A hierarchical early warning method and system for water-soil coupled disasters in tunnels in weak strata based on mechanism inversion

Abstract

This invention relates to the field of tunnel and underground engineering safety monitoring technology, and discloses a method and system for graded early warning of water-soil coupled disasters in soft strata tunnels based on mechanism inversion. The method includes: using multi-scale coupling technology to parametrically couple a pre-constructed macroscopic continuous medium model and a microscopic constitutive model, and generating a mechanism inversion model based on the parameter coupling results to simulate the water-soil coupled disaster response of the tunnel surrounding rock; inputting pre-processed historical monitoring data into the mechanism inversion model, iteratively adjusting the parameters of the mechanism inversion model through inverse analysis to generate a combination of inversion parameters; and inputting pre-processed real-time monitoring data into the corrected mechanism inversion model. This invention, through local mesh refinement and erosion interface mapping based on level set functions, accurately embeds a microscopic constitutive model capable of describing soil particle migration and pore evolution into high-risk areas of the surrounding rock, achieving coupling of macroscopic and microscopic physical processes at key locations.

Description

Technical Field

[0001] This invention relates to the field of tunnel and underground engineering safety monitoring technology, specifically to a method and system for graded early warning of water-soil coupled disasters in soft strata tunnels based on mechanism inversion. Background Technology

[0002] When constructing tunnels in water-rich, soft strata, water-soil coupling is a major cause of surrounding rock instability, water inrush, mudslides, and even collapses. Existing early warning technologies for water-soil coupling disasters in tunnels in soft strata largely rely on the statistical characteristics of monitoring data or single parameter thresholds, which presents two major problems:

[0003] 1. Existing technologies cannot reveal the catastrophic physical processes driven by the coupling of seepage field and stress field, resulting in a lack of mechanism, a lack of predictability in early warning, and frequent false alarms and missed alarms.

[0004] 2. Existing technologies adopt a simple "normal-alarm" binary model, which cannot distinguish the severity and development stage of a hazard, resulting in a lack of targeted control measures, either overreacting or responding too slowly.

[0005] While numerical simulation techniques can reveal the underlying mechanisms, the forward modeling results often deviate significantly from measured values ​​due to uncertainties in geotechnical parameters, making them unsuitable for direct and accurate early warning. Therefore, there is an urgent need in this field for a graded early warning method for water-soil coupled disasters in tunnels in weak strata that can deeply integrate physical mechanisms with real-time data and achieve accurate, tiered early warning.

[0006] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0007] To address the problems in related technologies, this invention proposes a hierarchical early warning method and system for water-soil coupled disasters in soft strata tunnels based on mechanism inversion, in order to overcome the aforementioned technical problems existing in the current related technologies.

[0008] Therefore, the specific technical solution adopted by the present invention is as follows:

[0009] According to one aspect of the present invention, a method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion is provided, the method comprising:

[0010] S1. Collect historical and real-time monitoring data of the target tunnel and preprocess them to obtain preprocessed historical and real-time monitoring data.

[0011] S2. Based on the pre-acquired engineering geological exploration data and combined with the seepage-stress coupling theory, the pre-constructed macroscopic continuous medium model and microscopic constitutive model are parametrically coupled using multi-scale coupling technology, and a mechanism inversion model for simulating the water-soil coupled disaster response of tunnel surrounding rock is generated based on the parametric coupling results.

[0012] S3. Input the preprocessed historical monitoring data into the mechanism inversion model, iteratively adjust the parameters of the mechanism inversion model through the inversion analysis method and generate the inversion parameter combination, identify the key parameter combination from the inversion parameter combination to drive the model correction, and obtain the corrected mechanism inversion model.

[0013] S4. Input the preprocessed real-time monitoring data into the corrected mechanism inversion model to simulate the catastrophic evolution process of the target tunnel surrounding rock in future time periods and output the catastrophic evolution prediction results.

[0014] S5. Calculate early warning indicators based on the disaster evolution prediction results, and classify and issue early warnings for tunnel disaster risks according to the early warning indicators.

[0015] Preferably, the step of parametrically coupling a pre-constructed macroscopic continuous medium model and a microscopic constitutive model using multi-scale coupling technology based on pre-acquired engineering geological exploration data and seepage-stress coupling theory, and generating a mechanism inversion model for simulating the water-soil coupled catastrophic response of tunnel surrounding rock based on the parametric coupling results includes:

[0016] S21. Extract engineering geological survey reports, borehole data and seismic CT detection data from the pre-acquired engineering geological exploration data as input data, and use 3D modeling software to construct a 3D geological entity model containing stratigraphic interfaces, major faults, weak interlayers and the spatial distribution of water-rich areas.

[0017] S22. Further import the tunnel design axis, cross-section and geometric and material parameters of the support structure into the completed three-dimensional geological entity model, and generate a dynamic excavation model reflecting the interaction between the support structure and the surrounding rock through the simulation function of the three-dimensional modeling software.

[0018] S23. Introduce a micro-constitutive model based on erosion effect into the macro-continuous medium model composed of a three-dimensional geological entity model and a dynamic excavation model. Construct a mechanism inversion model for simulating the water-soil coupled disaster response of tunnel surrounding rock through macro-micro bidirectional coupling and linkage.

[0019] Preferably, the introduction of a micro-constitutive model based on erosion effect into the macro-continuous medium model jointly composed of the three-dimensional geological entity model and the dynamic excavation model, and the construction of a mechanism inversion model for simulating the water-soil coupled catastrophic response of tunnel surrounding rock through macro-micro bidirectional coupling and linkage, includes:

[0020] S231. Using the model integration function of 3D modeling software, the parameters of the 3D geological entity model and the dynamic excavation model are calibrated and docked sequentially to form a macroscopic continuous medium model.

[0021] S232. Use mesh densification technology to locally densify the high-risk area of ​​the tunnel surrounding rock in the macroscopic continuous medium model, and embed the pre-constructed micro-constitutive model based on erosion effect into the locally densified area.

[0022] S233. Establish a parameter transfer interface between the macroscopic continuous medium model and the microscopic constitutive model. Through the parameter transfer interface, the macroscopic continuous medium model provides boundary conditions and initial parameters to the microscopic constitutive model, while the microscopic constitutive model simulates soil particle migration and pore structure evolution and outputs macroscopic and microscopic bidirectional transfer parameters in real time.

[0023] S234. Based on the macro-micro bidirectional transfer parameters, the water-soil coupled catastrophic evolution process of the tunnel surrounding rock under the action of seepage-stress coupling was simulated, and finally the mechanism inversion model was constructed.

[0024] Preferably, the step of using mesh refinement technology to locally refine the high-risk area of ​​the tunnel surrounding rock within the macroscopic continuous medium model, and embedding a pre-constructed microscopic constitutive model based on erosion effects into the locally refined area includes:

[0025] S2321. Extract the rectangular computational domain of the high-risk area of ​​the tunnel surrounding rock from the macroscopic continuous medium model, and use a quadtree structure to discretize the rectangular computational domain until the preset mesh size is met, while storing the discretized information in the quadtree database.

[0026] S2322. Select boundary sub-rectangles and internal sub-rectangles that meet the grid densification termination conditions from the quadtree database and convert them into grid cells of the macroscopic continuous medium model and store them in the quadtree database. At the same time, define special geological points in high-risk areas of tunnel surrounding rock as grid cell nodes.

[0027] S2323. Calculate the calculation error of each grid cell using the posterior error estimation method, and select grid cells whose calculation error exceeds the preset threshold as grid cells that need to be refined and corrected.

[0028] S2324. Based on the sub-rectangle pointer information corresponding to the grid cell that needs to be encrypted and corrected, the corresponding source sub-rectangle block is matched in the quadtree database, the source sub-rectangle block and its adjacent sub-rectangle blocks that need to be encrypted are encrypted and corrected, and an invalid identifier is added to the corrected sub-rectangle block in the quadtree database.

[0029] S2325. Regenerate new mesh cells based on the encrypted and corrected sub-rectangular blocks, and embed the pre-built micro-constitutive model based on the erosion effect into the corresponding new mesh cells using mesh matching technology.

[0030] Preferably, the step of regenerating new mesh cells based on the encrypted and corrected sub-rectangular blocks, and embedding the pre-built micro-constitutive model based on the erosion effect into the corresponding new mesh cells using mesh matching technology includes:

[0031] S23251. Use the zero isosurface of the level set function to set the erosion interface of the micro-constitutive model, establish a spatial coordinate mapping between the erosion interface and the new grid cell, and obtain the corresponding level set function value at each node of the new grid cell.

[0032] S23252. The material parameters at the erosion interface determined by the micro-constitutive model are used as initial values. Combined with the velocity field related to the erosion physical mechanism, the initial values ​​are propagated to the new grid cell nodes as wavefront information, and each node is assigned a material property value.

[0033] S23253. Based on the material property values ​​obtained from all nodes on the new mesh element, interpolation integration is performed using the element shape function to calculate the equivalent material parameter tensor at the numerical integration point; the level set function value controlling the erosion interface is updated based on the equivalent material parameter tensor.

[0034] Preferably, the zero isosurface of the level set function is the zero isosurface of the level set function φ=0;

[0035] Propagating the initial value as wavefront information to all new grid cell nodes includes propagating the initial value as wavefront information from the erosion interface to all new grid nodes in both the φ>0 and φ<0 directions.

[0036] Preferably, the step of inputting preprocessed historical monitoring data into the mechanism inversion model, iteratively adjusting the mechanism inversion model parameters through inverse analysis and generating inversion parameter combinations, identifying key parameter combinations from the inversion parameter combinations to drive model correction, and obtaining the corrected mechanism inversion model includes:

[0037] S31. Extract the measured inversion period data and measured verification period data for a preset time period from the preprocessed historical monitoring data respectively;

[0038] S32. Input the preprocessed historical monitoring data into the mechanism inversion model to calculate the inversion period data, and build an objective function based on the mean square error between the inversion period data and the measured inversion period data.

[0039] S33. Set the maximum number of iterations, and in each iteration, adjust the model parameters of the mechanism inversion model by minimizing the objective function to obtain the inversion parameter combination;

[0040] S34. Identify key parameter combinations from the inversion parameter combinations, use the key parameter combinations to drive the mechanism inversion model to perform simulation prediction for a preset time period to obtain prediction verification period data, and compare the prediction verification period data with the measured verification period data.

[0041] S35. If the coefficient of determination between the predicted verification period data and the measured verification period data is greater than the preset verification threshold or the current iteration reaches the maximum number of iterations, the mechanism inversion model correction is completed and the corrected mechanism inversion model is obtained; otherwise, the next round of iteration is carried out until the coefficient of determination is greater than the preset verification threshold or the maximum number of iterations is reached, and the corrected mechanism inversion model is output.

[0042] Preferably, identifying key parameter combinations from the inversion parameter combinations includes:

[0043] Based on the combination of inversion parameters and their range of values ​​generated in each iteration, two sets of reference sample sets are generated by Latin hypercube sampling and are respectively denoted as the first reference sample set and the second reference sample set.

[0044] The numerical column of each inversion parameter in the second reference sample set is replaced with the corresponding column in the first basic sample set, while keeping the values ​​of other inversion parameters unchanged, to generate a perturbation sample set for the current inversion parameter.

[0045] Input all generated first baseline sample sets, second baseline sample sets, and perturbation sample sets into the mechanism inversion model to obtain the corresponding model output result sets;

[0046] Estimate the overall variability of the model output, and calculate the first-order sensitivity index of the impact of the change of each inversion parameter on the model output, as well as the total sensitivity index of the combined impact of the interaction between the inversion parameter and other inversion parameters;

[0047] The key parameter combinations that play a dominant role in the output of the mechanism inversion model are selected by combining the total sensitivity index and the first-order sensitivity index.

[0048] Preferably, the combination of key parameters includes elastic modulus, cohesion, internal friction angle, and permeability coefficient.

[0049] According to another aspect of the present invention, a graded early warning system for water-soil coupled disasters in tunnels in weak strata based on mechanism inversion is also provided, the system comprising:

[0050] The data acquisition module is used to collect historical and real-time monitoring data of the target tunnel and preprocess them to obtain preprocessed historical and real-time monitoring data.

[0051] The model building module is used to couple the pre-built macroscopic continuous medium model and microscopic constitutive model with parameters based on the pre-acquired engineering geological exploration data and the seepage-stress coupling theory, and generate a mechanism inversion model for simulating the water-soil coupled disaster response of the tunnel surrounding rock based on the parameter coupling results.

[0052] The model calibration module is used to input preprocessed historical monitoring data into the mechanism inversion model, iteratively adjust the parameters of the mechanism inversion model through inverse analysis and generate inversion parameter combinations, identify key parameter combinations from the inversion parameter combinations to drive model calibration, and obtain the calibrated mechanism inversion model.

[0053] The evolution module is used to input the preprocessed real-time monitoring data into the corrected mechanism inversion model, simulate the catastrophic evolution process of the target tunnel surrounding rock in future time periods, and output the catastrophic evolution prediction results.

[0054] The early warning module is used to calculate early warning indicators based on the disaster evolution prediction results, and to classify and issue early warnings for tunnel disaster risks according to the early warning indicators.

[0055] According to a third aspect of the present invention, a computer device is provided, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the above-described method.

[0056] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0057] The beneficial effects of this invention are as follows:

[0058] 1. This invention upgrades the existing data-driven early warning technology to a hybrid-driven early warning technology based on physical mechanisms and data, giving the early warning results a solid physical basis and significantly improving accuracy and reliability. By performing advanced simulation on the corrected mechanism inversion model, it achieves a fundamental shift from post-event alarm to pre-event early warning, gaining valuable time for risk prevention and control. Furthermore, the hierarchical early warning system makes safety management decisions more targeted and hierarchical, avoiding resource waste or insufficient response, and promoting the leap from experience-based decision-making to scientific decision-making in tunnel safety management.

[0059] 2. This invention constructs a refined three-dimensional geological entity model containing strata, faults, and water-rich areas based on real exploration data, and incorporates dynamic excavation and support to ensure the engineering authenticity of the macroscopic model. Furthermore, through local mesh refinement and erosion interface mapping based on level set functions, a microscopic constitutive model that can describe soil particle migration and pore evolution is accurately embedded into the high-risk area of ​​the surrounding rock, realizing the coupling of macroscopic and microscopic physical processes in key parts. Finally, through the parameter transfer interface, multi-field bidirectional linkage of stress, seepage, and erosion damage is realized, enabling the inversion model to not only reflect the macroscopic mechanical response, but also reveal the microscopic mechanism of seepage erosion leading to material degradation and thus inducing disaster.

[0060] 3. By quantifying the first-order and total sensitivity indices of each parameter, this invention can distinguish the primary and secondary effects of the parameters, thereby focusing computational and correction resources on the key mechanisms that dominate the disaster response. This not only significantly improves the efficiency and stability of model parameter inversion and correction, avoiding the dimensionality curse and local oscillations of global optimization, but also ensures the consistency between the core physical mechanism of the corrected model and engineering reality by removing secondary interferences. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart illustrating a method for graded early warning of water-soil coupled disasters in soft strata tunnels based on mechanism inversion according to an embodiment of the present invention.

[0063] Figure 2 This is a flowchart illustrating the process of constructing a mechanism inversion model in a water-soil coupled disaster classification and early warning method for tunnels in weak strata based on mechanism inversion according to an embodiment of the present invention.

[0064] Figure 3 This is a flowchart illustrating the verification of the mechanism inversion model in a water-soil coupled disaster classification and early warning method for tunnels in weak strata based on mechanism inversion according to an embodiment of the present invention.

[0065] Figure 4 This is a schematic diagram of a graded early warning system for water-soil coupling disasters in soft strata tunnels based on mechanism inversion, according to an embodiment of the present invention.

[0066] Figure 5 This is a schematic diagram of the structure of a computer device.

[0067] In the picture:

[0068] 1. Data acquisition module; 2. Model building module; 3. Model calibration module; 4. Evolution module; 5. Early warning module. Detailed Implementation

[0069] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0070] According to an embodiment of the present invention, a method and system for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion is provided.

[0071] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown in the embodiment of the present invention, a mechanistic inversion-based method for graded early warning of water-soil coupled disasters in tunnels in weak strata includes:

[0072] S1. Collect historical and real-time monitoring data of the target tunnel and preprocess them to obtain preprocessed historical and real-time monitoring data.

[0073] Specifically, historical and real-time monitoring data for the first 15 days after the excavation of the target tunnel are collected and preprocessed to obtain preprocessed historical and real-time monitoring data. The monitoring data includes data on crown settlement, horizontal convergence, arch foot stress, pore water pressure, and cracks. The preprocessing includes outlier identification and removal, and spatiotemporal alignment of data.

[0074] S2. Based on the pre-acquired engineering geological exploration data and combined with the seepage-stress coupling theory, the pre-constructed macroscopic continuous medium model and microscopic constitutive model are parametrically coupled using multi-scale coupling technology, and a mechanism inversion model for simulating the water-soil coupled disaster response of the tunnel surrounding rock is generated based on the parametric coupling results.

[0075] Among them, such as Figure 2 As shown, the process of parametrically coupling a pre-constructed macroscopic continuous medium model and a microscopic constitutive model using multi-scale coupling technology based on pre-acquired engineering geological exploration data and seepage-stress coupling theory, and generating a mechanism inversion model for simulating the water-soil coupled catastrophic response of tunnel surrounding rock based on the parametric coupling results includes:

[0076] S21. Extract engineering geological survey reports, borehole data and seismic CT detection data from the pre-acquired engineering geological exploration data as input data, and use 3D modeling software to construct a 3D geological entity model containing stratigraphic interfaces, major faults, weak interlayers and the spatial distribution of water-rich areas.

[0077] S22. Further import the tunnel design axis, cross-section and geometric and material parameters of the support structure into the completed three-dimensional geological entity model, and generate a dynamic excavation model reflecting the interaction between the support structure and the surrounding rock through the simulation function of the three-dimensional modeling software.

[0078] Specifically, the 3D modeling software is FLAC3D. In FLAC3D, a 3D geological solid model is generated. First, the engineering geological survey report, borehole data, and seismic CT detection data are converted into a format that the software can recognize. Through its built-in Griddle plugin, based on the spatial distribution data of stratigraphic interfaces, faults, weak interlayers, and water-rich areas, an accurate geological surface is established and a 3D solid mesh model is generated. At the same time, the wave velocity field obtained by inverting the seismic CT data is used to assign the initial rock mass mechanical parameter partitions to the model.

[0079] Based on this three-dimensional geological entity model, the tunnel design axis and cross-sectional geometry are imported through programming or built-in commands. Using FLAC3D's mesh reconstruction function, mesh cells in the excavation area are gradually deleted (model null) along the tunnel axis, and specific structural cells representing the support structure are activated (model mech) on the excavation boundary in real time. By setting the step-by-step excavation sequence and the timing of support, a dynamic excavation model that can dynamically simulate the stress release of the surrounding rock, the stress on the support structure, and the interaction process can be generated.

[0080] S23. Introduce a micro-constitutive model based on erosion effect into the macro-continuous medium model composed of a three-dimensional geological entity model and a dynamic excavation model. Construct a mechanism inversion model for simulating the water-soil coupled disaster response of tunnel surrounding rock through macro-micro bidirectional coupling and linkage.

[0081] Specifically, the core objective of step S23 is to overcome the fundamental limitations of traditional macroscopic continuous medium models in simulating water-soil coupled disasters. Traditional macroscopic models based on seepage-stress coupling theory treat the soil and rock mass as a homogeneous continuous medium in their governing equations (which are based on the effective stress principle to describe the stress field and Darcy's law to describe the seepage field). They can only simulate the interaction between stress and pore water pressure, but cannot describe or predict the key physical process of water flow eroding and migrating soil particles during seepage, which leads to damage to the microstructure and deterioration of macroscopic properties of the material. In contrast, water-soil coupled disasters in tunnels in weak strata are often caused by seepage erosion leading to particle loss, increased porosity, and softening of material strength, which in turn induces disasters such as water inrush and mudslides.

[0082] Therefore, the introduction of a micro-constitutive model based on erosion effect is precisely to realistically depict the initiation and migration of soil particles under seepage in key areas of the macro model, such as the excavation face and the inside of fault fracture zones, as well as the dynamic evolution of macroscopic parameters such as porosity, permeability, and strength. This is equivalent to embedding a digital microscope that can reveal the micro-mechanism of catastrophe in the macro-continuous medium framework.

[0083] The mechanism inversion model constructed in this invention still uses the effective stress principle and Darcy's law as its core governing equation framework. However, the key material parameters in the equations (such as the permeability tensor and strength parameters) are no longer fixed values ​​or simple empirical functions, but variables that are dynamically evolved with the erosion state, provided by the microscopic constitutive model in real time. This allows the model to simulate not only the stress-seepage interaction, but also the entire catastrophic evolution process of seepage-erosion-material degradation-stress redistribution, thus achieving a leap from describing the state to predicting the process and catastrophic events. The core equation set, based on Biot's consolidation theory and the modified Darcy's law, is as follows:

[0084] (1) Equilibrium equations (considering the effective stress principle):

[0085] ;

[0086] In the formula, ▽ represents the gradient operator; D is the elasticity matrix; This is the total strain tensor; α is the plastic strain tensor; α is the Biot coefficient (reflecting the contribution of pore water pressure to the total stress); p is the pore water pressure; F is the volume force vector.

[0087] (2) Seepage continuity equation (considering water-soil coupling effect):

[0088] ;

[0089] In the formula, u is the displacement vector; t is time; k is the permeability coefficient tensor (this is a key inversion parameter and can be expressed as a function of porosity and particle size distribution in the multi-scale sub-model); γ w ρ is the specific weight of water. w Let be the density of water; g be the gravitational acceleration vector; and Q be the source and sink terms.

[0090] (3) Elastoplastic constitutive equation (using a model that can reflect the rheological and dilatational characteristics of soft rock), with modification; κ is the plastic internal variable, σ is the stress tensor, p is the mean stress, where the yield function is F(σ,p,κ)=0; the plastic potential function is G(σ,p,κ).

[0091] Among them, the Mohr-Coulomb criterion is a linear strength theory based on the internal friction angle and cohesion of materials, and is widely used in judging the shear failure of soils and weak rocks. The Hoek-Brown criterion is a nonlinear empirical strength criterion based on the rock mass quality index GSI and the rock material constant mi, used to describe the failure behavior of jointed rock masses.

[0092] Specifically, the mechanism inversion model in this invention still uses the effective stress principle and Darcy's law as its core governing equation framework. However, the key material parameters in the equations are no longer fixed values ​​or simple empirical functions, but variables that are dynamically evolved with the erosion state, provided by the micro-constitutive model in real time. This step itself does not directly perform dynamic transformation calculations of parameters. By refining the local mesh in high-risk areas, smaller and more precise mesh cells are generated at specific locations in the macro model. These refined mesh cells are the pre-reserved embedding positions for the micro-constitutive model, which enables the micro model to run in a local area with higher spatial resolution that is seamlessly connected to the macro model, thereby accurately simulating the particle erosion process in that local area.

[0093] Subsequent embedding steps, such as defining the erosion interface using level set functions and performing spatial coordinate mapping, all rely on this refined mesh. The results of the micro-model, such as "real-time changing porosity and particle contact state," calculated within the refined mesh cells, are interpolated, averaged, or homogenized through the nodes and cells of these refined meshes, which then calculate the equivalent permeability tensor (k) and equivalent intensity parameter H (κ) of that local region at the macroscopic scale.

[0094] The process of introducing a micro-constitutive model based on erosion effects into the macro-continuous medium model composed of a three-dimensional geological entity model and a dynamic excavation model, and constructing a mechanism inversion model for simulating the water-soil coupled catastrophic response of tunnel surrounding rock through macro-micro bidirectional coupling, includes:

[0095] S231. Using the model integration function of 3D modeling software, the parameters of the 3D geological entity model and the dynamic excavation model are calibrated and connected sequentially to form a macroscopic continuous medium model.

[0096] Specifically, integrating a 3D geological entity model with a dynamic excavation model to form a macroscopic continuous medium model in FLAC3D software is achieved through a systematic modeling process. First, the established 3D geological entity model is imported into FLAC3D, and its built-in Griddle plugin or mesh generator is used to convert it into a 3D differential mesh that the software can recognize. At the same time, the mesh is grouped according to geological zones such as strata and faults, and initial material models and corresponding geomechanical parameters are assigned.

[0097] Subsequently, the design model, including the tunnel axis, cross-section, and support structure, was spatially aligned with the geological grid. Using FLAC3D's modeling language or command flow, the `zone null` command was used to delete grid cells within the excavation area according to the construction sequence to simulate the tunneling process. Simultaneously, support structure units such as lining and anchor bolts were activated in real-time at the excavation boundary using the `structure` command. Mechanical interaction parameters were defined for the contact surface between the support and the surrounding rock to simulate their coordinated stress. Based on this, the mechanical and seepage parameters of the surrounding rock and support were finely calibrated using the `zone property` command, and initial geostress fields, displacement boundary conditions, and pore water pressure boundaries were applied. Finally, initial equilibrium calculations were run to bring the integrated model to mechanical and seepage stability, thus forming a parameterized macroscopic continuous medium model that realistically reflects the stratum distribution, construction process, and support function.

[0098] S232. Use mesh densification technology to locally densify the high-risk area of ​​the tunnel surrounding rock in the macroscopic continuous medium model, and embed the pre-constructed micro-constitutive model based on erosion effect into the locally densified area.

[0099] Specifically, the micro-constitutive model based on erosion effects refers to the existing technology that uses the discrete element method (DEM) or a CFD-DEM coupled model that considers fluid action. Its architecture treats the soil as a collection of a large number of movable discrete particles. The particles interact with each other through contact forces, such as elastic force, friction force, and cohesion force. The fluid phase is considered by solving the Navier-Stokes equations. Its operating principle is to calculate the motion of each particle under the action of the net external force according to Newton's second law, and update the force by detecting particle contact in real time. When the fluid drag force exceeds the cohesion force or static friction force between particles, the particles initiate and migrate, thereby simulating the micro-erosion process and causing the dynamic evolution of micro-structural parameters such as porosity. The parameter selection mainly includes particle parameters, such as particle size distribution, density, stiffness, friction coefficient, bond strength, and contact model parameters, which can be calibrated through geotechnical tests.

[0100] The core of this invention's step of setting up local mesh refinement and embedding microscopic models lies in the creative construction of a locally embedded multi-scale coupling architecture, which solves the fundamental limitations of traditional single-scale models: traditional macroscopic continuous models treat parameters such as permeability coefficient and strength as fixed values ​​or empirical degradation functions, and cannot physically and mechanistically reflect the process of seepage-driven particle migration leading to dynamic deterioration of material parameters; while microscopic discrete element models have huge computational requirements and cannot be directly used for engineering-scale simulation.

[0101] This invention achieves real-time calculation of microstructure evolution based on actual particle migration physics in key areas by locally refining the mesh in high-risk regions of a macroscopic model and embedding a microscopic model. It also dynamically updates the permeability coefficient and intensity parameters in the macroscopic model, thus preserving the physical mechanism of the erosion process while avoiding the computational cost of full-scale microscopic simulation. This elevates the model from a static description of catastrophic states to a dynamic prediction of the seepage-erosion-deterioration chain process, significantly improving the mechanistic and reliable nature of early warning. Specifically, the method of locally refining the mesh in high-risk areas of tunnel surrounding rock within a macroscopic continuous medium model and embedding a pre-constructed microscopic constitutive model based on erosion effects into the locally refined region includes:

[0102] S2321. Extract the rectangular computational domain of the high-risk area of ​​the tunnel surrounding rock from the macroscopic continuous medium model, and use a quadtree structure to discretize the rectangular computational domain until the preset mesh size is met. At the same time, store the discretized information in the quadtree database.

[0103] Specifically, the outer rectangle of the high-risk area of ​​the tunnel surrounding rock is taken as the root node, and then it is recursively divided into four parts to generate four child nodes. Each child node represents a smaller sub-rectangle. This division process continues until the size of all sub-rectangles reaches the preset fine mesh size. At this point, each leaf node corresponds to a final discrete unit. The hierarchical structure formed by the entire division process, such as which parent rectangle the sub-rectangle is divided from, its size and position information, is efficiently stored in the quadtree database.

[0104] The quadtree structure discretizes a rectangular computational domain by recursively dividing it into four sub-rectangles until each sub-rectangle reaches a preset grid size. This enables fine-grained discretization of high-risk areas. Through the recursive subdivision capability of the quadtree, the extracted high-risk rectangular computational domain can be quickly discretized into a series of regular or irregular grid cells that reach a preset size, thus meeting the high spatial resolution requirements of micro-model embedding. Secondly, a hierarchical grid data structure is established. The naturally formed tree-like hierarchical structure of the quadtree, including the root node, internal nodes, and leaf nodes, clearly stores the refinement process of the grid from coarse to fine, the parent-child containment relationship between cells, and the geometric information of each cell. This greatly facilitates efficient spatial querying, grid attribute management, and subsequent calculations.

[0105] In practical applications, the quadtree structure acts as a mesh lifecycle manager: it not only automatically terminates subdivision to generate a refined mesh based on preset dimensions, but more importantly, its tree-like database provides a unique identifier for each mesh cell, enabling subsequent steps to quickly locate and manipulate specific cells. For example, when a cell is identified as needing correction, the program can directly trace its root and adjacent nodes through the quadtree node pointers recorded for that cell, achieving precise subdivision of the local area without manipulating the global mesh. This process is clearly traceable. The output of the quadtree is not an unparseable set of meshes, but a data structure with complete hierarchical metadata, which provides a spatial mapping basis for data docking between macro-mesh and micro-model.

[0106] S2322. Select boundary sub-rectangles and internal sub-rectangles that meet the grid densification termination conditions from the quadtree database and convert them into grid cells of the macroscopic continuous medium model and store them in the quadtree database. At the same time, define special geological points in high-risk areas of tunnel surrounding rock as grid cell nodes.

[0107] S2323. Calculate the computation error of each grid cell using the posterior error estimation method, and select grid cells whose computation error exceeds a preset threshold as grid cells that need to be refined and corrected.

[0108] Specifically, the posterior error estimation method is a mathematical method that quantitatively evaluates the error between the approximate solution and the true solution based on the obtained solution after numerical calculation. It solves the governing equations on the grid of an existing macroscopic continuous medium model to obtain a preliminary numerical solution. Then, based on the numerical solution, the posterior error estimator is applied to calculate the error index of each grid cell. The error value of each cell is compared with a preset accuracy threshold, and all grid cells with error values ​​exceeding the threshold are screened out. This is an existing technology and will not be elaborated on further here.

[0109] S2324. Based on the sub-rectangle pointer information corresponding to the grid cell that needs to be encrypted and corrected, the corresponding source sub-rectangle block is matched in the quadtree database, the source sub-rectangle block and its adjacent sub-rectangle blocks that need to be encrypted are encrypted and corrected, and an invalid identifier is added to the corrected sub-rectangle block in the quadtree database.

[0110] Specifically, the source region that needs to be corrected is precisely located in the quadtree database by using the subrectangle pointer, and it and the adjacent regions to be encrypted are refined in a coordinated manner. This ensures a smooth transition between the encrypted region and the surrounding coarse grid, avoiding numerical discontinuities. At the same time, an invalidation mark is added to the corrected block, which is essentially marking the end of the life cycle of these old grids in the quadtree structure, so that the data version can be clearly distinguished when generating new grids in the future.

[0111] S2325. Regenerate new mesh cells based on the encrypted and corrected sub-rectangular blocks, and embed the pre-built micro-constitutive model based on the erosion effect into the corresponding new mesh cells using mesh matching technology.

[0112] Specifically, by using quadtrees to discretize the rectangular computational domain and setting custom nodes based on geological points, the grid foundation of the encrypted area is ensured to be both regularly generated and conform to complex geological features, providing a structured spatial carrier for embedding micro-models. The introduction of posterior error estimation transforms the encryption behavior from a preset, static process to a dynamic, adaptive process based on actual computational accuracy feedback. Encryption correction is only triggered when the computational error of the grid cell exceeds a threshold, avoiding computational waste caused by uniform encryption and achieving optimized spatial allocation of computational resources. Finally, the entire process uses a quadtree database as its core hub, with each operation precisely recorded and associated in the tree structure through pointers and identifiers. This makes the generation and correction history of the macro-grid completely traceable, and ultimately, through grid matching technology, the micro-model is seamlessly and accurately embedded into these new, accuracy-verified grid cells.

[0113] Taking a high-risk area of ​​the surrounding rock of a tunnel (coordinate range X: 100-200m, Y: 50-150m, Z: -30-0m) as the research object, the preset mesh size is 0.5m, the refinement termination condition is that the side length of the sub-rectangle is ≤0.1m, the preset threshold of calculation error is 5%, and the micro-constitutive model based on the erosion effect adopts the particle flow PFC3D model, with particle size of 0.005-0.02m and bonding strength of 0.8MPa. During implementation, the rectangular computational domain (100×100×30m) of the high-risk area is first extracted, and it is discretized using a quadtree structure. The initial side length of the discretized sub-rectangle is 2m, and after 4 splitting operations, the preset mesh size of 0.5m is reached. The discretized information (sub-rectangle coordinates, side length, and level) is stored in the quadtree database.

[0114] 120 boundary sub-rectangles and 850 internal sub-rectangles were selected from the database. The selected sub-rectangles met the encryption termination conditions and were converted into macroscopic model grid units and stored. At the same time, the fault intersection point (X: 150m, Y: 100m, Z: -15m) and the water-rich core point (X: 145m, Y: 95m, Z: -12m) were defined as grid unit nodes.

[0115] The seepage stress coupling calculation error of each grid cell was calculated using a posterior error estimation method. Among them, 68 grid cells had calculation errors exceeding the 5% threshold (maximum error 8.2%) and were identified as cells requiring densification and correction. Based on the sub-rectangle pointer information of these cells (e.g., pointer address 0x0012 corresponds to sub-rectangle X: 140-145m, Y: 90-95m, Z: -10-15m), source sub-rectangle blocks were matched in the quadtree database. The source block and the adjacent 12 sub-rectangle blocks requiring densification were split to a side length of 0.1m, and split twice more. An invalid identifier was added to the original 0.5m side length sub-rectangle blocks. New grid cells were generated based on the corrected sub-rectangle blocks. The pre-constructed PFC3D micro-constitutive model was embedded into the corresponding new grid cells using grid coordinate matching technology.

[0116] The step of regenerating new mesh cells based on the encrypted and corrected sub-rectangular blocks, and embedding the pre-built micro-constitutive model based on the erosion effect into the corresponding new mesh cells using mesh matching technology includes:

[0117] S23251. Use the zero isosurface of the level set function to set the erosion interface of the micro constitutive model, establish a spatial coordinate mapping between the erosion interface and the new grid cell, and obtain the corresponding level set function value at each node of the new grid cell.

[0118] Wherein, the zero isosurface of the level set function is the zero isosurface of the level set function φ=0; propagating the initial value as wavefront information to all new grid cell nodes includes propagating the initial value as wavefront information from the erosion interface to all new grid cells in two directions, φ>0 and φ<0.

[0119] Specifically, the erosion interface of the micro-constitutive model is precisely defined by the zero isosurface of the level set function. The erosion interface is bound to the newly densified grid cells by spatial coordinate mapping. The level set function value is assigned to all nodes by the bidirectional propagation of wavefront information. This allows for a dynamic and accurate representation of the spatial location and evolution of the erosion interface in high-risk areas. The zero isosurface of the level set function can flexibly characterize the irregular erosion interface morphology, avoiding the limitations of traditional geometric modeling. Spatial coordinate mapping achieves precise matching between the erosion interface and the densified grid cells. The bidirectional propagation of wavefront information ensures that all grid nodes can obtain the function value.

[0120] S23252. The material parameters at the erosion interface, determined by the microscopic constitutive model, are used as initial values. Combined with the velocity field related to the erosion physical mechanism, the initial values ​​are propagated as wavefront information to the nodes of the new grid cell. Material property values ​​are assigned to each node. Specifically, the initial value of the erosion interface (e.g., the initial state of the erosion interface corresponding to the level set function φ=0) is determined as the wavefront information, with the zero isosurface of the level set function as the propagation starting point. Then, combined with the velocity field parameters, the wavefront information is gradually propagated to all nodes of the new grid cell in two directions: φ>0 (uneroded area) and φ<0 (eroded area).

[0121] S23253. Based on the material property values ​​obtained from all nodes on the new mesh element, interpolation integration is performed using the element shape function to calculate the equivalent material parameter tensor at the numerical integration point; the level set function value controlling the erosion interface is updated based on the equivalent material parameter tensor.

[0122] Specifically, the element shape function is an interpolation function defined within a grid cell to describe the variation of physical quantities at any point within the cell with nodal values. It satisfies the characteristic that the value is 1 at nodes and 0 at non-nodes, enabling continuous interpolation of physical quantities within the cell. The purpose of this step is to perform interpolation integration on the nodal material property values ​​using the element shape function, transforming the material properties of discrete nodes into continuous equivalent material parameter tensors at numerical integration points. This accurately characterizes the spatial distribution characteristics of material properties within the refined grid cell. Then, the level set function value is updated based on the equivalent material parameter tensor, realizing the dynamic coupling between the evolution of the erosion interface and the changes in material parameters, thereby improving the accuracy of the microscopic constitutive model in numerical simulation of the erosion process.

[0123] S233. Establish a parameter transfer interface between the macroscopic continuous medium model and the microscopic constitutive model. Through the parameter transfer interface, the macroscopic continuous medium model provides boundary conditions and initial parameters to the microscopic constitutive model, while the microscopic constitutive model simulates soil particle migration and pore structure evolution and outputs macroscopic and microscopic bidirectional transfer parameters in real time.

[0124] S234. Based on the macro-micro bidirectional transfer parameters, the water-soil coupled catastrophic evolution process of the tunnel surrounding rock under the action of seepage-stress coupling was simulated, and finally the mechanism inversion model was constructed.

[0125] Specifically, this step essentially involves establishing a bidirectional, real-time feedback coupled computational loop. The following is a simplified example illustrating how to build and run a mechanism inversion model based on this interface:

[0126] At the initial moment, the macro model sets the initial stress field and seepage field based on geological exploration data, and transmits the current stress, pore water pressure and hydraulic gradient of the high-risk densified grid area as boundary conditions to the embedded micro model through the parameter transfer interface.

[0127] The micro-model is activated under these boundary conditions to simulate the activation, migration, and deposition of soil particles in this local area under the action of seepage force. It statistically analyzes and outputs the micro-structural parameters that evolve due to particle loss in real time. These micro-parameters are instantly fed back to the macro-model through the same interface and used to dynamically update the macro-equivalent material parameters of the corresponding fine grid cells. For example, the permeability tensor is updated according to the porosity evolution, and the cohesion and internal friction angle are updated according to the coordination number and contact force network.

[0128] The macroscopic model performs a new round of stress-seepage coupling calculations based on the updated material parameters, obtaining new stress and pore water pressure distributions, which again serve as new boundary conditions to drive the microscopic erosion simulation. Thus, the steps of overall macroscopic stress / seepage field - microscopic particle migration - microstructure evolution - macroscopic material parameter update - macroscopic field redistribution form a closed, time-step-progressing dynamic feedback loop.

[0129] S3. Input the preprocessed historical monitoring data into the mechanism inversion model, iteratively adjust the parameters of the mechanism inversion model through the inversion analysis method and generate the inversion parameter combination, identify the key parameter combination from the inversion parameter combination to drive the model correction, and obtain the corrected mechanism inversion model.

[0130] Among them, such as Figure 3 As shown, the process involves inputting preprocessed historical monitoring data into the mechanism inversion model, iteratively adjusting the model parameters using inverse analysis, generating inversion parameter combinations, identifying key parameter combinations from these combinations to drive model correction, and obtaining the corrected mechanism inversion model, including:

[0131] S31. Extract the measured inversion period data and measured verification period data for a preset time period from the preprocessed historical monitoring data respectively;

[0132] S32. Input the preprocessed historical monitoring data into the mechanism inversion model to calculate the inversion period data, and build an objective function based on the mean square error between the inversion period data and the measured inversion period data.

[0133] S33. Set the maximum number of iterations, and in each iteration, adjust the model parameters of the mechanism inversion model by minimizing the objective function to obtain the inversion parameter combination;

[0134] S34. Identify key parameter combinations from the inversion parameter combinations, use the key parameter combinations to drive the mechanism inversion model to perform simulation prediction for a preset time period to obtain prediction verification period data, and compare the prediction verification period data with the measured verification period data.

[0135] The identification of key parameter combinations from the inversion parameter combinations includes:

[0136] Based on the combination of inversion parameters and their range of values ​​generated in each iteration, two sets of reference sample sets are generated by Latin hypercube sampling and are respectively denoted as the first reference sample set and the second reference sample set.

[0137] The numerical column of each inversion parameter in the second reference sample set is replaced with the corresponding column in the first basic sample set, while keeping the values ​​of other inversion parameters unchanged, to generate a perturbation sample set for the current inversion parameter.

[0138] Input all generated first baseline sample sets, second baseline sample sets, and perturbation sample sets into the mechanism inversion model to obtain the corresponding model output result sets;

[0139] Estimate the overall variability of the model output, and calculate the first-order sensitivity index of the influence of the change of each inversion parameter on the model output, as well as the total sensitivity index of the combined influence of the interaction between the inversion parameter and other inversion parameters.

[0140] Specifically, the overall variability of the model output is estimated by variance decomposition. This involves using the probability distribution of the model input inversion parameters as a basis, generating multiple sets of parameter combinations using Monte Carlo sampling or Sobol sampling, substituting them into the model to obtain multiple sets of output results, and then calculating the total variance of all output results. This total variance is the overall variability of the model output. Subsequently, based on the total variance, further normalization is performed to obtain the first-order sensitivity index of each parameter and the total sensitivity index.

[0141] The total sensitivity index measures the proportion of the total variance of the model output caused by the change of a parameter itself and its interaction with all other parameters. The first-order sensitivity index measures only the proportion of variance contributed by the independent change of the parameter itself, specifically including:

[0142] Calculate the overall sensitivity index of all parameters. Parameters with extremely low index values, such as 0.01, are considered negligible and excluded. Among the remaining parameters, prioritize sorting based on the overall sensitivity index, as parameters with higher overall indices have a greater impact on the output. Simultaneously, refer to the first-order sensitivity index. If a parameter has a high overall index but a very low first-order index, it indicates that its influence mainly comes from complex interactions with other parameters and it still needs to be considered a key parameter. Finally, select a set of parameters whose overall sensitivity index exceeds a preset threshold and which occupy a core position in the parameter interaction network as the key parameter combination.

[0143] The key parameter combinations that play a dominant role in the output of the mechanism inversion model are selected by combining the total sensitivity index and the first-order sensitivity index.

[0144] Specifically, this invention achieves efficient dimensionality reduction and core mechanism identification for complex mechanism inversion models. Through systematic global sensitivity analysis, it accurately identifies the key parameters and their interactions that play a dominant role in the model output from numerous uncertain parameters. This greatly narrows the research focus of subsequent model calibration, inversion optimization, and uncertainty quantification, avoiding resource consumption on secondary parameters, thereby significantly improving analysis efficiency and making the understanding and explanation of the dominant mechanisms of complex processes such as seepage-stress-erosion coupling catastrophe clearer and more reliable.

[0145] S35. If the coefficient of determination between the predicted verification period data and the measured verification period data is greater than the preset verification threshold or the current iteration reaches the maximum number of iterations, the mechanism inversion model correction is completed and the corrected mechanism inversion model is obtained; otherwise, the next round of iteration is carried out until the coefficient of determination is greater than the preset verification threshold or the maximum number of iterations is reached, and the corrected mechanism inversion model is output.

[0146] Specifically, the principle of step S3 in practical application will be explained in further detail below.

[0147] The data from the first 10 days of the preprocessed historical monitoring data were used as the inversion period data for actual measurement, and the data from the last 5 days were used as the verification period data for actual measurement.

[0148] The objective function is constructed based on the mean square error between the inversion period data calculated by the mechanism inversion model and the measured inversion period data. The formula is as follows:

[0149] ;

[0150] In the formula, m is the number of data categories; k=1 represents displacement monitoring data, including crown settlement and horizontal convergence; k=2 represents stress monitoring data, including arch foot stress; and k=3 represents seepage monitoring data, including pore water pressure. This represents the number of monitoring data of the kth category; is the weight coefficient for the k-th class; For the model with parameters The j-th monitoring data of the k-th category is calculated at time; This represents the measured j-th monitoring data of the k-th category;

[0151] Set a maximum number of iterations. In each iteration, use a parameter inversion method based on optimization algorithms (such as particle swarm optimization, Bayesian inversion, etc.) to iteratively adjust the parameters of the mechanism inversion model by minimizing the objective function. This includes elastic modulus E, cohesion c, and internal friction angle. Permeability coefficient k, etc.;

[0152] Using the parameter combination obtained by inversion The mechanism inversion model is simulated for 5 days to obtain the predicted validation period data, which is then compared with the measured validation period data. If the determination coefficient between the predicted validation period data and the measured validation period data is greater than the preset validation threshold, such as 90%, or the current iteration reaches the maximum number of iterations, then the mechanism inversion model correction is completed, and the corrected mechanism inversion model is obtained; otherwise, the next iteration is performed.

[0153] S4. Input the preprocessed real-time monitoring data into the corrected mechanism inversion model to simulate the catastrophic evolution process of the target tunnel surrounding rock in future periods, and output the catastrophic evolution prediction results, specifically including:

[0154] Specifically, a preset recording period is established. Preprocessed real-time monitoring data is input into a corrected mechanism inversion model for advanced numerical simulation, simulating the stability evolution of the surrounding rock over a future period (e.g., after three construction steps). The evolution results and predicted monitoring data are recorded for each period, and early warning indicators are calculated based on these results. These indicators include the expansion rate of the plastic zone in the surrounding rock, the energy dissipation rate of the surrounding rock system, and pre-disaster characteristics. The expansion rate of the plastic zone is the growth rate of the plastic zone volume per unit time. The energy dissipation rate is the rate at which the energy consumed by the plastic deformation of the surrounding rock is consumed. Pre-disaster characteristics include the crown settlement rate, crown settlement acceleration, horizontal convergence rate, horizontal convergence acceleration, arch foot stress change rate, pore water pressure change rate, and seepage pressure-stress linkage index. The expansion rate of the plastic zone and the energy dissipation rate of the surrounding rock system are calculated based on the evolution results, while the pre-disaster characteristics are calculated based on the predicted monitoring data.

[0155] S5. Calculate early warning indicators based on the disaster evolution prediction results, and classify and issue early warnings for tunnel disaster risks according to the early warning indicators.

[0156] Specifically, if any warning indicator data exceeds the first threshold of the corresponding warning indicator for the first time, a warning signal (blue) of concern level will be triggered.

[0157] If at least two warning indicator values ​​simultaneously exceed the second threshold of the corresponding warning indicator, or if the growth rate of any warning indicator data exceeds the preset growth threshold in an adjacent recording period, such as 10%, then a warning signal (yellow) is triggered; the second threshold is greater than the first threshold.

[0158] If the evolution results indicate that the plastic zone of the surrounding rock is about to be connected, a danger-level warning signal (red) will be triggered.

[0159] For example, if the settlement acceleration of the arch... And the volume expansion rate of the plastic zone If so, a warning at the attention level will be triggered;

[0160] If the settlement acceleration of the arch And the volume expansion rate of the plastic zone If the evolution results show that the plastic zone will penetrate to the tunnel outline in the next step, a danger-level warning will be triggered immediately, and the on-site audible and visual alarm system will be activated.

[0161] Table 1: Crack Data Acquisition Table

[0162] In addition, to comprehensively obtain on-site measured data of water-soil coupled disasters in the weak strata of the target tunnel and improve the monitoring data acquisition system, in addition to monitoring the crown settlement, horizontal convergence, arch foot stress and pore water pressure, this study also carried out special on-site measurements of cracks on the tunnel's up line, down line and B entrance / exit. The F71 crack width and depth comprehensive tester with a nominal accuracy of 0.01mm and a measurement range of 0-8mm was used. Crack measurements were completed according to the principle of "only counting the number of small cracks, prioritizing the measurement of wide cracks, and comparing multiple sets of wide crack rings". The measured results systematically reflect the crack development characteristics of different areas of the tunnel, providing real engineering measured data support for the subsequent delineation of high-risk areas of surrounding rock, input and verification of mechanism inversion models, and construction of a graded early warning index system. The specific measured statistical results are shown in Table 1.

[0163] According to another embodiment of the invention, such as Figure 4 As shown, a graded early warning system for water-soil coupled disasters in tunnels in weak strata based on mechanism inversion is also provided. This system includes:

[0164] Data acquisition module 1 is used to collect historical and real-time monitoring data of the target tunnel and preprocess them to obtain preprocessed historical and real-time monitoring data.

[0165] Model building module 2 is used to couple the pre-built macroscopic continuous medium model and microscopic constitutive model with parameters based on the pre-acquired engineering geological exploration data and combined with the seepage-stress coupling theory, and generate a mechanism inversion model for simulating the water-soil coupled disaster response of tunnel surrounding rock based on the parameter coupling results.

[0166] Model calibration module 3 is used to input preprocessed historical monitoring data into the mechanism inversion model, iteratively adjust the parameters of the mechanism inversion model through inverse analysis and generate inversion parameter combinations, identify key parameter combinations from the inversion parameter combinations to drive model calibration, and obtain the calibrated mechanism inversion model.

[0167] Evolution module 4 is used to input the preprocessed real-time monitoring data into the corrected mechanism inversion model, simulate the catastrophic evolution process of the target tunnel surrounding rock in future time periods, and output the catastrophic evolution prediction results.

[0168] Early warning module 5 is used to calculate early warning indicators based on the disaster evolution prediction results, and to classify and issue early warnings for tunnel disaster risks according to the early warning indicators.

[0169] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores static and dynamic information data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0170] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0171] In addition, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0172] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0173] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0174] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A hierarchical early warning method for water-soil coupled disasters in tunnels in weak strata based on mechanism inversion, characterized in that, The method includes: S1. Collect historical and real-time monitoring data of the target tunnel and preprocess them to obtain preprocessed historical and real-time monitoring data. S2. Based on the pre-acquired engineering geological exploration data and combined with the seepage-stress coupling theory, the pre-constructed macroscopic continuous medium model and microscopic constitutive model are parametrically coupled using multi-scale coupling technology, and a mechanism inversion model for simulating the water-soil coupled disaster response of tunnel surrounding rock is generated based on the parametric coupling results. S3. Input the preprocessed historical monitoring data into the mechanism inversion model, iteratively adjust the parameters of the mechanism inversion model through the inversion analysis method and generate the inversion parameter combination, identify the key parameter combination from the inversion parameter combination to drive the model correction, and obtain the corrected mechanism inversion model. S4. Input the preprocessed real-time monitoring data into the corrected mechanism inversion model to simulate the catastrophic evolution process of the target tunnel surrounding rock in future time periods and output the catastrophic evolution prediction results. S5. Calculate early warning indicators based on the disaster evolution prediction results, and classify and issue early warnings for tunnel disaster risks according to the early warning indicators; The process involves using pre-acquired engineering geological exploration data and combining it with seepage-stress coupling theory, employing multi-scale coupling technology to parametrically couple a pre-constructed macroscopic continuous medium model and a microscopic constitutive model, and generating a mechanism inversion model based on the parametric coupling results to simulate the water-soil coupled catastrophic response of tunnel surrounding rock. This includes: S21. Extract engineering geological survey reports, borehole data and seismic CT detection data from the pre-acquired engineering geological exploration data as input data, and use 3D modeling software to construct a 3D geological entity model containing the spatial distribution of stratigraphic interfaces, faults, weak interlayers and water-rich areas. S22. Further import the tunnel design axis, cross-section and geometric and material parameters of the support structure into the completed three-dimensional geological entity model, and generate a dynamic excavation model reflecting the interaction between the support structure and the surrounding rock through the simulation function of the three-dimensional modeling software. S23. Introduce a micro-constitutive model based on erosion effect into the macro-continuous medium model composed of a three-dimensional geological entity model and a dynamic excavation model. Construct a mechanism inversion model for simulating the water-soil coupled disaster response of tunnel surrounding rock through macro-micro bidirectional coupling and linkage.

2. The method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion as described in claim 1, characterized in that, The introduction of a micro-constitutive model based on erosion effects into the macroscopic continuous medium model jointly composed of a three-dimensional geological entity model and a dynamic excavation model, and the construction of a mechanism inversion model for simulating the water-soil coupled catastrophic response of tunnel surrounding rock through macro-micro bidirectional coupling and linkage, includes: S231. Using the model integration function of 3D modeling software, the parameters of the 3D geological entity model and the dynamic excavation model are calibrated and docked sequentially to form a macroscopic continuous medium model. S232. Use mesh densification technology to locally densify the high-risk area of ​​the tunnel surrounding rock in the macroscopic continuous medium model, and embed the pre-constructed micro-constitutive model based on erosion effect into the locally densified area. S233. Establish a parameter transfer interface between the macroscopic continuous medium model and the microscopic constitutive model. Through the parameter transfer interface, the macroscopic continuous medium model provides boundary conditions and initial parameters to the microscopic constitutive model, while the microscopic constitutive model simulates soil particle migration and pore structure evolution and outputs macroscopic and microscopic bidirectional transfer parameters in real time. S234. Based on the macro-micro bidirectional transfer parameters, the water-soil coupled catastrophic evolution process of the tunnel surrounding rock under the action of seepage-stress coupling was simulated, and finally the mechanism inversion model was constructed.

3. The method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion according to claim 2, characterized in that, The method of using mesh refinement technology to locally refine the high-risk area of ​​tunnel surrounding rock within a macroscopic continuous medium model, and embedding a pre-constructed microscopic constitutive model based on erosion effects into the locally refined area includes: S2321. Extract the rectangular computational domain of the high-risk area of ​​the tunnel surrounding rock from the macroscopic continuous medium model, and use a quadtree structure to discretize the rectangular computational domain until the preset mesh size is met, while storing the discretized information in the quadtree database. S2322. Select boundary sub-rectangles and internal sub-rectangles that meet the grid densification termination conditions from the quadtree database and convert them into grid cells of the macroscopic continuous medium model and store them in the quadtree database. At the same time, define special geological points in high-risk areas of tunnel surrounding rock as grid cell nodes. S2323. Calculate the calculation error of each grid cell using the posterior error estimation method, and select grid cells whose calculation error exceeds the preset threshold as grid cells that need to be refined and corrected. S2324. Based on the sub-rectangle pointer information corresponding to the grid cell that needs to be encrypted and corrected, the corresponding source sub-rectangle block is matched in the quadtree database, the source sub-rectangle block and its adjacent sub-rectangle blocks that need to be encrypted are encrypted and corrected, and an invalid identifier is added to the corrected sub-rectangle block in the quadtree database. S2325. Regenerate new mesh cells based on the encrypted and corrected sub-rectangular blocks, and embed the pre-built micro-constitutive model based on the erosion effect into the corresponding new mesh cells using mesh matching technology.

4. The method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion according to claim 3, characterized in that, The step of regenerating new mesh cells based on the encrypted and corrected sub-rectangular blocks, and embedding the pre-built micro-constitutive model based on the erosion effect into the corresponding new mesh cells using mesh matching technology includes: S23251. Use the zero isosurface of the level set function to set the erosion interface of the micro-constitutive model, establish a spatial coordinate mapping between the erosion interface and the new grid cell, and obtain the corresponding level set function value at each node of the new grid cell. S23252. The material parameters at the erosion interface determined by the micro-constitutive model are used as initial values. Combined with the velocity field related to the erosion physical mechanism, the initial values ​​are propagated to the new grid cell nodes as wavefront information, and each node is assigned a material property value. S23253. Based on the material property values ​​obtained from all nodes on the new mesh element, interpolation integration is performed using the element shape function to calculate the equivalent material parameter tensor at the numerical integration point; the level set function value controlling the erosion interface is updated based on the equivalent material parameter tensor.

5. A method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion as described in claim 4, characterized in that, The zero isosurface of the level set function is the zero isosurface of the level set function φ=0; Propagating the initial value as wavefront information to all new grid cell nodes includes propagating the initial value as wavefront information from the erosion interface to all new grid nodes in the directions φ>0 and φ<0.

6. The method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion according to claim 1, characterized in that, The process involves inputting preprocessed historical monitoring data into the mechanism inversion model, iteratively adjusting the model parameters using inverse analysis, generating inversion parameter combinations, identifying key parameter combinations from these combinations to drive model correction, and obtaining the corrected mechanism inversion model, including: S31. Extract the measured inversion period data and measured verification period data for a preset time period from the preprocessed historical monitoring data respectively; S32. Input the preprocessed historical monitoring data into the mechanism inversion model to calculate the inversion period data, and build an objective function based on the mean square error between the inversion period data and the measured inversion period data. S33. Set the maximum number of iterations, and in each iteration, adjust the model parameters of the mechanism inversion model by minimizing the objective function to obtain the inversion parameter combination; S34. Identify key parameter combinations from the inversion parameter combinations, use the key parameter combinations to drive the mechanism inversion model to perform simulation prediction for a preset time period to obtain prediction verification period data, and compare the prediction verification period data with the measured verification period data. S35. If the coefficient of determination between the predicted verification period data and the measured verification period data is greater than the preset verification threshold or the current iteration reaches the maximum number of iterations, the mechanism inversion model correction is completed and the corrected mechanism inversion model is obtained; otherwise, the next round of iteration is carried out until the coefficient of determination is greater than the preset verification threshold or the maximum number of iterations is reached, and the corrected mechanism inversion model is output.

7. A method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion as described in claim 6, characterized in that, The identification of key parameter combinations from inversion parameter combinations includes: Based on the combination of inversion parameters and their range of values ​​generated in each iteration, two sets of reference sample sets are generated by Latin hypercube sampling and are respectively denoted as the first reference sample set and the second reference sample set. The numerical column of each inversion parameter in the second reference sample set is replaced with the corresponding column in the first basic sample set, while keeping the values ​​of other inversion parameters unchanged, to generate a perturbation sample set for the current inversion parameter. Input all generated first baseline sample sets, second baseline sample sets, and perturbation sample sets into the mechanism inversion model to obtain the corresponding model output result sets; Estimate the overall variability of the model output, and calculate the first-order sensitivity index of the impact of the change of each inversion parameter on the model output, as well as the total sensitivity index of the combined impact of the interaction between the inversion parameter and other inversion parameters; The key parameter combinations that play a dominant role in the output of the mechanism inversion model are selected by combining the total sensitivity index and the first-order sensitivity index.

8. A method for graded early warning of water-soil coupled disasters in tunnels in weak strata based on mechanism inversion as described in claim 7, characterized in that, The key parameter combination includes elastic modulus, cohesion, internal friction angle, and permeability coefficient.

9. A mechanism-based inversion-based hierarchical early warning system for water-soil coupled disasters in tunnels in weak strata, used to implement the mechanism-based inversion-based hierarchical early warning method for water-soil coupled disasters in tunnels in weak strata as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition module is used to collect historical and real-time monitoring data of the target tunnel and preprocess them to obtain preprocessed historical and real-time monitoring data. The model building module is used to couple the pre-built macroscopic continuous medium model and microscopic constitutive model with parameters based on the pre-acquired engineering geological exploration data and the seepage-stress coupling theory, and generate a mechanism inversion model for simulating the water-soil coupled disaster response of the tunnel surrounding rock based on the parameter coupling results. The model calibration module is used to input preprocessed historical monitoring data into the mechanism inversion model, iteratively adjust the parameters of the mechanism inversion model through inverse analysis and generate inversion parameter combinations, identify key parameter combinations from the inversion parameter combinations to drive model calibration, and obtain the calibrated mechanism inversion model. The evolution module is used to input the preprocessed real-time monitoring data into the corrected mechanism inversion model, simulate the catastrophic evolution process of the target tunnel surrounding rock in future time periods, and output the catastrophic evolution prediction results. The early warning module is used to calculate early warning indicators based on the disaster evolution prediction results, and to classify and issue early warnings for tunnel disaster risks according to the early warning indicators.

Images