Construction detection method and device based on digital twinning

Abstract

The application relates to the technical field of structural health monitoring, and discloses a structure detection method and device based on digital twinning, wherein the method comprises the following steps: acquiring depth coordinates, real-time water levels, strain and stress monitoring data of each monitoring point of an underwater pipe gallery, and calculating current water pressure values of the monitoring points; calculating water pressure response components based on a digital twinning body water pressure-structure coupling solver, generating data after water pressure compensation; calculating a deviation value matrix and generating a structure deviation gradient field and a pipeline deviation vector; representing the pipe gallery system as a heterogeneous coupling graph and calculating cross-medium influence coefficients to form a weighted heterogeneous coupling graph; identifying structure singular points and pipeline abnormal values and performing spatial correlation analysis; performing reverse tracing and forward prediction along the weighted heterogeneous coupling graph to generate a cross-medium damage causal chain; and performing stability screening on the damage causal chain of multiple tide level periods, and outputting a cross-medium damage positioning result.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring technology, and more specifically, to a method and apparatus for detecting structures based on digital twins. Background Technology

[0002] Cross-river or cross-sea tunnels often contain integrated utility tunnels, housing various types of pipelines, including power lines, communication lines, water supply lines, and heating lines. The tunnel structure is situated underwater and deforms under external water pressure. This deformation is transmitted to the individual pipelines through pipeline supports, forming a coupled system of water pressure, structure, and pipelines. The water pressure exhibits a linear distribution with water depth and also fluctuates periodically due to tidal forces.

[0003] In existing technologies, structural health monitoring of underwater utility tunnels typically involves deploying sensors separately on the tunnel structure and various pipelines to independently collect response data such as strain and stress, and then comparing these data with their respective benchmark values ​​or thresholds to determine the damage state. This method treats the tunnel structure and pipelines as independent monitoring objects, establishing separate damage discrimination models for each.

[0004] Existing technologies suffer from the following drawbacks: First, water pressure variations with depth generate depth-dependent additional stresses, which overlap with abnormal stresses caused by structural damage in the monitoring signals, leading to unstable and difficult-to-separate damage signals. Second, damage to the utility tunnel structure may cause deformation of pipeline supports, resulting in pipeline damage; pipeline leaks may also corrode the utility tunnel structure, causing structural damage. This cross-media coupled damage evolution relationship cannot be identified using traditional separate monitoring methods. Third, traditional methods monitor and analyze the utility tunnel structure and various pipelines separately, failing to trace the complete causal chain of damage transmitted from the utility tunnel structure to the pipelines through water pressure changes, making it impossible to determine the root cause of the damage. These drawbacks lead to the technical problem of accurately locating multi-media damage in underwater utility tunnels. Summary of the Invention

[0005] This invention provides a method and apparatus for detecting structures based on digital twins, solving the technical problem of accurately locating multi-media damage in underwater pipe corridors in related technologies.

[0006] This invention discloses a structure inspection method based on digital twins, comprising the following steps: Acquire the depth coordinates, real-time water level data, strain monitoring data of the underwater utility tunnel structure, and stress monitoring data of each pipeline at each monitoring point, and calculate the current water pressure value at each monitoring point. A water pressure-structure coupling solver based on a digital twin can calculate the structural response components of the pipe gallery caused by water pressure and the expected stress values ​​at each pipeline support by inputting the current water pressure distribution. Subtract the water pressure response component from the measured response data to generate water pressure compensated pipe gallery structure response data and pipeline stress data; Calculate the deviation matrix between the response after water pressure compensation and the expected response of the twin; perform spatial gradient calculation on the deviation matrix to generate the structural deviation gradient field; calculate the pipeline stress deviation to generate the pipeline deviation vector. The utility tunnel system is represented as a heterogeneous coupling diagram, with the structural nodes of the utility tunnel and each pipeline node marked with the medium type, and the physical transmission path of water pressure-structure-pipeline as the edge connecting each node. Based on twin calculations, the influence coefficients of structural deformation on pipeline stress and pipeline leakage on structural corrosion are used as edge weights to form a weighted heterogeneous coupling graph. Identify singular points in the structural deviation gradient field, detect outliers in the pipeline deviation vector, perform spatial correlation analysis on the weighted heterogeneous coupling graph, and identify associated anomaly combinations located on the same transmission path. By tracing back and predicting forward along the weighted heterogeneous coupling graph from the associated anomaly combinations, a cross-medium damage causal chain is generated; Stability screening is performed on the damage causal chain across multiple tidal cycles, and the cross-medium damage localization results are output.

[0007] Furthermore, the calculation of the current water pressure value at each monitoring point includes: Obtain the vertical depth of each monitoring point relative to the reference water level and the current water level height. Multiply the product of water density and gravitational acceleration by the difference between the current water level height and the depth of the monitoring point to obtain the current water pressure value of each monitoring point.

[0008] Furthermore, the hydraulic-structural coupled solver employs the finite element analysis method, including: The pipe gallery is divided into multiple calculation segments along the depth direction, and the water pressure value corresponding to the depth is applied to each calculation segment as a surface load. A water pressure surface load is applied to the outer surface of the pipe gallery structure, and the structural deformation field and stress field are obtained by solving the elasticity equilibrium equation. The displacement and stress transfer values ​​of each pipeline support location are calculated based on the structural deformation field.

[0009] Furthermore, the spatial gradient calculation of the deviation matrix includes: The first spatial derivative of the deviation value is calculated along the longitudinal and circumferential directions of the pipe gallery using the finite difference method. For each monitoring point, the longitudinal gradient value is the difference between the deviation values ​​of adjacent upstream and downstream monitoring points divided by twice the longitudinal distance between the two points; The longitudinal gradient and the circumferential gradient are combined to form a two-dimensional offset gradient field.

[0010] Furthermore, identifying the location of singularities in the structural deviation gradient field includes: Calculate the gradient magnitude at each location in the deviation gradient field; The location where the gradient magnitude reaches a local maximum in the spatial neighborhood is identified as a singular point; Generate a set of structural singularities containing the coordinates of all singular point locations.

[0011] Furthermore, the detection of outliers in the pipeline deviation vector includes: Set an abnormal judgment threshold for pipeline stress deviation; Pipeline monitoring points whose absolute values ​​deviate from the threshold are identified as anomalies. Generate a set of pipeline anomaly points containing all anomaly point numbers.

[0012] Furthermore, the coefficient for calculating the influence of structural deformation on pipeline stress includes: Apply a unit deformation to each structural node in the twin; The stress change at each pipeline node caused by the unit deformation was calculated using finite element analysis. The ratio of stress change to unit deformation is used as the influence coefficient of the structural node on each pipeline node.

[0013] Furthermore, the reverse tracing and forward prediction from the associated anomaly combinations along the weighted heterogeneous coupling graph includes: For each anomalous node in the associated anomaly combination, traverse upstream along the edges of the weighted heterogeneous coupling graph; Based on the edge weights and the order in which each node anomalies occur, the node with the earliest anomaly occurrence and a non-zero weight connection edge with the current node is identified as the upstream damage source node. Traverse downstream along the edge and calculate the damage risk value of the downstream node based on the edge weight and the anomaly level of the current node; By connecting the tracing and prediction results in the order of propagation, a cross-media damage causal chain with the propagation direction marked is generated.

[0014] Furthermore, the stability screening of the damage causal chain across multiple tidal cycles includes: The damage causal chain generation step is repeated over multiple complete tidal cycles. Calculate the frequency of occurrence of each damage causal chain in different tidal cycles; Causal chains whose frequency exceeds a preset proportion are retained as stable causal chains; The abnormal detection results of different tidal cycles are normalized to the same water pressure reference conditions for comparison.

[0015] This invention discloses a structure inspection device based on digital twins, comprising: The data acquisition module is used to acquire the depth coordinates, real-time water level data, strain monitoring data of the underwater utility tunnel structure, and stress monitoring data of each pipeline at each monitoring point, and to calculate the current water pressure value at each monitoring point. The water pressure compensation module is used to calculate the water pressure response component based on the water pressure-structure coupling solver of the digital twin, and to subtract the water pressure response component from the measured response data to generate the water pressure compensated data. The deviation analysis module is used to calculate the deviation matrix between the response after water pressure compensation and the expected response of the twin, and to generate the structural deviation gradient field and pipeline deviation vector. The heterogeneous coupling diagram construction module is used to represent the pipe gallery system as a heterogeneous coupling diagram and calculate the cross-medium influence coefficient to generate a weighted heterogeneous coupling diagram. The correlation analysis module is used to identify singularities in the structural deviation gradient field, detect outliers in the pipeline deviation vector, and perform spatial correlation analysis on the weighted heterogeneous coupling graph to identify associated anomaly combinations. The causal chain tracing module is used to generate cross-medium damage causal chains by tracing back and predicting forward from associated anomaly combinations along a weighted heterogeneous coupling graph. The stability screening module is used to screen the damage causal chain across multiple tidal cycles and output the cross-medium damage location results.

[0016] This invention employs a digital twin hydraulic-structural coupled solver to calculate and compensate for hydraulic pressure response components, eliminating the additional influence of depth-dependent hydraulic pressure on the tunnel structure and pipeline response. It establishes a heterogeneous coupling graph containing tunnel structural nodes and pipeline nodes, and calculates the influence coefficients of structural deformation on pipeline stress and pipeline leakage on structural corrosion as edge weights, enabling the tracing of the complete damage causal chain from hydraulic pressure anomalies to pipelines through the structure. Spatial correlation analysis is performed on the weighted heterogeneous coupling graph between singular points of structural deviation from the gradient field and pipeline anomalies, identifying associated anomalies located on the same physical transmission path. Multi-tidal periodic stability screening retains persistent damage causal chains while eliminating false anomalies caused by hydraulic pressure fluctuations. This invention solves the technical problem of accurately locating multi-media damage in underwater tunnels, achieving the technical effects of effectively separating hydraulic pressure-induced stress from damage stress, identifying cross-media coupled damage correlations, tracing the complete damage causal chain, and accurately locating the root cause of damage. Attached Figure Description

[0017] Figure 1 This is a flowchart of the digital twin-based structure detection method of the present invention. Detailed Implementation

[0018] Cross-river or cross-sea tunnels often contain integrated utility tunnels, housing various types of pipelines, including power lines, communication lines, water supply lines, and heating lines. The tunnel structure is situated underwater and deforms under external water pressure. This deformation is transmitted to the individual pipelines through pipeline supports, forming a coupled system of water pressure, structure, and pipelines. The water pressure exhibits a linear distribution with water depth and also fluctuates periodically due to tidal forces.

[0019] Currently, there are several technical challenges in monitoring the structural health of underwater utility tunnels: First, water pressure variations with depth generate depth-dependent additional stresses, which overlap with abnormal stresses caused by structural damage in the monitoring signals, leading to unstable and difficult-to-separate damage signals. Second, structural damage to the utility tunnel may cause deformation of pipeline supports, resulting in pipeline damage, and pipeline leaks may also corrode the utility tunnel structure, causing structural damage. This cross-media coupled damage evolution relationship cannot be identified using traditional separate monitoring methods. Third, traditional methods monitor and analyze the utility tunnel structure and various pipelines separately, making it impossible to trace the complete causal chain of damage transmitted from water pressure changes through the utility tunnel structure to the pipelines, thus making it impossible to determine the root cause of the damage.

[0020] The method of this embodiment includes the following steps: Step 100: Obtain monitoring data and environmental parameters of the underwater utility tunnel, and calculate the current water pressure value at each monitoring point.

[0021] Obtain depth coordinate data of each monitoring point in the underwater utility tunnel. ,in Number the monitoring points; obtain real-time water level data. ,in For the current moment; acquire strain monitoring data of the utility tunnel structure. and stress monitoring data of each pipeline ,in Number the pipeline monitoring points. Calculate the current water pressure value at each monitoring point based on its depth coordinates and real-time water level data. ,in The density of water, This is the acceleration due to gravity.

[0022] It should be noted that the above-mentioned depth coordinate data is obtained by extracting the elevation information of each monitoring point from the pipe gallery design drawings, or by obtaining the vertical distance of each monitoring point relative to the reference water level through on-site measurement.

[0023] Furthermore, the density of water in the above water pressure calculation formula... The values ​​range from 1000 kg / m³ to 1025 kg / m³, with 1000 kg / m³ for freshwater environments and 1025 kg / m³ for seawater environments; gravitational acceleration. The value is 9.8 m / s².

[0024] Furthermore, the above water pressure calculation formula is applicable when the real-time water level is higher than the depth coordinate of the monitoring point, i.e., it satisfies... When this condition is met, the water pressure value When the monitoring point is above the water surface, that is... This monitoring point is not affected by water pressure, so the water pressure value is zero.

[0025] Taking a cross-sea tunnel integrated utility tunnel as an example, the tunnel is 4.5 kilometers long, and the utility tunnel contains power lines, communication lines, and water supply lines. Five structural monitoring points are set up in the depth-varying sections of the utility tunnel, and six stress monitoring points are set up along the three pipelines. Current moment During high tide, the real-time water level Meters (relative to mean sea level). The depth coordinates, measured strain, and calculated water pressure for each monitoring point are shown in the table below.

[0026] Table 1. Data from structural monitoring points and water pressure calculation results; Taking monitoring point S3 as an example, the water pressure calculation process is as follows: kPa. Among the measured stress data of each pipeline monitoring point, the measured stresses of power pipeline monitoring points P1 and P2 are 45.6 MPa and 52.3 MPa, respectively; the measured stresses of communication pipeline monitoring points P3 and P4 are 12.8 MPa and 15.2 MPa, respectively; and the measured stresses of water supply pipeline monitoring points P5 and P6 are 38.7 MPa and 41.5 MPa, respectively.

[0027] Step 200: Based on the digital twin water pressure-structure coupled solver, calculate the structural response components caused by water pressure and the expected stress values ​​at the pipeline support.

[0028] The current water pressure distribution data is input into the water pressure-structure coupled solver in the digital twin. The water pressure-structure coupled solver is a finite element analysis module, and its input is the sequence of water pressure values ​​at each monitoring point. The output is the theoretical response value of each location of the pipe gallery structure under water pressure. Theoretical stress values ​​at each pipeline support The water pressure-structure coupled solver performs elasticity calculations based on the geometric model, material parameters, and boundary conditions of the pipe gallery structure to obtain the structural deformation field and stress field corresponding to the water pressure load.

[0029] It should be noted that the above-mentioned water pressure-structure coupled solver uses the finite element method, which applies water pressure as a surface load to the outer surface of the pipe gallery structure and obtains the structural response by solving the elastic equilibrium equation.

[0030] In this embodiment of the application, in order to improve the accuracy of water pressure response calculation, the depth gradient effect of water pressure is considered in the solver. The pipe gallery is divided into multiple calculation segments along the depth direction, and the water pressure value corresponding to the depth is applied to each calculation segment to avoid the calculation error introduced by the assumption of uniform water pressure.

[0031] The water pressure value sequence of each monitoring point in Table 1 is input into the water pressure-structure coupled solver of the digital twin. The solver performs finite element analysis based on the geometric model of the concrete lining of the pipe gallery and the material parameters of C50 concrete, and outputs the theoretical strain response of each structural monitoring point under the current water pressure. The solver simultaneously calculates the theoretical stress values ​​of water pressure transmitted to each pipeline support through the deformation of the pipe gallery structure. .

[0032] Step 300: Obtain the measured response data of each monitoring point, and generate the pipe gallery structure response data and pipeline stress data after water pressure compensation.

[0033] The measured response data at each monitoring point are obtained, and the water pressure response component calculated by the digital twin is subtracted from the measured strain data of the pipe gallery structure to obtain the water pressure compensated pipe gallery structure response data. Subtract the water pressure transmission stress component calculated by the digital twin from the measured pipeline stress data to obtain the water pressure compensated pipeline stress data. .

[0034] The measured strain data in Table 1 were compared with the water pressure response components calculated by the digital twin. The water pressure compensation results for each structural monitoring point are shown in the table below.

[0035] Table 2. Calculation results of water pressure compensation at structural monitoring points; Taking monitoring point S3 as an example, the strain after compensation This value is significantly higher than the 16.2 value at the adjacent monitoring point S2. This indicates that there may be a structural anomaly at point S3. After water pressure compensation, the compensated stresses of power lines P1 and P2 are 5.2 MPa and 6.8 MPa, respectively; the compensated stresses of communication lines P3 and P4 are 2.1 MPa and 2.5 MPa, respectively; and the compensated stresses of water supply lines P5 and P6 are 12.5 MPa and 7.3 MPa, respectively.

[0036] Step 400: Calculate the deviation matrix and generate the structural deviation gradient field and pipeline deviation vector.

[0037] Calculate the structural response of the pipe gallery after water pressure compensation and compare it with the expected response of the digital twin under no water pressure and no damage conditions. The deviation values ​​are used to generate a deviation matrix. Spatial gradient calculations are performed on the deviation matrix. The first-order spatial derivatives of the deviations are calculated along the longitudinal and circumferential directions of the pipe gallery using the finite difference method, generating the structural deviation gradient field. ,in The vertical axis is... Using circumferential coordinates. Calculate the pipeline stress after water pressure compensation and the expected value from the digital twin. The deviation is used to generate a pipeline deviation vector. .

[0038] Furthermore, the aforementioned structure deviates from the gradient field. Mid-time dimension The way it is reflected is: at each sampling time Based on the deviation matrix at that time respectively The corresponding spatial gradient field is calculated to form a gradient field sequence that evolves over time. Subsequent singularity detection is performed on the gradient field at the current moment.

[0039] Furthermore, the aforementioned expected response and The method for obtaining the data is as follows: In the digital twin, the external load is set to zero and no damage or defects are introduced. Static analysis is performed on the pipe gallery structure and pipelines to obtain the baseline response value of each monitoring point under the condition of no external disturbance. This baseline response value is used as the expected response for subsequent deviation calculation.

[0040] It should be noted that the spatial gradient calculation above uses the central difference scheme. For the numbered... The longitudinal gradient calculation formula for the monitoring points is as follows: ,in This represents the longitudinal spacing between adjacent monitoring points.

[0041] Furthermore, for monitoring points located at the boundary of the monitoring area, since data from adjacent monitoring points on one side are lacking, forward differencing or backward differencing schemes are used instead of central differencing schemes for gradient calculation. For the initial boundary point, forward differencing schemes are used. The backward difference scheme is used for the termination boundary points. ,in This represents the total number of monitoring points.

[0042] Expected strain values ​​of digital twins under water pressure-free and damage-free conditions All are zero, therefore the deviation value is zero. Equal to the strain value after compensation. Longitudinal spacing between adjacent monitoring points. Meters, circumferential spacing Meters. The longitudinal gradient of the deviation values ​​in Table 2 is calculated. Taking monitoring point S3 as an example, its longitudinal gradient is... / m. The deviation gradient calculation results for each structural monitoring point are shown in the table below.

[0043] Table 3 Calculation results of structural deviation from gradient field; In the calculation of pipeline deviation vector, the expected stress of the digital twin is... The baseline stress values ​​of each pipeline material under normal operating conditions are taken. Taking monitoring point P5 of the water supply pipeline as an example, its expected stress is 5.0 MPa, and the compensated stress is 12.5 MPa, with a deviation of [value missing]. MPa.

[0044] Step 500: Represent the utility tunnel system as a heterogeneous coupling diagram and establish the physical transmission path connection between water pressure, structure, and pipelines.

[0045] The utility tunnel system is represented as a heterogeneous coupling diagram. , where the set of nodes Includes pipe gallery structural nodes and each pipeline node Each node is labeled with its media type attribute. (Edge set) Includes internal connecting edges of the structure Internal connection edge of pipeline and cross-medium connection edge The cross-medium connection edge takes the pipeline support location as the connection point, representing the physical transmission path of water pressure-structure-pipeline.

[0046] Furthermore, the connecting edges inside the above structure The establishment rule is as follows: when two structural monitoring points belong to adjacent finite element elements or are located on the same structural component in the pipe gallery geometric model, a connection edge is established between the corresponding structural nodes; the above-mentioned internal connection edges of the pipeline The establishment rule is: when two pipeline monitoring points are located on adjacent pipe segments of the same pipeline, a connection edge is established between the corresponding pipeline nodes.

[0047] Step 600: Calculate the cross-medium influence coefficient and generate a weighted heterogeneous coupling diagram.

[0048] Based on the structure-pipeline interaction analysis module of the digital twin, the influence coefficient of structural deformation on the stress of each pipeline is calculated. , representing a structural node Unit deformation causes pipeline nodes Stress change; Calculate the influence coefficient of pipeline leakage on structural corrosion. , indicating pipeline node When a leak occurs, it affects the structural nodes. The degree of influence of corrosion rate. The influence coefficient is used as the weight of the cross-medium edge in the heterogeneous coupling diagram to form a weighted heterogeneous coupling diagram. ,in Let be the set of edge weights.

[0049] It should be noted that the above-mentioned influence coefficient is calculated by applying unit structural deformation and unit pipeline leakage in the digital twin, and obtaining the corresponding response change through finite element analysis. The influence coefficient is equal to the ratio of the response change to the excitation.

[0050] Furthermore, the aforementioned unit structural deformation is defined as the deformation at the structural nodes. A displacement load of 1 mm is applied at the point, and the above-mentioned unit pipeline leakage is defined as the leakage at the pipeline node. A leakage rate of 1 liter / hour was set at the point, and the response changes of other nodes under this unit excitation were obtained through digital twin simulation analysis.

[0051] In the heterogeneous coupling diagram of this utility tunnel, the set of structural nodes... Pipeline node set Structural node S3 is connected to power line node P2 and water supply line node P5 via pipeline supports, and S4 is connected to water supply line node P6 via pipeline supports. Some cross-media influence coefficients calculated using the digital twin are shown in the table below.

[0052] Table 4. Cross-medium influence coefficients; Influence coefficient Taking MPa / mm as an example, this value means that when structural node S3 undergoes a 1 mm deformation, the stress increment transmitted to water supply pipeline node P5 through the pipeline support is 3.12 MPa.

[0053] Step 700: Identify singularities in the structural deviation gradient field and perform spatial correlation analysis on the weighted heterogeneous coupling graph.

[0054] In a structure deviating from the gradient field, a local extremum detection algorithm is used to identify singular points. A singular point is defined as the location where the gradient magnitude reaches a local maximum in its spatial neighborhood, generating a set of structural singular points. ,in The first in the set of structural singularities A singularity, For singularity index, For spatial neighborhood any point in, For point The spatial neighborhood. Detecting deviations exceeding a preset threshold in the pipeline deviation vector. The outliers are used to generate a set of pipeline anomaly points. ,in The first point in the set of pipeline anomalies One anomaly, This is an index for outliers.

[0055] Furthermore, the aforementioned gradient magnitude The calculation method is the two-dimensional Euclidean norm, and the calculation formula is: ,in and Points The longitudinal gradient component and the circumferential gradient component at the location.

[0056] Furthermore, the aforementioned spatial neighborhood The range is determined by the neighborhood radius. Determine the neighborhood radius. The value is taken as 1 to 3 times the distance between adjacent monitoring points, and the spatial neighborhood includes the center point. The distance is less than or equal to All monitoring points.

[0057] Furthermore, the aforementioned preset threshold The value is determined based on the allowable stress and safety factor of the pipeline material. The value ranges from 5% to 20% of the allowable stress of the pipeline material. When the absolute value of the deviation exceeds this threshold, it is judged as abnormal.

[0058] It should be noted that, due to the different physical dimensions of structural deviation gradient magnitude and pipeline stress deviation, to ensure comparability of anomaly severity across different media types, the Z-score standardization method is used to render various anomaly severity values ​​dimensionless in subsequent correlation analysis and risk calculations. This converts the gradient magnitude of structural singularities and the stress deviation of pipeline anomalies into standardized anomaly severity values ​​relative to their respective historical statistical distributions. Spatial correlation analysis of singularities and pipeline anomalies is then performed on the weighted heterogeneous coupling diagram to identify anomaly combinations located along the same physical transmission path, generating a set of correlated anomaly combinations. ,in This indicates the structural nodes in the weighted heterogeneous coupling graph. To pipeline nodes The connection path, which consists of the sequence of edges connecting two nodes in the graph, when and When there is a cross-medium connection edge that is directly connected or indirectly connected through an intermediate node. It is a non-empty path.

[0059] The aforementioned local extremum detection algorithm takes as input the structural deviation gradient field. The output is a set of structural singularities. The local extremum detection algorithm traverses each spatial location in the gradient field, compares the gradient magnitude of the center point with that of the neighboring points within a preset spatial neighborhood window, and adds the locations with gradient magnitudes greater than all neighboring points to the singularity set.

[0060] The aforementioned spatial correlation analysis takes a set of structural singularities as its input. Pipeline anomaly point set Weighted heterogeneous coupling graph The output is a set of associated anomaly combinations. Spatial correlation analysis traverses cross-medium edges in a weighted heterogeneous coupling graph For each edge, check whether the connected structural nodes belong to the set of structural singularities and whether the connected pipeline nodes belong to the set of pipeline anomalous points. Add the node pairs that meet the conditions to the set of associated anomalous combinations.

[0061] Based on the gradient magnitude data in Table 3, the neighborhood radius is set. The distance is measured in meters (twice the distance between adjacent monitoring points). Within this neighborhood, the gradient magnitude of monitoring point S3 (0.168) is greater than that of its neighbors S2 (0.145) and S4 (0.085), satisfying the local maximum condition. Therefore, S3 is identified as a structural singularity. Preset thresholds for water supply pipeline anomaly detection. Taking 5% of the allowable stress of 140 MPa, which is 7 MPa, the deviation of P5 is 7.5 MPa, exceeding the threshold, while the deviation of P6 is 2.3 MPa, which does not exceed the threshold. The deviation of power line P2 is 1.8 MPa, which is below its threshold of 2.5 MPa, while the deviations of communication lines P3 and P4 are both below their respective thresholds. Therefore, the set of abnormal points for the pipelines is... Spatial correlation analysis examined the cross-medium edges in the weighted heterogeneous coupling diagram and found a cross-medium connection edge (influence coefficient) between structural singularity point S3 and pipeline anomaly point P5. Generate a set of associated anomaly combinations. .

[0062] Step 800: Perform reverse tracing and forward prediction along the heterogeneous coupling graph to generate a cross-medium damage causal chain.

[0063] Starting from each anomaly combination in the associated anomaly combination set, a reverse tracing is performed along the edges of the weighted heterogeneous coupling graph to determine the upstream node of damage propagation based on the edge weights and the chronological order of anomaly occurrence. A forward prediction is then performed along the edges, calculating the damage risk of downstream nodes based on the edge weights and the current anomaly severity. This generates a causal chain of water pressure-structure-pipeline cross-media damage. ,in For the first to the second in the causal chain of damage 1 node This represents the total number of nodes in the damage causal chain, and indicates the damage propagation order of each node in the damage causal chain. The arrow direction indicates the damage propagation direction.

[0064] It should be noted that the above-mentioned reverse tracing judgment rule is: if node The anomaly occurred earlier than the node. ,and and If there are connection edges with non-zero weights, then for Damage may originate from upstream nodes. spread to .

[0065] Furthermore, the method for determining the above-mentioned anomaly occurrence time is as follows: monitor the time series of deviation values ​​for each node, and record the moment when the deviation value of a node first exceeds the anomaly judgment threshold as the anomaly occurrence time of that node. For structural nodes, the moment when the gradient amplitude exceeds the neighborhood mean is used; for pipeline nodes, the moment when the stress deviation value exceeds a preset threshold is used. At that moment.

[0066] The aforementioned positive prediction takes as input the current anomaly level value of each node in the associated anomaly combination and the edge weights originating from each node in the weighted heterogeneous coupling graph, and outputs the damage risk value of the downstream adjacent node, which is the product of the current anomaly level value and the corresponding edge weight.

[0067] Furthermore, the criteria for determining the damage risk value are as follows: when the damage risk value of a downstream node is greater than zero, it indicates that the node is at risk of being damaged due to the propagation of damage from the upstream node, and the node is added to the positive extension path of the damage causal chain; the larger the damage risk value, the higher the probability of damage propagation.

[0068] From associated anomaly combinations We then traced back to find the time when the anomaly occurred at each node. Monitoring records showed that the gradient magnitude of S3 first exceeded the neighborhood mean at [time missing]. = 09:15:32, the stress deviation value of P5 exceeded the preset threshold for the first time. The time is = 09:47:18, the anomaly in S3 occurred approximately 32 minutes earlier than in P5. Based on the heterogeneous coupling graph, there is a non-zero weight connection edge between S3 and P5 ( S3 is determined to be an upstream node of P5, and the damage propagates from structural node S3 to pipeline node P5. During forward prediction, the standardized anomaly value of S3 is 2.8. The damage risk value of another downstream node P2 of S3 is calculated by multiplying this value by the edge weights. This indicates that P2 poses a risk of further damage. Ultimately, a trans-medium damage causal chain is generated. S3 is marked as the damage origin node and P5 is the damage propagation endpoint.

[0069] Step 900: Perform stability screening on the damage causal chain of multiple tidal cycles and output the cross-medium damage localization results.

[0070] Steps 100 to 800 are repeated over multiple tidal cycles to obtain a set of damage causal chains for each tidal cycle. Stability screening is performed on the damage causal chains across multiple tidal cycles, calculating the frequency of occurrence of each damage causal chain in different tidal cycles, and retaining chains with frequencies exceeding a preset proportion. The damage causal chain is taken as a stable damage causal chain. The output is the cross-media damage localization result, including the starting node position, propagation path and ending node position of the stable damage causal chain.

[0071] Furthermore, the calculation method for the above-mentioned frequency of occurrence is as follows: Let the total number of tidal cycles be... A certain causal chain of damage is involved. If the damage is detected in a tidal cycle, then the frequency of this causal chain is: ,when The causal chain of the damage is preserved at that time.

[0072] Furthermore, the number of the aforementioned multiple tidal cycles shall be no less than three complete tidal cycles to cover different water pressure conditions such as high tide, low tide, and intermediate tide, ensuring that the stability screening results are statistically significant.

[0073] Furthermore, the aforementioned preset ratio The range of values ​​is This value constraint ensures that the preserved stable damage causal chain is detected in most tidal cycles; when When the value is close to 1, the screening criteria are stricter, retaining only damage causal chains that appear in almost all tidal cycles; when... When the value is 0.5, the screening criteria are relatively lenient, retaining the damage causal chains that appear in more than half of the tidal cycles.

[0074] In this embodiment of the application, in order to eliminate the interference of false anomalies caused by water pressure fluctuations on stability screening, the water pressure fluctuation amplitude of each tidal cycle is standardized during screening, and the anomaly detection results of different tidal cycles are normalized to the same water pressure reference conditions for comparison.

[0075] Repeat steps 100 to 800 over four consecutive tidal cycles (approximately 50 hours) to set the stability screening ratio. The results of the tidal cycle measurements are summarized in the table below.

[0076] Table 5. Detection results of causal chain of multi-tidal periodic damage; Damage causal chain It was detected in all four tidal cycles, with a frequency of occurrence. This is determined to be a stable damage causal chain. Damage causal chain It only appears in period 1, and its frequency is... This was determined to be a false anomaly caused by water pressure fluctuations and was therefore ruled out. (Damage causal chain) Frequency of occurrence These are also excluded. The final output shows the cross-medium damage localization result: the stable damage causal chain is... The starting node is structural monitoring point S3 (longitudinal mileage station K2+340, depth -31.8 meters), the propagation path passes through pipeline support No. 3, and the ending node is water supply pipeline monitoring point P5 (12 meters downstream of the support).

[0077] This implementation method, by employing a digital twin water pressure-structure coupled solver to calculate and compensate for the water pressure response components, eliminates the additional influence of depth-dependent water pressure on the tunnel structure and pipeline response, thus overcoming the instability of damage signals caused by the overlap of water pressure-induced additional stress and damage stress. By establishing a heterogeneous coupling diagram including tunnel structure nodes and pipeline nodes, and calculating the influence coefficients of structural deformation on pipeline stress and pipeline leakage on structural corrosion as edge weights, the complete damage causal chain transmitted from water pressure anomalies through the structure to the pipeline can be traced, thus overcoming the inability of traditional separate monitoring methods to identify cross-media damage correlations. By performing spatial correlation analysis between structural deviation gradient field singularities and pipeline anomalies on the weighted heterogeneous coupling diagram, structural damage and pipeline damage located on the same physical transmission path can be identified as correlated anomalies, thus overcoming the lack of spatial correlation relationships for cross-media damage. By retaining persistent damage causal chains through multi-tidal periodic stability screening, pseudo-anomalies caused by water pressure fluctuations are eliminated, thus solving the technical problem of multi-media damage localization in underwater tunnels.

[0078] This embodiment has the following advantages over the prior art: First, a damage feature extraction method combining water pressure compensation and deviation gradient field. In existing technologies, damage detection of underwater structures typically involves directly comparing the measured response with the expected response, neglecting the additional influence of water pressure as a time-varying external load on the response. This implementation calculates and compensates for the water pressure response component using a digital twin water pressure-structure coupled solver, and then calculates the deviation gradient field on the compensated data, thus separating the local abnormal signals caused by damage from the global water pressure effect. This method combines the physical simulation capabilities of digital twins with gradient analysis methods in signal processing, forming a dedicated damage feature extraction scheme for underwater environments.

[0079] Second, a cross-medium damage correlation analysis method based on heterogeneous coupling graphs. In existing technologies, the monitoring of pipe gallery structures and pipelines is usually independent, using their own damage discrimination criteria. This implementation modeles the pipe gallery system as a heterogeneous coupling graph, using the physical transmission path of water pressure-structure-pipeline as edges connecting nodes of different media, and using digital twins to calculate cross-medium influence coefficients as edge weights. This graph structure allows for the explicit expression of the physical correlation between damage in different media, providing a data structure foundation for tracing the causal chain of cross-medium damage.

[0080] Third, a method for eliminating false anomalies through multi-tidal cycle stability screening. In existing technologies, structural monitoring under tidal conditions typically uses fixed thresholds for anomaly detection, neglecting the impact of tidal level changes on the stability of anomaly detection results. This implementation method, through repeated multi-tidal cycle detection and stability screening, retains the damage causal chain that persists under different tidal conditions, eliminating false anomalies caused by water pressure fluctuations. This method utilizes the difference between the temporal persistence characteristics of true damage and the temporal transient characteristics of false anomalies caused by water pressure fluctuations, improving the reliability of damage localization results.

Claims

1. A method for inspecting structures based on digital twins, characterized in that, Includes the following steps: Acquire the depth coordinates, real-time water level data, strain monitoring data of the underwater utility tunnel structure, and stress monitoring data of each pipeline at each monitoring point, and calculate the current water pressure value at each monitoring point. A water pressure-structure coupling solver based on a digital twin can calculate the structural response components of the pipe gallery caused by water pressure and the expected stress values ​​at each pipeline support by inputting the current water pressure distribution. Subtract the water pressure response component from the measured response data to generate water pressure compensated pipe gallery structure response data and pipeline stress data; Calculate the deviation matrix between the response after water pressure compensation and the expected response of the twin; perform spatial gradient calculation on the deviation matrix to generate the structural deviation gradient field; calculate the pipeline stress deviation to generate the pipeline deviation vector. The utility tunnel system is represented as a heterogeneous coupling diagram, with the structural nodes of the utility tunnel and each pipeline node marked with the medium type, and the physical transmission path of water pressure-structure-pipeline as the edge connecting each node. Based on twin calculations, the influence coefficients of structural deformation on pipeline stress and pipeline leakage on structural corrosion are used as edge weights to form a weighted heterogeneous coupling graph. Identify singular points in the structural deviation gradient field, detect outliers in the pipeline deviation vector, perform spatial correlation analysis on the weighted heterogeneous coupling graph, and identify associated anomaly combinations located on the same transmission path. By tracing back and predicting forward along the weighted heterogeneous coupling graph from the associated anomaly combinations, a cross-medium damage causal chain is generated; Stability screening is performed on the damage causal chain across multiple tidal cycles, and the cross-medium damage localization results are output.

2. The method according to claim 1, characterized in that, The calculation of the current water pressure value at each monitoring point includes: Obtain the vertical depth of each monitoring point relative to the reference water level and the current water level height. Multiply the product of water density and gravitational acceleration by the difference between the current water level height and the depth of the monitoring point to obtain the current water pressure value of each monitoring point.

3. The method according to claim 1, characterized in that, The water pressure-structure coupled solver employs the finite element analysis method, including: The pipe gallery is divided into multiple calculation segments along the depth direction, and the water pressure value corresponding to the depth is applied to each calculation segment as a surface load. A water pressure surface load is applied to the outer surface of the pipe gallery structure, and the structural deformation field and stress field are obtained by solving the elasticity equilibrium equation. The displacement and stress transfer values ​​of each pipeline support location are calculated based on the structural deformation field.

4. The method according to claim 1, characterized in that, The calculation of the spatial gradient of the deviation matrix includes: The first spatial derivative of the deviation value is calculated along the longitudinal and circumferential directions of the pipe gallery using the finite difference method. For each monitoring point, the longitudinal gradient value is the difference between the deviation values ​​of adjacent upstream and downstream monitoring points divided by twice the longitudinal distance between the two points; The longitudinal gradient and the circumferential gradient are combined to form a two-dimensional offset gradient field.

5. The method according to claim 1, characterized in that, The identification of singularity locations in the structural deviation gradient field includes: Calculate the gradient magnitude at each location in the deviation gradient field; The location where the gradient magnitude reaches a local maximum in the spatial neighborhood is identified as a singular point; Generate a set of structural singularities containing the coordinates of all singular point locations.

6. The method according to claim 1, characterized in that, The detection of outliers in the pipeline deviation vector includes: Set an abnormal judgment threshold for pipeline stress deviation; Pipeline monitoring points whose absolute values ​​deviate from the threshold are identified as anomalies. Generate a set of pipeline anomaly points containing all anomaly point numbers.

7. The method according to claim 1, characterized in that, The coefficient for calculating the influence of structural deformation on pipeline stress includes: Apply a unit deformation to each structural node in the twin; The stress change at each pipeline node caused by the unit deformation was calculated using finite element analysis. The ratio of stress change to unit deformation is used as the influence coefficient of the structural node on each pipeline node.

8. The method according to claim 1, characterized in that, The reverse tracing and forward prediction from the associated anomaly combination along the weighted heterogeneous coupling graph includes: For each anomalous node in the associated anomaly combination, traverse upstream along the edges of the weighted heterogeneous coupling graph; Based on the edge weights and the order in which each node anomalies occur, the node with the earliest anomaly occurrence and a non-zero weight connection edge with the current node is identified as the upstream damage source node. Traverse downstream along the edge and calculate the damage risk value of the downstream node based on the edge weight and the anomaly level of the current node; By connecting the tracing and prediction results in the order of propagation, a cross-media damage causal chain with the propagation direction marked is generated.

9. The method according to claim 1, characterized in that, The stability screening of the damage causal chain across multiple tidal cycles includes: The damage causal chain generation step is repeated over multiple complete tidal cycles. Calculate the frequency of occurrence of each damage causal chain in different tidal cycles; Causal chains whose frequency exceeds a preset proportion are retained as stable causal chains; The abnormal detection results of different tidal cycles are normalized to the same water pressure reference conditions for comparison.

10. A structure inspection device based on digital twins, used to execute the structure inspection method based on digital twins according to any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire the depth coordinates, real-time water level data, strain monitoring data of the underwater utility tunnel structure, and stress monitoring data of each pipeline at each monitoring point, and to calculate the current water pressure value at each monitoring point. The water pressure compensation module is used to calculate the water pressure response component based on the water pressure-structure coupling solver of the digital twin, and to subtract the water pressure response component from the measured response data to generate the water pressure compensated data. The deviation analysis module is used to calculate the deviation matrix between the response after water pressure compensation and the expected response of the twin, and to generate the structural deviation gradient field and pipeline deviation vector. The heterogeneous coupling diagram construction module is used to represent the pipe gallery system as a heterogeneous coupling diagram and calculate the cross-medium influence coefficient to generate a weighted heterogeneous coupling diagram. The correlation analysis module is used to identify singularities in the structural deviation gradient field, detect outliers in the pipeline deviation vector, and perform spatial correlation analysis on the weighted heterogeneous coupling graph to identify associated anomaly combinations. The causal chain tracing module is used to generate cross-medium damage causal chains by tracing back and predicting forward from associated anomaly combinations along a weighted heterogeneous coupling graph. The stability screening module is used to screen the damage causal chain across multiple tidal cycles and output the cross-medium damage location results.

Images