Anti-floating method and system for large-diameter shield tunnel based on anti-floating coefficient

Abstract

The application provides a large-diameter shield tunnel anti-floating method and system based on an anti-floating coefficient, and belongs to the technical field of tunnel anti-floating. The method comprises the following steps: obtaining static and dynamic multi-dimensional anti-floating parameters, calculating the anti-floating coefficient after preprocessing, setting an early warning value and selecting an anti-floating pile arrangement scheme according to the anti-floating coefficient, and optimizing the pile arrangement design by using a genetic algorithm; reinforcing the stratum by differential grouting in different zones, and optimizing the parameters of the steel frame and the prestressed anchor cable in combination with the genetic algorithm for the portal structure to construct a collaborative anti-floating reinforcement system; carrying out real-time anti-floating monitoring, establishing a tunnel anti-floating digital twin based on data, performing state prediction and system dynamic optimization, and outputting the final reinforcement result. The application effectively solves the problem that the static value of stratum resistance in the traditional method does not match the actual construction disturbance, and significantly improves the accuracy and engineering applicability of the anti-floating design.

Description

Technical Field

[0001] This invention relates to a method and system for anti-buoyancy of large-diameter shield tunnels based on an anti-buoyancy coefficient, belonging to the field of tunnel anti-buoyancy technology. Background Technology

[0002] Large-diameter shield tunnels are widely used in major engineering projects such as river-crossing and sea-crossing channels and underground utility tunnels due to their advantages such as strong crossing capacity, high construction efficiency, and minimal impact on the ground environment. However, in water-rich strata or under conditions of high groundwater head, the tunnel structure is prone to floating deformation due to buoyancy. This can lead to leakage at the segment joints and structural cracking, or even cause tunnel axis deviation and construction accidents. Therefore, anti-buoyancy design has become a core technical challenge in large-diameter shield tunnel engineering.

[0003] Traditional anti-buoyancy methods for large-diameter shield tunnels mainly rely on single or combined anti-buoyancy measures. For example, patent CN118309469A discloses a method for suppressing segment floatation during large-diameter shield tunneling. This method mainly addresses the segment floatation problem during shield tunneling, but it does not establish a complete anti-buoyancy coefficient calculation system, making it impossible to comprehensively assess anti-buoyancy safety. The anti-buoyancy measures are only for the tunneling stage and do not consider ground reinforcement, portal frame structure coordination, etc., making it difficult to meet the anti-buoyancy requirements throughout the tunnel's entire life cycle. In addition, the anti-buoyancy coefficient calculation often only considers basic parameters such as the self-weight of the segments and the self-weight of the overlying soil, ignoring dynamic factors such as the instantaneous buoyancy increase caused by construction vibration, leading to inaccurate calculations. The calculated results deviate significantly from the actual engineering situation. For example, patent CN112818565A calculates the total weight of components, the weight on the cantilever slab, buoyancy, and anti-buoyancy coefficient by creating a three-dimensional water tank model in Revit. However, this patent is only applicable to underground water tank scenarios. In large-diameter shield tunnels, traditional anti-buoyancy coefficient calculations often only consider basic parameters such as the self-weight of the tunnel segments and the self-weight of the overlying soil layer, ignoring dynamic factors such as temporary construction loads, additional forces from groundwater seepage, frictional resistance of tunnel segment joints, and instantaneous buoyancy increments caused by construction vibrations. This leads to a significant deviation between the calculated results and the actual engineering situation. At present, there is a need for an anti-buoyancy method and system for large-diameter shield tunnels based on the anti-buoyancy coefficient. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for anti-buoyancy of large-diameter shield tunnels based on the anti-buoyancy coefficient, which solves the problems of insufficient accuracy in calculating the anti-buoyancy coefficient and poor coordination of anti-buoyancy measures in the prior art.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] Firstly, a method for preventing buoyancy in large-diameter shield tunnels based on an anti-buoyancy coefficient is provided, including the following steps:

[0007] Obtain multi-dimensional anti-buoyancy parameters, including static anti-buoyancy parameters and dynamic anti-buoyancy parameters;

[0008] The static anti-buoyancy parameters are preprocessed, and the anti-buoyancy coefficient is calculated based on the preprocessed static anti-buoyancy parameters;

[0009] Based on the anti-buoyancy coefficient, an early warning value is set, an anti-buoyancy pile layout scheme is selected according to the early warning value, and the anti-buoyancy pile layout is optimized by a genetic algorithm.

[0010] Based on the optimized anti-buoyancy pile layout, a zoned differentiated grouting method was adopted to reinforce the stratum, and a collaborative anti-buoyancy reinforcement system was established by combining genetic algorithms to optimize the combination parameters of steel frame and prestressed anchor cable for portal frame structure.

[0011] Based on the aforementioned collaborative anti-buoyancy reinforcement system, anti-buoyancy data monitoring is conducted, and a digital twin image of the tunnel's anti-buoyancy is constructed based on the monitoring data, and state prediction is performed.

[0012] The collaborative anti-buoyancy reinforcement system is optimized based on the monitoring data and state prediction results output by the digital twin mirror, and the final reinforcement result is output.

[0013] Preferably, the static anti-buoyancy parameters include: the self-weight of the tunnel segments, the self-weight of the overlying soil layer, the shear zone stratum resistance, the frictional resistance of the tunnel segment joints, and the buoyancy force on the tunnel.

[0014] The dynamic anti-buoyancy parameters include: anti-buoyancy force of the anti-buoyancy pile, temporary construction load, additional force from groundwater seepage, and instantaneous buoyancy increment caused by construction vibration.

[0015] Preferably, the static anti-buoyancy parameter preprocessing involves dynamically correcting the initial resistance of the shear zone strata by introducing a calculation model for the radius of the disturbance zone during shield tunneling construction, as shown in the following formula:

[0016] ,

[0017] ,

[0018] in, For the corrected shear zone formation resistance, The initial resistance of the shear zone strata. Formation disturbance coefficient Distance from tunnel axis The internal friction angle of the strata, Distance from tunnel axis The cohesion of the strata, The area of ​​action of the shear band. Tunnel structure surface =0 initial disturbance coefficient The attenuation coefficient is... The radius of the disturbance zone.

[0019] Preferably, the anti-buoyancy coefficient is the ratio of dynamic total anti-buoyancy force to total buoyancy force, wherein the total buoyancy force includes tunnel buoyancy force and instantaneous buoyancy force increment due to construction vibration, and the dynamic total anti-buoyancy force includes the self-weight of the tunnel segment, the self-weight of the overlying soil layer, the corrected shear zone stratum resistance, the anti-buoyancy force of the anti-buoyancy pile, the temporary construction load, the frictional resistance of the tunnel segment joints, and deducts the additional force of groundwater seepage.

[0020] Preferably, the step of selecting the anti-buoyancy pile arrangement scheme based on the early warning value includes:

[0021] Set low warning value and high warning value ;

[0022] When the anti-buoyancy coefficient is greater than or equal to the low warning value and less than the high warning value, conventional anti-buoyancy pile arrangement shall be adopted.

[0023] When the anti-buoyancy coefficient is less than the low warning value, the buoyancy resistance of a single pile is calculated, and the total buoyancy resistance of the pile group is calculated based on the buoyancy resistance of the single pile and the interaction between the anti-buoyancy piles, as shown in the following formula:

[0024] ,

[0025] ,

[0026] ,

[0027] in, The interaction coefficient, The circumference of the pile body For the pile length, and These are the characteristic values ​​of pile side friction and pile end resistance, respectively. The cross-sectional area of ​​the pile tip. For the first The buoyancy resistance of a single pile, This represents the total number of piles. For the total buoyancy resistance of the pile group, The pile spacing is... The diameter of the pile;

[0028] The corrected dynamic total buoyancy is obtained by calculating the sum of the self-weight of the tunnel segment, the self-weight of the overlying soil layer, the corrected shear zone stratum resistance, the total buoyancy of the pile group, the temporary construction load, and the frictional resistance of the tunnel segment joints, and subtracting the additional force of groundwater seepage. The buoyancy coefficient is calculated based on the corrected dynamic total buoyancy. If the buoyancy coefficient is not in the range of greater than or equal to the low warning value and less than the high warning value, the design index of the piles related to the interaction coefficient is replanned until the buoyancy coefficient is greater than or equal to the low warning value and less than the high warning value.

[0029] When the anti-buoyancy coefficient is greater than or equal to the high warning value, the number of piles is reduced by 10% increments until the anti-buoyancy coefficient falls back below the high warning value.

[0030] Preferably, the specific method for optimizing the early warning value of the anti-buoyancy pile layout using a genetic algorithm includes:

[0031] The tunnel cross-section is divided into a grid, and the coordinates of the pile center are used as genes for binary encoding to form an initial population.

[0032] The dual objective functions are maximizing total buoyancy resistance and minimizing material usage. The constraints include a pile spacing greater than or equal to three times the pile diameter, a buoyancy resistance coefficient greater than or equal to the low warning value, and a pile top elevation not higher than the bottom elevation of the tunnel segment. The selection operator adopts the tournament selection method, the crossover operator adopts arithmetic crossover, and the mutation operator adopts Gaussian mutation. After iteration, the optimal pile position coordinate matrix that satisfies the dual objectives is output.

[0033] The pile positions are determined based on the optimal pile position coordinate matrix, and adaptive adjustments are made according to the stress characteristics of the strata in different areas of the tunnel cross section. The density of anti-buoyancy piles is increased in the arch area with a central angle between 60 and 90 degrees, the pile spacing is set at 3 to 4 times the pile diameter, and the pile body is lengthened, with an embedment depth of greater than or equal to 5 meters into stable strata. The pile bodies in the sidewall area with a central angle between 90 and 120 degrees are set outward from the tunnel according to a lateral dip angle of 5 to 10 degrees. The pile length is reduced in the bottom area with a central angle between 120 and 150 degrees, the embedment depth in stable strata is greater than or equal to 3 meters, and the pile spacing is set at 4 to 5 times the pile diameter.

[0034] Preferably, the method of reinforcing the strata using differentiated grouting in different zones, and the optimization of the combination parameters of the steel frame and prestressed anchor cables using a genetic algorithm for the portal frame structure, are as follows:

[0035] The ground reinforcement area is divided into a core area and a transition area. The core area is the area around the anti-buoyancy pile and the tunnel wall to the inside of the anti-buoyancy pile. The transition area is the area outside the core area to the boundary of the disturbance area. The core area adopts split grouting and the transition area adopts permeation grouting.

[0036] Grouting parameters are optimized using a backpropagation (BP) neural network. The input layer of the BP neural network comprises eight neurons, and its input features include soil physical parameters, anti-buoyancy pile layout parameters, a synergy coefficient, and interaction features. The soil physical parameters include initial unit weight and permeability coefficient. The anti-buoyancy pile layout parameters include average spacing and pile diameter. The synergy coefficient is the ratio of the buoyancy resistance of a single anti-buoyancy pile to the initial resistance of the corresponding soil layer. The output layer of the BP neural network comprises eight neurons, and its output features include key grouting parameters and predicted results. The key grouting parameters include grouting pressure, grouting flow rate, and water-cement ratio. The predicted results include the increase in soil unit weight and compressive strength after reinforcement. The feedback function for the output layer parameters is as follows:

[0037] ,

[0038] in, This is expressed as the grouting pressure after dynamic calibration. This is represented by the initial grouting pressure output by the neural network. This is represented as a correction factor. This is represented as the preset target value for soil weight increment. This is represented as the predicted value of soil weight increment output by the neural network.

[0039] The BP neural network employs an engineering constraint regularization term to improve the loss function. :

[0040] ,

[0041] in, This is represented as a penalty coefficient. These are actual observed values. These are the predicted values ​​from the BP neural network. This is the preset target value for soil weight increment, which is the minimum effective threshold for soil reinforcement required by the project. The predicted value of soil weight increment output by the BP neural network;

[0042] A genetic algorithm was used to optimize the combined parameters of the steel frame and prestressed anchor cables, with the dual objectives of maximizing pull-out resistance and minimizing material usage. The optimization variables included: the cross-sectional height and width of the steel frame, the web thickness, the number of bolts, the anchor cable diameter, and the tension control stress. The objective function was:

[0043] ,

[0044] ,

[0045] in, For total pull-out force, For the pull-out resistance of the steel frame, The total pull-out force of the anchor cable. For the quantity of steel, For the number of doors, For the number of anchor cables, This refers to the number of bolts.

[0046] The constraints are: the stress in the steel frame is less than or equal to the design strength of the steel, the tension control stress of the anchor cable is greater than or equal to 0.75 times the standard value of the tensile strength, and the bearing capacity of the connection node between the steel frame and the anchor cable is greater than or equal to 1.2 times the design pull-out force.

[0047] The selection operator uses the roulette wheel selection method, the crossover operator uses single-point crossover, and the mutation operator adds ±5% random perturbation to the variables.

[0048] Preferably, the step of monitoring anti-buoyancy data based on the collaborative anti-buoyancy reinforcement system and constructing a digital twin image of the tunnel's anti-buoyancy based on the monitoring data is carried out in the following manner:

[0049] Collect data on single pile axial force, pile top displacement, soil pressure and moisture content at different depths, steel frame stress, and anchor cable tension to form a time series dataset;

[0050] The collected data is fused with the calculated values ​​of the initial finite element model using the Kalman filter algorithm to achieve data assimilation.

[0051] The key parameters of the digital twin image are dynamically updated based on the assimilated data.

[0052] Predicting buoyancy resistance using long short-term memory networks.

[0053] Preferably, the optimization of the collaborative anti-buoyancy reinforcement system based on the monitoring data and state prediction results output by the digital twin mirror specifically includes:

[0054] When the anti-buoyancy coefficient is insufficient, a layered optimization algorithm is adopted. The first layer optimizes the load distribution ratio between the anti-buoyancy piles and the portal structure by adjusting the anchor cable tension control stress, so that the ratio of the total axial force of the anti-buoyancy piles to the total pull-out force of the portal structure is maintained in a reasonable range of 2:1 to 3:1. The calculation formula is as follows:

[0055] ,

[0056] in, This is the anchor cable tension adjustment value. As a safety threshold, To enhance the buoyancy resistance of anti-buoyancy piles, For the number of anchor cables, Given the cross-sectional area of ​​a single anchor cable, if the anti-buoyancy coefficient still does not meet the standard after adjusting the anchor cable, a second layer of optimization is performed. The number of anti-buoyancy piles is supplemented by a genetic algorithm, with the dual objectives of maximizing the new anti-buoyancy force and minimizing the new cost. The constraint is that the distance between the new pile position and the existing pile is greater than or equal to twice the pile diameter, and the optimal supplementation scheme is output.

[0057] When local stress concentration occurs, topology optimization algorithms are used to adjust the parameters of the portal frame structure, including:

[0058] For areas of steel frame where stress exceeds limits, stress can be reduced by adding stiffeners or adjusting cross-sectional dimensions. The corrected formula is as follows:

[0059] ,

[0060] in, To optimize the trailing edge width, The original wing width, These are the measured stress values ​​in the stress-over-limit area of ​​the steel frame. Design strength for steel frame stress;

[0061] For uneven anchor cable tension distribution, force flow balance is achieved by adjusting the tensioning sequence. A greedy algorithm is used to adjust the tension of each anchor cable in descending order of tension deviation rate until the standard deviation of the measured anchor cable tension is less than 0.1.

[0062] When the reinforcement effect of the stratum diminishes, the feedback correction mechanism based on the BP neural network initiates the optimization of secondary grouting parameters. The current stratum moisture content, soil pressure and primary grouting parameters are input into the model, and the pressure increment and water-cement ratio adjustment value of secondary grouting are output.

[0063] Secondly, a large-diameter shield tunnel anti-buoyancy system based on an anti-buoyancy coefficient is provided, including:

[0064] The data acquisition module is configured to acquire multi-dimensional anti-buoyancy parameters, including static anti-buoyancy parameters and dynamic anti-buoyancy parameters;

[0065] The preprocessing module is configured to preprocess the static anti-buoyancy parameters and calculate the anti-buoyancy coefficient based on the preprocessed static anti-buoyancy parameters, including introducing a calculation model for the radius of the shield tunneling disturbance zone to dynamically correct the shear zone stratum resistance.

[0066] The conversion module is configured to construct a support model of anti-buoyancy piles and strata based on the anti-buoyancy coefficient, including setting an anti-buoyancy safety threshold, calculating the anti-buoyancy force of a single pile based on the safety threshold, and optimizing the arrangement of anti-buoyancy piles using a genetic algorithm.

[0067] The reinforcement module is configured to use anti-buoyancy pile layout parameters to perform coordinated anti-buoyancy reinforcement of the strata and portal structure, including reinforcing the strata by using zoned differentiated grouting methods, and optimizing the combination parameters of the steel frame and prestressed anchor cables for the portal structure using a genetic algorithm.

[0068] The digital twin module is configured to monitor anti-buoyancy data based on the collaborative anti-buoyancy reinforcement system, and to construct a digital twin image of the tunnel's anti-buoyancy based on the monitoring data;

[0069] The output module is configured to optimize the monitoring data and status prediction results output by the digital twin mirror, and output the final hardening result.

[0070] The advantages of this invention are as follows: By constructing a dynamic correction mechanism for the shear zone stratum resistance, this invention introduces a calculation model for the radius of the shield tunneling disturbance zone and a decay function that varies with distance, and dynamically corrects the initial resistance of the shear zone stratum. This solves the problem of mismatch between the static value of stratum resistance and the actual construction disturbance in traditional anti-buoyancy calculations, making the calculation of the anti-buoyancy coefficient more consistent with engineering practice and providing a precise basis for subsequent anti-buoyancy design.

[0071] This invention constructs a support model for anti-buoyancy piles and strata based on the anti-buoyancy coefficient, optimizes the arrangement of anti-buoyancy piles using a genetic algorithm, and makes adaptive adjustments to address the differences in the stress characteristics of strata in different regions. Simultaneously, it utilizes the anti-buoyancy pile arrangement parameters to perform synergistic anti-buoyancy reinforcement of the strata and portal structure. Through zoned differentiated grouting and optimization of the combination parameters of the steel frame and prestressed anchor cables, a three-in-one synergistic anti-buoyancy system is formed, from the anti-buoyancy piles to the strata and finally to the portal structure, significantly improving the overall bearing capacity and stability of the anti-buoyancy system.

[0072] This invention constructs a digital twin image of tunnel anti-buoyancy based on monitoring data of a collaborative anti-buoyancy reinforcement system. Through data assimilation using the Kalman filter algorithm, dynamic parameter updates, and long short-term memory network state deduction, it achieves real-time mapping between physical entities and virtual models and accurate prediction of future anti-buoyancy states, providing digital decision support for the dynamic adjustment of the anti-buoyancy system.

[0073] The BP neural network constructed in this invention achieves intelligent optimization and dynamic calibration of grouting parameters through multi-dimensional input feature fusion, adaptive topology hidden layer, and dual output structure of parameters and effects. It overcomes the limitations of traditional empirical calculation, improves the adaptability and economy of stratum reinforcement parameters, and ensures that the reinforcement effect and the synergistic anti-buoyancy requirements are accurately matched.

[0074] This invention optimizes the output results of the digital twin mirror and adopts targeted measures such as hierarchical optimization algorithms. Under the premise of ensuring that the anti-buoyancy coefficient, component stress and ground reinforcement index meet the safety requirements, it achieves effective control of the whole life cycle cost, balances the safety and economy of the anti-buoyancy system, and significantly improves the risk resistance capability compared with traditional static design. Attached Figure Description

[0075] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0076] Figure 1 This is a schematic diagram of the overall process of an anti-buoyancy method for large-diameter shield tunnels based on an anti-buoyancy coefficient in an embodiment of the present invention.

[0077] Figure 2 This is a schematic diagram of the optimization logic of an anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient in an embodiment of the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Example 1

[0080] like Figure 1 As shown, a method for preventing buoyancy in large-diameter shield tunnels based on an anti-buoyancy coefficient includes the following steps:

[0081] S1: Obtain multi-dimensional anti-buoyancy parameters, including static anti-buoyancy parameters and dynamic anti-buoyancy parameters;

[0082] S2: Preprocess the static anti-buoyancy parameters and calculate the anti-buoyancy coefficient based on the preprocessed static anti-buoyancy parameters;

[0083] S3: Set an early warning value based on the anti-buoyancy coefficient, select an anti-buoyancy pile layout scheme according to the early warning value, and optimize the anti-buoyancy pile layout through a genetic algorithm;

[0084] S4: Based on the optimized anti-buoyancy pile layout, a zoned differentiated grouting method is adopted to reinforce the stratum, and a collaborative anti-buoyancy reinforcement system is established by combining genetic algorithm to optimize the combination parameters of steel frame and prestressed anchor cable for portal frame structure.

[0085] S5: Based on the aforementioned collaborative anti-buoyancy reinforcement system, conduct anti-buoyancy data monitoring, construct a digital twin image of the tunnel's anti-buoyancy based on the monitoring data, and perform state prediction;

[0086] S6: Based on the monitoring data and state prediction results output by the digital twin mirror, the collaborative anti-buoyancy reinforcement system is optimized, and the final reinforcement result is output.

[0087] As a refinement of the above embodiments, the static anti-buoyancy parameters in step S1 refer to the anti-buoyancy related parameters of the tunnel that do not change significantly with construction or time in a stable state, including the self-weight of the tunnel segments. The self-weight of the overlying soil layer Initial resistance of shear zone strata Frictional resistance of segment joints and the buoyancy of the tunnel Dynamic anti-buoyancy parameters refer to anti-buoyancy-related parameters affected by construction disturbances or time changes, mainly including the anti-buoyancy force of anti-buoyancy piles. Temporary construction loads Additional force from groundwater seepage and the instantaneous buoyancy increase caused by construction vibration .

[0088] The static anti-buoyancy parameters need to be determined comprehensively based on design data, geological surveys, and structural characteristics. Specifically, the calculation of the segment self-weight requires detailed breakdown using a BIM model, with each segment's weight calculated separately according to its ring sections. The self-weight of each segment is derived from the density of its material and the sum of the three-dimensional volumetric weights. ,in, For the first Material density of tunnel segments The volume of this tunnel segment is calculated using 3D dimensions extracted from a BIM model. The calculation of the overlying soil's self-weight requires consideration of strata characteristics. Thickness data for each soil layer is extracted from the geological survey report, combined with the corresponding natural unit weight of the soil layer, and then multiplied by the projected area of ​​the tunnel roof. The area of ​​a circular region calculated based on the tunnel's outer diameter is then calculated and summed layer by layer to obtain the total weight. This reflects the impact of differences in the unit weight of different soil layers on the overall self-weight. The shear zone resistance needs to be determined through geological surveys, typically within 1.5 times the tunnel's outer diameter. Then, based on the vertical effective stress, internal friction angle, and cohesion of each soil layer within this range, combined with the contact area between the soil layer and the tunnel segment, the resistance provided by each soil layer is calculated and summed to reflect the resistance of the strata around the tunnel to uplift within a specific area. ,in, For the first The effective vertical stress of the soil layer, and These represent the internal friction angle and cohesion of the soil layer, respectively. The contact area between the soil layer and the tunnel segment is given. The frictional resistance of the tunnel segment joints is calculated based on the joint type, taking into account the friction coefficient of the joint surface, which is usually taken as 0.3~0.4. This is added to the total preload of all bolts in each joint, and the frictional resistance of all joints is accumulated. The buoyancy force on the tunnel is calculated according to Archimedes' principle, by multiplying the specific weight of groundwater by the volume of water displaced by the tunnel, based on the outer contour dimensions of the tunnel, including the thickness of the tunnel segments and external protruding structures, to obtain the upward buoyancy force generated by groundwater on the tunnel under steady state.

[0089] Anti-buoyancy piles Before pile construction, the characteristic values ​​of pile perimeter, pile length, pile side friction, and pile end resistance are estimated based on design parameters. After pile construction, these are actually measured through static load tests to ensure that the data are dynamically updated with each construction stage. ,in, The circumference of the pile body For the pile length, and These are the characteristic values ​​of pile side friction and pile end resistance, respectively. This represents the cross-sectional area at the pile tip of the anti-buoyancy pile, and the anti-buoyancy force of the anti-buoyancy pile. This reflects the actual pull-out resistance that the anti-buoyancy piles can provide. Temporary construction loads include the weight of the tunnel boring machine (TBM), segment transport vehicles, hoisting equipment, and other temporary loads acting on the tunnel or its surroundings. These are calculated by accumulating the loads based on the equipment list and weight parameters in the construction organization design. The additional force from groundwater seepage needs to consider the effects of groundwater level fluctuations and seepage velocity, calculated according to Darcy's law, combined with the hydraulic gradient and the area of ​​the tunnel sidewall affected by seepage. ,in, For hydraulic gradient, The area affected by seepage through the tunnel sidewalls. The density of groundwater is taken as the buoyancy index. The additional force from groundwater seepage reflects the extra buoyancy generated by the flow of groundwater. The instantaneous buoyancy increment caused by construction vibration is obtained by collecting the peak vibration acceleration through vibration sensors installed on the tunnel boring machine, which is usually between 0.5 and 1.5 m / s². This is then multiplied by the total mass per unit length of the tunnel, including the mass of the tunnel segments, internal filling materials, and temporary storage materials, to obtain the instantaneous buoyancy increment caused by construction vibration. ,in, Mass per unit length of tunnel To measure the peak vibration acceleration and reflect the impact of dynamic disturbances on buoyancy, static and dynamic parameters are associated and stored according to spatial location, such as segment ring number, formation depth, and time node, forming a multi-dimensional anti-buoyancy parameter database that can be dynamically called, providing basic data for subsequent anti-buoyancy coefficient calculation.

[0090] As a refinement of the above embodiments, the core of step S2 in preprocessing the static anti-buoyancy parameters lies in dynamically correcting the initial resistance of the shear zone strata to eliminate the impact of construction disturbances on the mechanical properties of the strata, specifically including:

[0091] S201: Determine the radius R of the disturbance zone during shield tunneling based on the tunnel diameter D. In this embodiment, we take... Within this radius, the strata undergo stress redistribution due to construction activities such as shield tunneling and cutterhead cutting, resulting in a decrease in their mechanical parameters compared to their natural state. Therefore, a strata disturbance coefficient is defined within the disturbance radius. Used to characterize the degree of disturbance to strata at different locations:

[0092] ,

[0093] in, To calculate the horizontal distance of the point from the tunnel axis, The attenuation coefficient is... Tunnel structure surface =0 initial disturbance coefficient The value of follows the law of attenuation with increasing distance, and the disturbance is most significant at the surface of the tunnel structure where x=0. The maximum value is taken as 0.6~0.8; the disturbance is minimized at the boundary of the disturbance zone, i.e., when x equals the disturbance radius. The coefficient is reduced to 0.1-0.2, and the intermediate region is smoothly transitioned through linear interpolation or an exponential function, forming a decay function that varies with distance. This ensures that the disturbance coefficient can truly reflect the spatial distribution differences in the mechanical properties of the strata. Specifically, the strata disturbance coefficient uses an exponential function to simulate this decay process. The decay rate of the exponential function is initially fast and then slows down, consistent with the propagation law of actual construction disturbances. Near the tunnel area... The disturbance intensity decreases rapidly with increasing distance, especially in the far tunnel region. The disturbance tends to level off and eventually stabilizes at the residual disturbance level. From the perspective of the mechanism of action, pass This correction term is directly involved in the calculation of formation resistance.

[0094] S202: Based on the above disturbance coefficients, the calculation of the corrected shear zone formation resistance needs to cover the area affected by the shear zone, which is achieved through integration:

[0095] ,

[0096] in, This refers to the shear zone area, which is the contact area between the strata and the tunnel segments within the disturbance radius of the tunnel top and sides. Distance from tunnel axis The internal friction angle of the strata, Distance from tunnel axis The cohesion of the strata, For the initial resistance of the shear zone strata, at each point within this region, the initial resistance of the shear zone strata is compared with... Multiply to correct for the resistance reduction caused by the disturbance, and then combine with the internal friction angle of the formation at that point. With cohesion The resistance per unit area is calculated, and finally integrated over the entire area of ​​action to obtain the corrected shear zone stratum resistance. This correction process fully considers the weakening effect of construction disturbance on stratum resistance, making the static parameters more consistent with the actual stress state during the construction stage.

[0097] S203: Calculating the buoyancy resistance coefficient first requires clarifying the composition of total buoyancy and dynamic total buoyancy resistance. Total buoyancy includes two parts: one is still water buoyancy, calculated based on the groundwater head and the volume of water displaced by the tunnel, i.e., the product of the groundwater density and the displaced volume; the other is the instantaneous buoyancy increment due to construction vibration. ,in, This indicates the instantaneous buoyancy caused by construction vibration. The incremental dynamic total anti-buoyancy force represents the tunnel buoyancy. It is calculated by superimposing various anti-buoyancy forces onto the pre-processed static anti-buoyancy parameters, including the self-weight of the tunnel segments, the self-weight of the overlying soil, the corrected shear zone stratum resistance, the anti-buoyancy force of the anti-buoyancy piles, temporary construction loads, and the frictional resistance of the segment joints, while deducting the additional force from groundwater seepage. The specific formula is as follows:

[0098] ,

[0099] The buoyancy resistance coefficient is defined as the ratio of the dynamic total buoyancy resistance to the total buoyancy. This coefficient directly reflects the safety reserve of the anti-buoyancy system.

[0100] As a refinement of the above embodiments, step S3 is as follows: Figure 2 As shown, it includes:

[0101] S301: Based on the modified anti-buoyancy coefficient Set warning values and This forms a three-level response logic, when Conventional anti-buoyancy piles were used at that time. When initiating an enhanced scheme, it is necessary to calculate the buoyancy resistance of a single pile and consider the interaction between the buoyancy-resistant piles to ensure the accuracy of the pile group's buoyancy resistance effect. The calculation of the buoyancy resistance of a single pile requires the introduction of an interaction coefficient. This coefficient reflects the reduction in bearing capacity caused by the superposition of loads on adjacent piles. The calculation formula is as follows:

[0102] ,

[0103] in, The pile spacing For the pile diameter, when When the interaction is negligible, the corrected monopile buoyancy resistance is:

[0104] ,

[0105] in, The circumference of the pile body For the pile length, and These are the characteristic values ​​of pile side friction and pile end resistance, respectively. The correction is based on the pile tip cross-sectional area. This correction makes the calculation of the buoyancy resistance of a single pile more closely reflect the actual stress under the working condition of a pile group. The corrected buoyancy resistance of a single pile is then accumulated according to the actual number of piles. The total buoyancy resistance of the pile group is obtained as follows:

[0106] ,

[0107] Special attention needs to be paid to the pile spacing at this time. In the region of interaction coefficient The buoyancy resistance of a single pile has been reduced, and the cumulative result directly reflects the actual total pull-out resistance of the pile group under mutual interference. The area The buoyancy resistance of a single pile is calculated based on its natural state, and the summation yields the total buoyancy resistance of the pile group in the area. Next, the total buoyancy resistance of the integrated support model is calculated. The total buoyancy resistance of the pile group is then... When superimposed with other anti-buoyancy forces and after deducting the additional force from groundwater seepage, the dynamic total anti-buoyancy force is obtained:

[0108] ,

[0109] Finally, the overall model was verified to meet the standard by using the anti-buoyancy coefficient.

[0110] S302: When using a genetic algorithm to optimize the layout of anti-buoyancy piles, the tunnel cross-section is first divided into 1m×1m grids, and the center coordinates of the pile positions are used as genes for binary encoding to form an initial population. The population size is 50 to 100. Then, a dual objective function is constructed.

[0111] One is to maximize the total buoyancy resistance. ;

[0112] Secondly, minimize material usage. ;

[0113] in, For the first Root length, For cross-sectional area, For material density, The unit cost is specified, and the constraint condition is set as pile spacing. After optimization, the anti-buoyancy coefficient is greater than or equal to 1.1, and the pile top elevation is not higher than the bottom elevation of the tunnel segment. In the genetic operation, the selection operator adopts the tournament selection method, randomly selecting 3 individuals each time and choosing the one with the highest fitness, with a selection probability of 65%. The crossover operator adopts arithmetic crossover, and the selected parent pile coordinates are calculated according to... ,in, Generate offspring using random numbers between 0 and 1, with a crossover rate of 0.7. The coordinate components of the pile location of the parent generation 1 anti-buoyancy pile are shown. The pile position coordinate components represent the parent generation 2 anti-buoyancy piles. The mutation operator uses Gaussian mutation, superimposing Gaussian random numbers with a mean of 0 and a standard deviation of 0.5 onto the pile position coordinates, with a mutation rate of 0.08. After 50-80 iterations, the output is the pile position coordinate matrix that satisfies both objectives. Decrease the number of piles in increments of 10% until... Falling back to The following approach aims to balance safety and economy, avoiding over-design or under-design.

[0114] S303: After optimization: Adaptability adjustments need to be made based on the geological stress characteristics of different areas of the tunnel cross-section: The tunnel cross-section is divided into the crown area (central angle 60° to 90°), the sidewall area (central angle 120° to 150°), and the bottom area (central angle 120° to 150°) according to the central angle. Due to the smaller overburden load in the crown area, the density of anti-buoyancy piles should be appropriately increased (pile spacing 3d-4d), and longer piles should be selected, with an embedding depth ≥ 5m into stable strata. In the sidewall area... Lateral water pressure has a significant impact, so the pile inclination angle is set at 5°-10° towards the outside of the tunnel to enhance horizontal restraint. The bearing capacity of the stratum at the bottom is relatively high, so the pile length can be appropriately reduced, the embedment depth in stable stratum is ≥3m, and the pile spacing is widened to 4d-5d. Through regional differentiation adjustments, the arrangement of anti-buoyancy piles is matched with the spatial distribution characteristics of stratum resistance, ultimately forming a support model of active anti-buoyancy piles resisting pull-out and passive stratum resistance, ensuring the balance between total anti-buoyancy force and total buoyancy, and avoiding overload of a single anti-buoyancy method.

[0115] As a refinement of the above embodiment, step S4 utilizes the anti-buoyancy pile arrangement parameters to perform coordinated anti-buoyancy reinforcement of the stratum and portal frame structure. In specific implementation, it is necessary to strengthen the stratum resistance through differentiated grouting methods in different zones, and combine this with intelligent algorithms to optimize the portal frame structure parameters, thereby achieving coordinated stress distribution among the anti-buoyancy piles, stratum, and portal frame structure, and improving the overall anti-buoyancy performance. This includes:

[0116] S401: Based on the anti-buoyancy pile layout parameters and the tunnel outer diameter, the stratum reinforcement zone is divided. The core zone is the area around the anti-buoyancy pile and the area from the tunnel outer wall to the inner side of the anti-buoyancy pile, which is usually 0.5 to 0.8 times the tunnel outer diameter. This area directly bears the load transmitted by the anti-buoyancy pile. The splitting grouting method is used to split the stratum and fill the fissures with high-pressure grout to significantly improve the stratum density. The transition zone is from the outside of the core zone to the boundary of the disturbance zone. The permeation grouting method is used to allow the grout to penetrate and fill the stratum pores to enhance the integrity of the stratum.

[0117] S402: A BP neural network is constructed to optimize grouting parameters, breaking through the limitations of traditional empirical calculations or single-parameter fitting. Intelligent adaptation of grouting parameters is achieved through multi-dimensional feature fusion and dynamic feedback mechanisms. In the design of the network input layer, not only are conventional geological physical parameters such as initial unit weight and permeability coefficient, and anti-buoyancy pile layout parameters such as average spacing and pile diameter, are incorporated, but a synergy coefficient is also introduced. This coefficient is calculated by the ratio of the buoyancy resistance of a single anti-buoyancy pile to the initial resistance of the corresponding stratum, and is used to quantify the force coupling relationship between the anti-buoyancy pile and the stratum. When the anti-buoyancy piles bear the main load, the network will automatically tend to output higher grouting parameters to enhance the stratum's cooperative bearing capacity; when At this time, the grouting intensity is weakened to avoid wasting resources, so that the input characteristics are more in line with the essential requirements of the anti-floating pile-soil synergistic stress.

[0118] From the perspective of network structure, the input layer contains 8 neurons, which correspond to 4 types of core features: stratum physical parameters (initial unit weight, permeability coefficient), anti-buoyancy pile layout parameters (average spacing, pile diameter), synergy coefficient, which is the ratio of the anti-buoyancy force of a single anti-buoyancy pile to the initial resistance of the corresponding stratum, and interactive features derived from the above parameters. The coupling relationship between the anti-buoyancy pile and the stratum is strengthened through feature cross-linking.

[0119] Secondly, regarding the hidden layer topology, an adaptive dynamic adjustment mechanism for the number of neurons is adopted: initially, two hidden layers are set, each with 32 neurons. During training, the contribution of each neuron is calculated in real time. Based on the absolute value ratio of the weights in the gradient descent method, redundant neurons with a contribution of less than 5% are removed. At the same time, the layers containing key neurons with a contribution of more than 30% are split, and a parallel sub-network is added to form a hybrid structure of backbone network and branch network. The backbone network handles the mapping of common parameters, such as the relationship between pressure and density, while the branch network specifically learns the specific laws of different strata types, such as cohesive soil and sandy soil, to solve the problem of insufficient adaptability of traditional fixed structure networks to complex geological conditions.

[0120] The output layer design breaks through the single-parameter output mode, adopting a "parameter-effect" dual-output structure: in addition to outputting grouting pressure P, flow rate Q, and water-cement ratio, it simultaneously predicts the soil's weight increment and compressive strength increment 12 hours after reinforcement, and achieves dynamic parameter calibration through built-in effect and parameter feedback functions. The expression of the feedback function is as follows:

[0121] ,

[0122] in, This is expressed as the grouting pressure after dynamic calibration. This is represented by the initial grouting pressure output by the neural network. This is represented as a correction factor. This is represented as the preset target value for soil weight increment. This represents the predicted soil weight increment output by the neural network. From a neuron configuration perspective, the output layer contains 8 neurons, divided into two independent output channels. The first group is the parameter output channel, containing 3 neurons, corresponding to key grouting parameters common to both the core and transition regions: grouting pressure, grouting flow rate Q, and water-cement ratio. Each parameter is output through a linear activation function, with its numerical range strictly constrained within the allowable range for engineering practice, ensuring direct applicability for construction guidance. The second group is the effect prediction channel, containing 5 neurons. Two of these are used to predict the soil weight increment 12 hours after reinforcement, corresponding to the core and transition regions respectively. The other three are used to predict the compressive strength increment, covering strata of different depths. The effect prediction value is mapped to a preset target range using a Sigmoid function, ensuring a quantitative correlation with the actual reinforcement effect.

[0123] Furthermore, in the training algorithm, an engineering constraint regularization term is introduced to improve the loss function:

[0124] ,

[0125] in, This is represented as a penalty coefficient. These are actual observed values. These are the predicted values ​​from the BP neural network. This is the preset target value for soil weight increment, which is the minimum effective threshold for soil reinforcement required by the project. This refers to the predicted soil unit weight increment output by the BP neural network; when predicting the reinforcement unit weight... Below the target value At this time, the penalty will significantly increase the loss value, forcing the network to prioritize ensuring that the hardening effect meets the standard, and avoiding sacrificing engineering safety due to excessive pursuit of prediction accuracy.

[0126] S403: For portal frame structures, a genetic algorithm is used to optimize the combined parameters of the steel frame and prestressed anchor cables, with the dual objectives of maximizing pull-out resistance and minimizing material usage. Optimization variables include: the cross-sectional height, width, and web thickness of the steel frame; the number of bolts; the anchor cable diameter; and the tension control stress. The objective function is set as follows:

[0127] ,

[0128] ,

[0129] in, For total pull-out force, For the pull-out resistance of the steel frame, The total pull-out force of the anchor cable. For the quantity of steel, For the number of doors, For the number of anchor cables, The number of bolts is set; constraints are set: the stress of the steel frame is less than or equal to the design strength of the steel, the tension control stress of the anchor cable is greater than or equal to 0.75 times the standard value of the tensile strength, and the bearing capacity of the connection node between the steel frame and the anchor cable is greater than or equal to 1.2 times the design pull-out force; in the genetic algorithm operation, real number encoding is used to represent the optimization variables, the initial population size is 80 to 100, the selection operator uses the roulette wheel selection method, individuals with high fitness have a high probability of being selected, the crossover operator uses single-point crossover with a crossover rate of 0.8, the mutation operator adds ±5% random perturbation to the variables with a mutation rate of 0.1, and the optimal parameter combination is output after 60 to 100 generations of iteration.

[0130] As a refinement of the above embodiments, step S5, based on the constructed collaborative anti-buoyancy reinforcement system, achieves real-time mapping and dynamic simulation of the tunnel's anti-buoyancy state through the deep integration of multi-dimensional anti-buoyancy data monitoring and digital twin technology, providing digital decision support for subsequent optimization and adjustment. Specifically, it includes:

[0131] S501: Establish a tiered monitoring indicator system, clarify the scope of monitoring data collection and calculation methods, and focus on monitoring the axial force of single piles for anti-buoyancy pile systems. With pile top displacement Fiber optic grating sensors are installed in groups of 5 piles. Axial force is calculated using a strain-force conversion formula, which is expressed as:

[0132] ,

[0133] in, To measure strain, The elastic modulus of the pile. This represents the cross-sectional area of ​​the pile.

[0134] For the ground reinforcement area, monitor the earth pressure at different depths. With moisture content Earth pressure is collected by a vibrating wire sensor and calculated using a frequency-pressure conversion formula:

[0135] ,

[0136] in, For calibration coefficients, It is the square of the measured frequency. The water content is obtained by inversion using the time-domain reflectometry (TDR) method, which is the square of the initial frequency.

[0137] For portal frame structures, monitor the stress in the steel frame. With anchor cable tension Stress is measured using resistance strain gauges, and tension is calculated based on the frequency change of the anchor cable force gauge according to the following formula:

[0138] ,

[0139] in, The measured frequency squared value is collected by the anchor cable force gauge at the monitoring time. When the anchor cable is subjected to changes in force, the vibration frequency of the vibrating wire of the force gauge will change accordingly. The squared value of the measured frequency can reflect the tension change more linearly. Therefore, the square form is chosen for calculation. The initial frequency squared value of the anchor cable force gauge refers to the reference value of the frequency squared collected after the anchor cable is initially installed and the preset tension control stress is applied, without being disturbed by subsequent loads. This value serves as a reference for calculating tension changes. The calibration coefficients for the anchor cable force gauges are used; all monitoring data are transmitted to the data platform in real time at a frequency of 10 minutes per transmission, forming a time-series dataset.

[0140] .

[0141] S502: The core of constructing a digital twin image for tunnel anti-buoyancy lies in realizing the dynamic mapping between physical entities and virtual models, specifically through a three-step method of data assimilation, model correction, and state inference.

[0142] The first step, data assimilation, employs the Kalman filter algorithm to fuse the monitoring data with the calculated values ​​from the initial finite element model. The formula is as follows:

[0143] ,

[0144] in, The corrected state vector represents the set of parameters in the digital twin model that describe the core characteristics of the anti-buoyancy system, including parameters such as the stiffness of the anti-buoyancy piles and the elastic modulus of the formation. Predicted state For Kalman gain, For monitoring data, namely the time series dataset mentioned above. The measured values ​​in the middle, The observation matrix, whose elements are derived using formulas from elasticity mechanics, establishes a mapping relationship between the state vector and the measured parameters. The state vector contains key mechanical parameters of the anti-buoyancy system, which cannot be directly measured by sensors but determine the stress state of the anti-buoyancy system. The monitoring data, on the other hand, are directly observable physical quantities, whose values ​​are derived from the state vector parameters using mechanical laws. The elements need to quantify this causal relationship, that is: how small changes in state parameters affect changes in measured physical quantities, for example: according to the stress calculation model of a semi-infinite body under uniformly distributed load in elasticity, the earth pressure at a certain point around the tunnel. With formation elastic modulus Poisson's ratio of the strata Distance from this point to the tunnel structure and tunnel deformation The relationship is:

[0145] ,

[0146] This allows us to obtain the elements in H corresponding to the formation's elastic modulus to earth pressure:

[0147] ,

[0148] Its physical meaning is: when the elastic modulus of the stratum changes by 1 unit, the change in earth pressure at the monitoring point directly reflects the quantitative correlation between the two. It is a rectangular matrix with the number of rows equal to the number of measured parameters and the number of columns equal to the number of parameters in the state vector. This algorithm controls the deviation between the calculated model value and the measured value to within 5%.

[0149] The second step, model correction, is based on the assimilated parameters and dynamically updates key parameters of the digital twin image, such as the water content of the strata. When the shear strength of the soil exceeds the warning value (30% for cohesive soil and 20% for sandy soil), the shear strength parameters of the corresponding area should be lowered. , ,in, The initial moisture content, To monitor the actual water content of the formation at a given time, This is the corrected internal friction angle of the formation. The initial internal friction angle of the formation. The corrected formation cohesion, The initial cohesion of the formation (unit: kPa) is the inherent shear resistance parameter of the formation when the water content is not exceeded, ensuring the consistency between the model and the actual state.

[0150] The third step, state extrapolation, uses a Long Short-Term Memory (LSTM) network to predict the buoyancy resistance status for the next 24 hours. The input layer is the monitoring data sequence of the past 12 hours, and the output layer is the buoyancy resistance coefficient change curve. The attention mechanism is used to strengthen the weight allocation of key parameters, thereby achieving early warning of potential buoyancy risks.

[0151] The visualization of the digital twin mirror uses 3D mesh modeling to digitize the tunnel structure, anti-buoyancy piles, portal frame structure, and strata at a 1:1 scale. Each mesh cell is linked to real-time monitoring data and calculation results, and stress distribution, displacement magnitude, and anti-buoyancy coefficient gradient are intuitively displayed through color rendering. It supports multi-view cross-sections and dynamic playback, enabling technicians to grasp the overall stress state of the collaborative anti-buoyancy system in real time. Simultaneously, the mirror has a built-in anti-buoyancy safety assessment module. By calculating the ratio of monitored values ​​to design values, it automatically triggers an early warning when the ratio of any parameter exceeds 1.1 and pushes possible cause analyses.

[0152] As a refinement of the above embodiment, step S6, based on the monitoring data and state prediction results output by the digital twin mirror, dynamically adjusts the collaborative anti-buoyancy reinforcement system through a multi-dimensional optimization algorithm, ultimately outputting a reinforcement result that satisfies a balance between safety and economy. Specifically, firstly, an optimization triggering mechanism is established, based on the early warning signal output by the safety assessment module of the digital twin mirror, such as the anti-buoyancy coefficient. Anchor cable tension ratio If the formation moisture content exceeds the standard, the corresponding optimization process will be initiated. When the anti-buoyancy coefficient is lower than the warning value, priority will be given to adjusting the coordinated force of the anti-buoyancy piles and the portal structure incrementally; when there is local stress concentration, such as stress in the steel frame... , To enhance design strength, the focus is on local optimization of structural parameters. When the reinforcement effect of the stratum diminishes, the secondary grouting parameters are adjusted.

[0153] (1) To address the core issue of insufficient anti-buoyancy coefficient, a hierarchical optimization algorithm is adopted. The first layer optimizes the load distribution ratio between the anti-buoyancy piles and the portal structure, which is achieved by adjusting the tension control stress of the anchor cables. The goal is to maintain the ratio of the total axial force of the anti-buoyancy piles to the total pull-out force of the portal structure within a reasonable range of 2:1 to 3:1. The calculation formula is as follows:

[0154] ,

[0155] in, This is the anchor cable tension adjustment value. As a safety threshold, To enhance the buoyancy resistance of anti-buoyancy piles, For the number of anchor cables, The cross-sectional area of ​​a single anchor cable is determined by verifying the strength of the pile material and should not exceed 10% of the design value. If the anti-buoyancy coefficient still does not meet the standard after adjusting the anchor cable, the second layer of optimization is entered. The number of anti-buoyancy piles is supplemented by a genetic algorithm, with the dual objectives of maximizing the new anti-buoyancy force and minimizing the new cost. The constraint is that the distance between the new pile location and the existing pile is ≥2d. After 30 iterations, the optimal supplementation scheme is output.

[0156] (2) To address the localized stress concentration problem, a topology optimization algorithm is used to adjust the portal frame structure parameters: for areas where the steel frame stress exceeds the limit, stress is reduced by adding stiffening ribs or adjusting the cross-sectional dimensions. The corrected formula is as follows: ,in, To optimize the rear flange width, and to address the uneven distribution of anchor cable tension, the tensioning sequence is adjusted to achieve force flow balance. A greedy algorithm is used to fine-tune the tension of each anchor cable sequentially in descending order of tension deviation rate, with each adjustment not exceeding 5% of the design value, until the standard deviation is less than 0.1.

[0157] (3) To address the attenuation of the formation reinforcement effect, secondary grouting parameter optimization is initiated, based on a feedback correction mechanism using a BP neural network: the current formation water content is adjusted. Earth pressure Input the initial grouting parameters into the model and output the pressure increment for secondary grouting. , For the second grouting pressure, The initial grouting pressure and the water-cement ratio adjustment value. The revised formula references the formation response characteristics derived from the monitoring data in step five, ensuring that the soil density in the core area recovers to above 18 kN / m³ after secondary reinforcement, and that the additional grouting volume does not exceed 30% of the initial grouting volume.

[0158] After all optimization measures are implemented, the results are verified using a digital twin mirror. The optimized parameters are input, and the anti-buoyancy coefficient is calculated through simulation. The stress of each component and the ground reinforcement indicators must meet the following requirements. After verification, the stress peak value is ≤0.8×design value, the stratum index meets the standard, and the reinforcement results are output, including the final layout drawing of the anti-buoyancy piles, the portal structure parameter table, the grouting parameter report and the anti-buoyancy coefficient verification curve, forming a closed loop from design to optimization.

[0159] Example 2

[0160] The difference between this embodiment and Embodiment 1 is that this embodiment provides a large-diameter shield tunnel anti-buoyancy system based on an anti-buoyancy coefficient, including:

[0161] The data acquisition module is configured to acquire multi-dimensional anti-buoyancy parameters, including static anti-buoyancy parameters and dynamic anti-buoyancy parameters;

[0162] The preprocessing module is configured to preprocess the static anti-buoyancy parameters and calculate the anti-buoyancy coefficient based on the preprocessed static anti-buoyancy parameters, including introducing a calculation model for the radius of the shield tunneling disturbance zone to dynamically correct the shear zone stratum resistance.

[0163] The conversion module is configured to construct a support model of anti-buoyancy piles and strata based on the anti-buoyancy coefficient, including setting an anti-buoyancy safety threshold, calculating the anti-buoyancy force of a single pile based on the safety threshold, and optimizing the arrangement of anti-buoyancy piles using a genetic algorithm.

[0164] The reinforcement module is configured to use anti-buoyancy pile layout parameters to perform coordinated anti-buoyancy reinforcement of the strata and portal structure, including reinforcing the strata by using zoned differentiated grouting methods, and optimizing the combination parameters of the steel frame and prestressed anchor cables for the portal structure using a genetic algorithm.

[0165] The digital twin module is configured to monitor anti-buoyancy data based on the collaborative anti-buoyancy reinforcement system, and to construct a digital twin image of the tunnel's anti-buoyancy based on the monitoring data;

[0166] The output module is configured to optimize the monitoring data and status prediction results output by the digital twin mirror, and output the final hardening result.

[0167] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 method for anti-buoyancy of large-diameter shield tunnels based on an anti-buoyancy coefficient, characterized in that, Includes the following steps: Obtain multi-dimensional anti-buoyancy parameters, including static anti-buoyancy parameters and dynamic anti-buoyancy parameters; The static anti-buoyancy parameters are preprocessed by introducing a calculation model for the radius of the disturbance zone during shield tunneling to dynamically correct the initial resistance of the shear zone strata, and the anti-buoyancy coefficient is calculated based on the preprocessed static and dynamic anti-buoyancy parameters. Based on the anti-buoyancy coefficient, an early warning value is set, an anti-buoyancy pile layout scheme is selected according to the early warning value, and the anti-buoyancy pile layout is optimized by a genetic algorithm. Based on the optimized anti-buoyancy pile layout, a zoned differentiated grouting method is adopted to reinforce the stratum. The stratum reinforcement area is divided into a core area and a transition area. The core area is the area around the anti-buoyancy pile and the tunnel outer wall to the inner side of the anti-buoyancy pile. The transition area is the area outside the core area to the boundary of the disturbance area. The core area adopts splitting grouting, and the transition area adopts permeation grouting. The grouting parameters are optimized by BP neural network. A collaborative anti-buoyancy reinforcement system was established by combining genetic algorithms to optimize the combined parameters of the steel frame and prestressed anchor cables for portal structures. Based on the aforementioned collaborative anti-buoyancy reinforcement system, anti-buoyancy data monitoring is conducted, and a digital twin image of the tunnel's anti-buoyancy is constructed based on the monitoring data, followed by state prediction; the specific method is as follows: Data were collected on single pile axial force, pile top displacement, soil pressure and moisture content at different depths, steel frame stress, and anchor cable tension to form a time-series dataset. The collected data is fused with the calculated values ​​from the initial finite element model using the Kalman filter algorithm to assimilate the data. The key parameters of the digital twin are dynamically updated based on the assimilated data. Predicting buoyancy resistance using long short-term memory networks; The collaborative anti-buoyancy reinforcement system is optimized based on the monitoring data and state prediction results output by the digital twin mirror, and the final reinforcement result is output; the optimization includes the following three cases: When the anti-buoyancy coefficient is insufficient, a hierarchical optimization algorithm is adopted, which optimizes the load distribution ratio between the anti-buoyancy piles and the portal structure by adjusting the anchor cable tension control stress or by supplementing the number of anti-buoyancy piles through a genetic algorithm. When local stress concentration occurs, a topology optimization algorithm is used to adjust the parameters of the portal structure. When the reinforcement effect of the stratum diminishes, a feedback correction mechanism based on a BP neural network is initiated to optimize the secondary grouting parameters.

2. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 1, characterized in that, The static anti-buoyancy parameters include: the self-weight of the tunnel segments, the self-weight of the overlying soil layer, the shear zone stratum resistance, the frictional resistance of the tunnel segment joints, and the buoyancy force on the tunnel. The dynamic anti-buoyancy parameters include: anti-buoyancy force of the anti-buoyancy pile, temporary construction load, additional force from groundwater seepage, and instantaneous buoyancy increment caused by construction vibration.

3. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 1, characterized in that, The initial resistance of the shear zone strata is dynamically corrected by introducing a calculation model for the radius of the disturbance zone during shield tunneling, as shown in the following formula: ， ， in, For the corrected shear zone formation resistance, The initial resistance of the shear zone strata. Formation disturbance coefficient Distance from tunnel axis The internal friction angle of the strata, Distance from tunnel axis The cohesion of the strata, The area of ​​action of the shear band. Tunnel structure surface =0 initial disturbance coefficient The attenuation coefficient is... The radius of the disturbance zone.

4. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 3, characterized in that, The anti-buoyancy coefficient is the ratio of dynamic total anti-buoyancy force to total buoyancy force, wherein the total buoyancy force includes tunnel buoyancy force and instantaneous buoyancy force increment due to construction vibration. The dynamic total anti-buoyancy force includes the self-weight of the tunnel segments, the self-weight of the overlying soil layer, the corrected shear zone stratum resistance, the anti-buoyancy force of the anti-buoyancy piles, temporary construction loads, and the frictional resistance of the tunnel segment joints, and deducts the additional force of groundwater seepage.

5. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 1, characterized in that, The selection of anti-buoyancy pile arrangement scheme based on early warning value includes: Set low warning value and high warning value ; When the anti-buoyancy coefficient is greater than or equal to the low warning value and less than the high warning value, conventional anti-buoyancy pile arrangement shall be adopted. When the anti-buoyancy coefficient is less than the low warning value, the buoyancy resistance of a single pile is calculated, and the total buoyancy resistance of the pile group is calculated based on the buoyancy resistance of the single pile and the interaction between the anti-buoyancy piles, as shown in the following formula: ， ， ， in, The interaction coefficient, The circumference of the pile body For the pile length, and These are the characteristic values ​​of pile side friction and pile end resistance, respectively. The cross-sectional area of ​​the pile tip. For the first The buoyancy resistance of a single pile, This represents the total number of piles. For the total buoyancy resistance of the pile group, The pile spacing The diameter of the pile; The corrected dynamic total buoyancy is obtained by calculating the sum of the self-weight of the tunnel segment, the self-weight of the overlying soil layer, the corrected shear zone stratum resistance, the total buoyancy of the pile group, the temporary construction load, and the frictional resistance of the tunnel segment joints, and subtracting the additional force of groundwater seepage. The buoyancy coefficient is calculated based on the corrected dynamic total buoyancy. If the buoyancy coefficient is not in the range of greater than or equal to the low warning value and less than the high warning value, the design index of the piles related to the interaction coefficient is replanned until the buoyancy coefficient is greater than or equal to the low warning value and less than the high warning value. When the anti-buoyancy coefficient is greater than or equal to the high warning value, the number of piles is reduced by 10% increments until the anti-buoyancy coefficient falls back below the high warning value.

6. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 5, characterized in that, The specific methods for optimizing the early warning value of anti-buoyancy pile layout using genetic algorithms include: The tunnel cross-section is divided into a grid, and the coordinates of the pile center are used as genes for binary encoding to form an initial population. The dual objective functions are maximizing total buoyancy resistance and minimizing material usage. The constraints include a pile spacing greater than or equal to three times the pile diameter, a buoyancy resistance coefficient greater than or equal to the low warning value, and a pile top elevation not higher than the bottom elevation of the tunnel segment. The selection operator adopts the tournament selection method, the crossover operator adopts arithmetic crossover, and the mutation operator adopts Gaussian mutation. After iteration, the optimal pile position coordinate matrix that satisfies the dual objectives is output. The pile positions are determined based on the optimal pile position coordinate matrix, and adaptive adjustments are made according to the stress characteristics of the strata in different areas of the tunnel cross section. The density of anti-buoyancy piles is increased in the arch area with a central angle between 60 and 90 degrees, the pile spacing is set at 3 to 4 times the pile diameter, and the pile body is lengthened, with an embedment depth of greater than or equal to 5 meters into stable strata. The pile bodies in the sidewall area with a central angle between 90 and 120 degrees are set outward from the tunnel according to a lateral dip angle of 5 to 10 degrees. The pile length is reduced in the bottom area with a central angle between 120 and 150 degrees, the embedment depth in stable strata is greater than or equal to 3 meters, and the pile spacing is set at 4 to 5 times the pile diameter.

7. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 1, characterized in that, The optimization of grouting parameters using a BP neural network is specifically implemented as follows: The input layer of the BP neural network comprises eight neurons, and the input features include geological physical parameters, anti-buoyancy pile layout parameters, synergy coefficient, and interaction features. The geological physical parameters include initial unit weight and permeability coefficient; the anti-buoyancy pile layout parameters include average spacing and pile diameter; and the synergy coefficient is the ratio of the buoyancy resistance of a single anti-buoyancy pile to the initial resistance of the corresponding geological stratum. The output layer of the BP neural network comprises eight neurons, and the output features include key grouting parameters and prediction results. The key grouting parameters include grouting pressure, grouting flow rate, and water-cement ratio; and the prediction results include the increase in soil unit weight and compressive strength after reinforcement. The output layer also achieves dynamic parameter calibration through a parameter feedback function, which is as follows: ， in, This is expressed as the grouting pressure after dynamic calibration. This is represented by the initial grouting pressure output by the neural network. This is represented as a correction factor. This is represented as the preset target value for soil weight increment. This is represented as the predicted value of soil weight increment output by the neural network. The BP neural network employs an engineering constraint regularization term to improve the loss function. : ， in, This is represented as a penalty coefficient. These are actual observed values. These are the predicted values ​​from the BP neural network. This is the preset target value for soil weight increment, which is the minimum effective threshold for soil reinforcement required by the project. The predicted value of soil weight increment output by the BP neural network; The specific method for optimizing the combined parameters of the steel frame and prestressed anchor cables using a genetic algorithm for portal structures is as follows: A genetic algorithm was used to optimize the combined parameters of the steel frame and prestressed anchor cables, with the dual objectives of maximizing pull-out resistance and minimizing material usage. The optimization variables included: the cross-sectional height and width of the steel frame, the web thickness, the number of bolts, the anchor cable diameter, and the tension control stress. The objective function was: ， ， in, For total pull-out force, For the pull-out resistance of the steel frame, The total pull-out force of the anchor cable. For the quantity of steel, For the number of doors, For the number of anchor cables, This refers to the number of bolts. The constraints are: the stress in the steel frame is less than or equal to the design strength of the steel, the tension control stress of the anchor cable is greater than or equal to 0.75 times the standard value of the tensile strength, and the bearing capacity of the connection node between the steel frame and the anchor cable is greater than or equal to 1.2 times the design pull-out force. The selection operator uses the roulette wheel selection method, the crossover operator uses single-point crossover, and the mutation operator adds ±5% random perturbation to the variables.

8. The anti-buoyancy method for large-diameter shield tunnels based on the anti-buoyancy coefficient according to claim 1, characterized in that, The monitoring data and state prediction results based on the digital twin mirror output are used to optimize the collaborative anti-buoyancy reinforcement system, specifically including: When the anti-buoyancy coefficient is insufficient, a layered optimization algorithm is adopted. The first layer optimizes the load distribution ratio between the anti-buoyancy piles and the portal structure by adjusting the anchor cable tension control stress, so that the ratio of the total axial force of the anti-buoyancy piles to the total pull-out force of the portal structure is maintained in a reasonable range of 2:1 to 3:

1. The calculation formula is as follows: ， in, This is the anchor cable tension adjustment value. As a safety threshold, To enhance the buoyancy resistance of anti-buoyancy piles, For the number of anchor cables, Given the cross-sectional area of ​​a single anchor cable, if the anti-buoyancy coefficient still does not meet the standard after adjusting the anchor cable, a second layer of optimization is performed. The number of anti-buoyancy piles is supplemented by a genetic algorithm, with the dual objectives of maximizing the new anti-buoyancy force and minimizing the new cost. The constraint is that the distance between the new pile position and the existing pile is greater than or equal to twice the pile diameter, and the optimal supplementation scheme is output. When local stress concentration occurs, topology optimization algorithms are used to adjust the parameters of the portal frame structure, including: For areas of steel frame where stress exceeds limits, stress can be reduced by adding stiffeners or adjusting cross-sectional dimensions. The corrected formula is as follows: ， in, To optimize the trailing edge width, The original wing width, These are the measured stress values ​​in the stress-over-limit area of ​​the steel frame. Design strength for steel frame stress; For uneven anchor cable tension distribution, force flow balance is achieved by adjusting the tensioning sequence. A greedy algorithm is used to adjust the tension of each anchor cable in descending order of tension deviation rate until the standard deviation of the measured anchor cable tension is less than 0.

1. When the reinforcement effect of the stratum diminishes, the feedback correction mechanism based on the BP neural network initiates the optimization of secondary grouting parameters. The current stratum moisture content, soil pressure and primary grouting parameters are input into the model, and the pressure increment and water-cement ratio adjustment value of secondary grouting are output.

9. A large-diameter shield tunnel anti-buoyancy system based on an anti-buoyancy coefficient, characterized in that, The method for anti-buoyancy of large-diameter shield tunnels based on the anti-buoyancy coefficient as described in any one of claims 1-8 includes: The data acquisition module is configured to acquire multi-dimensional anti-buoyancy parameters, including static anti-buoyancy parameters and dynamic anti-buoyancy parameters; The preprocessing module is configured to preprocess the static anti-buoyancy parameters and calculate the anti-buoyancy coefficient based on the preprocessed static and dynamic anti-buoyancy parameters, including introducing a calculation model for the radius of the shield tunneling disturbance zone to dynamically correct the shear zone stratum resistance. The conversion module is configured to construct a support model of anti-buoyancy piles and strata based on the anti-buoyancy coefficient, including setting an anti-buoyancy safety threshold, calculating the anti-buoyancy force of a single pile based on the safety threshold, and optimizing the arrangement of anti-buoyancy piles using a genetic algorithm. The reinforcement module is configured to use anti-buoyancy pile layout parameters to perform coordinated anti-buoyancy reinforcement of the strata and portal structure, including reinforcing the strata by using zoned differentiated grouting methods, and optimizing the combination parameters of the steel frame and prestressed anchor cables for the portal structure using a genetic algorithm. The digital twin module is configured to monitor anti-buoyancy data based on the collaborative anti-buoyancy reinforcement system, and to construct a digital twin image of the tunnel's anti-buoyancy based on the monitoring data; The output module is configured to optimize the monitoring data and status prediction results output by the digital twin mirror, and output the final hardening result.

Images