Shield tunnel full life structure deformation monitoring and early warning method based on digital twinning

Abstract

The application discloses a shield tunnel full-life structure deformation monitoring and early warning method based on digital twinning, which comprises the following steps: S1, constructing a tunnel-geology complex model integrating BIM and three-dimensional point cloud technology according to tunnel completion section point cloud data; S2, monitoring the deformation of the tunnel in the full life cycle in real time by using FBG and incremental sensors; S3, constructing an LGBM proxy model to update the tunnel structure deformation prediction in the construction period; S4, constructing a GAT-LSTM model to predict the tunnel structure deformation in the service period; S5, constructing a tunnel structure deformation grading early warning method based on strength reduction and cusp mutation theory; and S6, constructing a tunnel monitoring and early warning system based on digital twinning. The application realizes the reconstruction of the tunnel-geology complex structure by integrating stratum information and high-precision spatial data based on the digital twinning technology, proposes a real-time deformation monitoring and early warning method and a twinning system for the tunnel construction and service period, and guarantees the service safety and failure risk early warning of the tunnel structure in the full life cycle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel engineering, and in particular to a shield tunnel full-life structure deformation monitoring and early warning method based on digital twinning. BACKGROUND

[0002] With the continuous development of transportation construction, the number and capital investment of shield tunnel construction have increased significantly. As a key node connecting the modern transportation network, shield tunnels bear important tasks such as vehicle traffic and personnel safety, and their safety problems in the full life cycle need to be paid attention to. However, the closed and special environmental characteristics of the tunnel bring great challenges to real-time deformation monitoring. Therefore, it is necessary to monitor and risk warn the deformation condition of shield tunnels in the construction and service life cycle in real time.

[0003] At present, there is a lack of effective real-time monitoring and risk warning scheme for the deformation condition of shield tunnels, and traditional manual monitoring and experience warning methods cannot achieve real-time monitoring and are prone to misjudgment.

[0004] Chinese invention patent: announcement number "CN116128290A", name "urban built-up area shield method construction environment risk intelligent evaluation digital twinning system", discloses a kind of urban built-up area shield method construction environment risk intelligent evaluation digital twinning system, including: environmental information perception module, construction information perception module, construction influence information perception module, digital twinning simulation deduction module, risk loss evaluation module and risk assessment and countermeasure module.This application can improve the scientificity of environmental risk assessment, realize the visualization, quantification of risk assessment, and make real-time decision for different risk levels, reduce the safety risk problems caused by damage to existing buildings in the construction process.The technical scheme is to evaluate the risk level of the influence of shield tunnel pre-construction on the surrounding environment, and does not consider the evaluation of the deformation condition of shield tunnel full life cycle after construction. SUMMARY

[0005] In order to solve the above-mentioned problems in the prior art that there is a lack of effective real-time monitoring and risk warning scheme for the deformation condition of shield tunnels, and traditional manual monitoring and experience warning methods cannot achieve real-time monitoring and are prone to misjudgment, the present application proposes a shield tunnel full-life structure deformation monitoring and early warning method based on digital twinning, which builds a tunnel-geology complex structure model, collects and transmits tunnel full life cycle deformation data, construction state data and the like in real time, and studies a shield tunnel full life cycle full space deformation prediction model and hierarchical early warning mechanism.

[0006] The present application is realized by the following technical scheme:

[0007] S1. Based on the point cloud data of the completed tunnel section, construct a tunnel-geological complex model that integrates BIM and 3D point cloud technologies;

[0008] S2. Employ FBG and incremental sensing to monitor tunnel deformation in real time throughout its entire life cycle.

[0009] S3. Construct an LGBM proxy model and update the tunnel structure deformation prediction during the construction period;

[0010] S4. Construct a GAT-LSTM model to predict the structural deformation of the tunnel during its service life;

[0011] S5. Construct a graded early warning method for tunnel structure deformation based on strength reduction and cusp catastrophe theory;

[0012] S6. Construct a tunnel monitoring and early warning system based on digital twins.

[0013] As a further preferred embodiment, the specific steps of step S1 are as follows:

[0014] S11. Obtain point cloud data of the completed section of the tunnel using the vehicle-mounted laser detection system;

[0015] S111. Install a lidar scanner above the inspection vehicle. Based on the lidar scanner, establish a local coordinate system with the center of the lidar scanner as the origin, and obtain the three-dimensional point cloud data P of the tunnel under this local coordinate system. local ;

[0016] S112. Install GPS on the inspection vehicle and obtain the vehicle's position coordinates T in the global coordinate system based on the GPS on the inspection vehicle;

[0017] S113. Install an inertial navigation system on the inspection vehicle to keep the data synchronized with the LiDAR scanner.

[0018] S114. Install an inclinometer inside the inspection vehicle to obtain accurate three-axis attitude angle data and construct a rotation matrix R;

[0019] S115. Based on the three-dimensional point cloud data P of the tunnel in the local coordinate system obtained in step S111. local The position coordinates T of the vehicle body in the global coordinate system obtained in step S112 and the rotation matrix R obtained in step S114 are used to obtain accurate three-dimensional point cloud coordinate information, that is, the three-dimensional point cloud data P of the tunnel in the global coordinate system. global The formula is as follows:

[0020] P global =R·P local +T

[0021] S12. Based on the point cloud data of the completed section of the tunnel obtained in step S11, reconstruct the shield tunnel structure.

[0022] S121, Point cloud data preprocessing;

[0023] S122. Construct a triangulation network for flat tunnel point cloud data based on the Delaunay triangulation algorithm, and reconstruct the main outline of the tunnel by combining the three-dimensional point cloud data of the tunnel.

[0024] S123. Project the point cloud data onto the yox and zox planes respectively. Construct points on the central axis by fitting the boundary functions obtained from the boundary points on both sides of the projected point cloud data. Use a cubic polynomial as the fitting function to obtain the fitting equations of the tunnel axis in each plane.

[0025] S124. Expand the point cloud data along the tunnel axis and determine the pixel value of the point based on its geometric features to create a binary image. Use the Laplacian operator to detect the boundary of the binary image. Based on morphological processing of the boundary image, divide it into individual ring linings along the ring joint. Use connected component analysis to set the same pixel value region to the same value and divide it into different segments. Obtain the center point and radius parameters by fitting the segment point cloud data. Use the CATIA platform to develop the as-built model of the lining ring. Import the fitting parameters of the center point and radius of different segments, the tunnel axis, and material properties as input parameters into the platform, establish geometric dimension constraint relationships, create a parameterized lining ring model, and thus reconstruct the shield tunnel structure.

[0026] S13. Adaptively align the shield tunnel structure reconstructed in step S12 with the geological environment model to build a tunnel-geological complex model.

[0027] As a further preferred embodiment, the specific steps of step S2 are as follows:

[0028] S21. Fiber optic grating multi-point displacement gauges are installed longitudinally on each deformation monitoring section of the shield tunnel to obtain the longitudinal deformation U during the tunnel construction period. l_c and longitudinal deformation during service life U l_s And transmit it to the receiving instrument;

[0029] S22. On each deformation monitoring section of the shield tunnel, pulleys, several data acquisition modules with incremental encoders, and weights are sequentially installed along the tunnel circumference. The pulleys, several data acquisition modules with incremental encoders, and weights are connected by flexible wires to form a non-closed polygon convergence monitoring system.

[0030] S221. Data acquisition modules with incremental encoders are installed at the arch crown, both sides of the arch waist, and the middle position of the tunnel deformation monitoring section. The locations of the data acquisition modules with incremental encoders are named P1, P2, ..., P1 from left to right.n Point P1 is fixed using pulleys. n A weight is connected at the point to maintain the tension of the conductor, and the pulley acts as a ball-and-socket joint;

[0031] S222. Using a total station, obtain the precise initial coordinates of point P1, record the initial calibration values ​​of the angles and traverse lengths at each point, and use trigonometric functions to sequentially obtain the coordinates of points P2, P3, ..., P... n The initial coordinates of the point;

[0032] S223. Inside each group of data acquisition modules with incremental encoders, a hollow shaft magnetic sensor A is installed on the left side, and a hollow shaft magnetic sensor B is installed on the right side. Hollow shaft magnetic sensor A monitors the angle increment of the left-side traverse at each point between time t and t-1, and hollow shaft magnetic sensor B monitors the angle increment of the right-side traverse at each point between time t and t-1. Combining this with the initial calibration values ​​obtained in step S222, P1, P2, ..., P... are obtained. n Angle value at point t and The formula is as follows:

[0033]

[0034]

[0035]

[0036] In the formula: i is the serial number of the data acquisition module deployment point, ranging from 1 to n; P at time t i The included angle between the two guide wires at the point; P at time t i The angle between the right-side guide wire and the vertical direction at the point; P measured by hollow shaft magnetic sensor A i The angle increment of the traverse to the left of the point between time t and time t-1; P measured by hollow shaft magnetic sensor B i The angle increment of the traverse to the right of the point between time t and time t-1;

[0037] S224. A wire support pulley and a solid shaft sensor C are installed at the connection point of two flexible wires inside each data acquisition module with an incremental encoder. The solid shaft sensor C is used to obtain the angle increment Δδ of the wire support pulley, combined with P... i-1 Point and P i Initial length of the conductor between points Obtain the length of each conductor segment at time t. The formula is as follows:

[0038]

[0039] In the formula: P at time t i Point and P i-1 The length of the conductor between the points; Let P be the initial time. i Point and P i-1 The length of the conductor between the points; r is the radius of the pulley supporting the conductor; and P at time t i-1 Point and P i The angle increment of the guide rail support pulley at the point;

[0040] S225. Based on the angle value at time t obtained in step S223 The wire length value at time t obtained in step S224 Using trigonometric functions in sequence, we obtain P2, P3, ..., P... n The location coordinates of each point at time t are determined, and finally, the tunnel convergence U at different times during the construction period is obtained based on the changes in the coordinate information of each point. c_c U-shaped arch sinking s_c And tunnel convergence U at different times during the service life c_s U-shaped arch sinking s_s .

[0041] As a further preferred embodiment, the specific steps of step S3 are as follows:

[0042] S31. Several pressure monitoring sections are evenly set along the longitudinal direction of the tunnel boring machine (TBM). Several pressure monitoring points are arranged on each pressure monitoring section. Miniature load cells are installed outside the TBM's protective cover at each pressure monitoring point. The miniature load cells are connected to a data logger using shielded data cables to measure the surrounding rock contact pressure F acting on each pressure monitoring point of the TBM at time t. mn Wireless accelerometers and strain gauge torque sensors are installed on different cutter mounting radii on the back of the cutterhead. The data is transmitted to the server through a wireless gateway near the cutterhead to obtain the average cutterhead rotation speed N and the average cutterhead torque T. Linear displacement sensors are installed on the shield machine shell to obtain the shield advance speed V, thereby obtaining the distance x between the different deformation monitoring surfaces and the tunnel face. Pressure sensors and flow meters are installed in the shield machine grouting system to obtain the grouting pressure P and grouting volume Q.

[0043] S32. The surrounding rock contact pressure F obtained in step S31 mnThe average cutterhead rotation speed N, average cutterhead torque T, shield advance speed V, tunnel depth H, distance x between deformation monitoring face and excavation face, internal friction angle φ, and cohesion c are used as inputs to the LGBM model. The ground parameters need to be weighted considering the location and thickness of the overlying strata, with the longitudinal deformation U of the tunnel during construction being the primary factor. l_c Peripheral convergence U c_c and the U-shaped arch sinking s_c As the output of the LGBM model, an LGBM proxy model is established;

[0044] S33. Preprocess the input and output data of the LGBM proxy model and divide it into a test set and a training set, as shown in the following formula:

[0045]

[0046] In the formula: x i The original data; max x i and min x i These are the minimum and maximum values ​​among the data points, respectively; by default, x p and x q They are 1 and -1 respectively;

[0047] S34. Based on the Bayesian optimization method based on Gaussian processes, the longitudinal deformation U corresponding to the tunnel construction period is obtained respectively. l_c Peripheral convergence U c_c With the U-shaped arch sinking s_c The hyperparameters for optimal prediction performance, and the formulas for the optimization process, are as follows:

[0048]

[0049] R 1:t ={(x1,y1),(x2,y2)…(x t ,y t )}

[0050] y t =f(x) t )+ε t

[0051] f(x t )~GP(m(x),k(x,x′))

[0052] In the formula: f is the objective function; y t x is the observation value at step t; t Let ε be the hyperparameter of the t-th step; t R represents the observation error. 1:t Aggregate the observations from the first t steps; p(R) 1:t |f) is R under a given objective function1:t The likelihood distribution of f; p(f) is the prior distribution of f; p(f|R) 1:t To observe R 1:t The posterior distribution of f; m(x) is the mean function; k(x,x') is the covariance function;

[0053] S35. Using the hyperparameters of the best prediction effect obtained in step S34 and the test set and training set divided in step S33, the LGBM surrogate model established in step S32 is trained and evaluated. Using the trained model, the deformation of the tunnel at the unknown section of the tunnel is predicted based on the burial depth, stratum parameters and distance from the tunnel face at the unknown section of the tunnel. Dynamic monitoring data during the tunneling process is collected to incrementally train the model and update the tunnel deformation state prediction.

[0054] As a further preferred embodiment, the specific steps of step S4 are as follows:

[0055] S41. The service-period longitudinal deformation U of the s front and s rear sections of the target cross-section collected and transmitted in step S2 within a continuous time period of l is calculated. l_s Peripheral convergence U c_s and the U-shaped arch sinking s_s The spatiotemporal data is preprocessed and reconstructed into window structures as model input data, represented as [2s+1, l]. The preprocessing formula is as follows:

[0056]

[0057] In the formula: x i The original data; max x i and min x i These are the minimum and maximum values ​​among the data points, respectively; by default, x p and x q They are 1 and -1 respectively;

[0058] S42. Represent the tunnel segment centered on the deformation monitoring section as different nodes and establish edges between the nodes of adjacent tunnel segments to construct graphical structural data. Use the spatial data at the same time as a graph and construct the input data into a graph sequence.

[0059] S43. Input the graphics at different times into the corresponding GAT units in the GAT model layer, extract the deformation space features at each time step, arrange and stitch them in chronological order.

[0060] S44. Extract features from the data output by the GAT layer using a convolutional layer;

[0061] S45. Use the LSTM model layer to extract the temporal features of the deformation data and obtain the predicted value.

[0062] As a further preferred embodiment, the specific steps of step S5 are as follows:

[0063] S51. Establish a three-dimensional geological model in Abaqus software. Define the location and shape of the tunnel and shield shell based on the tunnel depth, excavation radius, and three-dimensional spatial coordinate information. Simultaneously, assign material properties to the tunnel, shield shell, and strata based on material properties and stratum parameters. During the construction period, dynamically adjust and simulate the construction process in real time based on the actual construction status of the tunnel boring machine. For specific tunnel sections at different stages, use different strength reduction coefficients K = {K1, K2, K3, K4…} at equal intervals for simulation analysis, and obtain the corresponding increase in plastic deformation energy ΔU of the surrounding rock. P And the mutation eigenvalue Δ, plot and fit ΔU P The relationship curve with K; ΔU P The state in which a sudden change occurs and Δ changes from a positive value to a negative value is determined as the limit state, and the corresponding longitudinal deformation U of the tunnel is... l_c / U l_s Peripheral convergence U c_c / U l_s and the U-shaped arch sinking s_c / U l_s That is, the deformation limit value U of the cross section under the corresponding state. o ={U l_o U c_o U s_o Based on the average trisection points of the deformation limit values, four warning levels—red, orange, yellow, and green—are defined, with each warning level having a range of [U... o (+∞) and The corresponding increase in plastic deformation energy of the surrounding rock, ΔU, is obtained. P The formula for the mutation eigenvalue Δ is as follows:

[0064]

[0065] Where: m is the number of reduction cycles; n is the number of plastic units; σ ij The stress at the j-th strength reduction of plastic element i; Δε ij The strain increment is the strain increment at the j-th strength reduction of plastic element i; a1, a2, a3, and a4 are derived from ΔU P The undetermined coefficients of the fourth-order polynomial obtained by fitting the reduction factor K using the least squares method;

[0066] S52. Calculate the tunnel depth H, the internal friction angle φ, the cohesion c, and the surrounding rock contact pressure F at different construction stages. mnAverage cutterhead rotation speed N, average cutterhead torque T, shield tunneling speed V, distance x between deformation monitoring face and tunneling face, and cross-sectional deformation limit value U. l_o U c_o and U s_o The data is constructed into a dataset, a BP model is built and trained, the trained BP model is transferred to a new dataset at other sections of the tunnel for training, the first few layers of the BP model are frozen, and the parameters of the last few layers of the BP model are trained or new model layers are added.

[0067] S53. Obtain the graded early warning range of each monitoring section during the construction period from steps S51 and S52, and combine the tunnel section deformation monitoring value or prediction value to carry out risk early warning during the tunnel construction period.

[0068] S54. Delete the construction period parameters in steps S51 and S52, and re-execute steps S51 and S52 to obtain the graded early warning range of each monitoring section during the service period. Combine the tunnel section deformation monitoring values ​​or predicted values ​​to conduct real-time risk early warning for the tunnel service period. The construction period parameters include: surrounding rock contact pressure F. mn The average cutterhead rotation speed N, average cutterhead torque T, shield tunneling speed V, and the distance x between the deformation monitoring face and the tunneling face;

[0069] S55. Conduct real-time risk warning for the entire life cycle of the tunnel based on steps S53 and S54.

[0070] As a further preferred embodiment, the specific steps of step S6 are as follows:

[0071] S61. Collect three-dimensional point cloud data of the tunnel, structural deformation information throughout the entire life cycle, and shield construction status through sensors. Standardize the acquired data and encode it into JSON format. Transmit it to the server via a communication transmission module. The communication transmission module consists of an interface transmission base station, power supply equipment, GPRS module, Zigbee module, and ARM9 processor arranged in the tunnel to provide Zigbee, CAN bus, and TCP / IP protocols.

[0072] S62. The shield tunnel digital twin management platform includes a shield tunnel design and modeling module, a structural deformation prediction and early warning module, and a user interface and interaction module. Based on the shield tunnel digital twin management platform, it accesses the database in the server and processes various data to realize the construction of a virtual model of the tunnel-geological complex. Based on the structural deformation prediction model and graded early warning method during the tunnel construction and service periods, it realizes full-life-cycle full-space deformation prediction and risk early warning of the shield tunnel. The tunnel digital twin management platform provides a three-dimensional visualization interface, using the BIM model as the data carrier, to transmit, update, and synchronously render the tunnel deformation status, shield construction status, and early warning status assessment in real time.

[0073] As a further preferred option, the number of data acquisition modules with incremental encoders in step S22 is 5.

[0074] As a further preferred option: the number of pressure monitoring sections in step S31 is 5, and the number of pressure monitoring points arranged on each pressure monitoring section is 5.

[0075] As a further preferred option, the ratio of dividing the test set and the training set in step S33 is 1:4.

[0076] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0077] 1. Based on the shield tunnel-geological complex model that integrates stratigraphic information and high-precision spatial data, this invention proposes a method for monitoring and early warning of structural deformation throughout the entire life cycle of shield tunnels based on digital twins.

[0078] 2. Based on the constructed tunnel-geological complex model, this invention dynamically adjusts the model in real time according to the monitoring values ​​to form a digital twin model system, studies the real-time early warning mechanism and method of tunnel deformation, and realizes tunnel deformation monitoring and graded early warning.

[0079] 3. This invention acquires real-time structural deformation data through a sensor network composed of FBG and incremental encoder, and dynamically adjusts the prediction model. It considers the tunnel instability state from an energy perspective, and determines the early warning threshold for tunnel monitoring section deformation during the construction and service periods based on the strength reduction and cusp catastrophe theory, thereby realizing the full life cycle deformation monitoring, prediction and early warning of shield tunnels. Attached Figure Description

[0080] Figure 1 This is an overall flowchart of the early warning method of the present invention.

[0081] Figure 2 This is a schematic diagram of the vehicle-mounted laser detection system in this invention.

[0082] Figure 3 This is a diagram showing the arrangement of the longitudinal deformation monitoring sensors for tunnels in this invention.

[0083] Figure 4 This is a diagram showing the arrangement of the tunnel circumferential deformation monitoring sensors in this invention.

[0084] Figure 5 This is a schematic diagram of the installation of the circumferential deformation monitoring sensor in this invention.

[0085] Figure 6 This is a schematic diagram of the overall framework of the early warning system of the present invention.

[0086] The image shows:

[0087] 1. Inspection vehicle; 2. LiDAR scanner; 3. Inertial navigation system; 4. Inclinometer; 5. Computer; 6. GPS; 7. Deformation monitoring section; 8. Fiber optic grating multi-point displacement meter; 9. Hollow shaft magnetic sensor A; 10. Hollow shaft magnetic sensor B; 11. Solid shaft sensor C; 12. Wire support pulley; 13. Antenna; 14. Flexible wire; 15. Pulley; 16. Weight. Detailed Implementation

[0088] The advantages and features of the present invention will be illustrated and explained by the following non-limiting description of preferred embodiments, which are given by way of example only with reference to the accompanying drawings.

[0089] like Figures 1-6 As shown, this invention proposes a method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins. This method uses digital twins as its core, integrating geological information and high-precision spatial data to reconstruct the tunnel-geological complex structure. It proposes a real-time deformation monitoring and early warning method and twin system for tunnel construction and service life, ensuring the safety of the tunnel structure throughout its entire life cycle and providing early warning of potential failure risks. Specifically, it includes the following steps:

[0090] S1. Based on the point cloud data of the completed tunnel section, construct a tunnel-geological complex model that integrates BIM and 3D point cloud technologies.

[0091] S11. Based on the vehicle-mounted laser detection system, obtain the point cloud data of the completed section of the tunnel.

[0092] S111. A lidar scanner 2 is installed above the inspection vehicle 1. Based on the lidar scanner 2 above the inspection vehicle 1, a local coordinate system is established with the center of the lidar scanner 2 as the origin, and the three-dimensional point cloud data P of the tunnel under this local coordinate system is obtained. local ;

[0093] S112. Install GPS6 on the inspection vehicle 1, and obtain the position coordinates T of the vehicle body in the global coordinate system based on GPS6 on the inspection vehicle 1.

[0094] S113. An inertial navigation system 3 is installed on the inspection vehicle 1 to keep the data synchronized with the LiDAR scanner 2.

[0095] S114. An inclinometer 4 is installed inside the inspection vehicle 1 to obtain accurate three-axis attitude angle data and construct a rotation matrix R;

[0096] S115. Based on the three-dimensional point cloud data P of the tunnel in the local coordinate system obtained in step S111. localThe position coordinates T of the vehicle body in the global coordinate system obtained in step S112 and the rotation matrix R obtained in step S114 are used to obtain accurate three-dimensional point cloud coordinate information, that is, the three-dimensional point cloud data P of the tunnel in the global coordinate system. global The formula is as follows:

[0097] P global =R·P local +T

[0098] like Figure 2 As shown, a vehicle-mounted laser inspection system is used to acquire point cloud data of the completed tunnel section. A laser radar scanner 2 is installed above the inspection vehicle 1. The inspection vehicle 1 moves forward along the tunnel axis in conjunction with the rotation of the laser radar scanner 2 to establish a local coordinate system with the center of the laser radar scanner 2 as the origin. The three-dimensional point cloud data P of the tunnel under this local coordinate system is then acquired. local A GPS 6 is installed below the LiDAR scanner 2 on the inspection vehicle 1 to obtain the vehicle's position coordinates T in the global coordinate system; an inertial navigation system 3 is deployed on the right side of the LiDAR scanner 2 on the inspection vehicle 1 to synchronize its data with the LiDAR scanner 2; an inclinometer 4 is deployed inside the inspection vehicle 1 to assist in correcting the attitude angles of the inertial navigation system, record accurate three-axis attitude angle data, and construct a rotation matrix R; using formula P global =R·P local +T transforms point cloud data in the local coordinate system to the global coordinate system, obtaining accurate 3D point cloud coordinate information.

[0099] The steps involved in constructing the tunnel-geological complex model are as follows:

[0100] S12. Based on the point cloud data of the completed section of the tunnel obtained in step S11, reconstruct the shield tunnel structure.

[0101] S121, Point cloud data preprocessing;

[0102] The open-source point cloud processing platform CloudCompare was used to manually remove obvious abnormal point cloud data. Then, the KNN algorithm was used to eliminate trajectory noise and linear noise. The point cloud data was smoothed using the conditional random field method, and the data compression was achieved by compressing octree structured data using voxel context.

[0103] S122. Construct a triangulation network for flat tunnel point cloud data based on the Delaunay triangulation algorithm, and reconstruct the main outline of the tunnel by combining the three-dimensional point cloud data of the tunnel.

[0104] This step involves extracting the main outline of the tunnel.

[0105] S123. Project the point cloud data onto the yox and zox planes respectively. Construct points on the central axis by fitting the boundary functions obtained from the boundary points on both sides of the projected point cloud data. Use a cubic polynomial as the fitting function to obtain the fitting equations of the tunnel axis in each plane.

[0106] This step involves fitting the tunnel axis, resulting in fitting equations for the tunnel axis in each plane, namely x, y(x), and z(x).

[0107] S124. Expand the point cloud data along the tunnel axis and determine the pixel values ​​of the points based on their geometric features to create a binary image. Use the Laplacian operator to detect the boundaries of the binary image. Based on morphological processing of the boundary image, divide it into individual annular linings along the annular joint. Use connected component analysis to set regions with the same pixel value to the same value and divide them into different segments. Obtain the center point and radius parameters by fitting the segment point cloud data. Develop the as-built model of the lining ring using the CATIA platform. Import the fitting parameters of the center points and radii of different segments, the tunnel axis, and material properties as input parameters into the platform, establish geometric dimensional constraints, create a parametric lining ring model, and thus reconstruct the shield tunnel structure.

[0108] This step involves the reconstruction of shield tunnel segments: First, a point cloud is unfolded along the tunnel axis. The unfolded point cloud is then fitted using the least squares method, and the plane normal vector is calculated. The height and width of the binary image are determined based on the point cloud extent, and then the angle between the normal vector of each point in the unfolded point cloud and the plane normal vector is calculated. when The pixel value is set to 0 when the pixel is active and 255 otherwise, a binary image of the point cloud is obtained. The Laplacian operator is used to detect the boundary of the binary image. Based on morphological processing, the boundary image is divided into individual ring-shaped linings along the ring joint. Finally, connected component analysis is used to set the same pixel value to the same value and divide them into different segments. By fitting the point cloud of the segments, geometric parameters such as the center point and radius are obtained. Finally, the as-built model of the lining ring is developed using the CATIA platform. The fitting parameters of the center point and radius of different segments, the tunnel axis, material properties, etc. are imported into the platform as input parameters to establish geometric dimensional constraints and create a parametric lining ring model, thereby reconstructing the shield tunnel structure.

[0109] S13. Adaptively align the shield tunnel structure reconstructed in step S12 with the geological environment model to build a tunnel-geological complex model.

[0110] Geotechnical engineering information of the tunnel project, such as geometric, topological and attribute information, is integrated and stored in the GIS model. GIS data is introduced by extending the IFC standard and the geological entity relationships and attributes are described in detail. The conditional random field method is used to model the spatial variation of geological parameters. The axis data and key nodes, such as the start point, end point and turning point, of the shield tunnel completed section structural model are registered and aligned with the corresponding positions in the geological environment model to realize the construction of the tunnel-geological complex model.

[0111] S2. FBG and incremental sensing are used to monitor the deformation of the tunnel in real time throughout its entire life cycle.

[0112] S21. Fiber optic grating multi-point displacement gauges 8 are installed longitudinally on each deformation monitoring section 7 of the shield tunnel to obtain the longitudinal deformation U of the tunnel during construction. l_c and longitudinal deformation during service life U l_c And transmit it to the receiving instrument;

[0113] S22. On each deformation monitoring section 7 of the shield tunnel, pulleys 15, several data acquisition modules with incremental encoders and weights 16 are sequentially set along the tunnel circumference. The pulleys 15, several data acquisition modules with incremental encoders and weights 16 are connected by flexible wires 14 to form a non-closed polygon convergence monitoring system.

[0114] S221. Data acquisition modules with incremental encoders are installed at the arch crown, both sides of the arch waist, and the middle position of the tunnel deformation monitoring section. The locations of the data acquisition modules with incremental encoders are named P1, P2, ..., P1 from left to right. n Point P1 is fixed using pulley 15. n A weight is connected at the point to maintain the tension of the conductor, and the pulley 15 acts as a ball-and-socket joint;

[0115] The purpose of this step is to deploy a non-closed polygon convergence monitoring system. The preferred number of data acquisition modules with incremental encoders is five.

[0116] S222. Using a total station, obtain the precise initial coordinates of point P1, record the initial calibration values ​​of the angles and traverse lengths at each point, and use trigonometric functions to sequentially obtain the coordinates of points P2, P3, ..., P... n The initial coordinates of the point;

[0117] S223. Inside each group of data acquisition modules with incremental encoders, a hollow shaft magnetic sensor A9 is installed on the left and a hollow shaft magnetic sensor B10 is installed on the right. Hollow shaft magnetic sensor A9 is used to monitor the angle increment of the left-side conductor at each point between time t and time t-1, and hollow shaft magnetic sensor B10 is used to monitor the angle increment of the right-side conductor at each point between time t and time t-1. Combined with the initial calibration values ​​obtained in step S222, P1, P2, ..., P... are obtained. n Angle value at point t and The formula is as follows:

[0118]

[0119]

[0120]

[0121] In the formula: i is the serial number of the data acquisition module deployment point, ranging from 1 to n; P at time t i The included angle between the two guide wires at the point; P at time t i The angle between the right-side guide wire and the vertical direction at the point; P measured by hollow shaft magnetic sensor A i The angle increment of the traverse to the left of the point between time t and time t-1; P measured by hollow shaft magnetic sensor B i The angle increment of the traverse to the right of the point between time t and time t-1.

[0122] S224. A wire support pulley 12 and a solid shaft sensor C11 are installed at the connection point of the two flexible wires 14 inside each data acquisition module with an incremental encoder. The angle increment Δδ of the wire support pulley 12 is obtained using the solid shaft sensor C11, combined with P... i-1 Point and P i Initial length of the conductor between points Obtain the length of each conductor segment at time t. The formula is as follows:

[0123]

[0124] In the formula: P at time t i Point and P i-1 The length of the conductor between the points; Let P be the initial time. i Point and P i-1 The length of the conductor between the points; r is the radius of the conductor support pulley 12; and P at time t i-1 Point and P i The angle increment of the guide rail support pulley 12 at the point.

[0125] S225. Based on the angle value at time t obtained in step S223 The wire length value at time t obtained in step S224 Using trigonometric functions in sequence, we obtain P2, P3, ..., P... n The location coordinates of each point at time t are determined, and finally, the tunnel convergence U at different times during the construction period is obtained based on the changes in the coordinate information of each point. c_c U-shaped arch sinking s_c And tunnel convergence U at different times during the service life c_s U-shaped arch sinking s_s .

[0126] Deformation monitoring sections were set up every 50m along the entire shield tunnel, as well as at tunnel entrances and exits, joints, and locations with complex geological conditions.

[0127] like Figure 3 As shown, fiber optic multi-point displacement gauges 8 are installed longitudinally at the left and right arch waists of each deformation monitoring section 7 of the shield tunnel. The average longitudinal deformation at the left and right arch waists is used as the longitudinal deformation value of the deformation monitoring section. The instruments are connected to the network, and the longitudinal deformation U during the tunnel construction period is transmitted to the network. l_c and longitudinal deformation during service life U l_s Transmitted over long distances to the receiving instrument;

[0128] like Figure 4 and Figure 5 As shown, a non-closed polygon convergence monitoring system, consisting of flexible conductors 14, pulleys 15, and several data acquisition modules with incremental encoders, is deployed along the circumference of the tunnel at each deformation monitoring section 7 of the shield tunnel. Preferably, there are five incremental encoder data acquisition modules, located at the arch crown, the two sides of the arch waist, and the middle position of each section 7. The five incremental encoder data acquisition module locations are named P1, P2, P3, P4, and P5 from left to right. At point P1, a pulley 15 acting as a ball joint is used for fixation. At point P5, a weight 16 is connected to maintain the tension of the flexible wire 14. Inside each group of data acquisition modules with incremental encoders, a hollow shaft magnetic sensor A9 is set on the left and a hollow shaft magnetic sensor B10 is set on the right. The antennae 13 of the hollow shaft magnetic sensor A9 and the hollow shaft magnetic sensor B10 are respectively attached to the left and right flexible wires 14. Inside each group of data acquisition modules with incremental encoders, a wire support pulley 12 and a solid shaft sensor C11 are set at the connection of the two flexible wires 14.

[0129] First, the precise initial position coordinates of point P1 are obtained using a total station. The initial calibration values ​​of the angles and traverse lengths at each point are recorded. Trigonometric functions are then used to calculate the initial position coordinates of points P2, P3, P4, and P5 sequentially. During deformation monitoring, hollow shaft magnetic sensors A9 and B10 are used to monitor the angle increments of the left and right traverse lines at each point between time t and time t-1, respectively. Based on the initial calibration values ​​and the angle increment values ​​at time t recorded by the system, the angle values ​​at time t at points P1, P2, P3, P4, and P5 are obtained. and Simultaneously, the angle increment Δδ of the wire-supported pulley 12 is obtained using a solid shaft sensor C11, combined with P i-1 Point and P i Initial length of the conductor between points Obtain the length of the conductor segment The angle value at time t obtained based on the above calculations and wire length value The position coordinates of points P2, P3, P4, and P5 at time t are obtained sequentially using trigonometric functions. Finally, based on the changes in the coordinate information of each point, the tunnel convergence U at different times during the construction period is obtained. c_c U-shaped arch sinking s_c And tunnel convergence U at different times during the service life c_s U-shaped arch sinking s_s .

[0130] S3. Construct an LGBM proxy model to update the tunnel structure deformation prediction during the construction period.

[0131] S31. Several pressure monitoring sections are evenly set along the longitudinal direction of the tunnel boring machine (TBM), and several pressure monitoring points are arranged on each pressure monitoring section. Preferably, there are 5 pressure monitoring sections, with 5 pressure monitoring points on each section. Miniature load cells are installed outside the TBM's protective cover at each pressure monitoring point. These load cells are connected to a data logger using shielded data cables to measure the surrounding rock contact pressure F acting on each pressure monitoring point of the TBM at time t. mn Wireless accelerometers and strain gauge torque sensors are installed on different cutter mounting radii on the back of the cutterhead. The data is transmitted to the server through a wireless gateway near the cutterhead to obtain the average cutterhead rotation speed N and the average cutterhead torque T. Linear displacement sensors are installed on the shield machine shell to obtain the shield advance speed V, thereby obtaining the distance x between the different deformation monitoring surfaces and the tunnel face. Pressure sensors and flow meters are installed in the shield machine grouting system to obtain the grouting pressure P and grouting volume Q.

[0132] S32. The surrounding rock contact pressure F obtained in step S31 mnThe average cutterhead rotation speed N, average cutterhead torque T, shield advance speed V, tunnel depth H, distance x between deformation monitoring face and excavation face, internal friction angle φ, and cohesion c are used as inputs to the LGBM model. The ground parameters need to be weighted considering the location and thickness of the overlying strata, with the longitudinal deformation U of the tunnel during construction being the primary factor. l_c Peripheral convergence U c_c and the U-shaped arch sinking s_c As the output of the LGBM model, an LGBM proxy model is established.

[0133] S33. Preprocess the input and output data of the LGBM proxy model and divide it into a test set and a training set, as shown in the following formula:

[0134]

[0135] In the formula: x i The original data; max x i and min x i These are the minimum and maximum values ​​among the data points, respectively; by default, x p and x q The values ​​are 1 and -1, respectively. The preferred ratio for splitting the test set and the training set is 1:4.

[0136] S34. Based on the Bayesian optimization method based on Gaussian processes, the longitudinal deformation U corresponding to the tunnel construction period is obtained respectively. l_c Peripheral convergence U c_c With the U-shaped arch sinking s_c The hyperparameters for optimal prediction performance, and the formulas for the optimization process, are as follows:

[0137]

[0138] R 1:t ={(x1,y1),(x2,y2)…(x t ,y t )}

[0139] y t =f(x) t )+ε t

[0140] f(x t )~GP(m(x),k(x,x′))

[0141] In the formula: f is the objective function; y t x is the observation value at step t; t Let ε be the hyperparameter of the t-th step; t R represents the observation error. 1:t Aggregate the observations from the first t steps; p(R) 1:t|f) represents R under a given objective function. 1:t The likelihood distribution of f; p(f) is the prior distribution of f; p(f|R) 1:t To observe R 1:t The posterior distribution of f; m(x) is the mean function; k(x,x') is the covariance function.

[0142] S35. Using the hyperparameters of the best prediction effect obtained in step S34 and the test set and training set divided in step S33, the LGBM surrogate model established in step S32 is trained and evaluated. Using the trained model, the deformation of the tunnel at the unknown section of the tunnel is predicted based on the burial depth, stratum parameters and distance from the tunnel face at the unknown section of the tunnel. Dynamic monitoring data during the tunneling process is collected to incrementally train the model and update the tunnel deformation state prediction.

[0143] S4. Construct a GAT-LSTM model to predict structural deformation during the tunnel's service life.

[0144] S41. The service-period longitudinal deformation U of the s front and s rear sections of the target cross-section collected and transmitted in step S2 within a continuous time period of l is calculated. l_s Peripheral convergence U c_s and the U-shaped arch sinking s_s The spatiotemporal data is preprocessed and reconstructed into window structures as model input data, represented as [2s+1, l]. The preprocessing formula is as follows:

[0145]

[0146] In the formula: x i The original data; max x i and min x i These are the minimum and maximum values ​​among the data points, respectively; by default, x p and x q They are 1 and -1 respectively.

[0147] S42. Represent the tunnel segment centered on the deformation monitoring section as different nodes and establish edges between the nodes of adjacent tunnel segments to construct graphical structural data. Use the spatial data at the same time as a graph and construct the input data into a graph sequence.

[0148] S43. Input the graphics at different times into the corresponding GAT units in the GAT model layer, extract the deformation space features at each time step, arrange and stitch them in chronological order.

[0149] S44. Extract features from the data output by the GAT layer using a convolutional layer;

[0150] S45. Use the LSTM model layer to extract the temporal features of the deformation data and obtain the predicted value.

[0151] In this step, the deformation characteristics of the target deformation monitoring section and two deformation monitoring sections before and after it over five consecutive time periods are used to predict the future deformation of the target monitoring section. The longitudinal deformation U of the tunnel during its service life, collected and transmitted in step S2, is then used. l_s Peripheral convergence U c_s and the U-shaped arch sinking s_s The spatiotemporal data, according to the formula The data is preprocessed and reconstructed into window structures. The model input data has dimensions of {5, 5, 3}. The tunnel segment centered on the deformation monitoring section is represented as different nodes, and edges are established between the nodes of adjacent tunnel segments to construct graphical structure data. Spatial data at the same time are used as a graph with data dimensions of {5, 3}.

[0152] The images at five different time points are input into the GAT model layer, where different GAT units perform node attention weighting to adaptively extract the deformation spatial features at each time step. These features are then arranged and stitched together in chronological order. Subsequently, convolutional layers are used to extract features from the data output by the GAT layer. Finally, the LSTM model layer is used to extract the temporal features of the deformation data and obtain the predicted values.

[0153] S5. Construct a graded early warning method for tunnel structure deformation based on the theory of strength reduction and cusp catastrophe.

[0154] S51. Establish a three-dimensional geological model in Abaqus software. Define the location and shape of the tunnel and shield shell based on the tunnel depth, excavation radius, and three-dimensional spatial coordinate information. Simultaneously, assign material properties to the tunnel, shield shell, and strata based on material properties and stratum parameters. During the construction period, dynamically adjust and simulate the construction process in real time based on the actual construction status of the tunnel boring machine. For specific tunnel sections at different stages, use different strength reduction coefficients K = {K1, K2, K3, K4…} at equal intervals for simulation analysis, and obtain the corresponding increase in plastic deformation energy ΔU of the surrounding rock. P And the mutation eigenvalue Δ, plot and fit ΔU P The relationship curve with K; ΔU P The state in which a sudden change occurs and Δ changes from a positive value to a negative value is determined as the limit state, and the corresponding longitudinal deformation U of the tunnel is... l_c / U l_s Peripheral convergence U c_c / U l_s and the U-shaped arch sinking s_c / U l_s That is, the deformation limit value U of the cross section under the corresponding state. o ={U l_o Uc_o U s_o Based on the average trisection points of the deformation limit values, four warning levels—red, orange, yellow, and green—are defined, with each warning level having a range of [U... o (+∞) and The corresponding increase in plastic deformation energy of the surrounding rock, ΔU, is obtained. P The formula for the mutation eigenvalue Δ is as follows:

[0155]

[0156] Where: m is the number of reduction cycles; n is the number of plastic units; σ ij The stress at the j-th strength reduction of plastic element i; Δε ij The strain increment is the strain increment at the j-th strength reduction of plastic element i; a1, a2, a3, and a4 are derived from ΔU P The undetermined coefficients of the fourth-order polynomial are obtained by fitting the reduction coefficient K using the least squares method.

[0157] S52. Calculate the tunnel depth H, the internal friction angle φ, the cohesion c, and the surrounding rock contact pressure F at different construction stages. mn Average cutterhead rotation speed N, average cutterhead torque T, shield tunneling speed V, distance x between deformation monitoring face and tunneling face, and cross-sectional deformation limit value U. l_o U c_o and U s_o The data is constructed into a dataset, a BP model is built and trained, the trained BP model is transferred to a new dataset at other sections of the tunnel for training, the first few layers of the BP model are frozen, and the parameters of the last few layers of the BP model are trained or new model layers are added.

[0158] The BP model in this step is used to predict the limit value of tunnel cross-section deformation. The first few layers of the BP model are frozen, and the parameters of the last few layers of the BP model are trained or new model layers are added to achieve dynamic updates of the tunnel's full-space deformation warning value and improve computational efficiency.

[0159] S53. Obtain the graded early warning range of each monitoring section during the construction period from steps S51 and S52, and combine the tunnel section deformation monitoring value or prediction value to carry out risk early warning during the tunnel construction period.

[0160] S54. Delete the construction period parameters in steps S51 and S52, and re-execute steps S51 and S52 to obtain the graded early warning range of each monitoring section during the service period. Combine the tunnel section deformation monitoring values ​​or predicted values ​​to conduct real-time risk early warning for the tunnel service period. The construction period parameters include: surrounding rock contact pressure F. mn The average cutterhead rotation speed N, average cutterhead torque T, shield tunneling speed V, and the distance x between the deformation monitoring face and the tunneling face;

[0161] S55. Conduct real-time risk warning for the entire life cycle of the tunnel based on steps S53 and S54.

[0162] like Figure 6 As shown, this invention constructs a tunnel monitoring and early warning system based on digital twins, comprising the following steps:

[0163] S6. Construct a tunnel monitoring and early warning system based on digital twins.

[0164] S61. Collect three-dimensional point cloud data of the tunnel, structural deformation information throughout the entire life cycle, and shield construction status through sensors. Standardize the acquired data and encode it into JSON format. Transmit the data to the server via a communication transmission module. The communication transmission module consists of an interface transmission base station provided with Zigbee, CAN bus, and TCP / IP protocols, power supply equipment, GPRS module, Zigbee module, and ARM9 processor arranged in the tunnel.

[0165] S62. The shield tunnel digital twin management platform includes a shield tunnel design and modeling module, a structural deformation prediction and early warning module, and a user interface and interaction module. Based on the shield tunnel digital twin management platform, it accesses the database on the server and processes various data to construct a virtual model of the tunnel-geological complex. Based on the structural deformation prediction model and graded early warning method during the tunnel construction and service periods, it achieves full-lifecycle, full-space deformation prediction and risk early warning for the shield tunnel. The platform provides a 3D visualization interface, using the BIM model as the data carrier, to transmit, update, and synchronously render the tunnel deformation status, shield construction status, and early warning status assessment in real time.

[0166] In addition to the above embodiments, the present invention may have other implementation methods. All technical solutions formed by equivalent substitution or equivalent transformation fall within the protection scope claimed by the present invention.

Claims

1. A method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins, characterized in that: It includes the following steps: S1. Based on the point cloud data of the completed tunnel section, construct a tunnel-geological complex model that integrates BIM and 3D point cloud technologies; S2. Employ FBG and incremental sensing to monitor tunnel deformation in real time throughout its entire life cycle. S3. Construct an LGBM proxy model and update the tunnel structure deformation prediction during the construction period; S4. Construct a GAT-LSTM model to predict the structural deformation of the tunnel during its service life; S5. Construct a graded early warning method for tunnel structure deformation based on strength reduction and cusp catastrophe theory; S51. Establish a three-dimensional geological model in Abaqus software. Define the location and shape of the tunnel and shield shell based on the tunnel depth, excavation radius, and three-dimensional spatial coordinate information. Simultaneously, assign material properties to the tunnel, shield shell, and strata based on material properties and stratum parameters. During the construction period, dynamically adjust and simulate the construction process in real time based on the actual construction status of the tunnel boring machine. For specific tunnel sections at different stages, use different strength reduction coefficients K = {K1, K2, K3, K4…} at equal intervals for simulation analysis, and obtain the corresponding increase in plastic deformation energy ΔU of the surrounding rock. P And the mutation eigenvalue Δ, plot and fit ΔU P The relationship curve with K; ΔU P The state in which a sudden change occurs and Δ changes from a positive value to a negative value is determined as the limit state, and the corresponding longitudinal deformation U of the tunnel is... l_c / U l_s Peripheral convergence U c_c / U l_s and the U-shaped arch sinking s_c / U l_s That is, the deformation limit value U of the cross section under the corresponding state. o ={U l_o U c_o U s_o Based on the average trisection points of the deformation limit values, four warning levels—red, orange, yellow, and green—are defined, with each warning level having a range of [U... o (+∞) and The corresponding increase in plastic deformation energy of the surrounding rock, ΔU, is obtained. P The formula for the mutation eigenvalue Δ is as follows: Where: m is the number of reduction cycles; n is the number of plastic units; σ ij The stress at the j-th strength reduction of plastic element i; Δε ij The strain increment is the strain increment at the j-th strength reduction of plastic element i; a1, a2, a3, and a4 are derived from ΔU P The undetermined coefficients of the fourth-order polynomial obtained by fitting the reduction factor K using the least squares method; S52. Calculate the tunnel depth H, the internal friction angle φ, the cohesion c, and the surrounding rock contact pressure F at different construction stages. mn Average cutterhead rotation speed N, average cutterhead torque T, shield tunneling speed V, distance x between deformation monitoring face and tunneling face, and cross-sectional deformation limit value U. l_o U c_o and U s_o The data is constructed into a dataset, a BP model is built and trained, the trained BP model is transferred to a new dataset at other sections of the tunnel for training, the first few layers of the BP model are frozen, and the parameters of the last few layers of the BP model are trained or new model layers are added. S53. Obtain the graded early warning range of each monitoring section during the construction period from steps S51 and S52, and combine the tunnel section deformation monitoring value or prediction value to carry out risk early warning during the tunnel construction period. S54. Delete the construction period parameters in steps S51 and S52, and re-execute steps S51 and S52 to obtain the graded early warning range of each monitoring section during the service period. Combine the tunnel section deformation monitoring values ​​or predicted values ​​to conduct real-time risk early warning for the tunnel service period. The construction period parameters include: surrounding rock contact pressure F. mn The average cutterhead rotation speed N, average cutterhead torque T, shield tunneling speed V, and the distance x between the deformation monitoring face and the tunneling face; S55. Conduct real-time risk warning for the entire life cycle of the tunnel based on steps S53 and S54. S6. Construct a tunnel monitoring and early warning system based on digital twins; S61. Collect three-dimensional point cloud data of the tunnel, structural deformation information throughout the entire life cycle, and shield construction status through sensors. Standardize the acquired data and encode it into JSON format. Transmit it to the server via a communication transmission module. The communication transmission module consists of an interface transmission base station, power supply equipment, GPRS module, Zigbee module, and ARM9 processor arranged in the tunnel to provide Zigbee, CAN bus, and TCP / IP protocols. S62. The shield tunnel digital twin management platform includes a shield tunnel design and modeling module, a structural deformation prediction and early warning module, and a user interface and interaction module. Based on the shield tunnel digital twin management platform, it accesses the database in the server and processes various data to realize the construction of a virtual model of the tunnel-geological complex. Based on the structural deformation prediction model and graded early warning method during the tunnel construction and service periods, it realizes full-life-cycle full-space deformation prediction and risk early warning of the shield tunnel. The tunnel digital twin management platform provides a three-dimensional visualization interface, using the BIM model as the data carrier, to transmit, update, and synchronously render the tunnel deformation status, shield construction status, and early warning status assessment in real time.

2. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 1, characterized in that: The specific steps of step S1 are as follows: S11. Obtain point cloud data of the completed section of the tunnel using the vehicle-mounted laser detection system; S111. Install a lidar scanner above the inspection vehicle. Based on the lidar scanner, establish a local coordinate system with the center of the lidar scanner as the origin, and obtain the three-dimensional point cloud data P of the tunnel under this local coordinate system. local ; S112. Install GPS on the inspection vehicle and obtain the vehicle's position coordinates T in the global coordinate system based on the GPS on the inspection vehicle; S113. Install an inertial navigation system on the inspection vehicle to keep the data synchronized with the LiDAR scanner. S114. Install an inclinometer inside the inspection vehicle to obtain accurate three-axis attitude angle data and construct a rotation matrix R; S115. Based on the three-dimensional point cloud data P of the tunnel in the local coordinate system obtained in step S111. local The position coordinates T of the vehicle body in the global coordinate system obtained in step S112 and the rotation matrix R obtained in step S114 are used to obtain accurate three-dimensional point cloud coordinate information, that is, the three-dimensional point cloud data P of the tunnel in the global coordinate system. global The formula is as follows: P global =R·P local +T S12. Reconstruct the shield tunnel structure based on the point cloud data of the completed section of the tunnel obtained in step S11. S121, Point cloud data preprocessing; S122. Construct a triangulation network for flat tunnel point cloud data based on the Delaunay triangulation algorithm, and reconstruct the main outline of the tunnel by combining the three-dimensional point cloud data of the tunnel. S123. Project the point cloud data onto the yox and zox planes respectively. Construct points on the central axis by fitting the boundary functions obtained from the boundary points on both sides of the projected point cloud data. Use a cubic polynomial as the fitting function to obtain the fitting equations of the tunnel axis in each plane. S124. Expand the point cloud data along the tunnel axis and determine the pixel value of the point based on its geometric features to create a binary image. Use the Laplacian operator to detect the boundary of the binary image. Based on morphological processing of the boundary image, divide it into individual ring linings along the ring joint. Use connected component analysis to set the same pixel value region to the same value and divide it into different segments. Obtain the center point and radius parameters by fitting the segment point cloud data. Use the CATIA platform to develop the as-built model of the lining ring. Import the fitting parameters of the center point and radius of different segments, the tunnel axis, and material properties as input parameters into the platform, establish geometric dimension constraint relationships, create a parameterized lining ring model, and thus reconstruct the shield tunnel structure. S13. Adaptively align the shield tunnel structure reconstructed in step S12 with the geological environment model to build a tunnel-geological complex model.

3. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 2, characterized in that: The specific steps of step S2 are as follows: S21. Fiber optic grating multi-point displacement gauges are installed longitudinally on each deformation monitoring section of the shield tunnel to obtain the longitudinal deformation U during the tunnel construction period. l_c and longitudinal deformation during service life U l_s And transmit it to the receiving instrument; S22. On each deformation monitoring section of the shield tunnel, pulleys, several data acquisition modules with incremental encoders, and weights are sequentially installed along the tunnel circumference. The pulleys, several data acquisition modules with incremental encoders, and weights are connected by flexible wires to form a non-closed polygon convergence monitoring system. S221. Data acquisition modules with incremental encoders are installed at the arch crown, both sides of the arch waist, and the middle position of the tunnel deformation monitoring section. The locations of the data acquisition modules with incremental encoders are named P1, P2, ..., P1 from left to right. n Point P1 is fixed using pulleys. n A weight is connected at the point to maintain the tension of the conductor, and the pulley acts as a ball-and-socket joint; S222. Using a total station, obtain the precise initial coordinates of point P1, record the initial calibration values ​​of the angles and traverse lengths at each point, and use trigonometric functions to sequentially obtain the coordinates of points P2, P3, ..., P... n The initial coordinates of the point; S223. Inside each group of data acquisition modules with incremental encoders, a hollow shaft magnetic sensor A is installed on the left side, and a hollow shaft magnetic sensor B is installed on the right side. Hollow shaft magnetic sensor A monitors the angle increment of the left-side traverse at each point between time t and time t-1, and hollow shaft magnetic sensor B monitors the angle increment of the right-side traverse at each point between time t and time t-1. Combining this with the initial calibration values ​​obtained in step S222, P1, P2, ..., P... are obtained. n Angle value at point t and The formula is as follows: In the formula: i is the serial number of the data acquisition module deployment point, ranging from 1 to n; P at time t i The included angle between the two guide wires at the point; P at time t i The angle between the right-side guide wire and the vertical direction at the point; P measured by hollow shaft magnetic sensor A i The angle increment of the traverse to the left of the point between time t and time t-1; P measured by hollow shaft magnetic sensor B i The angle increment of the traverse to the right of the point between time t and time t-1; S224. A wire support pulley and a solid shaft sensor C are installed at the connection point of two flexible wires inside each data acquisition module with an incremental encoder. The angle increment Δδ of the wire support pulley is obtained using the solid shaft sensor C, combined with P... i-1 Point and P i Initial length of the conductor between points Obtain the length of each conductor segment at time t. The formula is as follows: In the formula: P at time t i Point and P i-1 The length of the conductor between the points; Let P be the initial time. i Point and P i-1 The length of the conductor between the points; r is the radius of the pulley supported by the wire; and P at time t i-1 Point and P i The angle increment of the guide rail support pulley at the point; S225. Based on the angle value at time t obtained in step S223 The wire length value at time t obtained in step S224 Using trigonometric functions in sequence, we obtain P2, P3, ..., P... n The location coordinates of each point at time t are determined, and finally, the tunnel convergence U at different times during the construction period is obtained based on the changes in the coordinate information of each point. c_c U-shaped arch sinking s_c And tunnel convergence U at different times during the service life c_s U-shaped arch sinking s_s .

4. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 3, characterized in that: The specific steps of step S3 are as follows: S31. Several pressure monitoring sections are evenly set along the longitudinal direction of the tunnel boring machine (TBM). Several pressure monitoring points are arranged on each pressure monitoring section. Miniature load cells are installed outside the TBM's protective cover at each pressure monitoring point. The miniature load cells are connected to a data logger using shielded data cables to measure the surrounding rock contact pressure F acting on each pressure monitoring point of the TBM at time t. mn Wireless accelerometers and strain gauge torque sensors are installed on different cutter mounting radii on the back of the cutterhead. The data is transmitted to the server through a wireless gateway near the cutterhead to obtain the average cutterhead rotation speed N and the average cutterhead torque t. Linear displacement sensors are installed on the shield machine shell to obtain the shield advance speed V, thereby obtaining the distance x between the different deformation monitoring surfaces and the tunnel face. Pressure sensors and flow meters are installed in the shield machine grouting system to obtain the grouting pressure P and grouting volume Q. S32. The surrounding rock contact pressure F obtained in step S31 mn The average cutterhead rotation speed N, average cutterhead torque T, shield advance speed V, tunnel depth H, distance x between deformation monitoring face and excavation face, internal friction angle φ, and cohesion c are used as inputs to the LGBM model. The ground parameters need to be weighted considering the location and thickness of the overlying strata, with the longitudinal deformation U of the tunnel during construction being the primary factor. l_c Peripheral convergence U c_c and the U-shaped arch sinking s_c As the output of the LGBM model, an LGBM proxy model is established; S33. Preprocess the input and output data of the LGBM proxy model and divide it into a test set and a training set, as shown in the following formula: In the formula: x i The original data; max x i and min x i These are the minimum and maximum values ​​among the data points, respectively; by default, x p and x q They are 1 and -1 respectively; S34. Based on the Bayesian optimization method based on Gaussian processes, the longitudinal deformation U corresponding to the tunnel construction period is obtained respectively. l_c Peripheral convergence U c_c With the U-shaped arch sinking s_c The hyperparameters for optimal prediction performance, and the formulas for the optimization process, are as follows: R 1:t ={(x1,y1),(x2,y2)…(x t ,y t )} y t =f(x t )+ε t f(x t )~GP(m(x),k(x,x′)) In the formula: f is the objective function; y t x is the observation value at step t; t Let ε be the hyperparameter of the t-th step; t R represents the observation error. 1:t Aggregate the observations from the first t steps; p(R) 1:t |f) represents R under a given objective function 1:t The likelihood distribution of f; p(f) is the prior distribution of f; p(f|R) 1:t To observe R 1:t The posterior distribution of f; m(x) is the mean function; k(x,x′) is the covariance function; S35. Using the hyperparameters of the best prediction effect obtained in step S34 and the test set and training set divided in step S33, the LGBM proxy model established in step S32 is trained and evaluated. Using the trained model, the deformation of the tunnel at the unknown section of the tunnel is predicted based on the burial depth, stratum parameters and distance from the tunnel face at the unknown section of the tunnel. The operating state is the real-time operating state of the tunnel boring machine. Collect dynamic monitoring data during the shield tunneling process, and perform incremental training on the model to update and predict tunnel deformation status.

5. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 4, characterized in that: The specific steps of step S4 are as follows: S41. The service-period longitudinal deformation U of the s front and s rear sections of the target cross-section collected and transmitted in step S2 within a continuous time period of l is calculated. l_s Peripheral convergence U c_s and the U-shaped arch sinking s_s The spatiotemporal data is preprocessed and reconstructed into window structures as model input data, represented as [2s+1, l]. The preprocessing formula is as follows: In the formula: x i The original data; max x i and min x i These are the minimum and maximum values ​​among the data points, respectively; by default, x p and x q They are 1 and -1 respectively; S42. Represent the tunnel segment centered on the deformation monitoring section as different nodes and establish edges between the nodes of adjacent tunnel segments to construct graphical structural data. Use the spatial data at the same time as a graph and construct the input data into a graph sequence. S43. Input the graphics at different times into the corresponding GAT units in the GAT model layer, extract the deformation space features at each time step, arrange and stitch them in chronological order. S44. Extract features from the data output by the GAT layer using a convolutional layer; S45. Use the LSTM model layer to extract the temporal features of the deformation data and obtain the predicted value.

6. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 3, characterized in that: In step S22, the number of data acquisition modules with incremental encoders is 5.

7. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 4, characterized in that: In step S31, there are 5 pressure monitoring sections, and 5 pressure monitoring points are arranged on each pressure monitoring section.

8. The method for monitoring and early warning of structural deformation throughout the entire life cycle of a shield tunnel based on digital twins as described in claim 4, characterized in that: In step S33, the ratio of dividing the test set to the training set is 1:4.

Images