Method for evaluating excavation stability of shallow-buried unsymmetrical-pressure multi-arch tunnel on flood plain

Abstract

The invention discloses a flood plain shallow-buried unsymmetrical-pressure multi-arch tunnel excavation stability evaluation method, and relates to the technical field of tunnel stability evaluation, and the method comprises the following steps: obtaining the geological conditions of a tunnel excavation area through on-site survey and drilling, and evaluating the groundwater emission amount and the inrush probability in combination with historical data; and basic data support is provided for subsequent risk modeling. According to the method, through geological and hydrological survey and a numerical model, the high-risk inrush area and the water pressure distribution are accurately predicted, the risk identification and early warning precision is effectively improved, and a scientific basis is provided for support design. Optimized design of the supporting structure is combined with real-time monitoring and dynamic adjustment, construction safety is guaranteed, and meanwhile structural reliability and economical efficiency are both considered.

Description

Technical Field

[0001] The invention relates to the technical field of tunnel stability evaluation, and in particular to a method for evaluating the excavation stability of a shallow-buried biased multi-arch tunnel in a floodplain. Background Art

[0002] "Stability Evaluation of Shallow Buried and Unevenly Loaded Multi-arch Tunnel Excavation in Floodplains" is a stability analysis method specifically used to evaluate the tunnel excavation process. This method is aimed at the tunnel excavation conditions of shallow buried and unevenly loaded (i.e., uneven forces on both sides) multi-arch structures in floodplain terrain, and comprehensively considers multiple factors such as strata, groundwater level, and uneven loads to evaluate the structural stability and safety. Through this evaluation, engineering designers can predict the structural deformation, stress concentration, and displacement that may occur during tunnel excavation, so as to determine reasonable support measures and construction methods to ensure the safety and quality of tunnel projects.

[0003] The prior art has the following deficiencies:

[0004] Under shallow buried conditions, floodplain terrain is usually accompanied by high-water sand or pebble layers. If not properly handled during tunnel excavation, the natural balance of shallow groundwater may be destroyed, causing groundwater to suddenly surge into the excavation surface. This sudden surge phenomenon may cause the excavation surface to become unstable, large-scale soil collapse, and even erode the overall structure of the tunnel, causing the support to fail, posing a major threat to the progress of the project and the safety of personnel. If the stability assessment fails to identify such risks, the consequences may be extremely serious.

[0005] First, the sudden influx of groundwater will quickly soften and wash away the soil on the excavation surface, significantly reducing the strength of the soil, leading to support failure and local collapse. Due to the high pressure of groundwater, the tunnel support structure is subjected to extremely high lateral water pressure in a short period of time, which is prone to failure or damage, and even causes the overall collapse of the tunnel. Secondly, the water flow brings in a large amount of mud and pebbles, which may cause the construction site to be flooded, equipment to be damaged, and the drainage system to be blocked, resulting in construction interruptions and huge drainage, dredging and equipment maintenance costs. More seriously, the sudden influx of groundwater may form large-scale cavities in the surrounding strata, causing surface subsidence or collapse, endangering the safety of surrounding buildings, roads and other facilities, and causing irreversible damage to the project progress, cost and surrounding environment.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method for evaluating the excavation stability of shallow-buried biased arch tunnels in floodplains. Through detailed geological surveys and hydrological data combined with historical sudden surge records, a numerical seepage model is constructed to accurately predict areas with high incidence of sudden surges and water pressure distribution, significantly improve the accuracy of risk identification and the timeliness of early warning, provide a scientific basis for the design of support structures, and avoid hidden dangers caused by sudden water surges. At the same time, the hydraulic bearing simulation and optimization design of the support structure can identify and reinforce weak points before construction to ensure the stability of the support in sudden surges or high-pressure water environments. The support thickness and material strength are dynamically adjusted in combination with real-time monitoring data to effectively ensure safety and avoid material waste, thereby achieving a support structure that is both reliable and economical to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a method for evaluating the excavation stability of a shallow-buried biased double-arch tunnel in a floodplain, comprising the following steps:

[0009] Through on-site survey and drilling, the geological conditions of the tunnel excavation area are obtained, and the groundwater outburst volume and probability of sudden inrush are evaluated in combination with historical data, providing basic data support for subsequent risk modeling;

[0010] Based on the collected geological condition data, a groundwater seepage numerical model was constructed to simulate the dynamic behavior of shallow groundwater during tunnel excavation. The water pressure distribution at different strata was obtained through model analysis, providing data support for the judgment of sudden surge risks during excavation.

[0011] According to the tunnel design parameters and excavation plan, simulate the stress of the support structure under different water pressures, analyze the ultimate water pressure that the support structure can withstand, and provide a basis for support selection and optimization in tunnel excavation;

[0012] Using simulation data, we built a risk assessment model based on machine learning, dynamically predicted the risk of sudden surges based on real-time monitoring data, identified high-risk sudden surge points, and provided early warning for tunnel construction;

[0013] Based on the results of sudden surge risk prediction, a multivariable optimization algorithm is used to dynamically optimize the support structure to determine the support strength and material configuration. At the same time, it is linked with the on-site monitoring data to adjust the support structure in real time to ensure the safety and stability of the tunnel structure.

[0014] Preferably, the specific steps of evaluating groundwater outburst volume and inrush probability in combination with historical data to provide basic data support for subsequent risk modeling are as follows:

[0015] First, a comprehensive geological survey of the tunnel excavation area is carried out;

[0016] After the risk areas are identified, a detailed drilling layout plan is developed based on the results of the geological survey;

[0017] Gain in-depth understanding of groundwater flow direction and pressure changes, install groundwater level monitoring devices in drilling holes, regularly record water level elevation and flow rate data, and analyze the dynamic changes of groundwater;

[0018] Based on the field survey and drilling data, combined with historical records of sudden surge events, the conditions and frequency of sudden surges in the tunnel excavation area are analyzed.

[0019] Preferably, the specific steps of obtaining the water pressure distribution at different strata by model analysis to provide data support for the judgment of the risk of sudden surge during excavation are as follows:

[0020] Before starting the model construction, the collected geological and hydrological data are first processed;

[0021] After the data processing is completed, a matching numerical model is selected for seepage simulation;

[0022] After the model is constructed, the calculation phase begins, simulating the flow dynamics of groundwater during tunnel excavation;

[0023] After the calculation is completed, the model results are calibrated with the field monitoring data to ensure the consistency of the model output with the actual situation.

[0024] Preferably, the specific steps for analyzing the ultimate water pressure borne by the support structure and providing a basis for support selection and optimization in tunnel excavation are as follows:

[0025] First, detailed data collection of the tunnel support structure is carried out. The acquired data will provide basic input for the model.

[0026] After obtaining the key parameters of the support structure, the force model of the support structure is established using the finite element analysis method;

[0027] After the model is established, the limit state analysis is performed by applying different water pressure states to calculate the maximum bearing capacity and limit pressure of the support structure when the water pressure changes;

[0028] According to the results of the limit state analysis, the design parameters of the support structure are adjusted, and the support materials and structural forms are optimized to ensure the safety of tunnel excavation under sudden surge conditions.

[0029] Preferably, the steps of dynamically predicting the surge risk based on the real-time monitoring data, identifying the high-risk surge points, and providing early warning for tunnel construction are as follows:

[0030] The acquired geological parameters, support structure stress data, and sudden surge event data collected in historical projects are preprocessed. To ensure the sensitivity of the model to the high-risk areas of sudden surge, a multidimensional feature matrix X is defined. The acquired water pressure distribution is calibrated as φ(x, y, z, t), the support structure force distribution is calibrated as σ(x, y, z), and the formation parameters are calibrated as ψ(x, y, z). The matrix characteristics are as follows:

[0031] X=[φ(x,y,z,t)σ(x,y,z)ψ(x,y,z)] T

[0032] , where X is a multidimensional characteristic matrix, φ(x, y, z, t) is the water pressure distribution, which indicates the water pressure value at each location in the tunnel excavation area, and its coordinates are represented by (x, y, z), t is the time point, σ(x, y, z) is the force distribution of the support structure, which indicates the stress state at each point of the tunnel support structure, and its three-dimensional coordinates are defined by (x, y, z), σ is in tensor form, ψ(x, y, z) is the formation parameter, which indicates the characteristic parameter matrix of the formation around the tunnel, and the spatial position is represented by (x, y, z), ψ is a variety of physical parameters of the formation, and T is the matrix transpose;

[0033] In order to improve the model's ability to identify sudden surge risks under complex geological conditions, the tensor decomposition method is used to extract high-order features, and the multidimensional feature matrix X is expressed as a third-order tensor: Where X is the original third-order tensor, represents the real number domain, indicating that all data in the tensor are real numbers, M is the spatial position dimension, N is the feature type dimension, representing the second dimension of the tensor, and P is the time dimension, representing the third dimension of the tensor;

[0034] The multi-level feature combination is obtained by tensor decomposition, and the expression is as follows:

[0035]

[0036] , where is the third-order tensor decomposition representation, λ k is the weight coefficient, u k is the space dimension vector, v k is the feature type dimension vector, w k is the time dimension vector;

[0037] The machine learning model is trained using an adaptive weighted iterative method. The weights are adjusted to optimize the recognition accuracy of high-risk areas for sudden surges. The prediction error deviation term is defined as follows: In the formula, ΔR t is the forecast error bias term, y t is the actual surge risk label, is the model prediction:

[0038] The model updates the loss function through an adaptive weight factor, and the calculation expression is as follows:

[0039]

[0040] , where L(w) is the loss function, which is used to measure the deviation between the accuracy of the model prediction and the actual value, w is the weight vector of the model, ω t is the weighting factor, f(ΔR t ) is the error function, which represents the prediction deviation at measurement time t, λ||w|| 2 is the regularization term, λ is the regularization coefficient;

[0041] In the model deployment stage, the surge risk is dynamically predicted by inputting real-time monitoring data to generate the surge risk probability, which is calculated as follows:

[0042]

[0043] , where is the surge risk probability, X real (t) is the real-time feature matrix, X real (t) = [φ real (t)σ real (t) ψ real (t)] T , including water pressure, stress and formation parameters at time t, b is the bias term of the model;

[0044] Based on surge risk probability Define the dynamic risk threshold θ(t), when When , the warning is triggered, the model parameters are updated, and the dynamic risk threshold calculation expression is as follows:

[0045]

[0046] , where θ(t) is the dynamic risk threshold, β is the smoothing factor, and β·θ(t-1) represents the smooth continuation of the threshold at the previous moment. Represents the dynamic update of the current predicted risk probability.

[0047] Preferably, according to the results of the sudden surge risk prediction, the specific steps of adjusting the support structure in real time to ensure the safety and stability of the tunnel structure are as follows:

[0048] Initialize the cost of supporting materials, strength of supporting materials, surge risk pressure and construction feasibility, and mark them as C i , S i , P i and D i , Ci Represents the cost coefficient of unit support material, S i is the strength coefficient of the unit support material, P i is the surge risk pressure coefficient, which indicates the intensity of groundwater pressure, D i is the construction feasibility coefficient, which reflects the difficulty of supporting structure installation. The variable matrix F = {f i}={f 1 , f 2 , ..., f n}, where f i represents the strength and performance combination of the i-th layer of support materials;

[0049] According to the actual support requirements, the objective function is constructed, and the cost, strength, risk coefficient and other parameters of the support structure are associated. The nonlinear multi-objective optimization function is adopted, and the expression is as follows:

[0050]

[0051] , where G(F) is the optimization objective function, representing the total cost and risk index of the tunnel support structure, and C i f i is the total cost of the i-th layer of support materials, ω is the risk penalty coefficient, which is used to adjust the influence weight of the sudden surge risk in the optimization process, n is the total number of layers of strength and performance combination of support materials, τ is the construction difficulty penalty coefficient, which is used to adjust the influence of construction difficulty in optimization, and ∈ is a constant;

[0052] In order to ensure the stability of the support structure, the following nonlinear constraints must be met:

[0053] g i (f i ) = f i S i -P i ≥0, i=1, 2, ..., n

[0054] , where g i (f i ) is the stability constraint function of the i-th support layer, f i S i Indicates the actual support strength provided by the i-th layer of support material;

[0055] The constraint function ensures that the support structure does not lose stability under the action of water pressure. In addition, in order to reduce material costs, the total amount of constrained support materials is calculated as follows:

[0056]

[0057] , where T budget is the total budget for support costs;

[0058] Surge risk data R based on real-time feedback from on-site monitoring data i (t), dynamically adjust the support strength matrix G(F), combine the Lagrange multiplier method and the dynamic weight adjustment strategy, update the optimization scheme, and the calculation expression is as follows:

[0059]

[0060] , where f i (t+1) is the value of the support strength at the i-th position at the next time step t+1, f i (t) is the support strength variable at the i-th position, its value at the current time point t, μ is the learning rate, a parameter used to control the optimization step size, is the risk feedback coefficient, which is used to control the influence of real-time monitoring data in the optimization process. i (t) is the real-time surge risk data of the ith location at time t.

[0061] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0062] The present invention can effectively improve the accuracy of sudden surge risk identification during tunnel excavation through detailed geological surveys and hydrological surveys combined with historical sudden surge data. The seepage numerical model is constructed using high-precision geological and hydrological data collected on site, so that the groundwater seepage path, pressure concentration areas and sudden surge high-incidence locations can be accurately predicted. This risk identification method based on field data can fully evaluate the sudden surge risk of shallow buried high-water level sand layers and pebble layers before excavation, and provide data support and risk warning for subsequent support structure design. Compared with traditional empirical methods or simple model analysis methods, this method intuitively displays the hydraulic characteristics of the excavation area through numerical simulation, greatly improving the accuracy and effectiveness of sudden surge prediction. In addition, the risk assessment data can be updated in real time when the hydrological conditions or excavation area change, ensuring the timeliness and scientificity of the prediction, so as to formulate sudden surge protection measures in advance and avoid the safety hazards caused by sudden water surge events.

[0063] The present invention can significantly enhance the adaptability of tunnel support structures in complex hydrological environments through the simulation and optimization design of water pressure bearing of support structures. Water pressure bearing simulation can accurately calculate the stress distribution of support structures under different water pressure conditions, clarify the ultimate bearing capacity and design boundaries of each layer of support materials, and ensure the stability of support structures under high-pressure water or surge conditions. This optimization design method can identify support weaknesses and adjust material strength before excavation construction, thereby avoiding surge pressure exceeding the design limit of the support structure and preventing collapse accidents caused by support failure. Real-time monitoring data also supports dynamic adjustment of support design. In the event of an unexpected increase in surge pressure, the impact of surge pressure can be dealt with by increasing support thickness or temporary reinforcement measures. This intelligent and flexible support design not only improves construction safety, but also effectively avoids material waste caused by excessive support design, making the support structure both reliable and economical. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0065] Figure 1 The present invention is a flow chart of the method for evaluating the excavation stability of a shallow-buried biased multi-arch tunnel in a floodplain. DETAILED DESCRIPTION

[0066] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0067] The present invention provides Figure 1 The excavation stability evaluation method of shallow-buried biased double-arch tunnel in floodplain shown in the figure includes the following steps:

[0068] Through on-site survey and drilling, the geological conditions of the tunnel excavation area are obtained, and the groundwater outburst volume and probability of sudden inrush are evaluated in combination with historical data, providing basic data support for subsequent risk modeling;

[0069] The specific steps to evaluate groundwater outburst volume and probability of sudden inrush in combination with historical data and provide basic data support for subsequent risk modeling are as follows:

[0070] First, a comprehensive geological survey of the tunnel excavation area is carried out;

[0071] The survey content includes stratigraphic structure, geological structure, hydrogeological conditions, soil layer composition, groundwater distribution, etc. In view of the floodplain terrain, special attention is paid to the location and thickness of shallow buried sand layers, pebble layers and saturated layers, and the permeability of the soil is checked. Through comprehensive analysis of geological maps, topographic maps and other data, the distribution of high-water-level sand layers and pebble layers is preliminarily determined, and the physical properties of these layers are recorded. This step not only lays the foundation for the selection of the best drilling location for subsequent drilling and surveying, but also helps to identify potential areas of sudden surge risk and build a macro-cognition of the factors affecting excavation.

[0072] After the risk areas are identified, a detailed drilling layout plan is developed based on the results of the geological survey;

[0073] The arrangement of drilling points should take into account the changes in the formation and the distribution of groundwater levels to ensure that the boreholes can fully cover the areas where high-water-level sand and gravel layers may exist. During the drilling process, layered sampling technology is used to collect soil and water samples at different depths to accurately analyze key parameters such as the permeability, moisture content, and particle composition of each soil layer. Every time a hole is drilled, data such as water level elevation, soil layer thickness, and permeability must be measured in real time to provide a real geological and hydrological data foundation for the sudden surge risk area. Drilling data should be classified and organized, and combined with analysis tools such as geological maps and profiles to construct a multi-level geological distribution map.

[0074] Gain in-depth understanding of groundwater flow direction and pressure changes, install groundwater level monitoring devices in drilling holes, regularly record water level elevation and flow rate data, and analyze the dynamic changes of groundwater;

[0075] During this process, it is necessary to focus on the changes in water flow velocity and water pressure in the sand and gravel layers in the stratum to assess the possibility of sudden surges. If the water level fluctuates significantly or the water pressure suddenly increases, further hydraulic analysis is required to simulate possible sudden surge conditions during excavation. At the same time, by comparing the monitored water level and water pressure changes with historical data, a verification basis is provided for the modeling of sudden surge risks. These data will provide detailed parameter support for the subsequent construction of groundwater seepage models and sudden surge pressure analysis.

[0076] Analyze the conditions and frequency of sudden surges in the tunnel excavation area based on field survey and drilling data combined with historical sudden surge event records;

[0077] By referring to the sudden surge data under similar geological conditions and water levels, the sudden surge probability in high-risk areas such as high-water sand layers and pebble layers is quantitatively evaluated. By comparing the water level, flow rate, pressure and other parameters of historical events with the current monitoring data, a sudden surge probability model is established. This analysis provides accurate basic data for subsequent risk modeling and can predict the sudden surge probability under different formation conditions during excavation.

[0078] Based on the collected geological condition data, a groundwater seepage numerical model was constructed to simulate the dynamic behavior of shallow groundwater during tunnel excavation. The water pressure distribution at different strata was obtained through model analysis, providing data support for the judgment of sudden surge risks during excavation.

[0079] The specific steps to obtain the water pressure distribution at different strata through model analysis and provide data support for the judgment of sudden surge risk during excavation are as follows:

[0080] Before starting the model construction, the collected geological and hydrological data are first processed;

[0081] These data include parameters such as the particle composition of the soil layer, hydrological water level, groundwater flow direction, porosity, and permeability coefficient. To ensure the accuracy and applicability of the model, the data needs to be standardized and screened to remove noise and abnormal data. At the same time, according to the characteristics of different strata and soil components, appropriate permeability parameters (such as the permeability coefficient in Darcy's law) are selected to reflect the actual water flow conditions. These parameters determine the flow speed and direction of groundwater in different strata around the tunnel and are the core basis of model simulation.

[0082] After the data processing is completed, a matching numerical model is selected for seepage simulation;

[0083] Commonly used models include finite element method and finite difference method, which can accurately describe the flow state of groundwater around the tunnel. When building the model, it is necessary to set boundary conditions, define the entry and exit positions of the water flow, and set boundary parameters such as regional water level. In addition, it is also necessary to set the tunnel boundary water pressure under different working conditions, such as the initial pressure distribution before excavation and the transient pressure change after excavation. The setting of boundary conditions directly affects the calculation results of the model, and should be verified and calibrated multiple times based on field data to ensure the applicability and accuracy of the model.

[0084] After the model is constructed, the calculation phase begins, simulating the flow dynamics of groundwater during tunnel excavation;

[0085] During the calculation, the water pressure changes and water flow paths generated during the excavation process are analyzed for different locations around the tunnel. Through pressure distribution analysis, high-risk areas with concentrated water pressure and concentrated seepage paths can be identified. This analysis result can help locate possible surge points and pressure points with support failure near the tunnel. The parameters in the calculation process need to be continuously adjusted and optimized to ensure the calculation accuracy, and finally obtain the pressure distribution map and water flow path map that conform to the actual situation, providing data support for the subsequent surge risk judgment.

[0086] After the calculation is completed, the model results are calibrated with the field monitoring data to ensure the consistency of the model output with the actual situation;

[0087] If the water pressure or flow direction predicted by the model deviates greatly from the actual monitoring data, it is necessary to further optimize the model parameters and adjust the permeability coefficient or boundary conditions until the calculated results match the actual data. After the model has been calibrated multiple times, the output pressure distribution and flow direction data can be used to determine the risk of sudden surges. By comparing the water pressure data in different areas, high-risk points that may cause sudden surges can be accurately identified and quantitative pressure values ​​can be provided to ensure that the model accurately predicts the behavior of groundwater throughout the tunnel construction process and prevent construction risks caused by sudden surges.

[0088] According to the tunnel design parameters and excavation plan, simulate the stress of the support structure under different water pressures, analyze the ultimate water pressure that the support structure can withstand, and provide a basis for support selection and optimization in tunnel excavation;

[0089] The specific steps to analyze the ultimate water pressure of the support structure and provide a basis for support selection and optimization in tunnel excavation are as follows:

[0090] First, detailed data collection of the tunnel support structure is carried out. The acquired data will provide basic input for the model.

[0091] The acquired data include the tunnel excavation section size, the type of support structure, material strength parameters (such as the compressive and tensile strength of reinforced concrete, etc.) and the geological and hydrological conditions of the construction area. These data will provide basic input for the model and determine the boundary conditions and physical properties of the model. The tunnel external water pressure, soil pressure and stratum displacement data collected in real time by multi-point monitoring devices will also be used as key inputs to the model. Comprehensive and accurate data collection is the basis for ensuring the reliability of model output. Therefore, various possible variables in tunnel design should be covered as much as possible to ensure that the model can truly simulate the actual stress conditions of the support structure during excavation.

[0092] After obtaining the key parameters of the support structure, the force model of the support structure is established using the finite element analysis (FEA) calculation method;

[0093] The model has the function of simulating the stress of the support structure at different excavation stages, and can dynamically adjust the boundary of the support structure to simulate the changes in its force during tunnel excavation and support construction. The model needs to define the physical parameters of the support material such as stress-strain relationship, elastic modulus, yield strength, etc., so as to accurately calculate the bearing capacity and deformation behavior of the support structure under different water pressure conditions. By loading the groundwater pressure during tunnel excavation, the stress distribution, displacement and potential failure mode of the support structure under extreme pressure can be calculated, providing a reference for subsequent optimization design.

[0094] After the model is established, the limit state analysis is performed by applying different water pressure states to calculate the maximum bearing capacity and limit pressure of the support structure when the water pressure changes;

[0095] The limit state analysis involves simulating the support structure to the design limit and observing the stress concentration area, deformation degree and failure mode of the structure under the condition of gradually increasing water pressure. By simulating the stress of the support under the extreme conditions of sudden water pressure, the weak areas and potential failure modes of the structure can be identified. For example, under the sudden high water pressure, the support structure may experience excessive deformation, buckling or shear failure. The limit state analysis can quantify the safety margin of the support structure, which is convenient for ensuring that the support strength is sufficient and reliable during the actual excavation process.

[0096] According to the limit state analysis results, the design parameters of the support structure are adjusted to optimize the support materials and structural forms to ensure the safety of tunnel excavation under sudden surge conditions;

[0097] The analysis shows that when certain support components are subjected to greater forces at specific locations, higher strength support materials can be selected, or the support layout can be adjusted to increase the support density in the area. In addition, the compressive performance of weak areas can be enhanced and material waste can be reduced by changing the support thickness, structural cross-sectional shape and other measures. The optimized support design will ensure that the tunnel maintains structural stability under extreme water pressure conditions, reduce construction risks and economic losses caused by sudden surges, and provide higher safety guarantees for the tunnel construction process.

[0098] Using simulation data, we built a risk assessment model based on machine learning, dynamically predicted the risk of sudden surges based on real-time monitoring data, identified high-risk sudden surge points, and provided early warning for tunnel construction;

[0099] The steps to dynamically predict the surge risk based on real-time monitoring data, identify high-risk surge points, and provide early warning for tunnel construction are as follows:

[0100] The obtained geological parameters (geological parameters include water pressure data, water flow direction, density of sand and gravel layers), stress data of support structures, and data of sudden surge events collected in historical projects are preprocessed. To ensure the sensitivity of the model to high-risk areas of sudden surge, a multidimensional feature matrix X is defined. The obtained water pressure distribution is calibrated as φ(x, y, z, t), the support structure force distribution is calibrated as σ(x, y, z), and the formation parameters are calibrated as ψ(x, y, z). The matrix characteristics are as follows:

[0101] X=[φ(x,y,z,t)σ(x,y,z)ψ(x,y,z)] T

[0102] , where X is a multidimensional characteristic matrix, φ(x, y, z, t) is the water pressure distribution, which indicates the water pressure value at each location in the tunnel excavation area, and its coordinates are represented by (x, y, z), t is the time point, σ(x, y, z) is the force distribution of the support structure, which indicates the stress state at each point of the tunnel support structure, and its three-dimensional coordinates are defined by (x, y, z), σ is a tensor form, usually including vertical, horizontal and lateral stress components, used to indicate the force condition of the support structure, ψ(x, y, z) is the formation parameter, which indicates the characteristic parameter matrix of the formation around the tunnel, and (x, y, z) indicates the spatial position, ψ is a variety of physical parameters of the formation, such as density, porosity, permeability and shear strength, etc., T is the matrix transpose, which transforms the multidimensional characteristic matrix X from row vector form to column vector form;

[0103] In order to improve the model's ability to identify sudden surge risks under complex geological conditions, the tensor decomposition method is used to extract high-order features, and the multidimensional feature matrix X is expressed as a third-order tensor: Where X is the original third-order tensor, a multidimensional data structure containing different parameters. represents the real number domain, indicating that all data in the tensor are real numbers. M is the spatial position dimension, which represents the first dimension in the tensor and is used to record the spatial position of the tunnel excavation area, such as the geological parameters at the three-dimensional coordinate point (x, y, z). N is the feature type dimension, which represents the second dimension of the tensor and represents different feature types. φ, σ, ψ are several main features in N dimensions, representing water pressure distribution φ, support stress σ, formation parameter ψ, etc., respectively. P is the time dimension, which represents the third dimension of the tensor and represents the time series of data. During the tunnel construction process, water pressure, stress and formation characteristics will change dynamically over time, so the time dimension is needed to capture these changes.

[0104] The multi-level feature combination is obtained through tensor decomposition, and the expression is as follows:

[0105]

[0106] , where is the third-order tensor decomposition representation, λ k is the weight coefficient, which represents the kth feature weight coefficient of the decomposed tensor. The weight controls the influence of each group of outer products. u k is a spatial dimension vector, which is a feature vector related to the first dimension of the tensor, the spatial position dimension M, and represents the feature information of the data in the spatial position dimension. k It is the feature type dimension vector, which represents the second dimension of the tensor, that is, the feature vector of the feature type dimension N, reflecting the relationship between different feature data. kIt is a time dimension vector and a characteristic vector of the time dimension P. It captures the characteristics of the tensor changing with time and reflects the dynamic changes of geological parameters and water pressure at different times.

[0107] The machine learning model is trained using an adaptive weighted iterative method. The weights are adjusted to optimize the recognition accuracy of high-risk areas for sudden surges. The prediction error deviation term is defined as follows: In the formula, ΔR t is the forecast error bias term, y t is the actual surge risk label, is the model prediction value;

[0108] The model updates the loss function through an adaptive weight factor, and the calculation expression is as follows:

[0109]

[0110] , where L(w) is the loss function, which is used to measure the deviation between the accuracy of the model prediction and the actual value, w is the weight vector of the model, which contains all the model parameters used for prediction, and ω t is the weighting factor, which is the weighting factor at time t, used to adjust the contribution of the prediction error at different time points to the loss function, f(ΔR t ) is the error function, which represents the prediction deviation at measurement time t, λ||w|| 2 is the regularization term, λ is the regularization coefficient, which is used to control the size of the weight w to prevent the model from overfitting;

[0111] In the model deployment stage, the surge risk is dynamically predicted by inputting real-time monitoring data to generate the surge risk probability, which is calculated as follows:

[0112]

[0113] , where is the surge risk probability, X real (t) is the real-time feature matrix, X real (t) = [φ real (t)σ real (t) ψ real (t)] T , including water pressure, stress and formation parameters at time t. After these data are input into the model, a linear combination of the surge risk is performed through the action of w and the bias term b. b is the bias term of the model, which is used to adjust the baseline value of the prediction result to improve the flexibility and accuracy of the prediction.

[0114] The bias of the model is an additional constant parameter used to adjust the output of the model so that the predicted value does not depend on the zero value state of the input feature. It plays a "translation" role in the model, and can adjust the model's prediction results upward or downward to better match the actual data. For example, in the prediction of tunnel surge risk, even when all input features (such as water pressure, stress, etc.) are zero, the bias term can still allow the model to generate a basic risk probability, reflecting the basic risk level that is not affected by the input data. The bias term enables the model to adapt to a wider variety of data distributions, thereby improving the accuracy and flexibility of the prediction.

[0115] Based on surge risk probability Define the dynamic risk threshold θ(t), when When , the warning is triggered, the model parameters are updated, and the dynamic risk threshold calculation expression is as follows:

[0116]

[0117] , where θ(t) is the dynamic risk threshold, β is the smoothing factor, and β·θ(t-1) represents the smooth continuation of the threshold at the previous moment. Indicates the dynamic update of the current predicted risk probability;

[0118] Based on the results of the sudden surge risk prediction, a multivariable optimization algorithm is used to dynamically optimize the support structure to determine the support strength and material configuration. At the same time, the support structure is adjusted in real time in conjunction with the on-site monitoring data to ensure the safety and stability of the tunnel structure.

[0119] According to the results of the sudden surge risk prediction, the specific steps for real-time adjustment of the support structure to ensure the safety and stability of the tunnel structure are as follows:

[0120] Initialize the cost of supporting materials, strength of supporting materials, surge risk pressure and construction feasibility, and mark them as C i , S i , P i and D i , C i Represents the cost coefficient of unit support material, S i is the strength coefficient of the unit support material, P j is the surge risk pressure coefficient, which indicates the intensity of groundwater pressure, D j is the construction feasibility coefficient, which reflects the difficulty of supporting structure installation. The variable matrix F = {f i}={f 1 , f 2 , ..., f n}, where f i represents the strength and performance combination of the i-th layer of support materials;

[0121] In the above optimization model, the key parameter (cost coefficient C per unit support material) j , strength coefficient S of unit support material i , surge risk pressure coefficient P i , Construction feasibility coefficient D i ) and the variable f i The relationship between determines the performance and optimization target of the support structure. The specific relationship is as follows:

[0122] 1. Cost coefficient C of unit support material i With f i Relationship:

[0123] Cost coefficient C of unit support material j It represents the cost of each unit of support material and directly affects the total cost. i is a combination of support strength and material selection, so C i f i reflects the total cost of the selected support materials. i f i Incorporating the objective function, the model will consider the support cost during the optimization process, so that the final f i The combination meets strength requirements while remaining economical.

[0124] 2. Strength coefficient S of unit support material i With f i Relationship:

[0125] Strength coefficient S of unit support material i Indicates the unit strength of the material, which is similar to f i The relationship is that f i S i It can be regarded as the effective strength of the support structure at this position. In order to ensure the stability of the tunnel support, the nonlinear constraint condition g i (f i ) = f i S i -P i ≥0 needs to be satisfied. This indicates that f i S i Must be greater than or equal to the surge risk pressure coefficient P i , that is, the surge risk pressure coefficient P i The choice must be high enough to cope with the risk of sudden surge pressure, thereby ensuring the reliability of the support structure.

[0126] 3. Surge risk pressure coefficient P i With f i Relationship:

[0127] Surge risk pressure coefficient Pi Indicates the maximum water pressure that the location may bear. In order to avoid structural instability, f i S i Must be greater than or equal to P i Therefore, the surge risk pressure P i The increase in f i This relationship ensures that a higher f value is selected in high-risk areas. i Value to enhance support.

[0128] 4. Construction feasibility coefficient f i With f i Relationship:

[0129] Construction feasibility factor f i Indicates the difficulty of installing the support material. This reflects the influence of material strength selection on construction difficulty. i As the strength of the support material increases, the feasibility of the support material decreases (for example, higher strength materials may be more difficult to install). Therefore, during the optimization process, the system will balance f i The selection and construction difficulty of the support should be considered to ensure that the support meets the strength requirements while maintaining the feasibility of construction.

[0130] Through these parameters and f i The optimization model can select the best f under multiple constraints. i combination to achieve a balance between cost, strength and construction feasibility.

[0131] According to the actual support requirements, the objective function is constructed, and the cost, strength, risk coefficient and other parameters of the support structure are associated. The nonlinear multi-objective optimization function is adopted, and the expression is as follows:

[0132]

[0133] , where G(F) is the optimization objective function, representing the total cost and risk index of the tunnel support structure, and C i f i is the total cost of the i-th layer of support materials, ω is the risk penalty coefficient, which is used to adjust the influence weight of the sudden surge risk in the optimization process, n is the total number of layers of strength and performance combination of support materials, τ is the construction difficulty penalty coefficient, which is used to adjust the influence of construction difficulty in optimization, and ∈ is a constant used to avoid When the denominator is zero,

[0134] In order to ensure the stability of the support structure, the following nonlinear constraints must be met:

[0135] g i (f i) = f i S i -P i ≥0, i=1, 2, ..., n

[0136] , where g i (f i ) is the stability constraint function of the i-th support layer, f i S i Indicates the actual support strength provided by the i-th layer of support material;

[0137] The constraint function ensures that the support structure does not lose stability under the action of water pressure. In addition, in order to reduce material costs, the total amount of constrained support materials is calculated as follows:

[0138]

[0139] , where T budget is the total budget for support costs;

[0140] Surge risk data R based on real-time feedback from on-site monitoring data i (t), dynamically adjust the support strength matrix G(F), combine the Lagrange multiplier method and the dynamic weight adjustment strategy, update the optimization scheme, and the calculation expression is as follows:

[0141]

[0142] , where f i (t+1) is the value of the support strength at the i-th position at the next time step t+1, f i (t) is the support strength variable at the i-th position, and its value at the current time point t. μ is the learning rate, which is used to control the optimization step size. The learning rate determines the support strength f in each optimization step. i The update range, is the risk feedback coefficient, which is used to control the influence of real-time monitoring data in the optimization process. i (t) is the real-time surge risk data of the ith location at time t, usually indicating risk indicators such as surge pressure or groundwater flow velocity;

[0143] Implementation method 1:

[0144] In the early stage of tunnel excavation, in order to ensure the safety of tunnel engineering under shallow buried conditions, it is necessary to identify the potential risks of sudden surges in the tunnel area through detailed geological surveys and hydrological surveys. The process first requires geological drilling and sampling to obtain important data such as the stratigraphic structure, lithological characteristics and water content of each layer in the tunnel excavation area. These survey data not only include the distribution of sand and pebble layers, but also need to accurately determine the elevation and flow direction of groundwater to form a complete set of groundwater level distribution maps. In addition, for strata such as sand and pebble layers that may cause sudden surges, special attention should be paid to their water saturation and porosity, and their hydrodynamic pressure should be measured to facilitate the assessment of the pressure and scale of potential sudden surges. The flow rate and direction of groundwater are particularly critical. Hydrological information at different depths is obtained through multi-point measurement, the infiltration path along the tunnel is determined, and areas with unstable hydrological conditions are identified.

[0145] After completing the data collection, the data are integrated into the surge risk identification model, and compared and analyzed using historical geological data and records of surge accidents. The model is established based on numerical analysis methods, based on the seepage path and water pressure distribution in different strata, combined with possible excavation depths and tunnel design parameters, to predict the location and amount of water where surges may occur. By identifying the surge risk points, the model can provide the construction team with risk warnings during the excavation process and recommend possible protective measures. In addition, the model analysis can also evaluate the pressure range of the surrounding soil and support structures once a surge occurs, providing a basis for the subsequent support structure design and construction plan optimization. In the actual construction process, this surge risk identification model can be updated at any time, especially when the hydrological conditions change, the risk can be re-evaluated in time, so as to adjust the surge protection plan in real time.

[0146] Implementation method 2:

[0147] Tunnel support structures often face complex and high-intensity water pressure shocks under shallow burial and high water level conditions. Therefore, before the excavation of the project, it is necessary to simulate and optimize the water pressure bearing of the tunnel support structure design based on geological survey data. This process first simulates the stress of the tunnel support structure under different water pressure conditions according to the tunnel design parameters (such as excavation section, support form, tunnel span, etc.) and the construction plan, especially under the extreme water pressure conditions when the sudden surge occurs, to analyze the bearing capacity of the support structure. For the main support materials in the tunnel, including steel arches, concrete linings, shotcrete layers, etc., the stress and deformation of the support structure are calculated through numerical simulation, and the safety factor of the structure is verified to determine whether it can withstand the expected water pressure load. The results of numerical simulation can help determine the optimal configuration of the tunnel support structure and take reinforcement measures in weak areas.

[0148] In areas with high risk of sudden surges, water pressure is concentrated and varies greatly. If the support design is insufficient, tunnel excavation may cause sudden surges and collapse. For this reason, special attention should be paid to the ultimate water pressure bearing range around the tunnel when optimizing the support design, and appropriate support materials and thickness should be selected in advance to ensure the stability of the structure in a high-pressure water environment. During the optimization process, the geological structure and support material parameters of the tunnel are imported into the simulation program using finite element analysis technology to analyze the support stress under various extreme conditions. In the support optimization design, reinforcement measures such as high-strength shotcrete and thickened arch frames can be considered to improve the bearing capacity of the support structure. If necessary, the pipe shed method or freezing method can also be used to reinforce the soil and support in areas with high risk of sudden surges to prevent support instability caused by sudden surges.

[0149] In addition, during the actual construction process, the optimal design of the support structure needs to be combined with the real-time monitoring data of hydrological conditions. In particular, when the pressure or water volume of the sudden surge exceeds expectations, the support design needs to be adjusted in a timely manner based on the real-time monitoring feedback. For example, if a sharp rise in water pressure in the tunnel is detected, the support thickness can be increased or temporary reinforcement facilities can be added to ensure the overall stability of the tunnel structure. Through the hydraulic bearing simulation and design optimization of the support structure, the most suitable support scheme can be accurately determined before construction, which can control construction costs and improve construction efficiency while ensuring the safety of the tunnel structure.

[0150] Implementation method three:

[0151] During the tunnel excavation process, in order to dynamically grasp the groundwater conditions and the bearing conditions of the support structure, the risk assessment accuracy during the tunnel construction process can be further improved by installing a real-time monitoring system and an intelligent risk prediction system. The real-time monitoring system includes water level monitors, water pressure sensors, soil displacement sensors and other devices. These devices can accurately monitor the groundwater level, seepage velocity and stress changes of the support structure around the tunnel, and provide real-time feedback on the changing trends of hydrological conditions. These monitoring data will be transmitted to the intelligent risk prediction system in real time, which will analyze the data, combine the previous geological model and support structure simulation results, and use machine learning algorithms to determine whether there is a sudden surge risk at present.

[0152] The risk prediction system is data-driven. It combines real-time monitoring data with historical data through machine learning algorithms to automatically identify potential high-risk areas for sudden surges and predict water pressure fluctuations. Especially when groundwater pressure suddenly increases or flow velocity changes abnormally, the algorithm can instantly capture these changes and predict their impact on the support structure. Once it detects that the sudden surge risk value exceeds the threshold, the system will automatically issue an early warning to remind construction personnel to take protective measures in advance. For example, the system can prompt the addition of temporary support structures or strengthening grouting in areas prone to sudden surges to prevent collapse caused by sudden surges. In addition, the risk prediction system also has a self-learning function that can continuously optimize model parameters based on real-time feedback, thereby improving the accuracy and timeliness of sudden surge predictions.

[0153] The intelligent system also integrates a multivariate decision-making algorithm to optimize the support structure in real time based on monitoring feedback. For example, when the risk prediction system issues an early warning, the system will dynamically adjust the support plan, including support thickness, material strength, reinforcement method, etc., based on the current hydrological monitoring data and support material strength to cope with sudden changes in water pressure and soil surges. By linking the support design with real-time data, the system can enhance the support strength when the risk of surges increases, ensuring the safety of tunnel construction. Ultimately, this real-time monitoring and intelligent risk prediction system can effectively reduce the incidence of surge accidents during construction, achieve dynamic optimization and adjustment of tunnel structures, and provide scientific support for the long-term stability of tunnel projects.

[0154] The present invention can effectively improve the accuracy of sudden surge risk identification during tunnel excavation through detailed geological surveys and hydrological surveys combined with historical sudden surge data. The seepage numerical model is constructed using high-precision geological and hydrological data collected on site, so that the groundwater seepage path, pressure concentration areas and sudden surge high-incidence locations can be accurately predicted. This risk identification method based on field data can fully evaluate the sudden surge risk of shallow buried high-water level sand layers and pebble layers before excavation, and provide data support and risk warning for subsequent support structure design. Compared with traditional empirical methods or simple model analysis methods, this method intuitively displays the hydraulic characteristics of the excavation area through numerical simulation, greatly improving the accuracy and effectiveness of sudden surge prediction. In addition, the risk assessment data can be updated in real time when the hydrological conditions or excavation area change, ensuring the timeliness and scientificity of the prediction, so as to formulate sudden surge protection measures in advance and avoid the safety hazards caused by sudden water surge events.

[0155] The present invention can significantly enhance the adaptability of tunnel support structures in complex hydrological environments through the simulation and optimization design of water pressure bearing of support structures. Water pressure bearing simulation can accurately calculate the stress distribution of support structures under different water pressure conditions, clarify the ultimate bearing capacity and design boundaries of each layer of support materials, and ensure the stability of support structures under high-pressure water or surge conditions. This optimization design method can identify support weaknesses and adjust material strength before excavation construction, thereby avoiding surge pressure exceeding the design limit of the support structure and preventing collapse accidents caused by support failure. Real-time monitoring data also supports dynamic adjustment of support design. In the event of an unexpected increase in surge pressure, the impact of surge pressure can be dealt with by increasing support thickness or temporary reinforcement measures. This intelligent and flexible support design not only improves construction safety, but also effectively avoids material waste caused by excessive support design, making the support structure both reliable and economical.

[0156] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for evaluating the excavation stability of shallow-buried biased double-arch tunnels in floodplains, characterized in that: The following steps are involved: Through on-site survey and drilling, the geological conditions of the tunnel excavation area are obtained, and the groundwater outburst volume and probability of sudden inrush are evaluated in combination with historical data, providing basic data support for subsequent risk modeling; Based on the collected geological condition data, a groundwater seepage numerical model was constructed to simulate the dynamic behavior of shallow groundwater during tunnel excavation. The water pressure distribution at different strata was obtained through model analysis, providing data support for the judgment of sudden surge risks during excavation. According to the tunnel design parameters and excavation plan, simulate the stress of the support structure under different water pressures, analyze the ultimate water pressure that the support structure can withstand, and provide a basis for support selection and optimization in tunnel excavation; Using simulation data, we built a risk assessment model based on machine learning, dynamically predicted the risk of sudden surges based on real-time monitoring data, identified high-risk sudden surge points, and provided early warning for tunnel construction; Based on the results of sudden surge risk prediction, a multivariable optimization algorithm is used to dynamically optimize the support structure to determine the support strength and material configuration. At the same time, it is linked with the on-site monitoring data to adjust the support structure in real time to ensure the safety and stability of the tunnel structure.

2. The method for evaluating excavation stability of shallow-buried bias-loaded double-arch tunnels in floodplains according to claim 1 is characterized in that: The specific steps to evaluate groundwater outburst volume and probability of sudden inrush in combination with historical data and provide basic data support for subsequent risk modeling are as follows: First, a comprehensive geological survey of the tunnel excavation area is carried out; After the risk areas are identified, a detailed drilling layout plan is developed based on the results of the geological survey; Gain in-depth understanding of groundwater flow direction and pressure changes, install groundwater level monitoring devices in drilling holes, regularly record water level elevation and flow rate data, and analyze the dynamic changes of groundwater; Based on the field survey and drilling data, combined with historical records of sudden surge events, the conditions and frequency of sudden surges in the tunnel excavation area are analyzed.

3. The method for evaluating excavation stability of shallow-buried bias-loaded double-arch tunnels in floodplains according to claim 1 is characterized by: The specific steps to obtain the water pressure distribution at different strata through model analysis and provide data support for the judgment of sudden surge risk during excavation are as follows: Before starting the model construction, the collected geological and hydrological data are first processed; After the data processing is completed, a matching numerical model is selected for seepage simulation; After the model is constructed, the calculation phase begins, simulating the flow dynamics of groundwater during tunnel excavation; After the calculation is completed, the model results are calibrated with the field monitoring data to ensure the consistency of the model output with the actual situation.

4. The method for evaluating excavation stability of shallow-buried bias-loaded double-arch tunnels in floodplains according to claim 1 is characterized by: The specific steps to analyze the ultimate water pressure of the support structure and provide a basis for support selection and optimization in tunnel excavation are as follows: First, detailed data collection of the tunnel support structure is carried out. The acquired data will provide basic input for the model. After obtaining the key parameters of the support structure, the force model of the support structure is established using the finite element analysis method; After the model is established, the limit state analysis is performed by applying different water pressure states to calculate the maximum bearing capacity and limit pressure of the support structure when the water pressure changes; According to the results of the limit state analysis, the design parameters of the support structure are adjusted, and the support materials and structural forms are optimized to ensure the safety of tunnel excavation under sudden surge conditions.

5. The method for evaluating excavation stability of shallow-buried bias-loaded double-arch tunnels in floodplains according to claim 1 is characterized by: The steps to dynamically predict the surge risk based on real-time monitoring data, identify high-risk surge points, and provide early warning for tunnel construction are as follows: The acquired geological parameters, support structure stress data, and sudden surge event data collected in historical projects are preprocessed. To ensure the sensitivity of the model to the high-risk areas of sudden surge, a multidimensional feature matrix X is defined. The acquired water pressure distribution is calibrated as φ(x, y, z, t), the support structure force distribution is calibrated as σ(x, y, z), and the formation parameters are calibrated as ψ(x, y, z). The matrix characteristics are as follows: X=[ϕ(x,y,z,t) σ(x,y,z) ψ(x,y,z)] T , Where X is a multidimensional characteristic matrix, φ(x, y, z, t) is the water pressure distribution, which indicates the water pressure value at each location in the tunnel excavation area, and its coordinates are represented by (x, y, z), t is the time point, σ(x, y, z) is the force distribution of the support structure, which indicates the stress state at each point of the tunnel support structure, and its three-dimensional coordinates are defined by (x, y, z), σ is in tensor form, ψ(x, y, z) is the formation parameter, which indicates the characteristic parameter matrix of the formation around the tunnel, and the spatial position is represented by (x, y, z), ψ is a variety of physical parameters of the formation, and T is the matrix transpose; In order to improve the model's ability to identify sudden surge risks under complex geological conditions, the tensor decomposition method is used to extract high-order features, and the multidimensional feature matrix X is expressed as a third-order tensor: Where X is the original third-order tensor, represents the real number domain, indicating that all data in the tensor are real numbers, M is the spatial position dimension, N is the feature type dimension, representing the second dimension of the tensor, and P is the time dimension, representing the third dimension of the tensor; The multi-level feature combination is obtained by tensor decomposition, and the expression is as follows: ， In the formula, is the third-order tensor decomposition representation, λ k is the weight coefficient, u k is the space dimension vector, v k is the feature type dimension vector, w k is the time dimension vector; The machine learning model is trained using an adaptive weighted iterative method. The weights are adjusted to optimize the recognition accuracy of high-risk areas for sudden surges. The prediction error deviation term is defined as follows: In the formula, ΔR t is the forecast error bias term, y t is the actual surge risk label, is the model prediction value; The model updates the loss function through an adaptive weight factor, and the calculation expression is as follows: ， Where L(w) is the loss function, which is used to measure the deviation between the accuracy of the model prediction and the actual value, w is the weight vector of the model, and ω t is the weighting factor, f(ΔR t ) is the error function, which represents the prediction deviation at measurement time t, λ||w|| 2 is the regularization term, λ is the regularization coefficient; In the model deployment stage, the surge risk is dynamically predicted by inputting real-time monitoring data to generate the surge risk probability, which is calculated as follows: ， In the formula, is the surge risk probability, X real (t) is the real-time feature matrix, X real (t) = [φ real (t) σ real (t) ψ real (t)] T , including water pressure, stress and formation parameters at time t, b is the bias term of the model; Based on surge risk probability Define the dynamic risk threshold θ(t), when When , the warning is triggered, the model parameters are updated, and the dynamic risk threshold calculation expression is as follows: ， In the formula, θ(t) is the dynamic risk threshold, β is the smoothing factor, and β·θ(t-1) represents the smooth continuation of the threshold at the previous moment. Represents the dynamic update of the current predicted risk probability.

6. The method for evaluating excavation stability of shallow-buried bias-loaded double-arch tunnels in floodplains according to claim 1 is characterized by: According to the results of the sudden surge risk prediction, the specific steps for real-time adjustment of the support structure to ensure the safety and stability of the tunnel structure are as follows: Initialize the cost of supporting materials, strength of supporting materials, surge risk pressure and construction feasibility, and mark them as C i , S i , P i and D i , C i Represents the cost coefficient of unit support material, S i is the strength coefficient of the unit support material, P i is the surge risk pressure coefficient, which indicates the intensity of groundwater pressure, D i is the construction feasibility coefficient, which reflects the difficulty of supporting structure installation. The variable matrix F = {f i }={f1,f2,……,f n }, where f i represents the strength and performance combination of the i-th layer of support materials; According to the actual support requirements, the objective function is constructed, and the cost, strength, risk coefficient and other parameters of the support structure are associated. The nonlinear multi-objective optimization function is adopted, and the expression is as follows: ， Where G(F) is the optimization objective function, representing the total cost and risk index of the tunnel support structure, and C i f i is the total cost of the i-th layer of support materials, ω is the risk penalty coefficient, which is used to adjust the influence weight of the sudden surge risk in the optimization process, n is the total number of layers of strength and performance combination of support materials, τ is the construction difficulty penalty coefficient, which is used to adjust the influence of construction difficulty in optimization, and ∈ is a constant; In order to ensure the stability of the support structure, the following nonlinear constraints must be met: g i (f i )=f i S i -P i ≥0,i=1,2,……,n, In the formula, g i (f i ) is the stability constraint function of the i-th support layer, f i S i Indicates the actual support strength provided by the i-th layer of support material; The constraint function ensures that the support structure does not lose stability under the action of water pressure. In addition, in order to reduce material costs, the total amount of constrained support materials is calculated as follows: ， Where, T budget is the total budget for support costs; Surge risk data R based on real-time feedback from on-site monitoring data i (t), dynamically adjust the support strength matrix G(F), combine the Lagrange multiplier method and the dynamic weight adjustment strategy, update the optimization scheme, and the calculation expression is as follows: ， In the formula, f i (t+1) is the value of the support strength at the i-th position at the next time step t+1, f i (t) is the support strength variable at the i-th position, its value at the current time point t, μ is the learning rate, which is used to control the optimization step length parameter, θ is the risk feedback coefficient, which is used to control the influence of real-time monitoring data in the optimization process, R i (t) is the real-time surge risk data of the ith location at time t.

Images