ; in: is the base of the natural logarithm; This is a parameter in the kernel function that controls the "width" of the RBF kernel. It is generally a positive number, and the specific value is determined by the provided training samples. The sample is numbered i; Let j be the sample numbered j; Furthermore, the input parameters include modeling parameters and boundary condition parameters.

[0014] Furthermore, the modeling parameters include the radius array and angle array corresponding to multiple sets of arc segments of the lining, the lining thickness, the lining width, and the equivalent stiffness coefficient.

[0015] Furthermore, the value ranges of each parameter in the modeling parameters are as follows: The radius array corresponding to the multiple sets of arc segments in the lining has a range of 5m to 10m. The range of the angle array corresponding to the multiple arc segments of the lining is 30°~60°; The lining thickness ranges from 1.5m to 2m; The lining width ranges from 0.3m to 0.6m; The equivalent stiffness coefficient ranges from 0.7 to 0.9.

[0016] The equivalent stiffness coefficient comprehensively reflects the overall stiffness of the lining structure, taking into account the effects of joints, circumferential joints, etc.

[0017] Furthermore, the boundary condition parameters include the top load of the lining, the bottom load of the lining, the minimum horizontal load, the maximum horizontal load, and the horizontal foundation stiffness.

[0018] Furthermore, the value ranges of each parameter in the boundary condition parameters are as follows: The range of the load on the top of the lining is 100kPa to 800kPa; The range of the load at the bottom of the lining is ~800 kPa, which is the same as the load at the top of the lining. The minimum horizontal load ranges from 20 kPa to 500 kPa. The maximum horizontal load ranges from the minimum horizontal load to 500 kPa. The range of horizontal foundation stiffness is 5MPa / m to 100MPa / m.

[0019] The top load of the lining refers to the vertical load acting on the top of the lining, mainly including rock pressure or soil pressure. The bottom load of the lining refers to the load acting on the bottom of the lining. The minimum and maximum horizontal loads are determined by the lateral soil pressure. The horizontal foundation stiffness reflects the foundation's resistance to horizontal loads and is usually calculated through tests or standard formulas.

[0020] Furthermore, the sampling rules for the input parameters, when generating several sets of lining structure models, regarding the random values ​​of the input parameters within a certain range are as follows: The modeling parameters are based on the uniform probability distribution rule. A uniform distribution is a distribution in which all values ​​have equal probability of occurring within a range, i.e., an equally likely distribution. The boundary condition parameters follow the Gaussian probability distribution rule, with the mean as the center. The data are distributed in a bell-shaped symmetry, with the probability density being higher the closer to the mean, and the peak value being at the mean.

[0021] Furthermore, the step of using the input parameters of the lining structure model as model input includes: The relevant parameters of the geometric properties of the tunnel profile are used as the first type of input; The relevant parameters of the mechanical properties of the tunnel profile are used as the second type of input; The relevant parameters of the external load properties of the tunnel profile are used as the third type of input; The relevant parameters of the soil and rock properties of the tunnel outline are used as the fourth type of input.

[0022] Furthermore, the calculation results of the structural analysis kernel include axial force, bending moment, shear force, and displacement.

[0023] After training the data on the four types of results—axial force, bending moment, shear force, and displacement—the four corresponding stored multi-input neural network models are encapsulated and embedded in the program's backend. This allows the proxy analysis calculation to be executed after inputting modeling parameters and boundary condition parameters, outputting the four types of results: axial force, bending moment, shear force, and displacement.

[0024] Furthermore, the calculation results are obtained by rendering the model using vertex shading and then visualizing the three-dimensional numerical results of the structural analysis using RGB.

[0025] The model is rendered by vertex shading, where vertices are selected from mesh nodes in the lining structure model. The RGB shading parameters corresponding to each node number are mapped to the color gradient corresponding to that number based on the distribution range of each array result, thus defining the RGB parameters. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of this drawing or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this drawing. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the steps of the present invention; Figure 2 This is a process diagram illustrating the generation of a horseshoe-shaped cross-sectional lining structure model for a tunnel, as described in this embodiment of the invention. Figure 3 This is a diagram showing the result of automatic grouping of grid nodes using an SVM model in an embodiment of the present invention. Figure 4 This is a structural diagram of the multi-input neural network model in an embodiment of the present invention; Figure 5 This is a hierarchical structure diagram of a multi-input neural network model; Figure 6 This is a diagram showing the output results in an embodiment of the present invention; Wherein, a) is the displacement output result diagram; b) is a graph showing the axial force output results; c) is a diagram showing the shear force output results; d) is a graph showing the bending moment output results; Figure 7 The results of the same case calculated using existing commercial finite element analysis software are shown in the figure. Wherein, a) is the displacement output result diagram; b) is a graph showing the axial force output results; c) is a diagram showing the shear force output results; d) is a graph showing the output results of the bending moment.

[0028] The purpose, features, and advantages of this accompanying drawing will be further explained in conjunction with the embodiments and with reference to the accompanying drawing. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments provided by this invention without inventive effort are within the scope of protection of this invention.

[0030] Obviously, the accompanying drawings described below are merely some examples or embodiments of the present invention. Those skilled in the art can apply the present invention to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this invention, modifications to design, manufacturing, or production based on the technical content disclosed in this invention are merely conventional technical means and should not be construed as insufficient disclosure of the present invention.

[0031] Unless otherwise specified, all embodiments and optional embodiments of the present invention can be combined with each other to form new technical solutions.

[0032] Unless otherwise specified, all technical features and optional technical features of this invention can be combined to form new technical solutions.

[0033] Unless otherwise specified, all steps of the present invention may be performed sequentially or randomly, preferably sequentially. For example, the method includes steps (a) and (b), indicating that the method may include steps (a) and (b) performed sequentially, or it may include steps (b) and (a) performed sequentially. For example, the mention that the method may also include step (c) indicates that step (c) may be added to the method in any order. For example, the method may include steps (a), (b), and (c), or it may include steps (a), (c), and (b), or it may include steps (c), (a), and (b), etc.

[0034] Unless otherwise specified, the terms "comprising" and "including" as used in this invention can be open-ended or closed-ended. For example, "comprising" and "including" can mean that other components not listed may also be included, or that only the listed components may be included.

[0035] Unless otherwise specified, the term "or" is inclusive in this invention. For example, the phrase "A or B" means "A, B, or both A and B". More specifically, the condition "A or B" is satisfied by any of the following conditions: A is true (or exists) and B is false (or does not exist); A is false (or does not exist) and B is true (or exists); or both A and B are true (or exist).

[0036] To better understand the solutions of the embodiments of the present invention, some related terms and concepts that may be involved in the embodiments of the present invention will be introduced below.

[0037] (1) Artificial intelligence (AI), also known as intelligent machinery or machine intelligence, refers to machines created by humans that can exhibit intelligence. Generally, artificial intelligence refers to the technology of presenting human intelligence through ordinary computer programs.

[0038] (2) Machine Learning (ML): Machine learning is the core of artificial intelligence. Machine learning theory mainly involves designing and analyzing algorithms that allow computers to learn automatically. Machine learning algorithms are a class of algorithms that automatically analyze data to obtain patterns and use these patterns to predict unknown data. Therefore, the core of machine learning is data, algorithms (models), and computing power (computer processing ability). Machine learning has a wide range of applications, including data mining, data classification, computer vision, natural language processing (NLP), biometrics, search engines, medical diagnosis, credit card fraud detection, securities market analysis, DNA sequencing, speech and handwriting recognition, strategy games, and robotics. Machine learning involves designing an algorithm model to process data and output the desired results. Users can continuously optimize the algorithm model to achieve more accurate data processing capabilities.

[0039] (3) Neural networks (NNs) are mathematical models that mimic the behavioral characteristics of animal neural networks to perform distributed parallel information processing. They can process information by adjusting the connections between a large number of nodes within the neural network, and possess self-learning and adaptive capabilities. Neural networks are commonly used in artificial intelligence model training and data derivation processing.

[0040] Specifically, neural networks typically contain multiple interconnected layers, such as convolutional layers, fully connected layers (FC), activation layers, or pooling layers.

[0041] (4) Multi-input neural networks are a model architecture that can process multiple input data of different types or sources simultaneously. The core idea is to receive data of different modalities or features through independent input layers, and then perform information fusion and processing through subsequent network layers. For example, in TensorFlow, multiple input layers (each input corresponds to a data type) can be defined and connected to shared or independent hidden layers to finally output a comprehensive result.

[0042] In this embodiment, as Figure 1 As shown, a proxy analysis method for the lining structure of a highway tunnel cross section includes the following steps: Step 1: Construct several sets of lining structure models for the tunnel cross-section using parametric modeling methods; Step 2: Call the structural analysis kernel to perform calculations on the constructed lining structure model in sequence; Step 3: Construct a multi-input neural network model, using the input parameters of the lining structure model as the model input and the corresponding calculation results of the structural analysis kernel as the model output; Step 4: Train the multi-input neural network model using the input parameters and calculation results of each lining structure model; Step 5: Store the multi-input neural network model after training is completed, and encapsulate it for use in proxy computation.

[0043] The cross-section of the tunnel includes different cases such as horseshoe shape, semi-circle, circle and rectangle. The parametric modeling method includes command flow-based parametric modeling methods in commonly used software such as AutodeskCAD, CATIA, Rhino or other software with parametric modeling functions. The parametric modeling method used in this embodiment is Rhino+Grasshopper.

[0044] A structural analysis kernel is a program that performs finite element calculations of structures. Commercial versions include Abaqus and Ansys, while the structural analysis kernel used in this example is Karamba 3D.

[0045] In this embodiment, the cross-section is horseshoe-shaped, and the energy flow is determined by the lining radius array through modeling command. Angle array The tunnel cross-section lining structure model is automatically generated from shell elements of width B. In this embodiment, the process of generating the tunnel horseshoe-shaped cross-section lining structure model is as follows: Figure 2 As shown in the figure The parameters refer to the radius parameters of different arc segments in the inner contour. These parameters are used to generate the inner contour circle-1 automatically by compiling the modeling script in the parametric modeling software.

[0046] The specific command flow involved is as follows: Input: The radii and angles corresponding to the multiple arc segments of the horseshoe-shaped cross-sectional profile, respectively set as... , , and , , ;from Figure 2 It can be seen that, for The angle with the vertical direction, for and The angle between them for and The included angle between them, and the lining width of the horseshoe-shaped cross section is set as B; Output: A lining structure model composed of shell elements, denoted as Shell_model; 1. Based on the ZX plane coordinate system, a horseshoe-shaped contour line is automatically generated using modeling parameters R1, R2, R3 and θ1, θ2, θ3, B, and is designated as Circle_1; 2. Divide Circle_1 into 100 equal parts; 3. Move and copy Circle_1 in the positive and negative directions of the Y-axis respectively, with a movement length of ±B / 2, and set them as Circle_2 and Circle_3 respectively; 4. Assign all 100 equally divided points of Circle_1, Circle_2 and Circle_3 respectively, for a total of 300 points, to P_Group; 5. Set the 300 point objects in P_Group as a matrix object [P]. 3×100 ; 6. For matrix objects [P] 3×100 The 300 points are sorted in a clockwise circular order; 7. Based on the point numbering rules, generate a lining structure model composed of quadrilateral units, and set it as Q_Mesh_Model; 8. Further divide the mesh into a shell element structural model composed of triangular elements, and set it as Shell_Model; 9. Return Shell_Model.

[0047] Furthermore, the input parameters of the lining structure model are used as model inputs, including: The relevant parameters of the geometric properties of the tunnel profile are used as the first type of input; The relevant parameters of the mechanical properties of the tunnel profile are used as the second type of input; The relevant parameters of the external load properties of the tunnel profile are used as the third type of input; The relevant parameters of the soil and rock properties of the tunnel outline are used as the fourth type of input.

[0048] The advantage of this classification method is that it allows for differentiated setting and fitting of neurons and activation functions associated with input parameters, better extraction of the relationships between various input parameters, and improved training efficiency of multi-input neural networks.

[0049] In this embodiment, the structure of the multi-input neural network model is as follows: Figure 4 As shown, the input parameters of the model are the parameters of the lining structure model, including the modeling parameters and boundary condition parameters shown in the figure, with the radius array as the most significant parameter. Angle array Thickness H and width parameter B are used as the first type of input; equivalent stiffness coefficient η is used as the second type of input; top load P1, bottom load P2, minimum horizontal load Q1, and maximum horizontal load Q2 are used as the third type of input; and horizontal foundation stiffness S is used as the fourth type of input. In the figure, 18, 36, and 300 represent the number of neurons. The output calculation results include four [P]3×100 matrices for axial force, bending moment, shear force, and displacement.

[0050] In this embodiment, the hierarchical structure of the multi-input neural network model is as follows: Figure 5 As shown, in addition to the input and output layers, there are two hidden layers. The first hidden layer has 73 neurons, each with different neurons depending on the input type. The second hidden layer merges with the neurons in the first hidden layer and is uniformly set to 300 neurons. The Dropout parameter is 0.2. The activation function for the first hidden layer is Tanh, and the activation function for the second hidden layer is Softmax.

[0051] When training a multi-input neural network model, at least 2000 training sets need to be prepared for the calculation results of each type of structure.

[0052] Techniques for storing trained multi-input neural network models include using the pickle toolkit in Python to store them in HDF5 format, or using the built-in storage capabilities of neural network frameworks such as TensorFlow, Keras, and PyTorch to encapsulate the models.

[0053] Furthermore, the lining structure model is a three-dimensional load-structure model based on shell elements, containing u×v non-repeating mesh nodes. Every three mesh nodes can form a triangular element patch, and each triangular element patch represents a shell element. The u×v non-repeating mesh nodes are automatically grouped using a pre-trained support vector machine model.

[0054] The nodes are divided into u groups of u×v, with each group containing v nodes evenly distributed along the outline of the tunnel cross section.

[0055] The node numbering of the lining structure model can be arranged and expressed in text format. The numbering method includes the following rules: ; In this embodiment, the result of automatic grouping of grid nodes using the SVM model is shown in the figure below. Figure 3 As shown, by Figure 3As can be seen, when the u×v non-repeating grid nodes are automatically grouped by the pre-trained support vector machine model, the load arrangement nodes in the lining structure model will be selected according to three categories: vertical direction, horizontal direction, and lateral constraint (i.e., 90° angle range on both sides). The top load P1 and bottom load P2 of the lining are arranged vertically, the minimum horizontal load Q1 and the maximum horizontal load Q2 are arranged horizontally, and the horizontal foundation stiffness S is arranged as a lateral constraint.

[0056] For node grouping in the vertical and horizontal directions, the pre-trained support vector machine model selects a linear kernel function, as shown in the following formula: ; For laterally constrained node grouping, the pre-trained support vector machine model selects the radial basis kernel function, as shown in the following formula: ; Furthermore, the input parameters include modeling parameters and boundary condition parameters.

[0057] Furthermore, the modeling parameters include the radius array and angle array corresponding to multiple sets of arc segments of the lining, the lining thickness, the lining width, and the equivalent stiffness coefficient.

[0058] Furthermore, the value ranges of each parameter in the modeling parameters are as follows: The radius array corresponding to the multiple sets of arc segments in the lining has a range of 5m to 10m. The range of the angle array corresponding to the multiple arc segments of the lining is 30°~60°; The lining thickness ranges from 1.5m to 2m; The lining width ranges from 0.3m to 0.6m; The equivalent stiffness coefficient ranges from 0.7 to 0.9.

[0059] The equivalent stiffness coefficient comprehensively reflects the overall stiffness of the lining structure, taking into account the effects of joints, circumferential joints, etc.

[0060] Furthermore, the boundary condition parameters include the top load of the lining, the bottom load of the lining, the minimum horizontal load, the maximum horizontal load, and the horizontal foundation stiffness.

[0061] Furthermore, the value ranges of each parameter in the boundary condition parameters are as follows: The range of the load on the top of the lining is 100kPa to 800kPa; The range of the load at the bottom of the lining is ~800 kPa, which is the same as the load at the top of the lining. The minimum horizontal load ranges from 20 kPa to 500 kPa. The maximum horizontal load ranges from the minimum horizontal load to 500 kPa. The range of horizontal foundation stiffness is 5MPa / m to 100MPa / m.

[0062] The top load of the lining refers to the vertical load acting on the top of the lining, mainly including rock pressure or soil pressure. The bottom load of the lining refers to the load acting on the bottom of the lining. The minimum and maximum horizontal loads are determined by the lateral soil pressure. The horizontal foundation stiffness reflects the foundation's resistance to horizontal loads and is usually calculated through tests or standard formulas.

[0063] Furthermore, regarding the input parameters, the sampling rules for random values ​​of the input parameters within a certain range when generating several sets of lining structure models are as follows: The modeling parameters are based on the uniform probability distribution rule. A uniform distribution is a distribution in which all values ​​have equal probability of occurring within a range, i.e., an equally likely distribution. The boundary condition parameters follow the Gaussian probability distribution rule, with the mean as the center. The data are distributed in a bell-shaped symmetry, with the probability density being higher the closer to the mean, and the peak value being at the mean.

[0064] Furthermore, the calculation results of the structural analysis kernel include axial force, bending moment, shear force, and displacement.

[0065] After training the data on four types of results—axial force, bending moment, shear force, and displacement—the four corresponding stored multi-input neural network models are encapsulated and embedded in the program's backend. This allows for proxy analysis calculations to be executed after inputting modeling parameters and boundary condition parameters, outputting the four types of results: axial force, bending moment, shear force, and displacement. In this embodiment of the invention, displacement ( Figure 6 a) Axial force ( Figure 6 b) Shear force ( Figure 6 c) and bending moment ( Figure 6 d) The output results of the four types of results are shown in the figure below. Figure 6 As shown, the results of the same case calculated by existing commercial finite element analysis software are as follows: Figure 7 As shown (displacement ( Figure 7 a) Axial force ( Figure 7 b) Shear force ( Figure 7 c) and bending moment ( Figure 7 d) Figure 6 and Figure 7 The positive and negative signs in the equation only indicate different directions, and the magnitudes are compared by absolute value.

[0066] from Figure 6 and Figure 7It can be seen that the results of the embodiments of the present invention differ from the results of finite element analysis by less than 10% in the four values, which is within the reasonable range of calculation and can replace conventional structural calculation methods. However, the present invention combines a structural analysis kernel and a multi-input neural network, and learns the complex mapping relationship between input parameters and structural response by classifying inputs. After quickly training the multi-input neural network model, it can cope with complex and ever-changing lining structure proxy analysis situations. It does not rely on the structural analysis kernel. After training with a small number of finite element samples, it can replace high-cost finite element analysis calculations, providing efficiency, complex and ever-changing handling capabilities, and generalization capabilities for lining structure proxy analysis.

[0067] Furthermore, the calculation results are obtained by rendering the model as a cloud map through vertex shading and representing the three-dimensional numerical results of the structural analysis with RGB visualization.

[0068] The model is rendered by vertex shading, where vertices are selected from mesh nodes in the lining structure model. The RGB shading parameters corresponding to each node number are mapped to the color gradient corresponding to that number based on the distribution range of each array result, thus defining the RGB parameters.

[0069] It should be noted that the present invention is not limited to the above-described embodiments. The above embodiments are merely examples, and any embodiments that have the same structure and perform the same effects as the technical concept within the scope of the present invention are included within the scope of the present invention. Furthermore, various modifications that can be conceived by those skilled in the art to the embodiments, and other ways of constructing by combining some of the constituent elements of the embodiments, without departing from the spirit of the present invention, are also included within the scope of the present invention.

Claims

1. A proxy analysis method for the lining structure of a highway tunnel cross section, characterized in that, Includes the following steps: Step 1: Construct several sets of lining structure models for the tunnel cross-section using parametric modeling methods; Step 2: Call the structural analysis kernel to perform calculations on the constructed lining structure model in sequence; Step 3: Construct a multi-input neural network model, using the input parameters of the lining structure model as the model input and the corresponding calculation results of the structural analysis kernel as the model output; Step 4: Train the multi-input neural network model using the input parameters and calculation results of each lining structure model; Step 5: Store the multi-input neural network model after training is completed, and encapsulate it for use in proxy computation.

2. The method for proxy analysis of lining structure in the cross-section of highway tunnels according to claim 1, characterized in that, The lining structure model is a three-dimensional load-structure model based on shell elements, containing u×v non-repeating mesh nodes. Every three mesh nodes can form a triangular element patch, and each triangular element patch represents a shell element. The u×v non-repeating mesh nodes are automatically grouped using a pre-trained support vector machine model.

3. The method for proxy analysis of lining structure in the cross-section of highway tunnels according to claim 1, characterized in that, The input parameters include modeling parameters and boundary condition parameters.

4. The method for proxy analysis of lining structure in the cross-section of highway tunnels according to claim 3, characterized in that, The modeling parameters include the radius array and angle array corresponding to multiple sets of arc segments of the lining, the lining thickness, the lining width, and the equivalent stiffness coefficient.

5. The proxy analysis method for the lining structure of a highway tunnel cross section according to claim 4, characterized in that, The value ranges of each parameter in the modeling parameters are as follows: The radius array corresponding to the multiple sets of arc segments in the lining has a range of 5m to 10m. The range of the angle array corresponding to the multiple arc segments of the lining is 30°~60°; The lining thickness ranges from 1.5m to 2m; The lining width ranges from 0.3m to 0.6m; The equivalent stiffness coefficient ranges from 0.7 to 0.

9.

6. The proxy analysis method for the lining structure of a highway tunnel cross section according to claim 3, characterized in that, The boundary condition parameters include the top load of the lining, the bottom load of the lining, the minimum horizontal load, the maximum horizontal load, and the horizontal foundation stiffness.

7. The method for proxy analysis of lining structure in the cross-section of a highway tunnel according to claim 6, characterized in that, The value ranges of each parameter in the boundary condition parameters are as follows: The range of the load on the top of the lining is 100kPa to 800kPa; The range of the load at the bottom of the lining is ~800 kPa, which is the same as the load at the top of the lining. The minimum horizontal load ranges from 20 kPa to 500 kPa. The maximum horizontal load ranges from the minimum horizontal load to 500 kPa. The range of horizontal foundation stiffness is 5MPa / m to 100MPa / m.

8. The method for proxy analysis of lining structure in the cross-section of highway tunnels according to claim 3, characterized in that, The sampling rules for the input parameters, when generating several sets of lining structure models, regarding the random selection of input parameters within a range are as follows: The modeling parameters follow a uniform probability distribution rule; The boundary condition parameters follow the Gaussian probability distribution rule.

9. The proxy analysis method for the lining structure of a highway tunnel cross section according to claim 1, characterized in that, The method of using the input parameters of the lining structure model as model input includes: The relevant parameters of the geometric properties of the tunnel profile are used as the first type of input; The relevant parameters of the mechanical properties of the tunnel profile are used as the second type of input; The relevant parameters of the external load properties of the tunnel profile are used as the third type of input; The relevant parameters of the soil and rock properties of the tunnel outline are used as the fourth type of input.

10. A proxy analysis method for the lining structure of a highway tunnel cross section according to claim 1, characterized in that, The calculation results of the structural analysis kernel include axial force, bending moment, shear force, and displacement.

Images