Method for rapidly judging crack stability of tunnel lining under uniform stratum

Abstract

The invention discloses a method for rapidly judging the crack stability of a tunnel lining under a uniform stratum. The method comprises the following steps: step 1, obtaining geometrical characteristic parameters and the like of cracks at different spatial positions on the surface of the tunnel lining; 2, establishing a three-dimensional geometric model of a tunnel lining in a finite element calculation environment, and calculating a stress intensity factor of a crack tip; 3, based on the stress intensity factor of the crack tip, calculating to obtain a crack stability parameter f; 4, taking geometrical characteristic parameters and the like of the cracks at different spatial positions on the surface of the tunnel lining as input characteristics, taking crack stability parameters as output labels, and constructing a structured sample database for machine learning training; 5, setting the structure of the machine learning model, and training; and 6, acquiring crack data of a to-be-judged area, inputting the trained and optimized machine learning model, and quickly judging the lining stability.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering safety assessment and intelligent operation and maintenance technology, and in particular to a method for rapid determination of the stability of tunnel lining cracks under uniform strata. Background Technology

[0002] As a crucial component of transportation infrastructure, the structural safety of tunnels directly impacts the safety of people's lives and property and the stable operation of the transportation system. The tunnel lining, as the main load-bearing and protective component, is a key factor in ensuring the safe service life of the tunnel. During long-term service, due to various factors such as surrounding rock deformation, groundwater erosion, construction defects, temperature changes, and load effects, the lining structure inevitably develops defects such as cracks and spalling. Among these, cracks are one of the most common, typical, and potentially dangerous forms of defects. Therefore, timely and accurate stability assessment of the crack condition of tunnel linings is of great significance for ensuring the structural safety of tunnels, extending their service life, and guiding operation and maintenance decisions.

[0003] Currently, the detection and assessment of tunnel lining cracks mainly rely on a combination of manual inspection and localized instrument testing. Manual inspection is significantly affected by personnel experience, subjective judgment, and the site environment, resulting in low efficiency, poor repeatability, and insufficient consistency of results. While some non-destructive testing techniques can obtain geometric or internal information about cracks, they typically only provide data characterizing the damage and cannot directly reflect the impact of cracks on the overall structural stability. On the other hand, conducting detailed mechanical analysis on individual cracks often requires building complex numerical models and performing lengthy calculations, which is insufficient to meet the engineering needs of large-scale inspections and rapid screening. Therefore, existing technologies still have significant shortcomings in terms of efficiency, automation, and the ability to directly characterize the structural safety status.

[0004] With the development of artificial intelligence technology, machine learning methods have shown great application potential in the field of structural health monitoring and safety assessment. They can automatically learn the complex mapping relationship between crack characteristics and structural state from large amounts of detection data, achieving efficient state identification and risk prediction. Meanwhile, numerical simulation methods based on physical mechanisms can precisely characterize the impact of cracks on the structural stress state and stability, exhibiting significant advantages in theoretical reliability and result interpretability. Combining the high-reliability mechanical information provided by numerical simulation with the data-driven advantages of machine learning methods promises to simultaneously balance analytical accuracy and assessment efficiency, thereby constructing a more scientific and efficient tunnel lining stability assessment system.

[0005] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by proposing a rapid method for determining the stability of tunnel lining cracks in uniform strata. This method combines machine learning and numerical simulation, using the mechanical analysis results provided by numerical simulation as an important basis for the machine learning model. This allows the model to have both a physical mechanism foundation and efficient data processing capabilities, thereby enabling rapid and reliable determination of the stability of tunnel lining cracks. This meets the actual needs of efficiency and safety in daily inspection and operation and maintenance management of tunnel projects.

[0007] To achieve the above objectives, the present invention also employs the following technical solution: A rapid method for determining the stability of tunnel lining cracks under uniform strata includes the following steps: Step 1: Obtain the geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions. Step 2: Establish a three-dimensional geometric model of the tunnel lining in the finite element calculation environment, input the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, set various crack condition combinations, and calculate the stress intensity factor at the crack tip. Step 3: Calculate the crack stability parameter f based on the stress intensity factor at the crack tip; Step 4: Using the geometric feature parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions as input features, and the crack stability parameters as output labels, a structured sample database for machine learning training is constructed. Step 5: Set up the structure of the machine learning model; and train the machine learning model using a structured sample database to establish a mapping relationship from crack features to crack stability parameters, so as to achieve rapid determination of lining stability. Step 6: Obtain crack data of the area to be judged, input it into the trained and optimized machine learning model, and quickly judge the stability of the lining.

[0008] Furthermore, step 1 includes the following steps: Step 1.1: Collect images of cracks at different spatial locations on the tunnel lining surface, and obtain the geometric feature parameters of the cracks through image recognition algorithms; the spatial locations include the arch crown, arch shoulder, arch waist, arch foot, and bottom plate; the geometric feature parameters include crack length, crack width, and crack direction; Step 1.2: Obtain the physical and mechanical parameters of the surrounding rock and strata in the tunnel section through geological survey reports, borehole test reports and field test results generated during the tunnel construction or operation phases; the physical and mechanical parameters shall include at least the elastic modulus, Poisson's ratio, internal friction angle and cohesion; Step 1.3: Obtain tunnel operation condition data through the tunnel operation management system or traffic monitoring system, convert the discrete operation condition data into equivalent load parameters for structural analysis, and then superimpose them; the operation condition data includes vehicle type, axle load level, and traffic flow.

[0009] Furthermore, step 2 includes the following steps: Step 2.1: Establish a three-dimensional geometric model of the tunnel lining in the finite element calculation environment, and set the finite element solution method and boundary conditions; the three-dimensional geometric model shall include at least two parts: the lining structure and the surrounding rock calculation domain; Step 2.2: Set up a mechanical constitutive model in the three-dimensional geometric model set in Step 2.1, and input the physical and mechanical parameters of the surrounding rock and strata in the tunnel section; Step 2.3: In the three-dimensional geometric model set in Step 2.1, multiple sets of crack geometric models with crack length, crack depth and crack inclination are preset at the positions of the arch crown, arch shoulder, arch waist, arch foot and base plate to form a variety of crack condition combinations; Step 2.4: Perform finite element analysis to obtain the stress intensity factor at the crack tip under each crack condition combination; the stress intensity factor includes the stress intensity factor K for Type I cracks (opening mode) and Type II cracks (cutting mode). I and K II .

[0010] Furthermore, in step 2.1, the lining structure adopts a circumferential structure consistent with the actual tunnel cross-section, and extends outward along the tunnel axis to form a three-dimensional segment model of a preset length, so as to avoid the influence of end constraints on the local stress field at the crack tip. The surrounding rock mass computational domain encloses the lining structure. The outer boundary of the surrounding rock mass and the outer edge of the lining are kept at a preset distance so that the stress disturbance at the outer boundary has a negligible impact on the lining area. A triaxial displacement constraint (Ux=Uy=Uz=0) is applied to the bottom surface of the surrounding rock mass to simulate the support of deep strata. A normal displacement constraint (limiting normal displacement and allowing tangential displacement) is applied to the lateral outer surface of the surrounding rock mass to simulate the lateral constraint of the far-field surrounding rock on the model. The top surface of the surrounding rock mass is set as a free surface or an equivalent vertical load boundary is applied to simulate the action of overlying strata pressure and operational loads. The interaction between the lining and the surrounding rock mass adopts a binding constraint. The equivalent load corresponding to the operating condition information is applied in a multi-point loading manner, which equates the train axle load to a concentrated force at several discrete loading points. The loading points are arranged along the tunnel axis according to the distance between the wheels and the bogies, and are applied at the sleeper support position or the track bed support area, thus forming a multi-point loading condition. For different operating conditions, different train formations and axle load levels are simulated by changing the magnitude of the concentrated force, the number of points, and the spacing between points.

[0011] Furthermore, in step 2.1, the finite element solution method employs a quasi-static analysis step; an implicit solution framework is used, and its governing equations are the following nonlinear equilibrium forms after finite element discretization: R(u)=F ext (u)-F int (u)=0; In the formula, u is the nodal displacement vector, and F ext (u) is the equivalent nodal force vector of the external load, F int (u) is the internal force vector obtained by integrating the material constitutive stress with the element stress; The Newton-Raphson iterative solution is used, and its iterative update format is as follows: K t (u i )△u i =R(u i ), u i+1 =u i +△u i ; In the formula, K t Let u be the tangent stiffness matrix. i Let Δu be the approximate solution of the displacement vector in the i-th iteration. i This refers to the displacement increment that needs to be compensated under the current unbalanced force.

[0012] Furthermore, in step 2.1, the following control strategy is adopted: Geometric nonlinearity is enabled, and automatic incrementation is used for incremental control. The initial increment is 1%–5% of the total step size, the minimum increment is 1e-6–1e-4, and the maximum increment is 5%–20%. The maximum number of increments is set to 200–2000 to avoid premature termination of the nonlinear stage. Force and displacement dual criteria are used to control convergence. When the crack tip region is more sensitive to convergence, residual control is selected. When convergence difficulties are caused by local softening or contact, automatic stabilization is enabled, and the proportion of stabilization dissipation energy to external work is limited to the order of 1e-5–1e-3 to ensure that the physical results are not dominated by numerical damping.

[0013] Furthermore, in step 2.4: the extended finite element method (XFEM) is used to calculate the crack tip. XFEM crack description is enabled in the lining for the area where the preset crack is located. The lining domain where the crack is located is defined as the XFEM enrichment zone, and a high-density mesh is set in the crack tip region. When only the stress intensity factor under a given crack condition needs to be obtained, the crack is set to not propagate, and only the fracture parameters are calculated for the predetermined crack leading edge. When the crack growth trend needs to be considered, the crack propagation criterion is enabled in XFEM. In step 2.4: A fracture parameter output request is defined for the crack front. A fracture mechanics post-processing method based on J-integral and interaction integral is used to calculate the stress intensity factor at the crack tip. Specifically, an integration domain is established for the crack front, and several integration profiles are set to verify the path independence of the integration results. When the fracture parameter results under multiple profiles converge to a stable value, this stable value is output as the stress intensity factor corresponding to the crack condition. For the three-dimensional crack front, the stress intensity factor distribution is output along multiple sampling points along the crack front, and the maximum value or the value at a representative location is taken as the characteristic stress intensity factor of the crack condition for subsequent safety factor calculation and sample labeling. After the solution is completed, the stress intensity factor K of the type I crack containing the Mode I component is output. I Output the stress intensity factor K of type II cracks, including the Mode II component. II .

[0014] Furthermore, in step 3, the crack stability parameter f is calculated using the following formula: ; In the formula, K Ic and K IIc The fracture toughness of type I cracks and type II cracks, respectively, K I and K II The stress intensity factors λ for Type I and Type II cracks, respectively. 12 This is the compressive shear stability coefficient.

[0015] Furthermore, step 4 includes the following steps: Step 4.1: The geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions are uniformly formatted. Continuous parameters are dimensionally unified and numerically normalized, and discrete parameters are encoded to construct a multi-dimensional input feature vector. Step 4.2: Match the crack stability parameters with the working conditions corresponding to the crack geometric feature parameters to form a one-to-one "feature-label" sample pair; Step 4.3: Perform data cleaning and consistency verification on the constructed sample pairs, and remove sample data with missing key feature parameters, obviously abnormal values ​​or inconsistent labels; smooth or correct parameters with noise interference to ensure the integrity, accuracy and consistency of sample data, thereby constructing a high-quality machine learning training sample database. Step 4.4: Organize the processed sample pairs into a structured sample database for subsequent model training and validation.

[0016] Furthermore, in step 5, the machine learning model is an ensemble learning model based on gradient boosting decision trees as the stability determination model; specifically, the XGBoost model structure is adopted, and the XGBoost model parameters are set as follows: n_estimators is 300, learning_rate is 0.1, max_depth is 7, subsample is 0.6, colsample_bytree is 0.8, gamma is 3, and lambda is 6.

[0017] Compared with the prior art, the beneficial effects of this invention are as follows: This system achieves a closed-loop and highly automated process encompassing detection, simulation, and judgment, significantly improving the overall efficiency of tunnel lining defect assessment and reducing reliance on manual labor and the risk of subjective misjudgment. High-confidence, refined numerical simulation results are used as training labels for the machine learning model, ensuring the scientific rigor and accuracy of the model's learning objectives. The trained machine learning model enables "second-level" judgments, meeting the stringent time requirements of daily tunnel inspections and rapid screening, thus saving valuable time for preventative maintenance. Based on this, a standardized tunnel lining stability assessment process can be established, which can be widely applied to the intelligent operation and maintenance management of various tunnel projects, including railways, highways, and subways, demonstrating broad engineering application prospects. Attached Figure Description

[0018] Figure 1 The following is a flowchart of a method for rapidly determining the stability of tunnel lining cracks in uniform strata, as described in the embodiments below.

[0019] Figure 2 The following is a schematic diagram of the stress intensity factor solution setup in the embodiments below.

[0020] Figure 3 The following diagram shows the calculated stress intensity factor results in the embodiments below.

[0021] Figure 4 The following is a graph showing the results of Pearson correlation analysis in the examples below.

[0022] Figure 5 The following diagram shows the prediction results of the XGBoost model structure in the embodiments below.

[0023] Figure 6 The following diagram shows the model calculation results after merging the target area sample with the original eastern coastal engineering sample in the following embodiments. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Example

[0025] A rapid method for determining the stability of tunnel lining cracks under uniform strata, such as Figure 1 As shown, it includes the following steps: Step 1: Obtain the geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions.

[0026] In this embodiment, step 1 includes the following steps: Step 1.1: Collect images of cracks at different spatial locations on the tunnel lining surface, and obtain the geometric feature parameters of the cracks through image recognition algorithms; the spatial locations include the arch crown, arch shoulder, arch waist, arch foot, and bottom plate; the geometric feature parameters include crack length, crack width, and crack direction.

[0027] In this embodiment, crack length is used to characterize the extent of crack extension along the lining surface, crack width is used to characterize the degree of crack opening, and crack orientation is used to characterize the orientation of the crack relative to the tunnel axis or cross-sectional direction. By associating the geometric feature parameters with the spatial location of the crack, a structured crack dataset is constructed.

[0028] Step 1.2: Obtain the physical and mechanical parameters of the surrounding rock and strata in the tunnel section through the geological survey report, borehole test report and field test results generated during the tunnel construction or operation phase; the physical and mechanical parameters include at least the elastic modulus, Poisson's ratio, internal friction angle and cohesion.

[0029] In this embodiment, if there are multiple sets of experimental data for a certain physical and mechanical parameter, the multiple sets of experimental data can be statistically processed, and the average value or representative value can be taken as the value of the parameter.

[0030] Step 1.3: Obtain tunnel operation condition data through the tunnel operation management system or traffic monitoring system, convert the discrete operation condition data into equivalent load parameters for structural analysis, and then superimpose them; the operation condition data includes vehicle type, axle load level (single axle or multiple axles), and traffic flow.

[0031] In this embodiment, the superimposed equivalent load parameters are used to characterize the effect of the operating load on the lining structure, serving as an important input condition for subsequent stability analysis.

[0032] Step 2: Establish a three-dimensional geometric model of the tunnel lining in the finite element calculation environment, input the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, set various crack condition combinations, and calculate the stress intensity factor at the crack tip.

[0033] In this embodiment, step 2 includes the following steps: Step 2.1: Establish a three-dimensional geometric model of the tunnel lining in the finite element calculation environment, and set the finite element solution method and boundary conditions; the three-dimensional geometric model shall include at least two parts: the lining structure and the surrounding rock calculation domain.

[0034] In this embodiment, a three-dimensional geometric model of the tunnel lining, namely a three-dimensional lining-surrounding rock coupling calculation model, can be established using the finite element analysis software ABAQUS.

[0035] In this embodiment, in step 2.1, the lining structure adopts a circumferential structure consistent with the actual tunnel cross-section, extending outward along the tunnel axis to form a three-dimensional segment model of a preset length, in order to avoid the influence of end constraints on the local stress field at the crack tip; the surrounding rock calculation domain encloses the lining structure, and the outer boundary of the surrounding rock maintains a preset distance from the outer edge of the lining (approximately 3 times the lining diameter), so that the stress disturbance at the outer boundary has a negligible impact on the lining area; a three-dimensional displacement constraint (Ux=Uy=Uz=0) is applied to the bottom surface of the surrounding rock to simulate the support of deep strata, and a normal displacement constraint (limiting normal displacement and allowing tangential displacement) is applied to the lateral outer surface of the surrounding rock to simulate the lateral impact of the far-field surrounding rock on the model. Constraints are applied by setting the top surface of the surrounding rock mass as a free surface or applying an equivalent vertical load boundary to simulate the action of overlying stratum pressure and operational loads; the interaction between the lining and the surrounding rock is constrained by binding; the equivalent load corresponding to the operational condition information is applied by multi-point loading, which equates the train axle load to a number of discrete loading points. The loading points are arranged along the tunnel axis according to the wheel-to-bogie spacing and applied to the sleeper support position or the track bed support area, thus forming a multi-point loading condition; for different operational conditions (referring to the combination of train formation type, load capacity and other conditions that can affect the equivalent load), different train formations and axle load levels are simulated by changing the magnitude of the concentrated force, the number of points and the spacing between points.

[0036] In this embodiment, in step 2.1, the finite element solution method adopts a quasi-static analysis step; an implicit solution framework is used (considering material nonlinearity, geometric nonlinearity, or contact nonlinearity), and its governing equations are the following nonlinear equilibrium forms after finite element discretization: R(u)=F ext (u)-F int (u)=0; In the formula, u is the nodal displacement vector, and F ext (u) is the equivalent nodal force vector of the external load, F int (u) is the internal force vector obtained by integrating the material constitutive properties and the element stress.

[0037] The Newton-Raphson iterative solution is used, and its iterative update format is as follows: K t (u i )△ui =R(u i ), u i+1 =u i +△u i ; In the formula, K t Let u be the tangent stiffness matrix. i This is an approximate solution for the displacement vector at the i-th iteration, i.e., the displacement of each node of the structure in the current iteration state, Δu. i It refers to the displacement increment that needs to be compensated under the current unbalanced force, that is, the intermediate quantity in the iteration process (rather than the final physical displacement).

[0038] In this embodiment, in step 2.1, the following control strategy is adopted (the parameters can be adjusted according to the model size and the strength of nonlinearity): Geometric nonlinearity is enabled, and automatic incrementation is used for incremental control. The initial increment is 1%–5% of the total step size, the minimum increment is 1e-6–1e-4, and the maximum increment is 5%–20%. The maximum number of increments is set to 200–2000 to avoid premature termination of the nonlinear stage. Force and displacement dual criteria are used to control convergence. When the crack tip region is more sensitive to convergence, residual control is selected. When convergence difficulties are caused by local softening or contact, automatic stabilization is enabled, and the proportion of stabilization dissipation energy to external work is limited to the order of 1e-5–1e-3 to ensure that the physical results are not dominated by numerical damping.

[0039] The above solution setting is a preferred embodiment. This invention does not limit the specific finite element software or specific iterative algorithm. Other finite element platforms and solution strategies that satisfy the solution of nonlinear equilibrium equations can also implement this step.

[0040] Step 2.2: Set up a mechanical constitutive model in the three-dimensional geometric model set in Step 2.1, and input the physical and mechanical parameters of the surrounding rock and strata in the tunnel section.

[0041] In this embodiment, the physical and mechanical parameters of the surrounding rock and strata of the tunnel section obtained in step 1 are input; the parameters can be set to be assigned to different strata types.

[0042] Step 2.3: In the three-dimensional geometric model set in Step 2.1, multiple sets of crack geometric models with crack length, crack depth and crack inclination are preset at the positions of the arch crown, arch shoulder, arch waist, arch foot and base plate to form a variety of crack condition combinations.

[0043] In this embodiment, the geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface obtained in step 1 are input; different crack lengths, different crack depths and different crack inclination angles are combined and varied to form multiple sets of crack conditions, which lay the foundation for subsequent extended finite element crack analysis.

[0044] Step 2.4: Perform finite element analysis to obtain the stress intensity factor at the crack tip under each crack condition combination; the stress intensity factor includes the stress intensity factor K for Type I cracks (opening mode) and Type II cracks (cutting mode). I and K II .

[0045] In this embodiment, in step 2.4: the extended finite element method (XFEM) is used to calculate the crack tip. XFEM crack description is enabled in the lining for the area where the preset crack is located. The lining domain where the crack is located is defined as the XFEM enrichment zone. A high-density mesh is set in the crack tip area. When only the stress intensity factor under a given crack condition needs to be obtained, the crack is set to not propagate, and only the fracture parameters are calculated for the predetermined crack leading edge. When the crack growth trend needs to be considered, the crack propagation criterion is enabled in XFEM.

[0046] In this embodiment, in step 2.4: a fracture parameter output request is defined for the crack front, and a fracture mechanics post-processing method based on J-integral and interaction integral is used to calculate the stress intensity factor at the crack tip; specifically, an integration domain is established for the crack front, and several integration profiles are set to check the path independence of the integration results; when the fracture parameter results under multiple profiles converge to a stable value, this stable value is output as the stress intensity factor corresponding to the crack condition; for the three-dimensional crack front, the stress intensity factor distribution is output along multiple sampling points of the crack front, and the maximum value or the value at a representative position is taken as the characteristic stress intensity factor of the crack condition for subsequent safety factor calculation and sample labeling. After the solution is completed, the stress intensity factor K1 of the type I crack containing Mode I component is output, the stress intensity factor K2 of the type II crack containing Mode II component is output, and the stress intensity factor K3 of the type III crack containing Mode II component is output. The schematic diagram of the stress intensity factor solution setting and the stress intensity factor calculation results are shown below. Figure 2 , Figure 3 As shown.

[0047] Step 3: Calculate the crack stability parameter f based on the stress intensity factor at the crack tip.

[0048] In this embodiment, in step 3, the crack stability parameter f is calculated using the following formula.

[0049] ; In the formula, K Ic and K IIc The fracture toughness of type I cracks and type II cracks, respectively, K I and K II The stress intensity factors λ for Type I and Type II cracks, respectively. 12 This is the compressive shear stability coefficient.

[0050] In this embodiment, the crack stability parameter f is calculated based on the results of step 2. The stability of the cracked lining structure is positively correlated with f, and an increase in the f value indicates an improvement in structural stability. When f ≥ 1, it indicates that the lining crack is in a stable state, meaning that the lining structure is basically stable; when f < 1, the lining crack will undergo unsteady propagation, and the lining structure is in an unstable state, and the smaller the f value, the more significant the deterioration of structural stability.

[0051] Step 4: Using the geometric feature parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions as input features, and the crack stability parameters as output labels, a structured sample database for machine learning training is constructed.

[0052] In this embodiment, step 4 includes the following steps: Step 4.1: The geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions are uniformly formatted. Continuous parameters are dimensionally unified and numerically normalized, and discrete parameters are encoded to construct a multi-dimensional input feature vector.

[0053] Step 4.2: Match the crack stability parameters with the working conditions corresponding to the crack geometric feature parameters to form a one-to-one "feature-label" sample pair.

[0054] In this embodiment, the crack stability parameters obtained in step 3 are matched with the corresponding crack conditions (combinations of different crack lengths, different crack depths and different crack dip angles) so that each set of geometric feature parameters forms a one-to-one correspondence with its corresponding crack stability parameters, thereby constructing "feature-label" sample pairs.

[0055] Step 4.3: Perform data cleaning and consistency verification on the constructed sample pairs, removing sample data with missing key feature parameters, obviously abnormal values, or inconsistent labels; smooth or correct parameters with noise interference to ensure the integrity, accuracy, and consistency of the sample data, thereby constructing a high-quality machine learning training sample database. The Pearson feature correlation analysis results are as follows: Figure 4 As shown.

[0056] Step 4.4: Organize the processed sample pairs into a structured sample database for subsequent model training and validation.

[0057] Step 5: Set up the structure of the machine learning model; and use a structured sample database to train the machine learning model to establish a mapping relationship from crack features to crack stability parameters, so as to achieve rapid determination of lining stability.

[0058] In this embodiment, in step 5, the machine learning model is an ensemble learning model based on gradient boosting decision trees as the stability determination model; specifically, the XGBoost model structure is adopted, and the XGBoost model parameters are set as follows: n_estimators = 300, learning_rate = 0.1, max_depth = 7, subsample = 0.6, colsample_bytree = 0.8, gamma = 3, lambda = 6. The prediction results of the XGBoost model structure are as follows. Figure 5 As shown.

[0059] The XGBoost model, by constructing multiple weak classifiers and performing weighted ensemble, can effectively fit the complex nonlinear mapping relationship between fracture geometry, formation physical and mechanical characteristics, operational conditions, and stability levels. It also has strong generalization ability and computational efficiency, making it suitable for large-scale sample training and rapid inference needs in engineering scenarios.

[0060] When the method of this embodiment is extended to other regions, the following parameter tuning / transfer process can be adopted: Small sample calibration set construction: Select several representative sections of the target tunnel (referring to different excavation rings of the target tunnel), collect a small amount of crack geometric parameters and environmental information to form a calibration dataset; Supplementing numerical simulation annotation: Call steps 2 and 3 for typical crack conditions corresponding to the calibration set, calculate the stress intensity factor and crack stability parameters to supplement the "feature-label" samples of the target tunnel; Sample merging and retraining: Merge the target region samples with the original eastern coastal engineering samples, or incrementally train / retrain the existing XGBoost model using only the target region samples. The model calculation results after merging the target region samples with the original eastern coastal engineering samples are as follows: Figure 6 As shown.

[0061] Step 6: Obtain crack data of the area to be judged, input it into the trained and optimized machine learning model, and quickly judge the stability of the lining.

[0062] In this embodiment, when a safety assessment of a certain area is required, the geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions are obtained and organized to meet the input format requirements of the machine learning model. Then, the data is input into the machine learning model to obtain crack stability parameters, which are used to guide the treatment decisions of engineering operation and maintenance personnel.

[0063] This embodiment presents a rapid method for determining the stability of tunnel lining cracks in uniform strata. It achieves a closed-loop and highly automated process from detection to simulation to determination, significantly improving the overall efficiency of tunnel lining defect assessment and reducing reliance on manual labor and the risk of subjective misjudgment. High-confidence, refined numerical simulation results are used as training labels for the machine learning model, ensuring the scientific validity and accuracy of the model's learning objectives. The trained machine learning model achieves "second-level" determination, meeting the stringent time requirements of daily tunnel inspections and rapid screening, thus saving valuable time for preventative maintenance. Based on this, a standardized tunnel lining stability assessment process can be established, which can be widely applied to the intelligent operation and maintenance management of various tunnel projects such as railways, highways, and subways, demonstrating broad engineering application prospects.

[0064] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for rapidly determining the stability of tunnel lining cracks under uniform strata, characterized in that, Includes the following steps: Step 1: Obtain the geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions. Step 2: Establish a three-dimensional geometric model of the tunnel lining in the finite element calculation environment, input the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, set various crack condition combinations, and calculate the stress intensity factor at the crack tip. Step 3: Calculate the crack stability parameter f based on the stress intensity factor at the crack tip; Step 4: Using the geometric feature parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions as input features, and the crack stability parameters as output labels, a structured sample database for machine learning training is constructed. Step 5: Set up the structure of the machine learning model; and train the machine learning model using a structured sample database to establish a mapping relationship from crack features to crack stability parameters, so as to achieve rapid determination of lining stability. Step 6: Obtain crack data of the area to be judged, input it into the trained and optimized machine learning model, and quickly judge the stability of the lining.

2. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 1, characterized in that, Step 1 includes the following steps: Step 1.1: Collect images of cracks at different spatial locations on the tunnel lining surface, and obtain the geometric feature parameters of the cracks through image recognition algorithms; the spatial locations include the arch crown, arch shoulder, arch waist, arch foot, and bottom plate; the geometric feature parameters include crack length, crack width, and crack direction; Step 1.2: Obtain the physical and mechanical parameters of the surrounding rock and strata in the tunnel section through geological survey reports, borehole test reports and field test results generated during the tunnel construction or operation phases; the physical and mechanical parameters shall include at least the elastic modulus, Poisson's ratio, internal friction angle and cohesion; Step 1.3: Obtain tunnel operation condition data through the tunnel operation management system or traffic monitoring system, convert the discrete operation condition data into equivalent load parameters for structural analysis, and then superimpose them; the operation condition data includes vehicle type, axle load level, and traffic flow.

3. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 1, characterized in that, Step 2 includes the following steps: Step 2.1: Establish a three-dimensional geometric model of the tunnel lining in the finite element calculation environment, and set the finite element solution method and boundary conditions; the three-dimensional geometric model shall include at least two parts: the lining structure and the surrounding rock calculation domain; Step 2.2: Set up a mechanical constitutive model in the three-dimensional geometric model set in Step 2.1, and input the physical and mechanical parameters of the surrounding rock and strata in the tunnel section; Step 2.3: In the three-dimensional geometric model set in Step 2.1, multiple sets of crack geometric models with crack length, crack depth and crack inclination are preset at the positions of the arch crown, arch shoulder, arch waist, arch foot and base plate to form a variety of crack condition combinations; Step 2.4: finite element calculation is performed to obtain the stress intensity factor of the crack tip under each crack condition combination; the stress intensity factor includes the stress intensity factor K of the I-type crack and the II-type crack I and K II , wherein: the I-type crack is in an opening mode, and the II-type crack is in a tearing mode.

4. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 3, characterized in that, In step 2.1, the lining structure adopts a circumferential structure consistent with the actual tunnel cross-section, and extends outward along the tunnel axis to form a three-dimensional segment model of a preset length, so as to avoid the influence of end constraints on the local stress field at the crack tip. The surrounding rock mass computational domain encloses the lining structure. The outer boundary of the surrounding rock mass maintains a preset distance from the outer edge of the lining, so that the stress disturbance at the outer boundary has a negligible impact on the lining area. A three-dimensional displacement constraint is applied to the bottom surface of the surrounding rock mass: Ux=Uy=Uz=0, to simulate the support of deep strata. A normal displacement constraint is applied to the lateral outer surface of the surrounding rock mass, including limiting normal displacement and allowing tangential displacement, to simulate the lateral constraint of the far-field surrounding rock on the model. The top surface of the surrounding rock mass is set as a free surface or an equivalent vertical load boundary is applied to simulate the action of overlying strata pressure and operational loads. The interaction between the lining and the surrounding rock mass adopts a binding constraint. The equivalent load corresponding to the operating condition information is applied in a multi-point loading manner, which equates the train axle load to a concentrated force at several discrete loading points. The loading points are arranged along the tunnel axis according to the distance between the wheels and the bogies, and are applied at the sleeper support position or the track bed support area, thus forming a multi-point loading condition. For different operating conditions, different train formations and axle load levels are simulated by changing the magnitude of the concentrated force, the number of points, and the spacing between points.

5. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 3, characterized in that, In step 2.1, the finite element method employs a quasi-static analysis step; an implicit solution framework is used, and its governing equations are the following nonlinear equilibrium forms after finite element discretization: R(u) = F ext (u) - F int (u) = 0; In the formula, u is the nodal displacement vector, and F ext (u) is the equivalent nodal force vector of the external load, F int (u) is the internal force vector obtained by integrating the material constitutive stress with the element stress; The Newton-Raphson iterative solution is used, and its iterative update format is as follows: K t (in i )△u i =R(in i ),in i+1 =in i +△u i ; In the formula, K t Let u be the tangent stiffness matrix. i Let Δu be the approximate solution of the displacement vector in the i-th iteration. i This refers to the displacement increment that needs to be compensated under the current unbalanced force.

6. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 3, characterized in that, In step 2.1, the following control strategy is adopted: Geometric nonlinearity is enabled, and automatic incrementation is used for incremental control. The initial increment is 1%–5% of the total step size, the minimum increment is 1e-6–1e-4, and the maximum increment is 5%–20%. The maximum number of increments is set to 200–2000 to avoid premature termination of the nonlinear stage. Force and displacement dual criteria are used to control convergence. When the crack tip region is more sensitive to convergence, residual control is selected. When convergence difficulties are caused by local softening or contact, automatic stabilization is enabled, and the proportion of stabilization dissipation energy to external work is limited to the order of 1e-5–1e-3 to ensure that the physical results are not dominated by numerical damping.

7. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 3, characterized in that, In step 2.4: The extended finite element method is used to calculate the crack tip. XFEM crack description is enabled in the lining for the area where the preset crack is located. The lining domain where the crack is located is defined as the XFEM enrichment zone. A high-density mesh is set in the crack tip region. When only the stress intensity factor under a given crack condition needs to be obtained, the crack is set to not propagate, and only the fracture parameters of the predetermined crack leading edge are calculated. When the crack growth trend needs to be considered, the crack propagation criterion is enabled in XFEM. In step 2.4: A fracture parameter output request is defined for the crack front. A fracture mechanics post-processing method based on J-integral and interaction integral is used to calculate the stress intensity factor at the crack tip. Specifically, an integration domain is established for the crack front, and several integration profiles are set to verify the path independence of the integration results. When the fracture parameter results under multiple profiles converge to a stable value, this stable value is output as the stress intensity factor corresponding to the crack condition. For the three-dimensional crack front, the stress intensity factor distribution is output along multiple sampling points along the crack front, and the maximum value or the value at a representative location is taken as the characteristic stress intensity factor of the crack condition for subsequent safety factor calculation and sample labeling. After the solution is completed, the stress intensity factor K of the type I crack containing the Mode I component is output. I Output the stress intensity factor K of type II cracks, including the Mode II component. II .

8. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 1, characterized in that, In step 3, the crack stability parameter f is calculated using the following formula: ； In the formula, K Ic and K IIc The fracture toughness of type I cracks and type II cracks, respectively, K I and K II The stress intensity factors λ for Type I and Type II cracks, respectively. 12 This is the compressive shear stability coefficient.

9. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 1, characterized in that, Step 4 includes the following steps: Step 4.1: The geometric characteristic parameters of cracks at different spatial locations on the tunnel lining surface, the physical and mechanical parameters of the surrounding rock and strata in the tunnel section, and the equivalent load parameters corresponding to the tunnel operation conditions are uniformly formatted. Continuous parameters are dimensionally unified and numerically normalized, and discrete parameters are encoded to construct a multi-dimensional input feature vector. Step 4.2: Match the crack stability parameters with the working conditions corresponding to the crack geometric feature parameters to form a one-to-one "feature-label" sample pair; Step 4.3: Perform data cleaning and consistency verification on the constructed sample pairs, and remove sample data with missing key feature parameters, obviously abnormal values ​​or inconsistent labels; smooth or correct parameters with noise interference to ensure the integrity, accuracy and consistency of sample data, thereby constructing a high-quality machine learning training sample database. Step 4.4: Organize the processed sample pairs into a structured sample database for subsequent model training and validation.

10. The method for rapid determination of tunnel lining crack stability under uniform strata according to claim 1, characterized in that, In step 5, the machine learning model is an ensemble learning model based on gradient boosting decision trees as the stability determination model; specifically, the XGBoost model structure is adopted, and the XGBoost model parameters are set as follows: n_estimators is 300, learning_rate is 0.1, max_depth is 7, subsample is 0.6, colsample_bytree is 0.8, gamma is 3, and lambda is 6.

Images