Shield tunneling digital twin stratum construction method and system fusing multi-source data

The three-dimensional model was established through multi-layer perceptron and sparse Bayesian learning technology, and combined with finite element and Kalman filtering technology, the uncertainty of stratigraphic distribution and geotechnical parameters in shield tunnels was solved, achieving high-precision construction Forecasting and dynamic updates improve the intelligence and safety of construction.

WO2024229914A1PCT designated stage expired Publication Date: 2024-11-14NO 3 ENG CO LTD OF CCCC THIRD HARBOR ENG CO LTD +2

Patent Information

Application Number
PCT/CN2023/098765
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-09
Filing Date
2023-06-07
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

In shield tunnel engineering, it is difficult to fully detect the strata situation in front of the strata due to concealment, which leads to uncertainty in the distribution of strata and spatial variability of geotechnical parameters, resulting in deviations from the actual situation in construction mechanical analysis, affecting the intelligence, efficiency and safety of construction.

Method used

The multi-layer perceptron model and sparse Bayesian learning technology are used to establish a three-dimensional model of stratigraphic distribution and geotechnical parameters, combined with finite element digital twin model and dual lossless Kalman filtering technology, and fuse multi-source data for stratigraphic distribution and geotechnical parameters. Dynamic inversion and update to achieve high-precision stratigraphic distribution and prediction of geotechnical parameters.

Benefits of technology

It significantly improves the accuracy and safety of shield construction, reduces the uncertainty of formation deformation and mechanical parameters, and improves construction efficiency and engineering application value.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2023098765_14112024_PF_FP_ABST
    Figure CN2023098765_14112024_PF_FP_ABST
Patent Text Reader

Abstract

Disclosed in the present invention are a shield tunneling digital twin stratum construction method and system fusing multi-source data, and the method comprises: first using investigation data of drilling, geotechnical tests, in-situ tests, etc. and considering the characteristics of "sparsity", "multi-dimensional correlation" and "spatial variability" of stratum distribution and geotechnical parameters to construct a multi-layer perceptron-based stratum three-dimensional distribution model, and establish a geotechnical parameter multivariate correlation three-dimensional conditional random field based on sparse Bayesian learning technology; using the stratum distribution and the multivariate geotechnical parameter random field to establish a shield tunnel tunneling random finite element digital twin model; and, on the basis of coupling double-unscented Kalman filtering, fusing shield tunneling mechanical parameters and stratum deformation monitoring data, and performing construction feedback-based dynamic inversion on stratum distribution uncertainty and geotechnical parameter space correlation to achieve fine characterization of stratum distribution and geotechnical parameters in front of a shield tunneling surface, so as to obtain a shield tunneling dynamic digital twin stratum model.
Need to check novelty before this filing date? Find Prior Art

Description

Method and system for constructing digital twin strata for shield tunneling by integrating multi-source data Technical Field

[0001] The present application relates to the field of shield tunneling, and in particular to a method and system for constructing a digital twin stratum for shield tunneling by integrating multi-source data. Background Art

[0002] Due to the hidden nature of shield tunneling, it is difficult to fully understand the true conditions of the strata ahead. Shield tunneling in complex environments faces uncertainties in stratum distribution and spatial variability in geotechnical parameters. When an earth pressure balance shield tunnels through complex and variable strata, the geological conditions along the longitudinal section vary greatly. This can include alternating soft and hard soil layers, sudden changes in stratum thickness, and significant differences in physical and mechanical properties between soft and hard soil layers. Ignoring the rich geological information and uncertainties inherent in the formation of the geotechnical mass, simplifying the variable strata into a homogeneous mass, and neglecting the inherent spatial variability of its physical and mechanical properties will inevitably lead to significant deviations between the actual conditions and the mechanical analysis of shield tunneling construction. Therefore, fully utilizing borehole data, geotechnical test data, and in-situ testing data to quantitatively model the uncertainty in stratum distribution between adjacent boreholes, the spatial variability of geotechnical parameters, and the intercorrelations between these multiple parameters, and thus accurately reconstructing the true conditions of the strata, is key to the intelligent, efficient, and safe excavation of earth pressure balance shield tunnels. At the same time, the stratum deformation monitoring data and shield tunneling mechanical parameter data have temporal and spatial variability and random fluctuations. They are interrelated and influence each other, reflecting to a certain extent the degree of interaction between the shield and the stratum, and can provide feedback information on the geological conditions in front of the shield tunnel excavation face. Relying on powerful modern network technology and communication means, it is ensured that the construction site data is transmitted to the constructed numerical model in real time and quickly to ensure its high-fidelity simulation of the actual construction space. Therefore, by integrating the stratum deformation response and mechanical parameter information of shield tunneling, and conducting dynamic inversion of stratum distribution and geotechnical parameters based on construction feedback, it can effectively reduce its uncertainty, realize the accurate prediction of stratum distribution, geotechnical parameters and surface deformation by integrating multi-source data, and thus effectively guide construction.

[0003] Summary of the Invention

[0004] In response to the above problems and shortcomings, the present invention intends to start from the quantification of stratum distribution and geotechnical parameter variability, multi-source information fusion and inversion, and propose a shield tunneling digital twin stratum construction method and system that integrates multi-source data, providing new ideas and solutions for promoting the robust design of shield tunnel engineering, and has important engineering application value for improving and developing the safe construction of earth pressure balance shields and intelligent tunneling theory in complex urban environments.

[0005] The present invention provides a shield tunneling digital twin stratum construction method and system integrating multi-source data, comprising the following steps:

[0006] Step 1: Collect geotechnical survey data at the shield tunnel location, and summarize and statistically analyze the drilling data, in-situ test data, and indoor geotechnical test data;

[0007] Step 2: Using the stratigraphic attribute information from the borehole data, a multi-layer perceptron model is established to reflect the spatial location characteristics of the bedrock interface at the shield tunnel location, and a three-dimensional reconstruction model of the stratigraphic distribution based on the shield tunnel location is established.

[0008] Step 3: Using indoor geotechnical test data and in-situ test data, considering the mutual correlation of multidimensional geotechnical parameters and the spatial autocorrelation of a single geotechnical parameter, a sparse Bayesian learning technique is used to establish a multivariate correlation three-dimensional conditional random field of geotechnical physical and mechanical parameters, and to construct a refined spatial distribution model of multivariate geotechnical parameters.

[0009] Step 4: To ensure high model fidelity, a finite element digital twin model of the interaction between shield tunneling and strata is established. Based on the stratum distribution multilayer perceptron model and the three-dimensional conditional random field model of geotechnical parameters constructed in Steps 2 and 3, the stratum properties and physical and mechanical parameters of the soil units in the numerical model are automatically assigned. A non-intrusive stochastic finite element digital twin model of earth pressure balance shield construction, integrating the stratum distribution and geotechnical parameters, is then established as the initial digital twin stratum model for mechanical analysis of stratum deformation and response during shield construction.

[0010] Step 5: Coupled with the dual lossless Kalman filter technology, the measured ground deformation and real-time mechanical parameter data from shield tunneling are integrated to update the ground distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters to obtain the optimal inversion estimate of the ground and geotechnical parameters. The ground distribution and geotechnical parameters for the next excavation cycle are then updated to predict the ground deformation for the next excavation cycle.

[0011] Step 6: Use the real-time data from shield tunnel position monitoring to verify the inverted stratum and geotechnical parameter model and the deformation prediction results of the finite element digital twin model.

[0012] Preferably, the three-dimensional reconstruction model of the stratum distribution based on the shield tunnel position is established in step 2 to provide initial stratum distribution parameters, namely δ0, for the subsequent coupled double lossless Kalman filter model.

[0013] Preferably, in step 3, a three-dimensional random field model is established, assuming that x = [x, y, z] T Represents a point in the three-dimensional space of the site, t=[t1,t2,…,t K ] T is the result of indoor geotechnical test or in-situ test of the sampling point, where t Krepresents the experimental value of the Kth soil sample property at this point. Here, a total of K soil sample properties are considered. Assuming that there are N groups of experimental data in the stratum, the random variable matrix likelihood function is constructed for all data:

[0014] In the above formula, B represents the matrix composed of all data; W is the random variable matrix, which is represented by chaotic polynomial expansion; R is the column covariance matrix, that is, the autocorrelation coefficient matrix, which represents the spatial correlation of parameters; C is the row covariance matrix, that is, the mutual correlation coefficient matrix, which represents the mutual correlation of parameters;

[0015] Based on the sparse Bayesian learning technology, conjugate prior distributions are defined for the chaotic polynomial expansion coefficient matrix and autocorrelation distance parameters. The chaotic polynomial expansion is used to characterize the mean of the random variable matrix. In order to improve the robustness of the model, the sparse Bayesian learning technology is used to solve the coefficient matrix of the chaotic polynomial. The coefficient matrix is ​​vectorized and defined as a zero-mean multidimensional normal distribution. Based on the above sparse Bayesian learning technology, a multivariate correlation three-dimensional conditional random field of geotechnical parameters at any point in space is generated, thereby providing the initial parameters of the geotechnical parameter random field, i.e., θ0, for the subsequent construction of the coupled double lossless Kalman filter model.

[0016] Preferably, in step 4, the simplified scheme of the interaction between the mechanical parameters of shield tunneling and the stratum of the model is as follows: 1. The soil compartment pressure is simplified to a linear load acting on the excavation surface; 2. The shield machine is simulated by a rigid unit, and its weight acts on the soil unit at the bottom in the form of a uniformly distributed load; 3. The cutter head torque acts on the soil unit of the excavation surface in the form of shear stress; 4. The thrust acts on the pipe segment; 5. The friction force of the shield shell acts on the soil unit outside the shield machine in the form of uniformly distributed shear stress; 6. The slurry in the gap at the shield tail has not hardened, and the grouting pressure simulation is adopted in the form of normal stress. The circumferential action acts on the soil units around the tunnel, and the slurry in the lining segments behind the gap slowly solidifies and hardens. The linear elastic solid unit is used to simplify the simulation of the solidification effect of the slurry; 7. The lining segments are simulated by linear elastic lining units; through the above simplified conditions, finite element analysis and modeling are carried out. According to the generated three-dimensional model of the stratum distribution and the multivariate correlated three-dimensional conditional random field, the stratum properties and physical and mechanical parameters of the soil units in the numerical model are automatically assigned, and a random finite element refined model of the interaction between shield tunneling and the stratum is established, thereby completing the construction of the initial digital twin.

[0017] Preferably, the dual lossless Kalman filter information fusion in step five is dynamically updated with the formation parameters, because the sampling frequency of the shield machine's tunneling mechanical parameters is much greater than the surface deformation sampling frequency, the tunneling mechanical parameters in each tunneling ring are equivalently processed, and the tunneling parameter fluctuations of the soil cabin pressure, cutterhead torque, jacking force and grouting pressure are regarded as multidimensional normally distributed random variables, and the mean and variance of the multidimensional normally distributed random variables are calculated based on the recorded data.

[0018] Preferably, in step six, the stratum deformation and tunneling mechanical parameter information are integrated to design two coupled nonlinear Kalman filter models, and the parameters of the stratum distribution three-dimensional model and the multivariate correlation three-dimensional conditional random field model are inverted and analyzed respectively; according to the obtained stratum distribution model and the multivariate correlation three-dimensional conditional random field of geotechnical parameters, the initial conditions of the Kalman filter are set, including the state to be estimated, the horizontal state transfer matrix, the geotechnical parameter autocorrelation distance and the initial error covariance matrix; according to the surface deformation prediction error, the observation noise covariance matrix of the Kalman filter is set; in the first Kalman filter model, the stratum deformation data s is integrated k , calculate the Kalman gain, and the parameters of the formation distribution model δ k Update and correct; then use the updated formation distribution model parameter δ k Input into the second Kalman filter model and integrate the formation deformation data s k , calculate the corresponding Kalman gain matrix and estimate the updated multivariate correlation conditional random field parameters θ k ;

[0019] For the stratum distribution model, the detailed process of obtaining the parameter values ​​of the updated stratum distribution model is as follows:

[0020] δ k-1 is the parameter value of the stratum distribution model before updating, M k-1 is the corresponding covariance matrix, η is the empirical hyperparameter, 2m+1 is the total number of ε points, represents the ε point;

[0021] Further estimates are obtained using the estimation function t(x):

[0022] Where, α i is the weight coefficient, N k-1 is the noise, δ k|k-1 is an intermediate estimator;

[0023] By δ k|k-1 According to the conversion formula, we can get the corresponding εk|k-1 , further estimated values ​​are obtained with the help of the estimation function l(x):

[0024] e k|k-1 For further estimation;

[0025] Then two covariance matrices are calculated:

[0026] Where B k-1 For noise;

[0027] From this, the Kalman filter gain matrix is ​​calculated:

[0028] Thus, the parameter value of the updated stratum distribution model is obtained: δ k =δ k|k-1 +F k (z k -e k|k-1 )

[0029] Similarly, the Kalman filter update model of the rock and soil random field parameters can be obtained:

[0030] After obtaining new deformation monitoring data, the stratigraphic distribution and multivariate correlation conditional random field parameters are dynamically updated according to the above steps.

[0031] Preferably, a shield tunneling digital twin stratum construction system based on multi-source data fusion includes:

[0032] Module 1: collects geotechnical survey data at the shield tunnel location, and summarizes and statistically analyzes the drilling data, in-situ test data, and indoor geotechnical test data.

[0033] Module 2: Uses the stratigraphic attribute information from drilling data to establish a multi-layer perceptron model that reflects the spatial location characteristics of the bedrock interface at the shield tunnel location, and builds a three-dimensional reconstruction model of the stratigraphic distribution based on the shield tunnel location.

[0034] Module 3: Uses indoor geotechnical test data and in-situ test data to consider the intercorrelations of multidimensional geotechnical parameters and the spatial autocorrelations of single geotechnical parameters. Based on sparse Bayesian learning technology, it establishes a multivariate correlation three-dimensional conditional random field of geotechnical physical and mechanical parameters and constructs a refined spatial distribution model of multivariate geotechnical parameters.

[0035] Module 4: Used to establish a finite element digital twin model of the interaction between shield tunneling and strata. Based on the stratum distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters constructed in steps 2 and 3, the stratum properties and physical and mechanical parameters of the soil units in the digital twin model are automatically assigned. Then, a non-intrusive stochastic finite element digital twin model of earth pressure balance shield construction is established that integrates the stratum distribution and geotechnical parameters to conduct mechanical analysis of the stratum deformation and response during shield construction.

[0036] Module 5: Used to couple the dual lossless Kalman filter technology, integrate the measured ground deformation and real-time mechanical parameter data of shield tunneling, update the ground distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters, and obtain the optimal inversion estimate of the ground and geotechnical parameters. The ground distribution and geotechnical parameters for the next excavation cycle are updated to predict the ground deformation in the next excavation cycle.

[0037] Module 6: Used to verify the inverted stratum and geotechnical parameter model and the deformation prediction results of the finite element digital twin model using real-time data from shield tunnel position monitoring.

[0038] Preferably, the three-dimensional reconstruction model of the stratum distribution based on the shield tunnel position is established in module 2 to provide initial stratum distribution parameters, namely δ0, for the subsequent coupled dual lossless Kalman filter model.

[0039] Preferably, a three-dimensional random field model is established in module 3, assuming that x = [x, y, z] T Represents a point in the three-dimensional space of the site, t=[t1,t2,…,t K ] T is the indoor geotechnical test result of the sample taken at this point, where t K It represents the experimental value of the K-th soil sample property at this point. Here, a total of K soil sample properties are considered. Assuming that there are N groups of experimental data in the stratum, the random variable matrix likelihood function is constructed for all data.

[0040] In the above formula, B represents the matrix composed of all data; W is the random variable matrix, which is represented by chaotic polynomial expansion; R is the column covariance matrix, that is, the autocorrelation coefficient matrix, which represents the spatial correlation of parameters; C is the row covariance matrix, that is, the mutual correlation coefficient matrix, which represents the mutual correlation of parameters;

[0041] Based on the sparse Bayesian learning technology, conjugate prior distributions are defined for the chaotic polynomial expansion coefficient matrix and autocorrelation distance parameters. The chaotic polynomial expansion is used to characterize the mean of the random variable matrix. In order to improve the robustness of the model, the sparse Bayesian learning technology is used to solve the coefficient matrix of the chaotic polynomial. The coefficient matrix is ​​vectorized and defined as a zero-mean multidimensional normal distribution. Based on the above sparse Bayesian learning technology, a multivariate correlation three-dimensional conditional random field of geotechnical parameters at any point in space is generated, thereby providing the initial parameters of the geotechnical parameter random field, i.e., θ0, for the subsequent construction of the coupled double lossless Kalman filter model.

[0042] Preferably, the simplified scheme of the interaction between the mechanical parameters of shield tunneling and the stratum in the model in module 4 is as follows: 1. The soil compartment pressure is simplified to a linear load acting on the excavation surface; 2. The shield machine is simulated by a rigid unit, and its weight acts on the soil unit at the bottom in the form of a uniformly distributed load; 3. The cutter head torque acts on the soil unit at the excavation surface in the form of shear stress; 4. The thrust acts on the pipe segment; 5. The friction force of the shield shell acts on the soil unit outside the shield machine in the form of uniformly distributed shear stress; 6. The slurry in the gap at the shield tail has not hardened, and the grouting pressure is used for simulation. The positive stress acts circumferentially on the soil units around the tunnel, and the slurry in the lining segments behind the gap slowly solidifies and hardens. The linear elastic solid unit is used to simplify the simulation of the solidification effect of the slurry; 7. The lining segments are simulated by linear elastic lining units; through the above simplified conditions, finite element analysis and modeling are carried out, and according to the generated three-dimensional model of stratum distribution and multivariate correlated three-dimensional conditional random field, the stratum properties and physical and mechanical parameters of the soil units in the digital twin model are automatically assigned, and a random finite element digital twin model of the interaction between shield tunneling and stratum is established.

[0043] Preferably, the dual lossless Kalman filter information fusion in module five is dynamically updated with the formation parameters, because the sampling frequency of the shield machine's tunneling mechanical parameters is much greater than the surface deformation sampling frequency, and the tunneling mechanical parameters in each tunneling ring are equivalently processed, and the tunneling parameter fluctuations of the soil cabin pressure, cutterhead torque, jacking force and grouting pressure are regarded as multidimensional normally distributed random variables, and the mean and variance of the multidimensional normally distributed random variables are calculated based on the recorded data.

[0044] Preferably, in module six, the information of stratum deformation and excavation mechanics parameters is integrated to design two coupled nonlinear Kalman filter models, and the parameters of the stratum distribution three-dimensional model and the multivariate correlation three-dimensional conditional random field model are inverted and analyzed respectively; according to the obtained stratum distribution model and the multivariate correlation three-dimensional conditional random field of geotechnical parameters, the initial conditions of the Kalman filter are set, including the state to be estimated, the horizontal state transfer matrix, the geotechnical parameter autocorrelation distance and the initial error covariance matrix; according to the surface deformation prediction error, the observation noise covariance matrix of the Kalman filter is set; in the first Kalman filter model, the stratum deformation data s is integrated k , calculate the Kalman gain, and the parameters of the formation distribution model δ k Update and correct; then use the updated formation distribution model parameter δ k Input into the second Kalman filter model and integrate the formation deformation data s k , calculate the corresponding Kalman gain matrix and estimate the updated multivariate correlation conditional random field parameters θ k ;

[0045] For the stratum distribution model, the detailed process of obtaining the parameter values ​​of the updated stratum distribution model is as follows:

[0046] δ k-1 is the parameter value of the stratum distribution model before updating, M k-1 is the corresponding covariance matrix, η is the empirical hyperparameter, 2m+1 is the total number of ε points, represents the ε point;

[0047] Further estimates are obtained using the estimation function t(x):

[0048] Where, α i is the weight coefficient, N k-1 is the noise, δ k|k-1 is an intermediate estimator;

[0049] By δ k|k-1 According to the conversion formula, we can get the corresponding ε k|k-1 , further estimated values ​​are obtained with the help of the estimation function l(x):

[0050] e k|k-1 For further estimation;

[0051] Then two covariance matrices are calculated:

[0052] Where B k-1 For noise;

[0053] From this, the Kalman filter gain matrix is ​​calculated:

[0054] Thus, the parameter value of the updated stratum distribution model is obtained: δ k =δ k|k-1 +F k (z k -e k|k-1 )

[0055] Similarly, the Kalman filter update model of the rock and soil random field parameters can be obtained:

[0056] After obtaining new deformation monitoring data, the stratigraphic distribution and multivariate correlation conditional random field parameters are dynamically updated according to the above steps.

[0057] Beneficial effects: By using a multi-layer perceptron model to reconstruct the three-dimensional distribution of strata and sparse Bayesian learning technology to establish a multivariate correlation three-dimensional conditional random field model of geotechnical parameters, and solving the difficulty of traditional methods in using a small amount of survey data to simultaneously characterize the mutual correlation and spatial variability of geotechnical parameters, a high-level spatiotemporal interaction between the digital twin model and the construction site is achieved. The large amount of real-time effective data generated by engineering monitoring is fully utilized to establish a coupled dual-lossless Kalman filter inversion model that integrates multi-source data, realizing real-time dynamic updates of stratum distribution and geotechnical parameter information based on construction feedback. Based on the inclusion of rich multi-source data and the full utilization of available information sources, the accuracy of the inversion of geological conditions in the shield construction area has been significantly improved, and the local site geological information originally limited to limited specific points in the survey report has been expanded to the global scope. The precise geological conditions can be fully and clearly displayed to engineering personnel, helping shield drivers to better adapt to site conditions in their tunneling operations and be safer and more reliable. The uncertainty information of stratum distribution and geotechnical parameters is fully expressed in shield construction control, thereby reducing the risk of equipment wear, excessive stratum deformation, and even shutdown of shield machines due to unreasonable construction parameter settings and other factors, improving shield tunneling efficiency, bringing huge economic benefits to engineering project construction, and providing an effective solution for real-time mechanical analysis of shield construction and precise control of indicators such as settlement. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] FIG1 is an overall flow chart of the present invention;

[0059] Figure 2 is a simplified schematic diagram of the interaction between the earth pressure balance shield tunneling and the stratum;

[0060] Figure 3 is a schematic diagram of the principle of the multi-layer perceptron stratum distribution model;

[0061] Figure 4 is a schematic diagram of the coupled dual lossless Kalman filter joint estimation of stratum distribution and geotechnical parameters. DETAILED DESCRIPTION

[0062] The present invention will be further described below with reference to the accompanying drawings and examples;

[0063] For step 1 in the technical solution, geotechnical investigation reports for shield tunnel construction projects are collected. The corresponding engineering case of the present invention is the section between Pukou Wanhuicheng Station and Qilihe West Station on Nanjing Metro Line 11. The stratigraphic attribute information of the exploration drilling data is summarized, organized and analyzed. For all the available data collected, an area with dense data volume is selected as the research object. The drilling data in the area is preprocessed, and resampled at regular intervals in the depth direction to convert the drilling data into a series of sample points with spatial coordinates and stratigraphic category attributes. For the convenience of calculation, the data is standardized. The processed data is divided into training samples and test samples according to a certain ratio.

[0064] For step 2 of the technical solution, a three-dimensional model of the stratum distribution is established using the processed training sample data based on a multi-layer perceptron model containing weight parameters. The model input is the spatial coordinate value of the point, and the output is the probability that the point belongs to the corresponding stratum category (see Figure 3 for the principle). After training, the drilling information of the site survey is integrated to obtain the stratum distribution of the site. After strong verification by the test samples, a preliminary simulation of the unknown stratum interface between the boreholes is achieved, effectively completing the three-dimensional modeling task based on the stratum distribution of the engineering site. The stratum distribution model obtained in this link mainly provides the initial stratum distribution parameters, namely δ0, for the subsequent coupled dual lossless Kalman filter model.

[0065] For step 3 of the technical solution, the in-situ test data (standard penetration test, static penetration test data) and indoor geotechnical test data (density, saturation, compression coefficient, compression modulus, cohesion, friction angle, etc.) collected in the survey data are summarized, sorted and statistically analyzed. For engineering projects, according to the soil stratification of the three-dimensional stratum model, the in-situ test or indoor geotechnical test data in the same soil layer are used to establish a three-dimensional conditional random field model of multivariate geotechnical parameter cross-correlation. Assumptions Represents a point in the three-dimensional space of the site, t=[t1,t2,…,t K ] T The indoor geotechnical test results of the sampling at this point. Assuming that there are N groups of experimental data in the stratum, the random variable matrix likelihood function is constructed for all the data.

[0066] In the above formula, B represents the matrix composed of all data; W is the random variable matrix, which is represented by chaotic polynomial expansion in this study; R is the column covariance matrix, that is, the autocorrelation coefficient matrix, which represents the spatial correlation of parameters; C is the row covariance matrix, that is, the mutual correlation coefficient matrix, which represents the mutual correlation of parameters.

[0067] Based on sparse Bayesian learning techniques, conjugate prior distributions are defined for the chaotic polynomial expansion coefficient matrix and the autocorrelation distance parameter. The chaotic polynomial expansion is used to characterize the mean of the random variable matrix. To improve the robustness of the model, sparse Bayesian learning techniques are used to solve the chaotic polynomial coefficient matrix. This coefficient matrix is ​​vectorized and defined as a zero-mean multidimensional normal distribution.

[0068] Due to the conjugation of the likelihood function and the prior function, we try to solve the analytical expression of the integral of the marginal likelihood function, and then take the partial derivative of the model parameters to obtain the optimal solution.

[0069] Generating a multivariate correlated three-dimensional conditional random field is challenging. This study investigates methods for decoupling horizontal and vertical autocorrelations, simplifies the calculation of the chaotic polynomial coefficient matrix, and addresses the high dimensionality and low computational efficiency of generating a three-dimensional conditional random field. The optimal model parameters are then used to determine the posterior distribution of the chaotic polynomial coefficient matrix. Using the posterior distribution samples of the coefficient matrix obtained using analytical methods, a multivariate correlated three-dimensional non-stationary conditional random field of geotechnical parameters at any point in space is further generated. Based on the aforementioned sparse Bayesian learning technique, a multivariate correlated three-dimensional conditional random field of geotechnical parameters at any point in space is generated, providing the initial parameters of the geotechnical parameter random field, i.e., θ0, for the subsequent construction of the coupled dual lossless Kalman filter model.

[0070] For step 4 of the technical solution, after completing the three-dimensional stratum modeling of the site and constructing the random field of geotechnical parameters, appropriate finite element modeling software was considered to establish a digital twin model of the interaction between shield tunneling and strata. A simplified representation of the interaction between shield tunneling mechanical parameters and strata in this model is shown in Figure 2. The specific solution is as follows: 1. The soil compartment pressure is simplified as a linear load acting on the excavation face; 2. The shield machine is simulated using rigid elements, with its weight acting as a uniformly distributed load on the bottom soil elements; 3. The cutterhead torque acts as a shear stress on the excavation face soil elements; 4. The thrust acts on the segments; 5. The shield shell friction acts as a uniformly distributed shear stress on the soil elements outside the shield machine; 6. The grouting pressure in the gap at the shield tail, which has not yet hardened, is simulated using grouting pressure, which acts circumferentially on the soil elements surrounding the tunnel in the form of normal stress. The grout in the lining segments behind the gap slowly solidifies and hardens, and linear elastic solid elements are used to simplify the simulation of the grout solidification; 7. The lining segments are simulated using linear elastic lining elements. The project only studies the short-term impact caused by shield machine excavation and does not consider the ground deformation caused by long-term consolidation, so the undrained analysis method is used for numerical simulation calculations.

[0071] Based on the above reasonable simplified conditions, finite element analysis modeling is carried out. According to the generated three-dimensional model of the stratum distribution and the multivariate correlated three-dimensional conditional random field, the stratum properties and their physical and mechanical parameters of the soil units in the numerical model are automatically assigned. A stochastic finite element refined model of the interaction between shield tunneling and the stratum is established, that is, a high-fidelity digital twin.

[0072] For step 5 of the technical solution, to achieve dual-lossless Kalman filter information fusion and dynamic updating of stratum parameters, appropriate intervals were selected within the project site and ground deformation monitoring was conducted in accordance with the "Technical Specifications for Monitoring of Urban Rail Transit Projects." Monitoring sections were arranged at intervals of twice the tunnel diameter along the shield tunneling direction to monitor ground deformation, with the monitoring frequency dynamically adjusted based on research and analysis needs.

[0073] Since the sampling frequency of shield machine tunneling mechanical parameters is much greater than the sampling frequency of surface deformation, for the convenience of calculation, the tunneling mechanical parameters within each tunneling ring are treated equivalently. The fluctuations of tunneling parameters such as soil compartment pressure, cutterhead torque, jacking force, and grouting pressure are regarded as multidimensional normally distributed random variables, and their means and variances are calculated based on the recorded data.

[0074] Regarding step 6 of the technical solution, the equivalent tunneling mechanics parameters of the shield machine are input into the established refined finite element model to calculate stratum deformation, and the predicted deformation is compared with the measured deformation in situ. Considering the validity of the numerical model's prediction range, the present invention only analyzes and integrates deformation data within a range of twice the tunnel diameter before and after the tunnel excavation face. By integrating stratum deformation and tunneling mechanics information, two coupled nonlinear Kalman filter models are designed to perform inversion analysis on the parameters of the 3D stratum distribution model and the parameters of the multivariate correlation 3D conditional random field model, respectively.

[0075] The specific method of the coupled double lossless Kalman filter model (see Figure 3) is as follows: Based on the obtained stratum distribution model and the multivariate correlation three-dimensional conditional random field of geotechnical parameters, the initial conditions of the Kalman filter are set, including the estimated state, parameters (horizontal state transfer matrix, geotechnical parameter autocorrelation distance) and the initial error covariance matrix. Based on the surface deformation prediction error, the observation noise covariance matrix of the Kalman filter is set. In the first Kalman filter model, the multivariate correlation conditional random field parameter θ of the previous iteration step is used. k , integrating the formation deformation data s k , calculate the Kalman gain, and the parameters of the formation distribution model δ k Update and correct; then use the updated formation distribution model parameter δ k Input into the second Kalman filter model and integrate the formation deformation data s k , calculate the corresponding Kalman gain matrix and estimate the multivariate correlation conditional random field parameters θ k ;

[0076] Taking updating the stratum distribution model as an example, the detailed process is as follows:

[0077] δ k-1 is the parameter value of the stratum distribution model before updating, M k-1 is the corresponding covariance matrix, η is the empirical hyperparameter, 2m+1 is the total number of ε points, Represents the ε point.

[0078] Further estimates are obtained using the estimation function t(x):

[0079] Where, α i is the weight coefficient, N k-1 is the noise, δ k|k-1 is an intermediate estimate.

[0080] By δ k|k-1 According to the conversion formula, we can get the corresponding εk|k-1 , further estimated values ​​are obtained with the help of the estimation function l(x):

[0081] e k|k-1 For further estimation;

[0082] Then two covariance matrices are calculated:

[0083] From this, the Kalman filter gain matrix is ​​calculated:

[0084] Thus, the parameter value of the updated stratum distribution model is obtained: δ k =δ k|k-1 +F k (z k -e k|k-1 )

[0085] Similarly, the Kalman filter update model of the rock and soil random field parameters can be obtained:

[0086] After obtaining new deformation monitoring data, the stratigraphic distribution and multivariate correlation conditional random field parameters are dynamically updated according to the above steps.

[0087] At the project site, new borehole stratigraphic attribute data was obtained through drilling and compared with a three-dimensional stratigraphic distribution model. Soil samples obtained from the boreholes were used for indoor geotechnical testing, and the results were compared with a three-dimensional conditional random field model of geotechnical parameters. Furthermore, the predicted stratigraphic deformation values ​​from the updated finite element digital twin model were compared with the measured values. If the aforementioned error values ​​are consistently within a reasonable range, the digital twin model demonstrates high overall credibility and strong potential for engineering applications.

[0088] In summary, the present invention first constructs a finite element digital twin model of the interaction between shield tunneling and strata in densely built areas based on the established refined model of stratum distribution and multivariate geotechnical parameter spatial distribution, realizes the automatic assignment of stratum properties and physical and mechanical parameters of the soil units in the model, and establishes a digital twin of shield tunneling numerical calculation based on coupling technology to facilitate the verification of the subsequent update effect of the dual lossless Kalman filter model. Based on the current stratum deformation monitoring data and shield tunneling parameters, the coupled dual lossless Kalman filter method is used to dynamically update the stratum distribution and geotechnical parameters in the area in real time to achieve accurate characterization of the originally unknown site geological conditions, thereby better feedback and guidance of shield tunneling construction. Adjusting shield tunneling parameters to adapt to the latest global geological conditions effectively reduces uncertainty in the process and greatly improves safety and information levels.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for constructing digital twin strata for shield tunneling by integrating multi-source data, comprising the following steps: Step 1: Collect geotechnical survey data at the shield tunnel location, and summarize and statistically analyze the drilling data, in-situ test data, and indoor geotechnical test data; Step 2: Using the stratigraphic attribute information of the drilling data, a multi-layer perceptron model is established to reflect the spatial position characteristics of the bedrock interface of the shield tunnel location, and a three-dimensional reconstruction model of the stratigraphic distribution based on the shield tunnel location is established; Step 3: Using indoor geotechnical test data and in-situ test data, considering the intercorrelation of multidimensional geotechnical parameters and the spatial autocorrelation of single geotechnical parameters, based on sparse Bayesian learning technology, a multivariate correlation three-dimensional conditional random field of geotechnical physical and mechanical parameters is established, and a multivariate geotechnical parameter spatial distribution model is constructed; Step 4: Establish a finite element digital twin model of the interaction between shield tunneling and strata. Based on the stratum distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters constructed in steps 2 and 3, automatically assign values ​​to the stratum properties and physical and mechanical parameters of the soil units in the digital twin model, and then establish a non-intrusive random finite element digital twin model of earth pressure balance shield construction that integrates stratum distribution and geotechnical parameters to conduct mechanical analysis of stratum deformation and response during shield construction. Step 5: Couple the double lossless Kalman filter technology, integrate the measured stratum deformation and the real-time mechanical parameter data of shield tunneling, update the stratum distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters, and obtain the optimal inversion estimation of stratum and geotechnical parameters; update the stratum distribution and geotechnical parameters of the next excavation cycle to predict the stratum deformation of the next excavation cycle; Step 6: Use the real-time data of shield tunnel position monitoring to verify the inverted formation and geotechnical parameter model and the deformation prediction results of the finite element digital twin model.

2. According to the method for constructing digital twin strata for shield tunneling by integrating multi-source data in claim 1, it is characterized in that: In step 2, the three-dimensional reconstruction model of the stratum distribution based on the shield tunnel position is established to provide the initial stratum distribution parameters, namely δ0, for the subsequent coupled dual lossless Kalman filter model.

3. The method for constructing digital twin strata for shield tunneling by integrating multi-source data according to claim 1, characterized in that: In step 3, a three-dimensional random field model is established, assuming that x = [x, y, z] T represents a point in the three-dimensional space of the site, t=[t1,t2,…,t K ] T is the indoor geotechnical test result of the sample taken at this point, where t K It represents the test value of the Kth soil sample property at this point. Here, a total of K soil sample properties are considered. Assuming that there are N groups of experimental data in the stratum, the random variable matrix likelihood function is constructed for all data. In the above formula, B represents the matrix composed of all data; W is the random variable matrix, which is characterized by chaotic polynomial expansion; R is the column covariance matrix, that is, the autocorrelation coefficient matrix, which represents the spatial correlation of parameters; C is the row covariance matrix, i.e., the cross-correlation coefficient matrix, which represents the cross-correlation of parameters; Based on the sparse Bayesian learning technology, conjugate prior distribution is defined for the chaotic polynomial expansion coefficient matrix and autocorrelation distance parameter, and the chaotic polynomial expansion is used to characterize the mean of the random variable matrix. In order to improve the robustness of the model, the sparse Bayesian learning technology is used to solve the coefficient matrix of the chaotic polynomial, and the coefficient matrix is ​​vectorized and defined as a zero-mean multidimensional normal distribution. Based on the above sparse Bayesian learning technology, a multivariate correlated three-dimensional conditional random field of geotechnical parameters at any point in space is generated, thereby providing the initial parameters of the geotechnical parameter random field, i.e. θ0, for the subsequent construction of the coupled double lossless Kalman filter model.

4. The method for constructing digital twin strata for shield tunneling by integrating multi-source data according to claim 1, characterized in that: In step 4, the simplified scheme of the interaction between the mechanical parameters of shield tunneling and the stratum in this model is as follows:

1. The soil compartment pressure is simplified to a linear load acting on the excavation surface; 2. The shield machine is simulated by a rigid unit, and its weight acts on the soil unit at the bottom in the form of a uniformly distributed load; 3. The cutter head torque acts on the soil unit at the excavation surface in the form of shear stress; 4. The thrust acts on the segment; 5. The friction force of the shield shell acts on the soil unit outside the shield machine in the form of uniformly distributed shear stress; 6. The slurry in the gap at the tail of the shield is not hardened, and the grouting pressure is used for simulation, with positive stress. The force acts in the form of a ring on the soil units around the tunnel, and the slurry in the lining segments behind the gap slowly solidifies and hardens. The linear elastic solid unit is used to simplify the simulation of the solidification effect of the slurry; 7. The lining segments are simulated by linear elastic lining units. Finite element analysis and modeling are carried out under the above simplified conditions. According to the generated three-dimensional model of stratum distribution and multivariate correlated three-dimensional conditional random field, the stratum properties and physical and mechanical parameters of the soil units in the digital twin model are automatically assigned, and a random finite element digital twin model of the interaction between shield tunnel excavation and stratum is established.

5. The method for constructing digital twin strata for shield tunneling by integrating multi-source data according to claim 1, characterized in that: In step five, the dual lossless Kalman filter information fusion and the dynamic update of the formation parameters are carried out. Because the sampling frequency of the shield machine's excavation mechanical parameters is much greater than the sampling frequency of the surface deformation, the excavation mechanical parameters in each excavation ring are equivalently processed, and the excavation parameter fluctuations of the soil chamber pressure, cutter head torque, jacking force and grouting pressure are regarded as multidimensional normally distributed random variables, and the mean and variance of the multidimensional normally distributed random variables are calculated based on the recorded data.

6. A shield tunneling digital twin stratum construction method integrating multi-source data according to claim 1, characterized in that: In step six, the information of stratum deformation and excavation mechanical parameters is integrated to design two coupled nonlinear Kalman filter models, and the parameters of the stratum distribution three-dimensional model and the multivariate correlation three-dimensional conditional random field model are inverted and analyzed respectively; according to the obtained stratum distribution model and the multivariate correlation three-dimensional conditional random field of geotechnical parameters, the initial conditions of the Kalman filter are set, including the state to be estimated, the horizontal state transfer matrix, the geotechnical parameter autocorrelation distance and the initial error covariance matrix; According to the surface deformation prediction error, the observation noise covariance matrix of the Kalman filter is set; in the first Kalman filter model, the multivariate correlation conditional random field parameter θ of the previous iteration step is used. k , integrating the formation deformation data k , calculate the Kalman gain, and the formation distribution model parameter δ k Update and correct; then use the updated formation distribution model parameter δ k Input into the second Kalman filter model and incorporate the formation deformation data s k , calculate the corresponding Kalman gain matrix and estimate the updated multivariate correlation a conditional random field parameters θ k , the subscript k in the above parameters represents the kth iteration; For the stratum distribution model, the detailed process of obtaining the parameter values ​​of the updated stratum distribution model is as follows: δ k-1 is the parameter value of the stratum distribution model before updating, M k-1 is the corresponding covariance matrix, η is the empirical hyperparameter, m is a constant, 2m+1 is the total number of ε points, represents the ε point; Further estimates are obtained using the estimation function t(x): In the formula, α i is the weight coefficient, N k-1 is the noise, δ k|k-1 is an intermediate estimate; By δ k|k-1 According to the conversion formula, we can get the corresponding ε k|k-1 , and further estimate the value by using the estimation function l(x): e k|k-1 For further estimation; Then two covariance matrices are calculated: In the formula, B k-1 for noise; From this, the Kalman filter gain matrix is ​​calculated: Thus, the parameter values ​​of the updated stratum distribution model are obtained: d k =d k|k-1 +F k (z k -e k|k-1 ) Where z k is the measurement vector; Similarly, the Kalman filter update model of rock and soil random field parameters can be obtained: In the above formula, * is used to distinguish the parameters of the stratum distribution model. After obtaining new deformation monitoring data, the stratum distribution and multivariate correlation conditional random field parameters are dynamically updated in accordance with the above steps.

7. A shield tunnel excavation digital twin stratum construction system based on multi-source data fusion, comprising: Module 1: It is used to collect geotechnical survey data of shield tunnel locations, and summarize and statistically analyze the drilling data, in-situ test data and indoor geotechnical test data; Module 2: Use the stratigraphic attribute information of the drilling data to establish a multi-layer perceptron model that reflects the spatial position characteristics of the bedrock interface at the shield tunnel location, and establish a three-dimensional reconstruction model of the stratigraphic distribution based on the shield tunnel location; Module 3: It is used to use indoor geotechnical test data and in-situ test data, consider the mutual correlation of multidimensional geotechnical parameters and the spatial autocorrelation of single geotechnical parameters, establish the multivariate correlation three-dimensional conditional random field of geotechnical physical and mechanical parameters based on sparse Bayesian learning technology, and construct a refined model of multivariate geotechnical parameter spatial distribution; Module 4: Used to establish a finite element digital twin model of the interaction between shield tunneling and strata. Based on the strata distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters constructed in steps 2 and 3, the digital twin The stratum properties of the soil units and the physical and mechanical parameters of the land units in the model are automatically assigned, and then a non-intrusive stochastic finite element digital twin model of earth pressure balance shield construction integrating stratum distribution and geotechnical parameters is established to conduct mechanical analysis of stratum deformation and response during shield construction. Module 5: Used to couple the dual lossless Kalman filter technology, integrate the measured stratum deformation and the real-time mechanical parameter data of shield tunneling, update the stratum distribution multi-layer perceptron model and the three-dimensional conditional random field model of geotechnical parameters, and obtain the optimal inversion estimation of stratum and geotechnical parameters; update the stratum distribution and geotechnical parameters of the next excavation cycle to predict the stratum deformation of the next excavation cycle; Module 6: Used to verify the inverted formation and geotechnical parameter model and the deformation prediction results of the finite element digital twin model using real-time data from shield tunnel position monitoring.

8. A shield tunneling digital twin stratum construction system based on multi-source data fusion according to claim 7, characterized in that: In module 2, the three-dimensional reconstruction model of the stratum distribution based on the position of the shield tunnel is established to provide the initial stratum distribution parameters, namely δ0, for the subsequent coupled dual lossless Kalman filter model.

9. A shield tunnel excavation digital twin stratum construction system based on multi-source data fusion according to claim 7, characterized in that: In module 3, a three-dimensional random field model is established, assuming that x = [x, y, z] T represents a point in the three-dimensional space of the site, t=[t1,t2,…,t K ] T is the indoor geotechnical test result of the sample taken at this point, where t K It represents the test value of the Kth soil sample property at this point. Here, a total of K soil sample properties are considered. Assuming that there are N groups of experimental data in the stratum, the random variable matrix likelihood function is constructed for all data. In the above formula, B represents the matrix composed of all data; W is the random variable matrix, which is characterized by chaotic polynomial expansion; R is the column covariance matrix, that is, the autocorrelation coefficient matrix, which represents the spatial correlation of parameters; C is the row covariance matrix, i.e., the cross-correlation coefficient matrix, which represents the cross-correlation of parameters; Based on the sparse Bayesian learning technology, conjugate prior distribution is defined for the chaotic polynomial expansion coefficient matrix and autocorrelation distance parameter, and the chaotic polynomial expansion is used to characterize the mean of the random variable matrix. In order to improve the robustness of the model, the sparse Bayesian learning technology is used to solve the coefficient matrix of the chaotic polynomial, and the coefficient matrix is ​​vectorized and defined as a zero-mean multidimensional normal distribution. Based on the above sparse Bayesian learning technology, a multivariate correlated three-dimensional conditional random field of geotechnical parameters at any point in space is generated, thereby providing the initial parameters of the geotechnical parameter random field, i.e. θ0, for the subsequent construction of the coupled double lossless Kalman filter model.

10. A shield tunnel excavation digital twin stratum construction system based on multi-source data fusion according to claim 7, characterized in that: The simplified scheme of the interaction between the mechanical parameters of shield tunneling and the strata in the model in module 4 is as follows:

1. The soil compartment pressure is simplified to a linear load acting on the excavation surface; 2. The shield machine is simulated by a rigid unit, and its weight acts on the soil unit at the bottom in the form of a uniformly distributed load; 3. The cutter head torque acts on the soil unit at the excavation surface in the form of shear stress; 4. The thrust acts on the segments; 5. The friction force of the shield shell acts on the soil unit outside the shield machine in the form of uniformly distributed shear stress; 6. The slurry in the gap at the tail of the shield is not hardened, and the grouting pressure is used for simulation, with positive stress. The force acts in the form of a ring on the soil units around the tunnel, and the slurry in the lining segments behind the gap slowly solidifies and hardens. The linear elastic solid unit is used to simplify the simulation of the solidification effect of the slurry; 7. The lining segments are simulated by linear elastic lining units. Finite element analysis and modeling are carried out under the above simplified conditions. According to the generated three-dimensional model of stratum distribution and multivariate correlated three-dimensional conditional random field, the stratum properties and physical and mechanical parameters of the soil units in the digital twin model are automatically assigned, and a random finite element digital twin model of the interaction between shield tunnel excavation and stratum is established.

11. The shield tunneling digital twin stratum construction system based on multi-source data fusion according to claim 7, characterized in that: The dual lossless Kalman filter information fusion and dynamic update of formation parameters in module five are carried out. Because the sampling frequency of shield machine excavation mechanical parameters is much greater than the sampling frequency of surface deformation, the excavation mechanical parameters in each excavation ring are treated equivalently, and the excavation parameter fluctuations of soil cabin pressure, cutter head torque, jacking force and grouting pressure are regarded as multidimensional normally distributed random variables, and the mean and variance of the multidimensional normally distributed random variables are calculated based on the recorded data.

12. A shield tunnel excavation digital twin stratum construction system based on multi-source data fusion according to claim 7, characterized in that: In module six, the information of stratum deformation and excavation mechanics parameters is integrated to design two coupled nonlinear Kalman filter models, and the parameters of the stratum distribution three-dimensional model and the multivariate correlation three-dimensional conditional random field model are inverted and analyzed respectively; according to the obtained stratum distribution model and the multivariate correlation three-dimensional conditional random field of geotechnical parameters, the initial conditions of the Kalman filter are set, including the state to be estimated, the horizontal state transfer matrix, the geotechnical parameter autocorrelation distance and the initial error covariance matrix; According to the surface deformation prediction error, the observation noise covariance matrix of Kalman filter is set; In the first Kalman filter model, the formation deformation data s is incorporated k , calculate the Kalman gain, and the formation distribution model parameter δ k Update and correct; then use the updated formation distribution model parameter δ k Input into the second Kalman filter model and incorporate the formation deformation data s k , calculate the corresponding Kalman gain matrix and estimate the updated multivariate correlation conditional random field parameters θ k , the subscript k in the above parameters represents the kth iteration For the stratum distribution model, the detailed process of obtaining the parameter values ​​of the updated stratum distribution model is as follows: δ k-1 is the parameter value of the stratum distribution model before updating, M k-1 is the corresponding covariance matrix, η is the empirical hyperparameter, 2m+1 is the total number of ε points, represents the ε point; Further estimates are obtained using the estimation function t(x): In the formula, α i is the weight coefficient, N k-1 is the noise, δ k|k-1 is an intermediate estimate; By δ k|k-1 According to the conversion formula, we can get the corresponding ε k|k-1 , and further estimate the value by using the estimation function l(x): e k|k-1 For further estimation; Then two covariance matrices are calculated: In the formula, B k-1 for noise; From this, the Kalman filter gain matrix is ​​calculated: Thus, the parameter values ​​of the updated stratum distribution model are obtained: d k =d k|k-1 +F k (z k -e k|k-1 ) Where z k is the measurement vector; Similarly, the Kalman filter update model of rock and soil random field parameters can be obtained: In the above formula, * is used to distinguish the parameters of the stratum distribution model. After obtaining new deformation monitoring data, the stratum distribution and multivariate correlation conditional random field parameters are dynamically updated in accordance with the above steps.

Citation Information

Patent Citations

  • Shield tunnel construction early warning method integrating urban building information and stratum structure

    CN113362469A

  • Shield tunnel excavation process and earth surface deformation three-dimensional finite difference numerical simulation method

    CN114969884A

  • CAE (Computer Aided Engineering) simulation inversion method for multi-stratum parameters of shield tunnel and server

    CN115618697A

  • Shield tunnel geotechnical parameter inversion and tunneling parameter optimization method based on approximate model

    CN115659758A

  • Shield tunneling digital twin stratum construction method and system fusing multi-source data

    CN116227309A

Cited By

  • Design method and structure of U-shaped shell reinforced shield tunnel structure

    CN119167505A

  • Dredging waste slag resource utilization potential assessment method based on layering characteristics

    CN119250657A

  • Pose adjustment boundary solving method in shield tunneling process

    CN119337047A

  • Efficient and refined prediction method for indoor vibration of proposed building close to subway

    CN119378083A

  • Multi-source data fused large steel structure digital twin model construction method and system

    CN119378085A