A local structure hazard analysis method using digital twinning

Abstract

The present application relates to the technical field of local structure damage detection, and particularly relates to a local structure hazard analysis method using digital twinning. The method acquires measured strain signals and vehicle parameters when a standard vehicle passes over a bridge, and extracts a normalized measured feature vector. At the same time, an isomorphic simulation feature vector is generated based on the vehicle parameters and a finite element model. An optimal transmission coupling matrix between the two is calculated using the Sinkhorn algorithm, and a distribution pattern transmission cost and an adaptive sparse weight are obtained, solving the recognition error caused by the phase shift in the waveform space. By minimizing the transmission cost and applying sparse constraints, the equivalent cross-section reduction coefficient of each unit is inverted, and the precise positioning of local stiffness damage is realized. Finally, the model material parameters are updated synchronously, and the load bearing capacity reduction rate of the structure is calculated through nonlinear whole-process analysis. The present application can accurately quantify the influence of local damage on the overall safety of the structure under the conditions of unknown vehicle weight and waveform shift.

Description

Technical Field

[0001] This invention relates to the field of local structural damage detection technology, and more specifically to a method for analyzing the hazards of local structures using digital twins. Background Technology

[0002] In the operation and maintenance of long-span steel bridges, localized areas such as welds and U-rib connections in orthotropic steel bridge decks are highly susceptible to hidden fatigue damage (such as fatigue crack propagation or structural corrosion) due to the repeated effects of vehicle wheel loads over long periods. These localized damages initially manifest as a slight reduction in the effective load-bearing cross-section of the component, leading to a degradation of local stiffness. However, due to the complex structure and highly statically indeterminate nature of steel bridge decks, the contribution of these minor changes in local stiffness to the overall structural deformation is very limited, making it difficult for traditional assessment methods based on macroscopic monitoring data to accurately detect these early damages.

[0003] Current structural health monitoring technologies primarily assess structural condition by deploying strain sensors at key locations to acquire real-time response data. However, in practical engineering applications, traffic load parameters exhibit significant randomness. In conventional monitoring scenarios, the absolute total weight of vehicles crossing a bridge is unknown. Since the amplitude of structural strain response is directly linearly related to the magnitude of the external load, without obtaining the exact vehicle weight, changes in response amplitude alone cannot accurately distinguish between an increase in load and a decrease in local structural stiffness. This makes it difficult to identify damage using simple amplitude threshold judgments, easily leading to false alarms or missed alarms. Furthermore, local stiffness damage can cause spatial shifts in the response waveform. That is, when the local stiffness of a structure degrades, not only will the response value change, but the peak position of the influence line (or response curve) under moving load will also undergo a slight spatial phase shift. Traditional model correction methods usually compare measured data with model data based on Euclidean distance (such as mean square error). This point-to-point numerical comparison method is very sensitive to the spatial shift of the waveform. A slight peak misalignment can produce huge calculation errors, making it difficult for the inversion algorithm to converge to the correct damage location, or even get stuck in a local minimum. At the same time, there is a lack of quantitative mapping from microscopic damage to macroscopic safety. This is because existing monitoring methods can usually only identify the current stiffness state (such as the percentage by which stiffness has decreased), but it is difficult to directly quantify the specific impact of this local stiffness reduction on the overall ultimate bearing capacity of the structure under future extreme conditions. It is also difficult to intuitively determine whether the damage has endangered the overall safety of the structure, resulting in a lack of clear quantitative basis for maintenance decisions. Summary of the Invention

[0004] To address the challenges of accurately identifying damage locations using traditional amplitude or Euclidean distance comparison methods under complex conditions such as random traffic loads (e.g., unknown vehicle weight) and spatial shifts in response waveforms caused by local stiffness damage, and the lack of a means to quantitatively assess the safety of macroscopic structural ultimate bearing capacity from microscopic local stiffness reduction, this invention aims to provide a method for analyzing the local structural hazard using digital twins. The specific technical solution adopted is as follows: This invention provides a method for local structural hazard analysis using digital twins, the method comprising the following steps: Acquire the measured strain signals, wheelbase parameters, and axle load ratio of a standard vehicle crossing the bridge, which are simultaneously collected by multiple strain sensors placed at multiple key stress points on the target bridge. The measured strain second-order difference features are obtained based on the measured strain signals, and the measured feature vector is obtained based on the measured strain second-order difference features; the simulation feature vector is obtained based on the wheelbase parameter, wheelbase weight ratio and finite element model. Based on measured and simulated feature vectors, combined with the geometric cost matrix, the optimal transmission coupling matrix is ​​obtained; according to the distribution of elements in the optimal transmission coupling matrix, the distribution-shaped transmission cost and adaptive sparse weight are obtained; based on the iterative change of the distribution-shaped transmission cost and the penalty constraint of the adaptive sparse weight on each element of the target bridge, the equivalent cross-section reduction coefficient of each element is obtained. Based on the equivalent section reduction factor, the elastic modulus and yield strength of the finite element model are updated simultaneously to obtain the rate of decrease in structural bearing capacity.

[0005] Furthermore, the method for obtaining the measured strain second-order difference characteristics is as follows: The measured strain signal of each strain sensor is subjected to polynomial smoothing filtering to obtain the corrected measured strain signal; The time axis is scaled and aligned by using the bridge start point and bridge fall point of the corrected measured strain signal. Based on the passing time of the standard vehicle and the bridge length, combined with the average vehicle speed, the corrected measured strain signals of each strain sensor after time axis scaling and alignment are mapped to the preset longitudinal reference coordinate sequence by linear interpolation, and a spatial domain strain response matrix with the sensor as the row and the coordinate point as the column is constructed. The spatial domain strain response matrix is ​​subjected to a second-order central difference operation along a preset longitudinal reference coordinate sequence to obtain the second-order difference value of each coordinate point. For coordinate points in the preset longitudinal reference coordinate sequence that are not covered by effective sensors in their neighborhood, their second-order difference value is recorded as zero. The obtained second-order difference values ​​are used as the measured strain second-order difference characteristics.

[0006] Furthermore, the method for obtaining the measured feature vector is as follows: When the maximum amplitude of the measured second-order difference characteristic of strain is not greater than the preset noise threshold, it is determined to be an invalid signal and the analysis is terminated. When the maximum amplitude of the measured strain second-order difference feature is greater than the preset noise threshold, the measured strain second-order difference features corresponding to all strain sensors at each coordinate point are squared and summed to obtain the first feature value of each coordinate point. The sum of the first eigenvalues ​​of all coordinate points whose second-order difference values ​​are not zero, plus the first preset minimum positive number, is taken as the overall eigenvalue. The ratio of the first feature value of each coordinate point to the overall feature value is taken as the relative deformation energy percentage of each coordinate point; The relative deformation energy percentages of each coordinate point are arranged in the order of a preset longitudinal reference coordinate sequence to construct the measured feature vector.

[0007] Furthermore, the method for obtaining the simulation feature vector is as follows: Apply a unit moving load column to the finite element model of the target bridge, with the same wheelbase parameters and axle load ratio as the standard vehicle. The unit moving load column is controlled to move gradually along a preset longitudinal reference coordinate sequence to extract simulated strain data corresponding to the actual strain signal acquisition position; The simulated strain data were sequentially smoothed, filtered, and subjected to second-order central difference calculation. The second-order difference values ​​of all corresponding simulated measurement points were then summed by squares and normalized to obtain the simulated strain second-order difference characteristic distribution for each coordinate point. The simulated strain second-order difference feature distributions at each coordinate point are arranged in the order of a preset longitudinal reference coordinate sequence to construct the simulated feature vector.

[0008] Furthermore, the method for obtaining the optimal transmission coupling matrix is ​​as follows: A geometric cost matrix is ​​constructed using each coordinate point in the preset vertical reference coordinate sequence as the row and column, respectively; where each element in the geometric cost matrix is ​​the square of the Euclidean distance between each pair of coordinate points in the preset vertical reference coordinate sequence divided by the maximum value among all squares of Euclidean distances; Based on the Sinkhorn algorithm, using the measured feature vectors as the target distribution and the simulated feature vectors as the source distribution, we search for a non-negative transmission coupling matrix that minimizes the transmission cost. An entropy regularization constraint is introduced into the search process, and the diagonal scaling vector is updated alternately through a fixed-point iterative algorithm until the norm difference between two iterations is less than a preset convergence threshold. The optimal transmission coupling matrix is ​​obtained by multiplying the updated diagonal scaling vector with the kernel matrix based on the geometric cost matrix.

[0009] Furthermore, the method for obtaining the transmission cost of the distributed pattern is as follows: The sum of the products of each element in the optimal transmission coupling matrix and the corresponding element in the geometric cost matrix is ​​taken as the distributional transmission cost.

[0010] Furthermore, the method for obtaining the adaptive sparse weights is as follows: For any row element in the optimal transmission coupling matrix, calculate the negative of the sum of the product of the value of each element in that row and the natural logarithm of the sum of the values ​​of the elements in that row and the sum of the natural logarithms ... Determine the nearest matching coordinate point of the geometric center of each finite element of the target bridge in the preset longitudinal reference coordinate sequence; Calculate the product of the inverse of the mean of the dispersion of all coordinate points and the preset proportional coefficient to obtain the sensitivity adjustment coefficient; The reciprocal of multiplying the distribution dispersion of each matched coordinate point by the sensitivity adjustment coefficient and adding one is used as the adaptive sparse weight of the corresponding finite element.

[0011] Furthermore, the method for obtaining the equivalent cross-section reduction factor is as follows: Construct an objective function that includes the transmission cost of the distribution pattern and a sparse constraint term, wherein the sparse constraint term is the sum of the products of the absolute value of the equivalent cross-section reduction coefficient and the corresponding adaptive sparse weight; Using the proximal gradient descent method, based on the finite difference method, the gradient of the objective function is obtained by applying a preset fixed step size perturbation to the equivalent cross-section reduction coefficient of each unit, and a soft thresholding operation is introduced during the iteration process; the threshold parameter of the soft thresholding operation is the product of the preset learning rate, the preset sparse regularization coefficient and the adaptive sparse weight of the corresponding unit. The equivalent cross-section reduction coefficient of each unit is updated iteratively within a preset physical constraint interval. When the norm difference between two adjacent iterations of the equivalent cross-section reduction coefficient is less than a preset threshold, the equivalent cross-section reduction coefficient of each unit is obtained.

[0012] Furthermore, the method for obtaining the structural bearing capacity reduction rate is as follows: The elastic modulus and yield strength of each finite element in the finite element model are proportionally reduced based on the equivalent section reduction factor. A static load model with the same load as a standard vehicle was applied to the modified finite element model, and the critical load factor was calculated by nonlinear incremental loading throughout the process. A static load model with the same loading conditions is applied to a non-destructive ideal finite element model, and the reference critical load factor is calculated and obtained. Calculate the difference between the reference critical load factor and the critical load factor, and use the ratio of this difference to the reference critical load factor as the structural bearing capacity reduction rate.

[0013] Furthermore, the method for obtaining the critical load factor is as follows: Determine whether the determinant of the overall stiffness matrix of the corrected finite element model is less than or equal to zero during the nonlinear incremental loading calculation. If so, determine that the corrected finite element model has reached the ultimate bearing state, and take the load factor of the corresponding static load model as the critical load factor. Alternatively, it can be determined whether the displacement increment of the node with the largest displacement increment in each incremental step of the modified finite element model exceeds the preset plastic limit threshold, or whether the slope of the tangent line of its load-displacement curve drops to five percent of the initial slope, in the nonlinear incremental loading calculation throughout the entire process. If so, the modified finite element model is determined to have reached the ultimate bearing state, and the load multiple of the corresponding static load model is taken as the critical load factor.

[0014] The present invention has the following beneficial effects: This invention first obtains the second-order difference features of measured strain based on measured strain signals, significantly amplifying the structural deformation distortion features caused by local stiffness degradation. Then, based on the measured strain second-order difference features, it obtains measured feature vectors, and through normalization processing, removes amplitude interference caused by unknown vehicle weight, extracting a dimensionless morphological fingerprint determined solely by the current relative stiffness distribution of the structure. Further, based on wheelbase parameters, axle load ratio, and a finite element model, it obtains simulation feature vectors, rigorously reproducing the theoretical deformation morphology under the same vehicle loading conditions in digital space. To overcome the waveform spatial phase shift problem caused by local damage, it further obtains the optimal transmission coupling matrix based on the measured and simulation feature vectors, combined with the geometric cost matrix, quantifying the minimum spatial transport scheme required to completely reshape the simulated waveform energy distribution into the measured waveform energy distribution. To extract the dimensionless morphological fingerprint from this transmission scheme... The system uses quantitative morphological difference indicators to automatically identify suspected damage areas. Based on the distribution of elements in the optimal transmission coupling matrix, it obtains the morphological transmission cost and adaptive sparse weights, and uses transmission entropy values ​​to identify spatial locations of severe waveform mismatches. To achieve accurate inversion in the absence of prior knowledge of damage locations, it uses iterative changes in the morphological transmission cost and the penalty constraints imposed by the adaptive sparse weights on each element of the target bridge to obtain the equivalent cross-sectional reduction coefficient for each element. This coefficient is physically equivalent to the relative loss ratio of the effective load-bearing cross-section of the component. Based on the equivalent cross-sectional reduction coefficient, the elastic modulus and yield strength of the finite element model are updated simultaneously to obtain the structural bearing capacity reduction rate. Through nonlinear ultimate bearing capacity analysis, the system intuitively quantifies the degree of weakening of the overall structural safety reserve by local minor damage, providing a clear basis for bridge operation and maintenance decisions. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic flowchart illustrating a method for local structural hazard analysis using digital twins, provided as an embodiment of the present invention; Figure 2 A structural diagram of a local structural hazard analysis system using digital twins is provided as an embodiment of the present invention. Figure 3 This is a schematic diagram of a computer device provided according to an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a local structural hazard analysis method using digital twins proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a local structural hazard analysis method using digital twins provided by this invention.

[0020] Example 1: The application scenario of this embodiment is the assessment of local structural damage mainly caused by tensile fatigue cracking (such as the U-ribs and weld joints of orthotropic steel bridge decks), and does not involve overall instability caused by large-scale geometric deformation.

[0021] This invention proposes a method for local structural hazard analysis using digital twins. Please refer to [link / reference]. Figure 1 The diagram illustrates a schematic flowchart of a method for local structural hazard analysis using digital twins, provided by an embodiment of the present invention. The method includes the following steps: Step S1: Acquire the measured strain signals of a standard vehicle crossing the bridge, the wheelbase parameters of the standard vehicle, and the axle load ratio, which are simultaneously collected by multiple strain sensors placed at multiple key stress points of the target bridge.

[0022] Specifically, this embodiment uses a long-span suspension bridge with orthotropic steel bridge deck as the target bridge for analysis, and all subsequent bridges refer to this bridge. It is known that in reality, conventional traffic loads are highly random, multi-vehicle traffic generates complex signal interference, and the absolute total weight of vehicles crossing the bridge is often unknown, making it difficult to accurately distinguish whether changes in strain response amplitude are due to external load fluctuations or a decrease in local stiffness within the bridge structure. To avoid interference from random traffic flow and ensure the uniqueness and certainty of the input benchmark for subsequent damage identification, this embodiment sets the data collection scenario as a single standard vehicle (e.g., a standard five-axle heavy truck) crossing the bridge during periods of sparse traffic (such as nighttime). Multiple strain sensors deployed at key stress points on the target bridge (such as the bottom plate of the orthotropic slab mid-span section or the lower edge of the U-rib) simultaneously collect the measured strain signals of the standard vehicle crossing the bridge, recording the actual temporal deformation response of the structure during loading.

[0023] Furthermore, although the total weight of the standard vehicle is unknown, the wheelbase distribution and axle load distribution of the multi-axle vehicle directly determine the spatial geometric characteristics of the strain influence line. In order to ensure that the response waveform generated by the subsequent finite element simulation is rigorously comparable to the measured waveform in terms of topological morphology (such as the number of peaks and wavelength distribution), the vehicle type of the standard vehicle is identified by a video recognition system deployed at the bridge entrance and exit. The corresponding wheelbase parameters and axle load ratio are then extracted from the pre-stored vehicle standard parameter library. The wheelbase parameters and axle load ratio of the standard vehicle are obtained as an indispensable prior physical condition for the subsequent construction of the multi-axle moving load train.

[0024] Step S2: Obtain the second-order difference feature of the measured strain based on the measured strain signal, and obtain the measured feature vector based on the second-order difference feature of the measured strain; obtain the simulation feature vector based on the shaft distance parameter, shaft weight ratio and finite element model.

[0025] Specifically, the known measured strain signals reflect the temporal dynamic deformation response of the structure during the passage of a standard vehicle. However, the directly measured strain values ​​are often mixed with high-frequency environmental noise and are not sensitive enough to minor local damage to the structure. In order to filter out random noise interference, accurately map the temporal response to a unified spatial coordinate sequence, and significantly amplify the deformation distortion characteristics caused by local stiffness degradation, this embodiment obtains the second-order difference features of measured strain based on the measured strain signals. This accurately reflects the relative deformation intensity of each stress section along the longitudinal direction of the target bridge and highlights the abnormal response at potential damage locations. In order to completely eliminate the amplitude interference caused by the unknown absolute total weight of the vehicle and remove the direct influence of load size on damage identification, the measured feature vector is obtained based on the second-order difference features of measured strain. The dimensionless morphological index determined only by the current relative stiffness distribution of the target bridge is extracted, making it a spatial probability distribution target benchmark independent of the absolute value of the external load. Considering that only structural responses excited under the same multi-axle arrangement of vehicles have direct topological geometrical comparison significance, in order to construct a theoretical reference object in digital space that is in the same physical dimension and domain of definition as the measured feature vectors and has rigorous geometric comparability, simulation feature vectors are obtained based on wheelbase parameters, axle load ratios, and finite element models. In this embodiment, an initial finite element model is constructed based on the design parameters of the target bridge. This model rigorously reproduces the traversal loading process of the corresponding unit multi-axis moving load column and performs the same feature extraction and normalization operations, thereby generating a source distribution vector that dynamically updates with the model parameters. This lays a reliable data alignment foundation for subsequent calculation of waveform space offset costs. The specific construction of the finite element model is well-known and will not be elaborated further.

[0026] Preferably, in one feasible embodiment, the method for obtaining the second-order difference features of the measured strain is as follows: First, considering that high-frequency environmental noise such as vehicle excitation and wind load in the actual bridge monitoring environment will severely obscure the minute strain features, the measured strain signal of each strain sensor is smoothed by polynomial smoothing filtering algorithm through Savitzky-Golay to obtain the corrected measured strain signal. While preserving the overall smooth trend of the response waveform, it effectively suppresses random interference spikes, which is beneficial for subsequent stable high-order differential calculations. The Savitzky-Golay polynomial smoothing filtering algorithm is well known and will not be described in detail here. To eliminate the mismatch between the time-series signal and the spatial position of the structure caused by fluctuations in the speed of standard vehicles and the fixed time step of data acquisition, the time axis is scaled and aligned by correcting the starting point of the measured strain signal on the bridge and the landing point after the bridge. That is, the actual time of the signal on the bridge and the time of the bridge landing are mapped to the theoretical start and end times of the vehicle passing through the bridge at the average speed. The time-series coordinates between the two are linearly scaled proportionally. Then, based on the passing time of the standard vehicle and the bridge length, combined with the average vehicle speed, the corrected measured strain signals of each strain sensor after time axis scaling and alignment are mapped onto a preset longitudinal reference coordinate sequence through linear interpolation. A spatial domain strain response matrix is ​​constructed with the sensor as the row and the coordinate point as the column, ensuring that the data point corresponds one-to-one with the physical position of the bridge structure. The specific operation process is as follows: Based on the average speed of the standard vehicle and the aligned timestamp, the estimated longitudinal coordinates of the standard vehicle on the bridge at each sampling time are calculated. Then, through linear weighted interpolation between two adjacent sampling points, the corrected measured strain values ​​falling on each discrete point in the preset longitudinal reference coordinate sequence are extracted. It should be noted that the method for obtaining the preset longitudinal reference coordinate sequence is as follows: Based on the actual monitoring area length of the target bridge, the longitudinal position of the bridge is discretized with a fixed step size (e.g., 0.5 meters, without limitation), and a sequence containing K coordinate points is constructed. ,in, The starting position of the monitoring area, The termination position, and ;in, The step size is fixed. This sequence serves as a universal spatial reference throughout the entire process, used to unify the physical domain of measured and simulated data, ensuring strict alignment between the two in spatial dimensions. To significantly enhance the sensitivity of structural deformation curves to identifying local stiffness abrupt changes and overcome the sluggish response of single strain amplitude to damage, a second-order central difference operation is performed on the spatial domain strain response matrix along a preset longitudinal reference coordinate sequence (a well-known process, not elaborated here). Specifically, the difference operator is used to calculate the approximate second derivative of the measured strain value at three adjacent coordinate points, obtaining the corresponding second-order difference value for each coordinate point. It should be noted that for coordinate points in the preset longitudinal reference coordinate sequence without effective sensor coverage in their neighborhood, their second-order difference value is recorded as zero. The sequence of second-order difference values ​​for each coordinate point constructed in this way constitutes the measured strain second-order difference characteristic. This indicator highlights the spatial singularity of the strain field caused by local stiffness abrupt changes and has extremely high sensitivity to local stiffness reduction. The larger the absolute value of the measured strain second-order difference characteristic, the more severe the abrupt change in the spatial distribution of strain near that location, suggesting the presence of stiffness damage or cross-sectional weakening in that region.

[0027] Preferably, in one feasible method of this embodiment, the method for obtaining the measured feature vector is as follows: Considering that in a static state with no vehicle load passing by or weak signal, the directly calculated normalized features are easily filled by background noise, causing the subsequent inversion algorithm to be inaccurate or even diverge due to the extremely low signal-to-noise ratio, in order to ensure that only signal segments containing effective structural stress response participate in feature extraction, this embodiment sets a preset noise threshold of 5 microstrains to ensure that only effective strain signals that significantly exceed the background white noise level are captured. The implementer can set the preset noise threshold according to the base noise level of the bridge field sensor and the amplitude of environmental vibration, which is not limited here; When the maximum amplitude of the measured strain second-order difference feature is not greater than the preset noise threshold, it indicates that the current response signal is mainly composed of environmental random vibration or sensor base noise and does not contain effective vehicle load excitation information. In order to prevent the pure noise signal from being forcibly normalized and generating a false shape distribution without physical meaning, it is determined to be an invalid signal and the analysis is terminated, so as to avoid the subsequent Sinkhorn algorithm from diverging the inversion results or generating false alarms due to the alignment of false waveforms; When the maximum amplitude of the measured strain second-order difference feature is greater than the preset noise threshold, it indicates that a heavy vehicle has indeed passed by at the current moment and has caused a sufficiently significant structural deformation response. In order to further highlight the distortion characteristics of the local strain field and eliminate the difference between the positive and negative signs of the measured values, the measured strain second-order difference features corresponding to all strain sensors at each coordinate point are squared and summed to obtain the first feature value of each coordinate point, which accurately reflects the degree of local deformation energy concentration. Among them, the larger the first feature value, the smaller the relative stiffness near the coordinate position and the more significant the deformation anomaly. To completely eliminate the interference from the unknown absolute weight of the vehicle, a dimensionless morphological index determined solely by the stiffness distribution of the structure itself is constructed. The sum of the first eigenvalues ​​of all coordinate points with non-zero second-order difference values ​​is then added to a first preset minimum positive number, which is taken as the overall eigenvalue. The ratio of the first eigenvalue of each coordinate point to the overall eigenvalue is then taken as the relative deformation energy proportion of each coordinate point, ensuring that this eigenvector possesses the mathematical property of a probability distribution function (a sum equal to 1). In this embodiment, the first preset minimum positive number is set to... To ensure stable numerical calculations even in extremely small difference ranges, the denominator is always non-zero. The implementer can set the first preset minimum positive number based on sensor accuracy and signal-to-noise ratio; no limitation is imposed here. To facilitate subsequent global morphological alignment and topological comparison with finite element simulation results, the relative deformation energy proportions of each coordinate point are arranged in the order of a preset longitudinal reference coordinate sequence to construct a measured feature vector, which serves as a spatial distribution benchmark characterizing the current actual stiffness state of the target bridge.

[0028] Preferably, in one feasible embodiment of this method, the simulation feature vector is obtained as follows: To ensure that the simulated waveform is geometrically and topologically comparable to the measured waveform, a unit moving load column with the same wheelbase parameters and axle load ratio as the standard vehicle is applied to the finite element model of the target bridge, and the total weight of the moving load is normalized to 1 to strictly replicate the multi-axle loading mode of the measured vehicle; To accurately reproduce the entire structural stress process of the vehicle passing over the bridge in digital space, the unit moving load column is controlled to move step by step along a preset longitudinal reference coordinate sequence, and static analysis is performed at each loading position to extract the simulated strain data corresponding to the measured strain signal acquisition position, maintaining strict spatial alignment between the simulated sampling points and the measured working conditions. To extract dimensionless features that are in the same physical dimension as the measured feature vector and are not affected by absolute amplitude, the simulation strain data is sequentially smoothed, second-order central difference is calculated, and the second-order central difference values ​​of all corresponding simulation measurement points are squared and normalized according to the method of obtaining the relative deformation energy ratio. This yields the simulation strain second-order difference feature distribution at each coordinate point, accurately reflecting the spatial distribution of theoretical deformation energy under the current finite element model parameter settings. To facilitate subsequent calculation of the morphological difference cost between the simulation distribution and the measured distribution based on the optimal transmission theory, the simulation strain second-order difference feature distribution at each coordinate point is arranged in the order of a preset vertical reference coordinate sequence to construct the simulation feature vector.

[0029] Step S3: Based on the measured feature vector and the simulated feature vector, and combined with the geometric cost matrix, obtain the optimal transmission coupling matrix; according to the distribution of elements in the optimal transmission coupling matrix, obtain the distribution-shaped transmission cost and the adaptive sparse weight; based on the iterative change of the distribution-shaped transmission cost and the penalty constraint of the adaptive sparse weight on each element of the target bridge, obtain the equivalent cross-section reduction coefficient of each element.

[0030] Specifically, it is known that local stiffness damage not only alters the response amplitude but also causes a slight shift in the spatial phase of the response waveform. This leads to significant spurious errors in traditional point-to-point Euclidean distance calculations due to peak misalignment. The method for obtaining Euclidean distance is a well-known technique and will not be elaborated upon further. To overcome the inversion difficulties caused by this morphological difference and achieve accurate alignment of the overall topological features of the waveform, this embodiment obtains the optimal transmission coupling matrix based on the measured feature vector and the simulated feature vector, combined with the geometric cost matrix. This quantifies the minimum spatial transmission scheme required to "transfer" the simulated deformation energy distribution to the measured deformation energy distribution. To extract quantitative morphological difference indicators from this transmission scheme and identify potential damage-sensitive areas, the distribution morphological transmission cost and adaptive sparse weight are obtained based on the distribution of elements in the optimal transmission coupling matrix. The distribution morphological transmission cost directly reflects the overall waveform similarity, while the adaptive sparse weight automatically marks areas with severe waveform morphological mismatch using the entropy value (i.e., energy dispersion degree) of the optimal transmission coupling matrix. To accurately locate damaged units and quantify their stiffness loss degree through iterative optimization in the absence of prior knowledge of the damage location, the equivalent cross-sectional reduction coefficient of each unit is obtained based on the iterative change of the distribution morphological transmission cost and the penalty constraint of the adaptive sparse weight on each unit of the target bridge. This coefficient, by minimizing the distribution morphological transmission cost and following weighted sparse regularization, is physically equivalent to the relative reduction ratio of the effective stress-bearing cross-sectional area of ​​each unit.

[0031] Preferably, in one feasible embodiment, the optimal transmission coupling matrix is ​​obtained as follows: To quantify the basic topological cost of energy transfer between different physical locations, a geometric cost matrix is ​​first constructed using each coordinate point in a preset longitudinal reference coordinate sequence as a row and column, respectively; wherein each element in the geometric cost matrix is ​​the square of the Euclidean distance between each pair of coordinate points in the preset longitudinal reference coordinate sequence divided by the maximum value among all squares of Euclidean distances. This normalization operation constrains the elements of the cost matrix to... The interval effectively prevents numerical underflow (calculated as 0) due to excessively large exponent terms when calculating the kernel matrix in the subsequent calculation. At the same time, it clearly defines the distance penalty cost required to move waveform features from one spatial coordinate to another. To address the waveform spatial misalignment problem caused by local damage, an optimal mapping relationship for morphological transformation is established. Based on the Sinkhorn algorithm, using the measured eigenvectors as the target distribution and the simulated eigenvectors as the source distribution, and under the premise of conserving total energy in the probability distribution, a non-negative transmission coupling matrix that minimizes transmission cost is sought. This matrix represents the optimal energy transport scheme for reshaping the simulated waveform into the measured waveform. Entropy regularization constraints are introduced into the search process, transforming the computationally complex linear programming problem into a strictly convex optimization problem to improve solution efficiency and ensure the uniqueness of the solution. A fixed-point iterative algorithm is used to alternately update the diagonal scaling vector until the norm difference between two iterations is less than a preset convergence threshold, indicating that the diagonal scaling vector has reached a stable state and the algorithm has converged to the optimal solution. The product of the updated diagonal scaling vector and the kernel matrix based on the geometric cost matrix is ​​then calculated to obtain the optimal transmission coupling matrix. In this embodiment, the preset convergence threshold is set to... To ensure that fixed-point iteration achieves high-precision shape alignment without falling into an infinite loop due to minor floating-point operation errors, implementers can set a preset convergence threshold based on actual computing hardware resources and alignment accuracy requirements; no restrictions are imposed here. The Sinkhorn algorithm, entropy regularization, and kernel matrix acquisition methods are all well-known techniques and will not be elaborated further.

[0032] Preferably, in one feasible embodiment, the method for obtaining the distribution morphology transmission cost is as follows: considering that only by quantifying the global waveform morphology differences can the risk of misjudgment caused by local peak phase misalignment be overcome, the sum of the products of each element in the optimal transmission coupling matrix and the corresponding element in the geometric cost matrix is ​​taken as the distribution morphology transmission cost. The distribution morphology transmission cost numerically represents the minimum weighted distance total cost required to completely reshape the energy distribution of the simulated waveform into the energy distribution of the measured waveform under the optimal transmission scheme. Even when the waveform has spatial translation or scaling deformation, it can still maintain good monotonicity and convexity, providing a stable and reliable loss function for subsequent inversion iteration.

[0033] Preferably, in one feasible embodiment, the method for obtaining adaptive sparse weights is as follows: Considering that in order to automatically distinguish between damaged and healthy regions during damage inversion and suppress spurious parameter fluctuations in non-damaged regions, it is necessary to establish a spatial weighting mechanism that relies solely on the characteristics of the data form itself. Therefore, for any row element in the optimal transmission coupling matrix, the negative of the sum of the product of the values ​​of each element in that row and their natural logarithms plus a second preset minimum positive number is calculated to obtain the distribution dispersion of the coordinate point corresponding to that row, accurately reflecting the intensity of the spatial diffusion of simulated energy at that location to match the measured energy. In this embodiment, the second preset minimum positive number is set as... To prevent numerical overflow during logarithmic operations due to the presence of zero elements in the optimal transmission coupling matrix, implementers can set a second preset minimum positive number based on actual conditions; no limitation is imposed here. The formula for calculating the distribution dispersion is: In the formula, denoted by , where is the dispersion of the coordinates corresponding to the i-th row of the optimal transmission coupling matrix; K is the number of columns in the optimal transmission coupling matrix. This represents the element in the i-th row and k-th column of the optimal transmission coupling matrix; is the second preset minimum positive number; ln is the logarithmic function with the natural constant as the base. The greater the dispersion of the distribution, the more the energy at that location needs to be dispersed and transmitted to a farther or wider area to match the measured data, suggesting that there may be a damage source nearby that causes drastic changes in waveform morphology; To map the aforementioned coordinate-point-based distribution dispersion back to the physical elements of the finite element model, and further determine the nearest matching coordinate point of the geometric center of each finite element element of the target bridge within a preset longitudinal reference coordinate sequence, a correspondence is established from the mathematical feature space to the physical structure space. Simultaneously, the product of the inverse mean of the distribution dispersion of all coordinate points and a preset proportional coefficient is calculated to obtain a sensitivity adjustment coefficient. This coefficient is used to dynamically adjust the spatial resolution of the weights, ensuring that the weights... The interval has good distinguishability, which is conducive to forming low-weight valleys in high dispersion (suspected damage) areas and high-weight platforms in low dispersion (healthy) areas. In this embodiment, a preset scaling factor is set to a constant between 5 and 15 (e.g., 10) to ensure that the response of the weight function to the local dispersion peak is neither too flat nor too steep. The implementer can set the size of the preset scaling factor according to the resolution requirements of the inversion target. There is no limitation here. The larger the sensitivity adjustment coefficient, the more sensitive it is to the response to the fluctuation of the distribution dispersion. The higher the polarization of the weight, that is, the weight in the suspected damage area (slightly high dispersion) will be quickly reduced to close to zero, while the weight in the healthy area (low dispersion) will remain at a high level, thereby producing a sparser and more focused damage localization result in the inversion process. To construct the target weight vector for the final constraint inversion, the reciprocal of the product of the distribution dispersion of each matching coordinate point and the sensitivity adjustment coefficient plus one is used as the adaptive sparse weight of the corresponding finite element. This allows the penalty weight to be automatically reduced in the suspected damage area with large distribution dispersion, allowing for larger parameter updates; while in the healthy area with low distribution dispersion, the weight approaches 1, thereby effectively suppressing spurious fluctuations in the inversion parameters.

[0034] Preferably, in one feasible embodiment, the method for obtaining the equivalent cross-section reduction coefficient is as follows: In order to accurately locate the damaged unit and quantify its stiffness loss through optimization in the absence of prior knowledge of the damage location, an objective function containing the distribution pattern transmission cost and sparse constraint term is constructed so that the inversion result satisfies the spatial sparsity assumption while minimizing waveform morphology differences; wherein, the sparse constraint term is the sum of the product of the absolute value of the equivalent cross-section reduction coefficient and the corresponding adaptive sparse weight. This term uses the adaptive weight obtained above to reduce the penalty intensity in the suspected damaged area and apply a high penalty in the healthy area, thereby guiding the inversion algorithm to converge to the physically real sparse solution; To address the difficulty in gradient acquisition caused by the non-differentiability of the transmission cost term in the objective function, a near-end gradient descent method is employed. Based on the finite difference method, this involves applying a preset fixed-step perturbation to the equivalent cross-section reduction coefficient of each element (the preset fixed step size ranges from [value missing]). to To avoid computational deadlock caused by the perturbation always being zero when the initial value is zero, the forward simulation is rerun to obtain the change in the objective function, thereby approximating the gradient of the objective function. Gradient descent updates are then performed, and a soft thresholding operation is introduced during the iteration process. The threshold parameter of the soft thresholding operation is the product of a preset learning rate, a preset sparse regularization coefficient, and the adaptive sparse weights of the corresponding unit. In this embodiment, the preset learning rate is set to a range of 0.01 to 0.1, and the preset sparse regularization coefficient is set to a range of 0.001 to 0.01. The soft thresholding operation subtracts this threshold parameter from the value updated by gradient descent; if the result is less than zero, it is truncated to zero, thus achieving sparse constraints on the parameters and ensuring non-negativity in physical meaning. Simultaneously, to prevent computational collapse due to singularity in the stiffness matrix of the finite element model, the equivalent section reduction coefficient after the soft thresholding operation is forcibly restricted to a preset physical constraint range. Within this framework, ensure that the element retains at least 5% numerical stiffness even under extreme damage. This is achieved when the norm difference between two adjacent iterations of the equivalent section reduction factor is less than a preset threshold (e.g., ...). When the value reaches 0, it indicates that the algorithm has converged to a stable solution, and at this point, the equivalent cross-sectional reduction coefficient of each element is obtained.

[0035] Step S4: Based on the equivalent section reduction factor, simultaneously update the elastic modulus and yield strength of the finite element model to obtain the structural bearing capacity reduction rate.

[0036] Specifically, a larger equivalent section reduction factor indicates a more severe loss of effective load-bearing cross-sectional area at that unit location. This not only leads to an elastic decrease in the local stiffness of the structure but also significantly reduces the ultimate stress level at which the component enters plastic yielding due to the reduced load-bearing area. To accurately simulate the entire process of nonlinear mechanical performance degradation caused by local section damage, the elastic modulus and yield strength of the finite element model are updated synchronously based on the equivalent section reduction factor, constructing a modified finite element model that is highly equivalent to the actual damage state in terms of physical mechanism. To intuitively quantify the macroscopic impact of local minor damage on the overall structural safety reserve, nonlinear limit loading analysis is performed based on this modified model to obtain the rate of decrease in structural bearing capacity. This indicator comprehensively considers the redistribution capacity of the statically indeterminate internal forces of the structure, providing a quantitative basis with clear engineering physical significance for bridge operation and maintenance decisions.

[0037] Preferably, in one feasible embodiment, the method for obtaining the structural bearing capacity reduction rate is as follows: Considering that local stiffness damage (such as cracks or corrosion) is macroscopically equivalent to a reduction in the effective cross-sectional area of ​​the component, and that axial stiffness and ultimate bearing capacity are both linearly proportional to the cross-sectional area, in order to accurately simulate this damage mechanism while maintaining the finite element mesh topology, the elastic modulus and yield strength of each finite element in the finite element model are proportionally reduced based on the equivalent cross-sectional reduction factor, that is, the elastic modulus of the element is updated to... The yield strength has been updated to This introduces the dual physical effects of elastic stiffness degradation and plastic limit reduction into the model; where E is the elastic modulus. For the updated elastic modulus; Yield strength; The updated yield strength; d is the equivalent section reduction factor of the corresponding finite element obtained from the above steps, and its value represents the relative loss ratio of the effective stress-bearing cross-sectional area of ​​the element; To investigate the extent to which local damage weakens the overall structural safety under extreme load conditions, a static load model with the same load as a standard vehicle was applied to a modified finite element model. Through full-process nonlinear incremental loading (such as the arc-length method), the entire process of the structure from elastic operation, local yielding, to overall failure was simulated, and the critical load factor was calculated to accurately reflect the maximum ultimate load multiple that the structure can withstand under the current damage state. Furthermore, a static load model with the same loading conditions was applied to an undamaged ideal finite element model, and a reference critical load factor was calculated as a theoretical safety reserve benchmark for the structure under design conditions. To provide bridge maintenance departments with intuitive and quantitative structural safety early warning indicators, the difference between the reference critical load factor and the critical load factor was calculated. The ratio of this difference to the reference critical load factor was used as the structural bearing capacity reduction rate, accurately reflecting the relative proportion of the loss in the overall ultimate bearing capacity of the structure caused by local stiffness reduction. A larger structural bearing capacity reduction rate indicates a lower remaining safety reserve and a higher risk of brittle fracture or instability, requiring immediate load limiting or reinforcement measures.

[0038] Preferably, in one feasible method of this embodiment, the critical load factor is obtained as follows: First, it is determined whether the determinant value of the overall stiffness matrix of the corrected finite element model is less than or equal to zero during the nonlinear incremental loading calculation of the whole process; if so, it indicates that the structural system has lost static equilibrium stability and entered the state of mechanical motion or instability. Then, it is determined that the corrected finite element model has reached the ultimate bearing state, and the load multiple of the corresponding static load model at this time is taken as the critical load factor, and the mathematical limit point of the structural bearing capacity is recorded. Alternatively, in the nonlinear incremental loading calculation throughout the entire process, it can be determined whether the displacement increment of the node with the largest displacement increment in each incremental step of the corrected finite element model exceeds the preset plastic limit threshold, or whether the slope of the tangent line of its load-displacement curve drops to five percent of the initial slope. If so, it indicates that a plastic hinge has formed locally in the structure and the deformation rate has increased sharply, rendering it no longer suitable for bearing loads. In this case, the corrected finite element model is determined to have reached the ultimate bearing state, and the load multiple of the corresponding static load model is taken as the critical load factor, serving as the engineering control point for the structural bearing capacity. In this embodiment, the preset plastic limit threshold is set to 5 times the displacement response in the elastic stage to ensure that the judgment standard has a clear physical meaning and conforms to engineering practices. Implementers can set the preset plastic limit threshold themselves according to the structural ductility design requirements; this is not limited here.

[0039] In summary, this embodiment acquires the measured strain signal and vehicle parameters of a standard vehicle crossing a bridge, and extracts the normalized measured feature vector. Simultaneously, it generates isomorphic simulation feature vectors based on the vehicle parameters and the finite element model. The Sinkhorn algorithm is used to calculate the optimal transmission coupling matrix between the two, obtaining the distributional transmission cost and adaptive sparse weights to resolve the identification error caused by waveform spatial phase shift. By minimizing the transmission cost and applying sparse constraints, the equivalent cross-sectional reduction coefficient of each element is inverted, achieving accurate local stiffness damage location. Finally, the model material parameters are updated synchronously, and the structural bearing capacity reduction rate is calculated through nonlinear full-process analysis. This invention can accurately quantify the impact of local damage on the overall structural safety under conditions of unknown vehicle weight and waveform shift.

[0040] Example 2: This invention also proposes a local structural hazard analysis system using digital twins; please refer to [link / reference]. Figure 2 The diagram illustrates a structural diagram of a local structural hazard analysis system using digital twins provided by an embodiment of the present invention. The system includes: a data acquisition module 10, a data analysis module 20, an equivalent section reduction coefficient acquisition module 30, and a structural bearing capacity reduction rate acquisition module 40.

[0041] The data acquisition module 10 is used to acquire the measured strain signals of a standard vehicle crossing the bridge, the wheelbase parameters of the standard vehicle, and the axle load ratio, which are synchronously collected by multiple strain sensors arranged at multiple key stress points of the target bridge.

[0042] The data analysis module 20 is used to obtain the second-order difference features of the measured strain based on the measured strain signal, and to obtain the measured feature vector based on the second-order difference features of the measured strain; and to obtain the simulation feature vector based on the wheelbase parameter, the wheelbase weight ratio and the finite element model.

[0043] The equivalent cross-section reduction coefficient acquisition module 30 is used to obtain the optimal transmission coupling matrix based on the measured feature vector and the simulated feature vector, combined with the geometric cost matrix; to obtain the distribution morphology transmission cost and adaptive sparse weight according to the distribution of elements in the optimal transmission coupling matrix; and to obtain the equivalent cross-section reduction coefficient of each element based on the iterative change of the distribution morphology transmission cost and the penalty constraint of the adaptive sparse weight on each element of the target bridge.

[0044] The structural bearing capacity reduction rate acquisition module 40 is used to obtain the structural bearing capacity reduction rate by synchronously updating the elastic modulus and yield strength of the finite element model based on the equivalent section reduction factor.

[0045] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the local structural hazard analysis system using digital twins and the local structural hazard analysis method using digital twins provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0046] Example 3: This invention also proposes a device for local structural hazard analysis using digital twins. The device includes a memory and a processor. The memory stores executable program code, and the processor calls and executes the executable program code to perform a local structural hazard analysis method using digital twins provided in the embodiments of this application. Specifically, the device may be a chip, component, or module. The chip may include a connected processor and memory; the memory stores instructions, and when the processor calls and executes the instructions, the chip can perform the local structural hazard analysis method using digital twins provided in the above embodiments.

[0047] Furthermore, this application also protects a computer device; please refer to [link to relevant documentation]. Figure 3 The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402. When the processor 402 executes the computer program 403, the computer device is able to perform any of the aforementioned methods for local structural hazard analysis using digital twins.

[0048] Example 4: This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the aforementioned method steps to implement the local structural hazard analysis method using digital twins provided in the above embodiment.

[0049] Example 5: This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to implement the local structural hazard analysis method using digital twins provided in the above embodiment.

[0050] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0051] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0052] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for local structural hazard analysis using digital twins, characterized in that, The method includes the following steps: Acquire the measured strain signals, wheelbase parameters, and axle load ratio of a standard vehicle crossing the bridge, which are simultaneously collected by multiple strain sensors placed at multiple key stress points on the target bridge. The measured strain second-order difference features are obtained based on the measured strain signals, and the measured feature vector is obtained based on the measured strain second-order difference features; the simulation feature vector is obtained based on the wheelbase parameter, wheelbase weight ratio and finite element model. Based on measured and simulated feature vectors, combined with the geometric cost matrix, the optimal transmission coupling matrix is ​​obtained; according to the distribution of elements in the optimal transmission coupling matrix, the distribution-shaped transmission cost and adaptive sparse weight are obtained; based on the iterative change of the distribution-shaped transmission cost and the penalty constraint of the adaptive sparse weight on each element of the target bridge, the equivalent cross-section reduction coefficient of each element is obtained. Based on the equivalent section reduction factor, the elastic modulus and yield strength of the finite element model are updated simultaneously to obtain the rate of decrease in structural bearing capacity.

2. The method for local structural hazard analysis using digital twins as described in claim 1, characterized in that, The method for obtaining the measured strain second-order difference characteristics is as follows: The measured strain signal of each strain sensor is subjected to polynomial smoothing filtering to obtain the corrected measured strain signal; The time axis is scaled and aligned by using the bridge start point and bridge fall point of the corrected measured strain signal. Based on the passing time of the standard vehicle and the bridge length, combined with the average vehicle speed, the corrected measured strain signals of each strain sensor after time axis scaling and alignment are mapped to the preset longitudinal reference coordinate sequence by linear interpolation, and a spatial domain strain response matrix with the sensor as the row and the coordinate point as the column is constructed. The spatial domain strain response matrix is ​​subjected to a second-order central difference operation along a preset longitudinal reference coordinate sequence to obtain the second-order difference value of each coordinate point. For coordinate points in the preset longitudinal reference coordinate sequence that are not covered by effective sensors in their neighborhood, their second-order difference value is recorded as zero. The obtained second-order difference values ​​are used as the measured strain second-order difference characteristics.

3. The method for local structural hazard analysis using digital twins as described in claim 2, characterized in that, The method for obtaining the measured feature vector is as follows: When the maximum amplitude of the measured second-order difference characteristic of strain is not greater than the preset noise threshold, it is determined to be an invalid signal and the analysis is terminated. When the maximum amplitude of the measured strain second-order difference feature is greater than the preset noise threshold, the measured strain second-order difference features corresponding to all strain sensors at each coordinate point are squared and summed to obtain the first feature value of each coordinate point. The sum of the first eigenvalues ​​of all coordinate points whose second-order difference values ​​are not zero, plus the first preset minimum positive number, is taken as the overall eigenvalue. The ratio of the first feature value of each coordinate point to the overall feature value is taken as the relative deformation energy percentage of each coordinate point; The relative deformation energy percentages of each coordinate point are arranged in the order of a preset longitudinal reference coordinate sequence to construct the measured feature vector.

4. The method for local structural hazard analysis using digital twins as described in claim 3, characterized in that, The method for obtaining the simulation feature vector is as follows: Apply a unit moving load column to the finite element model of the target bridge, with the same wheelbase parameters and axle load ratio as the standard vehicle. The unit moving load column is controlled to move gradually along a preset longitudinal reference coordinate sequence to extract simulated strain data corresponding to the actual strain signal acquisition position; The simulated strain data were sequentially smoothed, filtered, and subjected to second-order central difference calculation. The second-order difference values ​​of all corresponding simulated measurement points were then summed by squares and normalized to obtain the simulated strain second-order difference characteristic distribution for each coordinate point. The simulated strain second-order difference feature distributions at each coordinate point are arranged in the order of a preset longitudinal reference coordinate sequence to construct the simulated feature vector.

5. The method for local structural hazard analysis using digital twins as described in claim 4, characterized in that, The method for obtaining the optimal transmission coupling matrix is ​​as follows: A geometric cost matrix is ​​constructed using each coordinate point in the preset vertical reference coordinate sequence as the row and column, respectively; where each element in the geometric cost matrix is ​​the square of the Euclidean distance between each pair of coordinate points in the preset vertical reference coordinate sequence divided by the maximum value among all squares of Euclidean distances; Based on the Sinkhorn algorithm, using the measured feature vectors as the target distribution and the simulated feature vectors as the source distribution, we search for a non-negative transmission coupling matrix that minimizes the transmission cost. An entropy regularization constraint is introduced into the search process, and the diagonal scaling vector is updated alternately through a fixed-point iterative algorithm until the norm difference between two iterations is less than a preset convergence threshold. The optimal transmission coupling matrix is ​​obtained by multiplying the updated diagonal scaling vector with the kernel matrix based on the geometric cost matrix.

6. The method for local structural hazard analysis using digital twins as described in claim 5, characterized in that, The method for obtaining the transmission cost of the distribution pattern is as follows: The sum of the products of each element in the optimal transmission coupling matrix and the corresponding element in the geometric cost matrix is ​​taken as the distributional transmission cost.

7. The method for local structural hazard analysis using digital twins as described in claim 6, characterized in that, The method for obtaining the adaptive sparse weights is as follows: For any row element in the optimal transmission coupling matrix, calculate the negative of the sum of the product of the value of each element in that row and the natural logarithm of the sum of the values ​​of the elements in that row and the sum of the natural logarithms ... Determine the nearest matching coordinate point of the geometric center of each finite element of the target bridge in the preset longitudinal reference coordinate sequence; Calculate the product of the inverse of the mean of the dispersion of all coordinate points and the preset proportional coefficient to obtain the sensitivity adjustment coefficient; The reciprocal of multiplying the distribution dispersion of each matched coordinate point by the sensitivity adjustment coefficient and adding one is used as the adaptive sparse weight of the corresponding finite element.

8. The method for local structural hazard analysis using digital twins as described in claim 1, characterized in that, The method for obtaining the equivalent cross-section reduction factor is as follows: Construct an objective function that includes the transmission cost of the distribution pattern and a sparse constraint term, wherein the sparse constraint term is the sum of the products of the absolute value of the equivalent cross-section reduction coefficient and the corresponding adaptive sparse weight; Using the proximal gradient descent method, based on the finite difference method, the gradient of the objective function is obtained by applying a preset fixed step size perturbation to the equivalent cross-section reduction coefficient of each unit, and a soft thresholding operation is introduced during the iteration process; the threshold parameter of the soft thresholding operation is the product of the preset learning rate, the preset sparse regularization coefficient and the adaptive sparse weight of the corresponding unit. The equivalent cross-section reduction coefficient of each unit is updated iteratively within a preset physical constraint interval. When the norm difference between two adjacent iterations of the equivalent cross-section reduction coefficient is less than a preset threshold, the equivalent cross-section reduction coefficient of each unit is obtained.

9. The method for local structural hazard analysis using digital twins as described in claim 1, characterized in that, The method for obtaining the structural bearing capacity reduction rate is as follows: The elastic modulus and yield strength of each finite element in the finite element model are proportionally reduced based on the equivalent section reduction factor. A static load model with the same load as a standard vehicle was applied to the modified finite element model, and the critical load factor was calculated by nonlinear incremental loading throughout the process. A static load model with the same loading conditions is applied to a non-destructive ideal finite element model, and the reference critical load factor is calculated and obtained. Calculate the difference between the reference critical load factor and the critical load factor, and use the ratio of this difference to the reference critical load factor as the structural bearing capacity reduction rate.

10. The method for local structural hazard analysis using digital twins as described in claim 9, characterized in that, The method for obtaining the critical load factor is as follows: Determine whether the determinant of the overall stiffness matrix of the corrected finite element model is less than or equal to zero during the nonlinear incremental loading calculation. If so, determine that the corrected finite element model has reached the ultimate bearing state, and take the load factor of the corresponding static load model as the critical load factor. Alternatively, it can be determined whether the displacement increment of the node with the largest displacement increment in each incremental step of the modified finite element model exceeds the preset plastic limit threshold, or whether the slope of the tangent line of its load-displacement curve drops to five percent of the initial slope, in the nonlinear incremental loading calculation throughout the entire process. If so, the modified finite element model is determined to have reached the ultimate bearing state, and the load multiple of the corresponding static load model is taken as the critical load factor.

Images