A method for coordinated control of shield tunnel face instability and ground surface settlement in sandy cobble strata

By constructing a random field of soil parameters and a proxy model for surface settlement, a mapping relationship between the support force of the shield tunnel excavation face and surface settlement was established. This solved the problem of coordinated control of excavation face instability and surface settlement during shield tunnel construction under complex geological conditions, achieving dual protection of safety and environmental protection.

CN123116392APending Publication Date: 2026-11-17HENAN UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611236733.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Under complex geological conditions, traditional methods are unable to accurately reflect the co-evolution of instability at the excavation face and surface subsidence during shield tunnel construction, leading to unreasonable support force design and increasing construction safety risks and potential environmental damage.

Method used

By constructing a random field mapping of soil parameters to a shield tunnel numerical model, a first mapping relationship between excavation face support force and stratum loss is established. A second mapping relationship between stratum loss and the probability of exceeding the stratum settlement limit is determined using a surface settlement proxy model. A safety envelope diagram of excavation face support force is constructed to provide a reasonable support force range.

Benefits of technology

It improves the safety and scientific nature of shield tunnel construction, reduces the risk of instability at the excavation face and excessive surface settlement, and achieves dual safety control for different engineering conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116392A_ABST
    Figure CN123116392A_ABST
Patent Text Reader

Abstract

This application discloses a method for the coordinated control of excavation face instability and surface settlement in shield tunnels in gravel strata, relating to the field of tunnel and underground engineering construction control technology. The method involves obtaining a random field of soil property parameters of the target stratum and mapping it to a shield tunnel numerical model; determining a first mapping relationship between excavation face support force and stratum loss based on the mapped numerical model; determining a second mapping relationship between stratum loss and the probability of exceeding surface settlement limits based on a pre-trained surface settlement surrogate model; determining an allowable stratum loss rate based on a preset surface settlement exceeding probability threshold and the second mapping relationship; and determining the excavation face support force range based on the allowable stratum loss rate and the first mapping relationship to construct a safety envelope diagram of the excavation face support force. This application achieves coordinated control of excavation face instability and surface settlement risks.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of tunnel and underground engineering construction control technology, specifically to a method for the coordinated control of instability at the excavation face and surface settlement in a shield tunnel in gravel strata. Background Technology

[0002] With the continuous expansion of urban rail transit networks and underground infrastructure construction in my country, the shield tunneling method has become the main construction method for urban subway projects due to its advantages of high construction safety, minimal environmental disturbance, and high construction efficiency. However, as shield tunnels gradually extend into complex geological conditions and densely built-up areas, the stability of the excavation face and the control of surface settlement during construction have become increasingly prominent issues, posing key technical challenges affecting project safety and the surrounding environmental protection.

[0003] During tunnel boring machine (TBM) excavation, changes in the support pressure at the excavation face directly affect the stress balance of the soil ahead. When the support pressure is insufficient, the soil ahead of the excavation face is prone to plastic deformation or even overall instability, causing soil to move into the excavation space and resulting in ground loss. This ground loss, further transmitted to the upper soil, will lead to surface settlement or even collapse and other engineering disasters. Therefore, excavation face instability and surface settlement are not two independent engineering problems, but rather a unified process closely linked and evolving collaboratively through the transmission path of "stress release—excavation face deformation—ground loss—surface settlement." Excavation face stability determines the degree of ground loss development, while surface settlement reflects the final response after ground loss propagates to the surface; both jointly affect the safety of TBM construction and the stability of the surrounding environment.

[0004] Furthermore, in heterogeneous strata such as sand and gravel, soil parameters exhibit significant spatial variability. Traditional deterministic analysis methods struggle to accurately reflect the impact of this randomness on the co-evolution of excavation face instability and surface settlement. Consequently, support force design often relies on experience or a single safety factor, failing to achieve simultaneous and precise control of excavation face stability and surface settlement risk in a probabilistic sense. This can easily lead to unreasonable support force settings, increasing construction safety risks or potential environmental damage. Summary of the Invention

[0005] This application provides a method for the coordinated control of excavation face instability and surface settlement in shield tunnels in gravel strata. It can construct a safety envelope diagram of excavation face support force that integrates excavation face stability constraints and surface settlement probability control constraints. It can provide a reasonable support force range that meets dual safety requirements for different engineering conditions, thereby reducing the risk of excavation face instability and excessive surface settlement.

[0006] This application provides a method for the coordinated control of instability at the excavation face and surface settlement in a shield tunnel in gravel strata, comprising:

[0007] Obtain the random field of soil parameters of the target stratum and map the random field of soil parameters to the numerical model of the shield tunnel;

[0008] Based on the mapped shield tunnel numerical model, the first mapping relationship between the excavation face support force and the ground loss is determined.

[0009] Based on a pre-trained surface subsidence proxy model, a second mapping relationship between the stratum loss and the probability of exceeding the surface subsidence limit is determined;

[0010] Based on the preset land subsidence exceedance probability threshold and the second mapping relationship, the allowable formation loss rate is determined;

[0011] Based on the allowable formation loss rate and the first mapping relationship, the excavation face support force range is determined to construct an excavation face support force safety envelope diagram. The excavation face support force safety envelope diagram is used to query the minimum support force that meets the collaborative control requirements under different engineering conditions.

[0012] Optionally, obtaining the random field of soil parameters of the target stratum and mapping the random field of soil parameters to the shield tunnel numerical model includes:

[0013] Obtain the statistical characteristics and autocorrelation function of the soil parameters of the target stratum;

[0014] Based on the statistical characteristics and the autocorrelation function, the random field of soil parameters is generated;

[0015] The soil properties are discretized and mapped to each grid cell of the shield tunnel numerical model.

[0016] Optionally, generating the random field of soil parameters based on the statistical characteristics and the autocorrelation function includes:

[0017] Based on the Gaussian autocorrelation function, calculate the eigenvalues ​​and eigenfunctions of the autocorrelation function;

[0018] The number of truncation terms in the random field expansion is determined based on a preset energy ratio threshold.

[0019] An independent standard normal random sample matrix is ​​generated using a sampling method, and the correlation coefficient matrix is ​​decomposed to obtain a correlated normal random sample matrix.

[0020] Based on the eigenvalues, the eigenfunctions, the number of truncation terms, and the relevant normal random sample matrix, the relevant Gaussian random field is discretized using the series expansion method, and the relevant Gaussian random field is transformed into a relevant non-Gaussian random field through equal probability transformation to obtain the soil parameter random field;

[0021] The Gaussian autocorrelation function characterizes the spatial correlation of soil parameters based on the fluctuation range in the horizontal and vertical directions.

[0022] Optionally, determining the first mapping relationship between the excavation face support force and ground loss based on the mapped shield tunnel numerical model includes:

[0023] Different excavation face support forces are applied to the mapped shield tunnel numerical model, and the corresponding excavation face ground loss data are extracted.

[0024] The first mapping relationship is obtained by fitting the formation loss data of the excavation face based on the surrounding rock characteristic curve model.

[0025] Optionally, the excavation face formation loss data includes the volumetric deformation of each grid cell at the excavation face.

[0026] Optionally, the determination of the second mapping relationship between the formation loss and the probability of exceeding the surface subsidence limit based on the pre-trained surface subsidence surrogate model includes:

[0027] Construct an input space that includes tunnel burial depth ratio, stratum rock content, stratum loss rate, and random field variables;

[0028] Training samples are generated based on the input space, and the surface subsidence response is predicted using the surface subsidence proxy model.

[0029] Monte Carlo simulations were performed based on the surface subsidence response to establish the second mapping relationship.

[0030] Optionally, generating training samples based on the input space includes:

[0031] The tunnel burial depth ratio and the rock content of the stratum are uniformly sampled, the stratum loss rate is sampled using log-space stratified sampling, and the random field variables are sampled using standard normal stratified sampling to obtain the training samples.

[0032] Optionally, the step of determining the allowable ground loss rate based on a preset ground subsidence exceedance probability threshold and the second mapping relationship, and determining the excavation face support force range based on the allowable ground loss rate and the first mapping relationship, to construct an excavation face support force safety envelope diagram, includes:

[0033] Substitute the above-mentioned surface subsidence excess probability threshold into the second mapping relationship to determine the allowable formation loss rate;

[0034] Substituting the allowable formation loss rate into the first mapping relationship, the minimum support force that meets the settlement control requirements is determined, thereby determining the support force range of the excavation face;

[0035] Based on the excavation face support force range, a safety envelope diagram of the excavation face support force is constructed.

[0036] Optionally, the target stratum is a sand and gravel stratum; the soil parameters corresponding to the random field of the soil parameters include the internal friction angle and the reference secant modulus, both of which follow a log-normal distribution, and there is a cross-correlation between the internal friction angle and the reference secant modulus.

[0037] Optionally, the threshold for the probability of exceeding the surface settlement limit is 5%; and / or, the allowable settlement value corresponding to the probability of exceeding the surface settlement limit is 20 mm.

[0038] As described above, the method for coordinated control of instability at the excavation face and surface settlement in a shield tunnel in gravel strata provided in this application has the following beneficial effects:

[0039] This application presents a method for the coordinated control of excavation face instability and surface settlement in shield tunnels. It establishes a correlation between excavation face support force and the probability of excessive surface settlement by analyzing ground loss, overcoming the limitations of related technologies that analyze these two factors separately. By mapping soil parameter random fields to a numerical model, the spatial variability of heterogeneous stratum parameters is considered, and combined with a surrogate model, the scientific rigor of construction safety assessment under complex geological conditions is improved. Furthermore, the constructed safety envelope diagram of excavation face support force integrates excavation face stability constraints and surface settlement probability control constraints, providing a reasonable support force range that meets dual safety requirements for different engineering conditions. This reduces the risk of excavation face instability and excessive surface settlement, thereby enhancing the safety of shield tunnel construction. Attached Figure Description

[0040] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings:

[0041] Figure 1 This is a flowchart illustrating a method for coordinated control of instability at the excavation face and surface settlement in a shield tunnel in gravel strata, as shown in an exemplary embodiment of this application.

[0042] Figure 2 This is a schematic diagram of a random field of soil parameters illustrated in an exemplary embodiment of this application;

[0043] Figure 3 This is a schematic diagram of a numerical model of a shield tunnel, illustrating an exemplary embodiment of this application;

[0044] Figure 4This is a graph illustrating the evolution relationship between the excavation face support ratio and the formation loss rate, as shown in an exemplary embodiment of this application.

[0045] Figure 5 This is a scatter plot showing the comparison between predicted and actual values ​​of a land subsidence surrogate model, as illustrated in an exemplary embodiment of this application.

[0046] Figure 6 This is an exemplary embodiment of the present application illustrating the relationship between the probability of exceeding the limit of surface subsidence and the formation loss rate.

[0047] Figure 7 This is an exemplary embodiment of the present application illustrating the safety envelope diagram of the excavation face support force. Detailed Implementation

[0048] To better understand the above-mentioned objectives, features, and advantages of this application, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, where there is no conflict, the embodiments and features described in the embodiments of this application can be combined with each other.

[0049] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below.

[0050] Please see Figure 1 This is a flowchart illustrating a method for coordinated control of instability at the excavation face and surface settlement in a shield tunnel in gravel strata, as shown in an exemplary embodiment of this application. (Reference) Figure 1 It can be seen that the method for coordinated control of instability of the excavation face and surface settlement of the shield tunnel in the gravel stratum can include steps S110 to S150.

[0051] Step S110: Obtain the random field of soil parameters of the target stratum and map the random field of soil parameters to the shield tunnel numerical model.

[0052] In one embodiment of this application, the target stratum may include a geological body with heterogeneous characteristics in parameter space, such as sand and gravel strata, composite strata, or heterogeneous soft and hard strata. This embodiment of the application uses sand and gravel strata as an example for illustration. The soil parameter random field is a mathematical model generated based on geological survey statistics using random field theory, which reflects the fluctuation characteristics of soil parameters (such as internal friction angle, elastic modulus, etc.) in three-dimensional space. The process of mapping the soil parameter random field to the shield tunnel numerical model can be seen as a process of assigning material property values ​​that can be directly read by numerical computation grid cells, transforming abstract statistical laws into such values.

[0053] Step S120: Based on the mapped shield tunnel numerical model, determine the first mapping relationship between the excavation face support force and the ground loss.

[0054] In one embodiment of this application, by applying a series of boundary conditions with different sizes of excavation face support forces to the shield tunnel numerical model, the disturbance effect of support pressure changes on the soil ahead during shield tunneling is simulated, and the corresponding ground loss data is extracted. Ground loss can characterize the volumetric deformation or mass loss caused by the movement of soil at the excavation face into the excavation space. The first mapping relationship can characterize the nonlinear suppression law of support force on ground loss under different geological conditions.

[0055] Step S130: Based on the pre-trained surface subsidence proxy model, determine the second mapping relationship between ground loss and the probability of exceeding the surface subsidence limit.

[0056] In one embodiment of this application, due to the variability of soil parameter space, the same formation loss may induce different surface settlement responses under different random fields of soil parameters. A pre-trained surface settlement surrogate model is a model that has learned and mastered the nonlinear mapping relationship between input parameters and surface settlement. The surface settlement surrogate model can be a gradient boosting model, a neural network, or a support vector machine, etc. Using a surface settlement surrogate model to replace time-consuming traditional numerical calculations allows for rapid large-scale Monte Carlo simulations, thereby quickly establishing the correspondence between the formation loss rate and the probability that surface settlement exceeds a certain allowable value.

[0057] Step S140: Determine the allowable formation loss rate based on the preset surface subsidence excess probability threshold and the second mapping relationship.

[0058] In one embodiment of this application, the surface subsidence exceedance probability threshold can be a risk control standard determined according to the engineering environmental protection level, such as 5% or other more stringent values ​​in sensitive urban areas. Substituting the surface subsidence exceedance probability threshold into the second mapping relationship, the maximum degree of loss that the stratum can withstand under the current geological conditions and risk tolerance can be derived in reverse, i.e., the allowable stratum loss rate.

[0059] Step S150: Based on the allowable formation loss rate and the first mapping relationship, determine the excavation face support force range to construct the excavation face support force safety envelope diagram.

[0060] Among them, the safety envelope diagram of the excavation face support force is used to query the minimum support force that meets the requirements of collaborative control under different engineering conditions.

[0061] In one embodiment of this application, the allowable ground loss rate can be substituted into the first mapping relationship to deduce the required excavation face support force range that ensures both the excavation face remains stable and the risk of surface settlement is controllable. After determining the excavation face support force range, a multi-dimensional excavation face support force safety envelope diagram can be constructed by comprehensively considering changes in key engineering parameters such as tunnel depth ratio and rock content in the strata. The excavation face support force safety envelope diagram can be a comprehensive design map integrating mechanical stability constraints and environmental probability constraints. In practical engineering applications, technicians can directly obtain the minimum support force design value that satisfies both safety objectives by simply consulting the excavation face support force safety envelope diagram based on the current geological conditions and environmental requirements.

[0062] In one embodiment of this application, step S110, which involves obtaining a random field of soil parameters of the target stratum and mapping the random field of soil parameters to a shield tunnel numerical model, may include: obtaining statistical characteristics and autocorrelation functions of the soil parameters of the target stratum; generating a random field of soil parameters based on the statistical characteristics and autocorrelation functions; and discretizing and mapping the random field of soil parameters to each grid cell of the shield tunnel numerical model.

[0063] It should be noted that the statistical characteristics of soil parameters can include the mean, variance, or coefficient of variation of the soil parameters. These statistical characteristics can characterize the overall spatial distribution level and dispersion of soil parameters. The autocorrelation function can characterize the dependence of soil parameter values ​​on spatial distance, and can determine the spatial continuity and fluctuation scale of the heterogeneity of the target stratum. Through the joint constraints of the statistical characteristics of soil parameters and the autocorrelation function, the generated random field of soil parameters not only conforms to the measured laws in probability distribution, but also realistically reproduces the texture characteristics of the target stratum in spatial structure. This avoids the simulation distortion problems caused by the traditional homogeneous stratum assumption or simple random sampling that ignores spatial correlation. The generation of the random field of soil parameters can be implemented in the MATLAB environment.

[0064] Furthermore, the process of generating a random field of soil parameters based on statistical characteristics and autocorrelation functions can include: calculating the eigenvalues ​​and eigenfunctions of the autocorrelation function based on a Gaussian autocorrelation function; determining the number of cutoff terms in the random field expansion according to a preset energy ratio threshold; generating an independent standard normal random sample matrix using a sampling method, and decomposing the correlation coefficient matrix to obtain a correlated normal random sample matrix; discretizing the correlated Gaussian random field using a series expansion method based on the eigenvalues, eigenfunctions, number of cutoff terms, and correlated normal random sample matrix, and transforming the correlated Gaussian random field into a correlated non-Gaussian random field through an equal probability transformation to obtain the random field of soil parameters; wherein, the Gaussian autocorrelation function characterizes the spatial correlation of soil parameters based on the fluctuation range in the horizontal and vertical directions.

[0065] The Gaussian autocorrelation function, by introducing two key parameters—horizontal and vertical fluctuation ranges—characterizes the correlation distance differences of the target soil layer in the horizontal extension and vertical deposition directions, respectively. The eigenvalues ​​and eigenfunctions obtained based on the Gaussian autocorrelation function constitute an orthogonal basis describing spatial variability patterns. The magnitude of the eigenvalue represents the energy weight contained in the corresponding spatial pattern, i.e., its contribution to the overall spatial variability.

[0066] The random field of soil parameters can be discretized into independent standard normal random variables using the KL series expansion method, resulting in a two-dimensional random field. The KL series expansion is shown below:

[0067] ;

[0068] in, Represents the coordinates of any point within a two-dimensional computational domain. Represents external space coordinates. Represents a random field of soil parameters The mean, Represents a random field of soil parameters standard deviation Represents the eigenvalues ​​of the autocorrelation function. This represents the characteristic function corresponding to the eigenvalue. Represents a series of orthogonal and uncorrelated random variables. This indicates the number of truncated terms. Substituting the relevant normally distributed random sample matrix into the series expansion... The relevant Gaussian random field is obtained, and then the relevant Gaussian random field is transformed into a relevant non-Gaussian random field through equal probability transformation, thus obtaining the soil parameter random field.

[0069] For example, the Gaussian autocorrelation function can be expressed as:

[0070] ;

[0071] in, This represents the horizontal distance between two points in a two-dimensional random field. This represents the vertical distance between two points in a two-dimensional random field. Indicates the range of fluctuation in the horizontal direction. Indicates the range of fluctuation in the vertical direction. This represents an exponential function.

[0072] By calculating the proportion of the sum of the first few eigenvalues ​​to the total sum of all eigenvalues ​​(i.e., the expected energy ratio factor), the number of terms corresponding to the expected energy ratio factor reaching a preset threshold (e.g., 90% or 95%) is the truncation term. Determining the number of truncation terms can preserve the main low-frequency components and macroscopic trends in the spatial variability of the stratigraphy, ensuring the physical realism of the simulation, while eliminating high-frequency noise components that have a weak impact on the engineering response, reducing the computational dimensionality and resource consumption of subsequent numerical calculations.

[0073] For example, for a two-dimensional computational domain Expected ratio factor It can be represented as:

[0074] ;

[0075] in, The expected energy ratio factor, The eigenvalues ​​are arranged in descending order. The number of truncated terms, and These represent the dimensions of the two-dimensional computational domain in the horizontal and vertical directions, respectively. Number of truncation terms. The value of needs to satisfy It should be noted that the eigenvalues ​​need to be arranged in descending order.

[0076] An independent standard normal random sample matrix is ​​generated using the Latin hypercube sampling method. Subsequently, the correlation coefficient matrix is ​​decomposed to transform the independent samples into a correlated normal random sample matrix that satisfies a specific cross-correlation relationship. In sandy and gravelly strata, the internal friction angle and deformation modulus are physically related. Introducing cross-correlation through matrix decomposition can avoid the deviation in the joint probability distribution caused by the assumption of parameter independence.

[0077] Because the series expansion method mathematically generates a Gaussian random field that follows a normal distribution, while actual geotechnical parameters (such as permeability coefficient and strength indicators) often follow a log-normal distribution, a Gamma distribution, or other skewed distributions, directly using the Gaussian field for calculations will result in negative values ​​or non-physical parameter values, violating the laws of geotechnical mechanics. Therefore, each sample point in the relevant Gaussian random field can be mapped to the corresponding quantile of the target non-Gaussian distribution. The transformation process can maintain the relative position of the sample points in the sorted sequence, and can correct the marginal distribution to the target non-Gaussian distribution while preserving the spatial correlation structure constructed by the series expansion method. Figure 2 As shown, Figure 2 This is a schematic diagram of a random field of soil parameters, illustrating an exemplary embodiment of this application. Figure 2The left subplot represents a random field diagram of the reference secant modulus, where the horizontal axis represents the horizontal direction of the two-dimensional computational domain, the vertical axis represents the vertical direction of the two-dimensional computational domain, and the vertical axis represents the reference secant modulus. Figure 2 The right-hand subplot represents a random field diagram of the internal friction angle, where the horizontal axis represents the horizontal direction of the two-dimensional computational domain, the vertical axis represents the vertical direction of the two-dimensional computational domain, and the vertical axis represents the internal friction angle.

[0078] like Figure 3 As shown, it is a schematic diagram of a shield tunnel numerical model illustrating an exemplary embodiment of this application. Figure 3 The left-hand subgraph represents a schematic diagram of the reference secant modulus in the numerical model of a shield tunnel. Figure 3 The right-hand subplot represents a schematic diagram of the internal friction angle in the shield tunnel numerical model. The center point coordinates of each grid cell in the shield tunnel numerical model are extracted and substituted into the generated continuous random field function, or by nearest neighbor interpolation, to assign continuous field values ​​to discrete grid cells. This allows each cell in the shield tunnel numerical model to possess material properties that reflect the true spatial variability of the geological formation.

[0079] In one embodiment of this application, step S120, which determines the first mapping relationship between the excavation face support force and the ground loss based on the mapped shield tunnel numerical model, may include: applying different excavation face support forces to the mapped shield tunnel numerical model and extracting the corresponding excavation face ground loss data; and fitting the excavation face ground loss data based on the surrounding rock characteristic curve model to obtain the first mapping relationship.

[0080] In the numerical model of a shield tunnel mapped with a random field of soil parameters, a series of gradient-varying excavation face support forces are applied by adjusting the boundary conditions of the excavation face. The range of values ​​for the excavation face support forces should cover the complete mechanical evolution process of active instability at the excavation face, which is the process by which the excavation face support forces gradually decrease from the initial ground stress P0 to the minimum support force. For example, multiple different support force ratio conditions can be set to fully capture the response characteristics of the soil under different stress levels. After each level of excavation face support force is applied and reaches computational equilibrium, the corresponding excavation face stratum loss data is extracted. Based on the acquired multiple sets of support force and stratum loss correspondence data, a nonlinear regression fitting is performed using a surrounding rock characteristic curve model to obtain the first mapping relationship that can continuously describe the evolution law of the two.

[0081] For example, the numerical model of a shield tunnel can be built using FLAC3D (Fast Lagrangian Analysis of Continua in 3 Dimensions) software. A program can be written using FLAC3D's built-in FISH (FLAC3D Scripting Language) to extract the center point coordinates of each grid cell within the region of interest in the numerical model, used for spatial matching with a random field. Furthermore, code can be written using FLAC3D's built-in Python interface to assign the parameter values ​​corresponding to each cell in the parametric random field generated by MATLAB to the corresponding grid cells in the numerical model, thus achieving coupling between the random field and the numerical model.

[0082] It should be noted that the excavation face ground loss data includes the volumetric deformation of each grid cell at the excavation face. Ground loss represents the amount of volumetric loss of soil within the stratum. Volumetric deformation directly reflects the volumetric shrinkage or expansion of soil cells under three-dimensional stress paths, accurately revealing at a microscale the irreversible mass loss caused by soil movement into the excavation space. Figure 4 As shown, this is an exemplary embodiment of the present application illustrating the evolution of the excavation face support ratio and the formation loss rate curve, with the horizontal axis... Indicates the ratio of excavation face support force, longitudinal axis The curve represents the soil loss rate caused by tunnel excavation. Blue dots indicate numerical results, and red curves indicate fitting results. The soil loss rate curve, obtained based on the volumetric deformation integral, reflects the effect of changes in support force on suppressing soil volume loss.

[0083] After determining the formation loss metric, the data can be fitted using a surrounding rock characteristic curve model to obtain the first mapping relationship. The surrounding rock characteristic curve model can be used to describe the evolution relationship between surrounding rock convergence deformation and support pressure. The expression for the first mapping relationship can be:

[0084] ;

[0085] in, It represents the soil loss rate caused by tunnel excavation. In three dimensions, the soil loss rate represents the ratio of the volume loss of soil caused by tunnel excavation to the theoretical excavation volume of the tunnel. In two dimensions, the soil loss rate represents the ratio of the area of ​​soil loss caused by tunnel excavation per unit tunnel length to the theoretical area of ​​the excavation face. This indicates the ratio of excavation face support force to total excavation face support force. Initial ground stress at the center point of the excavation face Dimensionless ratio; Denotes the first fitting coefficient. This represents the second fitting coefficient. The third fitting coefficient is represented by , and the fourth fitting coefficient is represented by d. The determination of each fitting coefficient is related to parameters such as the tunnel burial depth ratio and the rock content of the strata.

[0086] Combination Figure 4 Analysis of the curve shape of this first mapping relationship reveals that the effect of excavation face support force on suppressing ground loss exhibits a significant nonlinear characteristic. When the support force is relatively low, the curve slope is large, indicating that even a small increase in support force can significantly reduce ground loss; this is a key sensitive area for controlling excavation face stability. As the support force ratio gradually increases, the curve tends to flatten and asymptotically approaches a certain residual value, indicating that excessively high support force has diminishing marginal benefits in further reducing ground loss. Based on Figure 4 Analysis shows that collaborative control needs to find a reasonable support force range that can both ensure the excavation face does not become unstable and control the stratum loss within the allowable range, under the constraint of the first mapping relationship.

[0087] In one embodiment of this application, a second mapping relationship between stratum loss and the probability of exceeding the stratum subsidence limit is determined based on a pre-trained surface subsidence surrogate model, including: constructing an input space containing tunnel burial depth ratio, stratum rock content, stratum loss rate and random field variables; generating training samples based on the input space, predicting the surface subsidence response through the surface subsidence surrogate model; and performing Monte Carlo simulation based on the surface subsidence response to establish the second mapping relationship.

[0088] Due to the spatial variability of soil parameters, surface settlement response exhibits significant randomness. Traditional deterministic calculation methods struggle to capture its probability distribution characteristics, while large-scale probabilistic analysis based directly on numerical models faces excessive computational costs. Therefore, a surrogate model can be used as a computational accelerator. By constructing a multidimensional input space including tunnel depth ratio, soil rock content, soil loss rate, and random field variables, representative training samples are generated. The surface settlement surrogate model is then used to quickly predict the corresponding surface settlement response. Monte Carlo simulations are then performed based on a large number of prediction results, ultimately establishing a quantitative mapping relationship between soil loss rate and the probability of exceeding surface settlement limits.

[0089] Furthermore, the process of generating training samples based on the input space may include: uniformly sampling the tunnel burial depth ratio and the rock content of the stratum, using log-space stratified sampling for the stratum loss rate, and using standard normal stratified sampling for the random field variables to obtain training samples.

[0090] Since the values ​​of tunnel burial ratio and soil rock content are relatively limited and exhibit discrete distribution characteristics in engineering, uniform sampling ensures balanced coverage of the samples across the entire design domain, avoiding predictive blind spots under certain working conditions. Considering the strong nonlinear sensitivity between soil loss rate and surface settlement, linear uniform sampling could easily lead to insufficient sample density in this sensitive area, causing the model to miss crucial details of risk evolution. Therefore, logarithmic spatial stratified sampling can be used for soil loss rate, densifying sample points towards low-value sensitive areas through logarithmic transformation, ensuring the model accurately captures subtle gradient changes in settlement risk as it transitions from a safe to a critical state. Random field variables can characterize the spatial variability of soil parameters. Since random field variables are independent standard normal random variables generated by series expansion, standard normal stratified sampling not only ensures the representativeness of the samples near the mean but also effectively covers areas with low probability density.

[0091] It should be noted that each training sample includes two parts: input features and output labels. The input features include the following four types of parameters: (1) tunnel burial depth ratio, which is the ratio of tunnel burial depth to excavation diameter; (2) stratum stone content, which is the volume ratio of pebbles in sand and gravel strata; (3) stratum loss rate, which is the ratio of stratum volume loss caused by tunnel excavation to the theoretical excavation volume of the tunnel; and (4) random field variables, which are a set of independent standard normal random variables generated by the series expansion method. The output label is the surface settlement value corresponding to the above input features, which is calculated by the numerical model. By pairing the above four types of input parameters with the output labels, a complete training sample set is formed, which is used to train the surface settlement surrogate model.

[0092] Machine learning algorithms such as gradient boosting models can be used to construct land subsidence surrogate models. For example... Figure 5 As shown, this is a scatter plot comparing the predicted and actual values ​​of a surrogate model for land subsidence, illustrating an exemplary embodiment of this application. The horizontal axis represents the actual value of land subsidence, and the vertical axis represents the predicted value. Blue dots represent paired data points of the surrogate model's predicted values ​​and the actual values ​​of the numerical simulation, while the red dashed line represents the ideal prediction baseline. By comparing and verifying the predicted values ​​of the surrogate model with the actual values ​​of the numerical simulation, it can be found that the data points are closely clustered near the diagonal line, indicating a high degree of consistency between the two. The surrogate model trained with differential sampling can replace time-consuming numerical simulations for subsequent large-scale probability calculations. Monte Carlo simulations are conducted using the surrogate model to quickly generate a histogram of the probability distribution of land subsidence, thereby statistically obtaining the probability of exceeding the limit for land subsidence under different stratum loss rate levels.

[0093] For example, the XGBoost (eXtreme Gradient Boosting) algorithm can be used to build a proxy model.

[0094] Based on discrete probability data points obtained from Monte Carlo simulations of surface subsidence response, a sigmoid function was used to fit the relationship between the formation loss rate and the probability of exceeding the surface subsidence limit, constructing a continuous and smooth second mapping relationship. For example... Figure 6 As shown in the figure, this is an exemplary embodiment of the present application illustrating the relationship between the probability of exceeding the limit for surface subsidence and the formation loss rate. The horizontal axis represents the formation loss rate, the vertical axis represents the probability of exceeding the limit for surface subsidence, the blue dots represent the calculation results of the surrogate model, and the red lines represent the fitting results of the Logistic function. The second mapping relationship curve exhibits a typical "S"-shaped evolution characteristic: the probability of exceeding the limit for surface subsidence remains at an extremely low level only when the formation loss rate is below a certain critical threshold, representing a safe plateau zone; as the formation loss rate increases and crosses the sensitive transition zone, the probability of exceeding the limit rises rapidly; when the loss rate exceeds another upper limit threshold, the probability tends to saturate, representing an inevitable exceedance zone.

[0095] For example, the expression for the second mapping relationship may include:

[0096] ;

[0097] in, This indicates the probability of exceeding the limit for land subsidence. Indicates formation loss rate, Indicates the critical formation loss rate. The value of is P f =50% value, This represents the risk growth sensitivity coefficient; the larger k is, the higher P is. f The faster the growth.

[0098] In one embodiment of this application, the process of determining the allowable ground loss rate based on a preset ground subsidence exceedance probability threshold and a second mapping relationship, and determining the excavation face support force range based on the allowable ground loss rate and the first mapping relationship, to construct an excavation face support force safety envelope map, may include: substituting the ground subsidence exceedance probability threshold into the second mapping relationship to determine the allowable ground loss rate; substituting the allowable ground loss rate into the first mapping relationship to determine the excavation face support force range; and constructing the excavation face support force safety envelope map based on the excavation face support force range.

[0099] The ground settlement exceedance probability threshold can be used as the environmental constraint boundary. Substituting this into the second mapping relationship (i.e., the S-shaped function relationship between the formation loss rate and the exceedance probability), the maximum loss the formation can withstand under the current risk tolerance, i.e., the allowable formation loss rate, can be calculated in reverse. This allowable formation loss rate can then be used as the constraint boundary, substituting this into the first mapping relationship (i.e., the surrounding rock characteristic curve model), to calculate the minimum support force required to meet settlement control requirements, serving as the lower limit of the excavation face support force range. The initial ground stress P0 is determined as the upper limit of the support force for active failure. The lower and upper limits of the support force together constitute the excavation face support force range. Based on this range, different combinations of tunnel depth ratios and formation rock content are traversed, and the support force ranges calculated under each working condition are fitted in three-dimensional space to construct a safety envelope diagram of the excavation face support force. In actual construction, technicians do not need to repeat time-consuming numerical calculations. They can simply look up the drawings based on the current working conditions to obtain the design value of the support force that simultaneously meets the dual objectives of excavation face stability and surface settlement control.

[0100] Furthermore, the threshold for the probability of exceeding the surface settlement limit is 5%; and / or, the allowable settlement value corresponding to the probability of exceeding the surface settlement limit is 20 mm. The allowable settlement value of 20 mm corresponding to the probability of exceeding the surface settlement limit is a physical limit determined based on a comprehensive consideration of the protection requirements for surrounding buildings, underground pipelines and road surfaces; the 5% probability threshold represents a 95% reliability level.

[0101] It should be noted that in other engineering scenarios, depending on the level of environmental protection, the threshold for the probability of exceeding the limit of surface settlement can also be set to 1%, 3% or 10%, etc., and the corresponding allowable settlement value can also be adjusted to 10mm, 15mm or 30mm, etc. This application does not limit this.

[0102] For example, when the probability threshold for exceeding the surface subsidence limit is 5%, the formula for calculating the allowable formation loss rate can be expressed as:

[0103] ;

[0104] in, This indicates the allowable formation loss rate.

[0105] Combination Figure 7 The structure and application of the safety envelope diagram for excavation face support force are explained in detail. For example... Figure 7 As shown, it is an exemplary embodiment of this application illustrating the safety envelope diagram of the excavation face support force. Figure 7 The envelope diagram shown is a three-dimensional surface diagram with the tunnel burial depth ratio C / D and the rock content of the stratum VBP as independent variables and the excavation face support force as dependent variable. The horizontal axis represents the rock content of the stratum (VBP), the vertical axis represents the tunnel burial depth ratio (C / D), and the vertical axis represents the excavation face support force. Figure 7 The three-dimensional surface shown essentially divides the area into a safe zone and an unsafe zone: the area above the surface represents sufficient support, which can ensure that the excavation face does not become unstable and can control the probability of excessive surface settlement within the preset threshold; while the area below the surface represents insufficient support, which poses a dual risk of excavation face instability or excessive surface settlement.

[0106] In practical engineering applications, the safety envelope diagram of excavation face support force provides an intuitive quantitative design tool. For example, when a shield tunnel passes through a gravel stratum, and the survey determines that the current tunnel depth ratio is 2.0 and the stratum rock content is 50%, technicians can directly... Figure 7 The corresponding coordinate point is located in the center, and the vertical axis value corresponding to the surface is read. This value represents the minimum design support force required to meet the collaborative control requirements under this working condition. If the support force set on-site is higher than this value, it is considered safe; otherwise, construction parameters need to be adjusted or reinforcement measures need to be taken. Compared with the qualitative judgment method that relies on the engineer's personal experience or a single safety factor in traditional technologies, the method of determining the support force based on the envelope diagram not only eliminates the uncertainty of human factors but also considers the spatial variability of formation parameters and multi-objective collaborative constraints, thereby improving the scientific nature and accuracy of support force design.

[0107] In one embodiment of this application, the target stratum can be a sandy gravel stratum; the soil parameters corresponding to the random field of soil parameters may include the internal friction angle and the reference secant modulus. Both the internal friction angle and the reference secant modulus follow a log-normal distribution, and there is a cross-correlation between the internal friction angle and the reference secant modulus. The sandy gravel stratum is composed of a gravel and pebble skeleton and a sand filling body. Its mechanical properties are significantly affected by particle size distribution, pore structure, and skeleton contact state, exhibiting extremely strong spatial variability and nonlinear characteristics. In such strata, if the homogeneous soil layer assumption or a simple normal distribution model is used, it is often difficult to truly reflect the actual distribution law of soil parameters, which leads to a large deviation in the excavation face stability analysis and surface settlement prediction.

[0108] The internal friction angle and reference secant modulus are selected as key random variables characterizing the mechanical properties of gravel strata. These two parameters determine the shear strength and deformation stiffness of the soil, respectively, directly influencing the instability mode of the excavation face and the propagation mechanism of stratum losses. Considering the natural non-negativity and right-skewed distribution characteristics of soil parameters, this embodiment sets both the internal friction angle and the reference secant modulus to follow a log-normal distribution. Compared to a normal distribution, the log-normal distribution can effectively avoid non-physical results such as negative values ​​during sampling, and at the same time better fits the long-tailed effect commonly found in measured data.

[0109] For example, based on the statistical results of an engineering survey of a typical sand and gravel stratum, the mean value of the internal friction angle can be taken as 35°, with a coefficient of variation of 0.15; the mean value of the reference secant modulus can be taken as 80MPa, with a coefficient of variation of 0.25; the horizontal fluctuation range is set to 10m, and the vertical fluctuation range is set to 2m, in order to reflect the anisotropy characteristics of the sedimentary strata, where the horizontal stratification is stronger than the vertical anisotropy.

[0110] It should be noted that this embodiment considers the cross-correlation between the internal friction angle and the reference secant modulus, setting their cross-correlation coefficient to 0.6. In gravel strata, when the gravel content is high and the particles are tightly interlocked, the soil not only exhibits higher shear strength (increased internal friction angle) but also greater resistance to deformation (increased reference secant modulus). If this intrinsic correlation is ignored during the random field generation process, and the two parameters are simply treated as independent random variables, false parameter combinations such as "high strength and low stiffness" or "low strength and high stiffness," which rarely occur in actual geology, will be created, causing the numerical simulation results to deviate from the actual working conditions. By introducing the cross-correlation coefficient matrix and performing decomposition processing, the generated joint random field can more accurately reproduce the coupled fluctuation characteristics between parameters.

[0111] Under the aforementioned specific parameter settings, based on the log-normal marginal probability density function and Gaussian autocorrelation function, a series expansion method is used to generate a joint random field considering cross-correlation of the internal friction angle and the reference secant modulus, and this field is discretized and mapped to each grid cell of the shield tunnel numerical model. In this heterogeneous numerical model, excavation face support forces under multiple working conditions are applied successively, volumetric deformation data is extracted, and a surrounding rock characteristic curve suitable for sand and gravel strata is fitted to obtain the first mapping relationship. A training sample containing burial depth ratio, rock content, stratum loss rate, and cross-correlation random field variables is generated using a differentiated sampling strategy to train a high-precision surface settlement proxy model. A second mapping relationship between stratum loss rate and surface settlement exceedance probability is established through Monte Carlo simulation. With a 5% surface settlement exceedance probability threshold as a constraint, the allowable stratum loss rate is calculated inversely, and a three-dimensional excavation face support force safety envelope diagram is constructed.

[0112] It should be noted that considering parameter cross-correlation can improve the accuracy of settlement risk assessment. Under the same mean and coefficient of variation conditions, if the positive correlation between the internal friction angle and the reference secant modulus is ignored, the calculated standard deviation of surface settlement will be higher than when considering cross-correlation, resulting in an inflated design value of support force and a waste of construction resources.

[0113] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0114] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for coordinated control of shield tunnel face instability and ground surface settlement in a sandy cobble stratum, characterized in that, include: Obtain the random field of soil parameters of the target stratum and map the random field of soil parameters to the numerical model of the shield tunnel; Based on the mapped shield tunnel numerical model, the first mapping relationship between the excavation face support force and the ground loss is determined. Based on a pre-trained surface subsidence proxy model, a second mapping relationship between the stratum loss and the probability of exceeding the surface subsidence limit is determined; Based on the preset land subsidence exceedance probability threshold and the second mapping relationship, the allowable formation loss rate is determined; Based on the allowable formation loss rate and the first mapping relationship, the excavation face support force range is determined to construct an excavation face support force safety envelope diagram. The excavation face support force safety envelope diagram is used to query the minimum support force that meets the collaborative control requirements under different engineering conditions.

2. The method according to claim 1, characterized in that, The process of obtaining a random field of soil parameters for the target stratum and mapping the random field of soil parameters to a shield tunnel numerical model includes: Obtain the statistical characteristics and autocorrelation function of the soil parameters of the target stratum; Based on the statistical characteristics and the autocorrelation function, the random field of soil parameters is generated; The soil properties are discretized and mapped to each grid cell of the shield tunnel numerical model.

3. The method according to claim 2, characterized in that, The generation of the random field of soil parameters based on the statistical characteristics and the autocorrelation function includes: Based on the Gaussian autocorrelation function, calculate the eigenvalues ​​and eigenfunctions of the autocorrelation function; The number of truncation terms in the random field expansion is determined based on a preset energy ratio threshold. An independent standard normal random sample matrix is ​​generated using a sampling method, and the correlation coefficient matrix is ​​decomposed to obtain a correlated normal random sample matrix. Based on the eigenvalues, the eigenfunctions, the number of truncation terms, and the relevant normal random sample matrix, the relevant Gaussian random field is discretized using the series expansion method, and the relevant Gaussian random field is transformed into a relevant non-Gaussian random field through equal probability transformation to obtain the soil parameter random field; The Gaussian autocorrelation function characterizes the spatial correlation of soil parameters based on the fluctuation range in the horizontal and vertical directions.

4. The method according to claim 1, characterized in that, The first mapping relationship between the excavation face support force and ground loss is determined based on the mapped shield tunnel numerical model, including: Different excavation face support forces are applied to the mapped shield tunnel numerical model, and the corresponding excavation face ground loss data are extracted. The first mapping relationship is obtained by fitting the formation loss data of the excavation face based on the surrounding rock characteristic curve model.

5. The method according to claim 4, characterized in that, The excavation face stratum loss data includes the volumetric deformation of each grid cell at the excavation face.

6. The method according to claim 1, characterized in that, The pre-trained surface subsidence proxy model determines a second mapping relationship between the formation loss and the probability of exceeding the surface subsidence limit, including: Construct an input space that includes tunnel burial depth ratio, stratum rock content, stratum loss rate, and random field variables; Training samples are generated based on the input space, and the surface subsidence response is predicted using the surface subsidence proxy model. Monte Carlo simulations were performed based on the surface subsidence response to establish the second mapping relationship.

7. The method according to claim 6, characterized in that, The generation of training samples based on the input space includes: The tunnel burial depth ratio and the rock content of the stratum are uniformly sampled, the stratum loss rate is sampled using log-space stratified sampling, and the random field variables are sampled using standard normal stratified sampling to obtain the training samples.

8. The method according to claim 1, characterized in that, The method involves determining the allowable ground loss rate based on a preset ground subsidence exceedance probability threshold and the second mapping relationship, and determining the excavation face support force range based on the allowable ground loss rate and the first mapping relationship to construct an excavation face support force safety envelope diagram, including: Substitute the above-mentioned surface subsidence excess probability threshold into the second mapping relationship to determine the allowable formation loss rate; Substituting the allowable formation loss rate into the first mapping relationship, the minimum support force that meets the settlement control requirements is determined, thereby determining the support force range of the excavation face; Based on the excavation face support force range, a safety envelope diagram of the excavation face support force is constructed.

9. The method according to claim 1, characterized in that, The target stratum is a sandy gravel stratum; the soil parameters corresponding to the random field of the soil parameters include the internal friction angle and the reference secant modulus. Both the internal friction angle and the reference secant modulus follow a log-normal distribution, and there is a cross-correlation between the internal friction angle and the reference secant modulus.

10. The method according to claim 8, characterized in that, The threshold for the probability of exceeding the surface subsidence limit is 5%; and / or, the allowable subsidence value corresponding to the probability of exceeding the surface subsidence limit is 20 mm.