A method and system for dynamically identifying a main failure mode of a half-through arch bridge

Abstract

The application provides a method and system for dynamically identifying a main failure mode of a half-through arch bridge, which comprises: obtaining real-time monitoring signals and a prior sample library during a service period; extracting a main trend item of the signals through a heuristic swarm intelligence filtering algorithm, and reconstructing a high signal-to-noise ratio observation vector; calculating a spatial correlation matrix of the observation vector, and dividing a potential failure space into a plurality of physical failure sub-regions; calculating a numerical deviation and a cosine similarity of a measured vector and a predicted vector in a full time range, constructing a time average sequential likelihood function which fuses a numerical magnitude and a topological morphological feature, and updating posterior probabilities of each failure mode; locking a target sub-region with the highest aggregated confidence through two-stage hierarchical identification, and normalizing the posterior probabilities in the target sub-region to output a target failure component. The method can strip environmental drift and strong noise interference, overcome likelihood distortion and near neighbor confusion, and realize high-robust and accurate identification of a main failure mode in a complex environment.

Description

Technical Field

[0001] This application relates to the field of electrical data processing technology, specifically to a method and system for dynamic identification of main failure modes of a through-arch bridge. Background Technology

[0002] As a typical composite structural system, the cable system of a mid-span arch bridge is a key force transmission hub connecting the rigid arch ribs and the flexible bridge deck system. Under complex environmental conditions and loads, it is extremely prone to damage or even severe damage.

[0003] However, existing methods for dynamic identification of main failure modes in through-arch bridges have significant drawbacks in practical engineering: the original monitoring signals collected on-site are often mixed with low-frequency drift of ambient temperature, vehicle load impact, and non-Gaussian burst noise. Directly calculating the traditional Euclidean distance likelihood function based on the original non-stationary signal is prone to serious distortion of the likelihood function due to the overall translation or local distortion of the residual vector. Summary of the Invention

[0004] This application provides a method and system for dynamic identification of the main failure modes of a through-arch bridge. It can automatically isolate low-frequency drift caused by solar radiation temperature difference, vehicle impact, and high-frequency noise from sensors, effectively avoiding the numerical drift and likelihood distortion problems that are easily generated when the traditional Euclidean distance likelihood function is directly calculated based on strong noise non-stationary signals.

[0005] A first aspect of this application provides a method for dynamic identification of the main failure modes of a through-arch bridge, the method comprising: Acquire real-time monitoring signals of the bridge structure under service conditions and call up a pre-generated prior sample library of structural failure modes; Adaptive time-frequency decomposition is performed on the real-time monitoring signal to extract steady-state feature components that reflect the actual stress state of the structure and reconstruct a high signal-to-noise ratio observation vector; Based on the reconstructed observation vectors, the correlation spectrum of the structural response in the spatial distribution is calculated, and the potential failure space of the full bridge is divided into multiple physical failure sub-regions according to the correlation spectrum. A sequential likelihood function integrating numerical magnitude features and topological morphology features is constructed, and the posterior probability of each failure mode is updated by Bayesian inference in combination with the prior sample library. Based on the posterior probability and each of the physical failure sub-regions, the target physical failure sub-region with the highest aggregation confidence is locked in the first-level identification. Within the locked target physical failure sub-region, a second level of identification is performed, the posterior probability of each specific failure component within the sub-region is renormalized, and the target failure component with the highest probability is output.

[0006] In one possible implementation, the step of performing adaptive time-frequency decomposition on the real-time monitoring signal to extract steady-state feature components reflecting the true stress state of the structure includes: A heuristic swarm intelligence filtering algorithm is used to iteratively decompose the original signal; Separate low-frequency environmental drift components from high-frequency random noise components; The main trend term with concentrated energy and significant mean is automatically selected from the intrinsic modal components obtained by decomposition using comprehensive evaluation indicators, and is used as the steady-state characteristic component.

[0007] In one possible implementation, the iterative decomposition of the original signal using a heuristic swarm intelligence filtering algorithm includes: The original monitoring signal is discretized into an input sequence, and the iteration parameters are initialized. Perform energy spectral density analysis on the current input sequence to estimate the dominant frequency characteristics of the current input sequence; Based on the dominant frequency characteristics, the group size parameter and flexibility time interval parameter of the group filter are adaptively calculated; Based on the population size parameter and the flexibility time interval parameter, a population filtering operation is performed on the current input sequence to obtain the filtered output. Calculate the iteration deviation between the current filter output and the previous filter output. If the deviation is greater than a preset threshold, then use the current filter output as a new input sequence. When the iterative deviation converges to within a preset threshold, the current filtered output is extracted as a single oscillation component, and this component is stripped from the original signal. The above process is repeated on the residual signal until the energy of the residual signal is lower than the preset threshold.

[0008] In one possible implementation, calculating the correlation map of the structural response in spatial distribution based on the reconstructed observation vector includes: Calculate the Pearson correlation coefficient between the response vectors of each monitoring point after reconstruction, and construct a spatial correlation matrix; Extract the blocky diagonalized distribution features from the spatial correlation matrix; Based on the blocky diagonalized distribution characteristics, adjacent monitoring points with similar stress-cooperative behavior are aggregated into the same physical failure sub-region.

[0009] In one possible implementation, constructing the sequential likelihood function that integrates numerical magnitude features and topological morphological features includes: For full-time observation data, the absolute numerical deviation between the predicted value of the prior sample and the measured observation vector at each time point is calculated as a numerical magnitude feature; Calculate the cosine similarity between the two vectors at each time step and convert it into a distance metric as a topological feature; We construct a sequential likelihood function based on the time-averaged distance by performing weighted cumulative averaging of the numerical magnitude features and topological features over the entire time period.

[0010] In one possible implementation, the sequential likelihood function based on the time-averaged distance is as follows:

[0011] In the formula, For the measured observation vector, For the first Types of failure modes This represents the total number of frames observed throughout the entire time period. For the first The numerical magnitude of the feature distance of the frame, For the first The distance of the topological morphological features of the frame. This is the morphological weight adjustment factor.

[0012] In one possible implementation, the topological morphological features are measured using vector cosine similarity. By calculating the cosine value of the angle between the predicted vector and the measured vector, and mapping it to a distance interval within a preset range, the influence of overall vector scaling and translation on morphological comparison is eliminated.

[0013] In one possible implementation, the step of identifying the target physical failure sub-region with the highest aggregate confidence in the first-level identification based on the posterior probability includes: Extract the top-ranking candidate failure modes with the highest posterior probabilities across the entire bridge and form a candidate set; For each physical failure sub-region, the sum of the probabilities of failure modes falling within the candidate set is calculated and used as the aggregate score for that physical failure sub-region. The physical failure sub-region with the highest aggregation score is selected as the target physical failure sub-region.

[0014] In one possible implementation, the second-level identification within the locked target physical failure sub-region includes: Using the target physical failure sub-region as a condition, the posterior probabilities of all failure modes within the sub-region are calculated using conditional probability normalization, so that the sum of the conditional posterior probabilities of each failure mode within the sub-region is 1. Based on the normalized conditional posterior probability, the failure component with the highest probability value is determined as the final target failure component.

[0015] This example provides a dynamic identification method for the main failure modes of a through-arch bridge. First, real-time monitoring signals of the bridge structure under service conditions are acquired, and a pre-generated prior sample library of structural failure modes is called. Then, an adaptive time-frequency decomposition of the real-time monitoring signals is performed using a heuristic swarm intelligence filtering algorithm to remove low-frequency environmental drift and high-frequency random noise, extracting the main trend term with concentrated energy and significant mean as a steady-state feature component, and reconstructing a high signal-to-noise ratio observation vector. Next, based on the reconstructed observation vector, a spatial correlation matrix is ​​calculated, and block diagonalized distribution features are extracted to divide the potential failure space of the entire bridge into multiple [variables / modes]. The scheme identifies physical failure sub-regions exhibiting coordinated stress behavior. It then calculates the absolute numerical deviation and cosine similarity between measured and predicted vectors over the entire timescale, constructing a weighted time-averaged sequential likelihood function that integrates numerical magnitude and topological features. This function, combined with a prior sample library, updates the posterior probability of each failure mode using Bayesian inference. Finally, a two-stage hierarchical identification process is performed. The first stage identifies the target physical failure sub-region with the highest aggregation confidence by statistically summing the probabilities falling within the candidate set. The second stage re-normalizes the conditional posterior probability within this target physical failure sub-region and outputs the target failed component with the highest probability. This scheme offers the following advantages: 1. By employing a heuristic swarm intelligence filtering algorithm to perform adaptive iterative decomposition on the original monitoring signal and combining it with the energy-mean comprehensive index to screen the main trend term, it can automatically remove low-frequency drift caused by solar radiation temperature difference, vehicle impact and high-frequency noise from sensors. This effectively avoids the numerical drift and likelihood distortion problems that are easily generated when the traditional Euclidean distance likelihood function is directly calculated based on strong noise non-stationary signals, and significantly improves the signal-to-noise ratio of the input observation vector and the restoration of the true state of structural stress.

[0016] 2. By constructing a weighted sequential likelihood function that integrates numerical absolute deviation and cosine similarity distance, the geometric and topological features are captured by utilizing the invariance of cosine similarity to the overall numerical scaling and translation of vectors. Combined with the cumulative averaging effect of full-time evidence, this overcomes the defect that the magnitude of residual vectors is difficult to distinguish under strong noise background by a single Euclidean distance. It can effectively suppress the likelihood misleading caused by high-amplitude non-Gaussian abrupt changes and local noise interference, and significantly widen the posterior probability gradient gap between the target component and the highly correlated interference term.

[0017] 3. By calculating the block diagonalization feature of the spatial correlation matrix, the full bridge is divided into physical failure sub-regions. A two-stage hierarchical inference mechanism of "first-level macroscopic aggregation localization and second-level conditional normalization precise localization within sub-regions" is adopted to converge the high-dimensional global search to the target physical cluster. This breaks the computational bottleneck caused by the traditional blind global search, completely eliminates the probability confusion and residual interference of neighboring components in non-target regions, and achieves high-confidence, high-robust dynamic locking and accurate identification of the main failure mode under complex and noisy environments.

[0018] A second aspect of this application provides a dynamic identification system for the main failure modes of a through-arch bridge, the system comprising: The first module is used to acquire real-time monitoring signals of the bridge structure in service and call a pre-generated prior sample library of structural failure modes. The second module is used to perform adaptive time-frequency decomposition on the real-time monitoring signal, extract steady-state feature components that reflect the actual stress state of the structure, and reconstruct a high signal-to-noise ratio observation vector. The third module is used to calculate the correlation spectrum of the structural response in the spatial distribution based on the reconstructed observation vector, and to divide the potential failure space of the full bridge into multiple physical failure sub-regions according to the correlation spectrum. The fourth module is used to construct a sequential likelihood function that integrates numerical magnitude features and topological morphology features. Combined with the prior sample library, the posterior probability of each failure mode is updated through Bayesian inference. The fifth module is used to identify the target physical failure sub-region with the highest aggregate confidence in the first-level identification based on the posterior probability. The sixth module is used to perform a second-level identification within the locked target physical failure sub-region, renormalize the posterior probability of each specific failure component within the sub-region, and output the target failure component with the highest probability.

[0019] A third aspect of this application provides a terminal including a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute the steps of the method for dynamic identification of main failure modes of under-bearing arch bridges as described in the first aspect of this application.

[0020] A fourth aspect of this application provides a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the dynamic identification method for main failure modes of a through-arch bridge in the first aspect of this application.

[0021] A fifth aspect of this application provides a computer program product, comprising a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the dynamic identification method for main failure modes of a through-arch bridge in the first aspect of this application. The computer program product may be a software installation package. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This application provides a schematic diagram of the overall process for a dynamic identification method of the main failure modes of a through-arch bridge. Figure 2 This application provides a finite element simulation data cable force spatial correlation heatmap for a dynamic identification method of the main failure modes of a through-arch bridge. Figure 3 This application provides a method for dynamic identification of the main failure modes of a through-arch bridge, using a heat map of the spatial correlation of cable forces based on the original monitoring data. Figure 4 This application provides a heatmap of the spatial correlation between cable forces and the SWD reconstruction trend term in a dynamic identification method for the main failure modes of a through-arch bridge. Figure 5 This application provides a schematic diagram of the sub-region division of the physical failure of the suspension cables of a through-arch bridge, which is a method for dynamic identification of the main failure modes of a through-arch bridge. Figure 6 This application provides a schematic diagram of the overall structure of a dynamic identification system for the main failure modes of a through-arch bridge. Figure 7 This application provides a schematic diagram of the structure of a terminal. Figure label: Module 1-1, Module 2-2, Module 3-3, Module 4-4, Module 5-5, Module 6-6. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0026] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0027] A dynamic identification method for the main failure modes of through-arch bridges is applied to a dynamic identification system for the main failure modes of through-arch bridges. Figure 1 A schematic diagram of the overall process for a dynamic identification method of main failure modes in a through-arch bridge is shown. Figure 1 As shown, it includes: S1. Obtain real-time monitoring signals of the bridge structure under service conditions and call the pre-generated prior sample library of structural failure modes.

[0028] Specifically, the process involves acquiring real-time monitoring signals of the bridge structure under capture conditions and calling a pre-generated prior sample library of structural failure modes. First, a sensor network (such as cable force sensors, strain gauges, and negative pressure sensors) installed on the bridge structure is used to collect real-time monitoring signal sequences of the bridge under capture conditions, generating the original monitoring signal sequence. Simultaneously, a pre-generated prior sample library of structural modes based on the time-varying reliability model and Monte Carlo simulation is called. This sample library contains the structural physical state, theoretical response support, and failure mode labels of the bridge under different capture periods and damage states.

[0029] S2. Perform adaptive time-frequency decomposition on the real-time monitoring signal, extract steady-state feature components that reflect the actual stress state of the structure, and reconstruct a high signal-to-noise ratio observation vector.

[0030] Step S2 includes the following sub-steps: A1. The original signal is iteratively decomposed using a heuristic swarm intelligence filtering algorithm.

[0031] Step A1 includes the following sub-steps: The original monitoring signal is discretized into an input sequence, and the iteration parameters are initialized. A101. Perform energy spectral density analysis on the current input sequence to estimate the dominant frequency characteristics of the current input sequence.

[0032] First, the current input sequence is calculated. Energy spectral density To eliminate random high-frequency small fluctuations in the energy spectral density, a Savitzky-Golay (SG) filter is used. Smoothing is performed to obtain the smoothed energy spectral density. Subsequently, an extreme value search threshold was set. In smooth energy spectral density The search energy is greater than the preset extreme value search threshold. The maximum spectral peak is determined, and the frequency corresponding to this maximum spectral peak is defined as the dominant frequency feature (i.e., center frequency) of the current input sequence. Main frequency characteristics The calculation formula is as follows:

[0033] In the formula, The dominant frequency feature of the current input sequence, Angular frequency, The smoothed energy spectral density, This is the preset extreme value search threshold (0.1 in this example).

[0034] A102. Based on the dominant frequency characteristics, adaptively calculate the group size parameter and flexibility time interval parameter of the group filter.

[0035] Among them, based on the estimated main frequency characteristics By using a pre-defined empirical fitting relationship equation, the population size parameter required for the population filter at the current iteration level is adaptively mapped and calculated. With flexibility time interval parameter Group size parameters The calculation formula is as follows:

[0036] In the formula, This is the group size parameter (i.e., the number of group members) for the group filter. This is a function for rounding to the nearest odd number. The dominant frequency characteristics estimated in step A101.

[0037] Flexibility time interval parameter The calculation formula is as follows:

[0038] In the formula, The time interval parameter is the flexibility parameter for the group filter. It is a primary frequency characteristic.

[0039] A103. Based on the population size parameter and the flexibility time interval parameter, perform a population filtering operation on the current input sequence to obtain the filtered output.

[0040] Among them, the current input sequence As the driving force for the target prey, initialization Members of a group; at various times The velocity and position of each group member are updated by calculating the driving force and the cohesion / repulsion within the group. Finally, the positions of all group members are weighted and synthesized to obtain the output sequence of the current sub-group filtering. .

[0041] driving force The calculation formula is as follows:

[0042] In the formula, For the first Members of the group The driving force at all times For the current input sequence in The value at time, For the first Members of the group The position at any given moment.

[0043] Cohesion and Repulsion The calculation formula is as follows:

[0044]

[0045] In the formula, For the first Members of the group The combined forces of group cohesion and repulsion that it constantly experiences. For group size parameters, and The first The and the first Members of the group Location at any given moment The interaction force function between members, The relative distance between members For symbolic functions, This is the preset critical distance for interactions between members.

[0046] Group member speed The update formula is as follows:

[0047] In the formula, For the first Members of the group The speed of time For the first Members of the group The speed of time For the flexibility time interval parameter, As the driving force, It is the combined force of cohesion and repulsion.

[0048] Group member position The update formula is as follows:

[0049] In the formula, For the first Members of the group Location at any given moment For the first Members of the group Location at any given moment For the flexibility time interval parameter, For the first Members of the group The speed of time.

[0050] Group filter output sequence The synthesis formula is as follows:

[0051] In the formula, For internal use Subgroup filtering in The output sequence at time 10:00. For group size parameters, For the first Members of the group Location at any given moment This is the preset output weighted contribution coefficient.

[0052] A104. Calculate the iteration deviation between the current filter output and the previous filter output. If the deviation is greater than a preset threshold, then use the current filter output as a new input sequence.

[0053] Among them, the current filter output is calculated. Compared with the previous filter output The energy normalization relative error between them is used as the iteration bias. ; to adjust the iteration deviation With preset convergence threshold If a comparison is made, Then update the internal iteration count to And update the current filtered output to the new input sequence (i.e. Return to step A103 to continue the group filtering operation. Iteration bias The calculation formula is as follows:

[0054] In the formula, This represents the iteration deviation between the current filter output and the previous filter output. For internal use Subgroup filtering in The output sequence at time 10:00. For internal use Subgroup filtering in The output sequence at time (when) hour, This is the initial input for step A103. ), This is the index for the sampling time point.

[0055] A105. When the iterative deviation converges to within a preset threshold, the current filtered output is extracted as a single oscillation component, and this component is stripped from the original signal. The above process is repeated on the residual signal until the energy of the residual signal is lower than the preset threshold.

[0056] Among them, when the iteration deviation At that time, extract the current filtered output. As the first in the current iteration level Each intrinsic mode component Then the component is removed from the current input sequence. After subtraction, the updated residual signal sequence is obtained. ; for residual signal sequences The smoothed energy spectrum is evaluated, and if its maximum peak energy is still greater than the preset extreme value search threshold, the result is determined. Then the input sequence is updated as follows: The hierarchical index is updated to Reset the internal iteration count Return to step A101 to re-estimate the dominant frequency and begin the extraction of the next intrinsic mode component; if the maximum peak of the smoothed energy spectrum of the residual signal is less than Then the decomposition terminates. Each intrinsic mode component The extraction formula is as follows:

[0057] In the formula, For the extracted first A sequence of intrinsic mode components. For convergence, the innermost... The output sequence of the subgroup filter.

[0058] Residual signal sequence The update formula is as follows:

[0059] In the formula, The next residual signal sequence after stripping the current component. The input sequence of the current layer before stripping. For the first A sequence of intrinsic mode components.

[0060] A2. Separate low-frequency environmental drift components from high-frequency random noise components.

[0061] Specifically, separating low-frequency environmental drift components from high-frequency random noise components involves iteratively extracting all intrinsic mode components through step A1. (in After removing the final residual components, the high-frequency random noise components with extremely high frequency and low energy (such as the first few intrinsic mode components corresponding to the highest frequency band) and the extremely low-frequency residual components that characterize long-period trends such as changes in ambient temperature are removed from the signal, leaving the mode set with significant force response characteristics.

[0062] A3. The main trend term with concentrated energy and significant mean is automatically selected from the intrinsic modal components obtained by decomposition using comprehensive evaluation indicators, and is used as the steady-state characteristic component.

[0063] Specifically, the main trend term with concentrated energy and significant mean is automatically selected from the intrinsic modal components obtained by decomposition using a comprehensive evaluation index, and is used as the steady-state characteristic component. This involves calculating the intrinsic modal components extracted in step A1. absolute mean With energy norm Therefore, a comprehensive evaluation index is constructed. ; Traverse all intrinsic mode components obtained from the decomposition, and select the comprehensive evaluation index. The first to reach the maximum value Each intrinsic mode component is used as an effective principal trend term to reconstruct a high signal-to-noise ratio steady-state feature observation vector. .

[0064] Comprehensive evaluation indicators The calculation formula is as follows:

[0065] In the formula, For the first A comprehensive evaluation index for each intrinsic modal component; For the first The absolute mean of each intrinsic modal component; For the first The energy norm of each intrinsic mode component; The total time history length (number of sampling points) of the discrete signal; The first decomposition obtained The intrinsic mode components in the th eigenmode component at the th The value of each sampling point.

[0066] In this example, the energy-mean maximization criterion is introduced to break the limitations of traditional screening that relies solely on a single energy contribution rate. It can automatically and accurately lock the reference mode containing the main static stress information of the structure from non-stationary signals containing strong DC components, and completely eliminate high-frequency random interference with zero mean, providing a steady-state data foundation with high signal-to-noise ratio for subsequent Bayesian updates.

[0067] S3. Based on the reconstructed observation vector, calculate the correlation spectrum of the structural response in the spatial distribution, and divide the potential failure space of the full bridge into multiple physical failure sub-regions according to the correlation spectrum.

[0068] Figure 2 This application provides a finite element simulation data cable force spatial correlation heatmap for a dynamic identification method of the main failure modes of a through-arch bridge. Figure 3 This application provides a method for dynamic identification of the main failure modes of a through-arch bridge, using a heat map of the spatial correlation of cable forces based on the original monitoring data. Figure 4 This application provides a heatmap of the spatial correlation between cable forces and the SWD reconstruction trend term in a dynamic identification method for the main failure modes of a through-arch bridge. Figures 2 to 4 As shown, step S3 includes the following sub-steps: B1. Calculate the Pearson correlation coefficient between the response vectors of each monitoring point after reconstruction, and construct the spatial correlation matrix.

[0069] Specifically, the Pearson correlation coefficient between the response vectors of each monitoring point after reconstruction is calculated, and a spatial correlation matrix is ​​constructed. This involves selecting the data reconstructed from step A3. The full-time steady-state characteristic observation vectors of each measuring point are used to calculate the Pearson correlation coefficient between the time-history response vectors of different measuring points pairwise. Thus forming a Spatial correlation matrix of dimensions .

[0070] Pearson correlation coefficient The calculation formula is as follows:

[0071] In the formula, For monitoring points With monitoring points Pearson correlation coefficient between response vectors; and monitoring points and In the Steady-state characteristic response values ​​at each time step; and monitoring points and The average response over the entire time period; This represents the total number of sampling points throughout the entire time period.

[0072] B2. Extract the block diagonalized distribution features from the spatial correlation matrix.

[0073] Specifically, extracting the blocky diagonalized distribution features from the spatial correlation matrix involves: [the process described in step B1]... Perform spectral feature analysis and matrix heatmap imaging to identify local dense regions in the matrix with significantly high correlation coefficients (close to 1) and extract highly correlated sub-block structures distributed along the diagonal.

[0074] B3. Based on the blocky diagonalized distribution characteristics, adjacent monitoring points with similar stress-cooperative behavior are aggregated into the same physical failure sub-region.

[0075] Figure 5 This application provides a schematic diagram illustrating the sub-region division of physical failure of the suspension cables in a through-arch bridge, based on a dynamic identification method for the main failure modes of through-arch bridges. Figure 5 As shown, combining the blocky diagonalized distribution characteristics and the topology of the arch bridge cable, the potential failure space of the entire bridge is divided into multiple non-overlapping physical failure sub-regions, namely, the mid-span regions C1 and C2, the quarter-span regions C3 and C4, and the arch foot region C5.

[0076] Specifically, based on the blocky diagonalization distribution characteristics, adjacent monitoring points with similar stress-cooperative behavior are aggregated into the same physical failure sub-region. This involves combining the blocky diagonalization sub-block boundaries extracted in step B2 with the physical spatial topology of the mid-span arch bridge (such as the mid-span long cable zone, the 1 / 4-span mid-short cable zone, the arch foot short cable zone, etc.), and then spatially reducing the dimensionality of all potential failure modes of the entire bridge to [the desired area]. Non-overlapping physical failure sub-regions .

[0077] In this example, spatial partitioning based on physical correlation changes the traditional global blind search mode that treats each component as an independent classification object. It solves the engineering problem that neighboring components with adjacent physical locations and highly similar force patterns are easily confused in a noisy environment, and significantly reduces the inference search space.

[0078] S4. Construct a sequential likelihood function that integrates numerical magnitude features and topological morphology features, and update the posterior probability of each failure mode by Bayesian inference in combination with the prior sample library.

[0079] Step S4 includes the following sub-steps: C1. For full-time observation data, calculate the absolute numerical deviation between the predicted value of the prior sample and the measured observation vector at each time point, and use it as a numerical magnitude feature.

[0080] Specifically, the absolute numerical deviation between the predicted value of the prior sample and the observed vector at each time step is calculated: for the first time step over the entire time period... frame( ) Observation data, extract the measured steady-state characteristic observation vector reconstructed in step A3. And call the first sample from the prior sample library Failure modes Corresponding finite element simulation prediction vector ; Calculate the sum of squares of the differences between the measured and predicted values ​​at all monitoring points at each point, and then standardize and normalize the weighted average by introducing the measurement error variance of each monitoring point to obtain the result. Frame 1 Numerical magnitude characteristic distance of various failure modes .

[0081] Numerical scale feature distance The calculation formula is as follows:

[0082] In the formula, For the first Frame time number The numerical magnitude characteristic distance of each failure mode; For the first Frame time Measured steady-state characteristic response values ​​at each monitoring point; For the first The first failure mode Finite element simulation prediction values ​​for each monitoring point; This represents the total number of monitoring points (22 in this example). For the first The standard deviation of sensor measurement error at each monitoring point.

[0083] C2. Calculate the cosine similarity of the two vectors in each time step and convert it into a distance metric as a topological feature.

[0084] Among them, topological morphological features are measured using vector cosine similarity. This is achieved by calculating the cosine value of the angle between the predicted vector and the measured vector, and mapping it to a distance interval within a preset range, in order to eliminate the influence of overall vector scaling and translation on morphological comparison. Specifically, the cosine similarity between the two vectors at each time step is calculated and converted into a distance metric. At frame time, the measured steady-state characteristic observation vector With the Prediction vectors of various failure modes The vectors are treated as high-dimensional space vectors; the cosine of the angle between the two vectors is calculated to quantify their geometric similarity in spatial topological directions; then, they are mapped to a distance interval. In the middle, we obtained the first Frame 1 Topological morphological feature distance of various failure modes .

[0085] Topological morphological feature distance The calculation formula is as follows:

[0086] In the formula, For the first Frame time Topological morphological feature distance of each failure mode; For the first The measured steady-state characteristic observation vector at frame time; For the first Finite element simulation prediction vectors for various failure modes; This is the vector dot product operator; Let be the Euclidean norm (modulus) of the vector.

[0087] C3. Perform a weighted time cumulative average on the numerical magnitude features and topological features over the entire time period to construct a sequential likelihood function based on the time average distance.

[0088] Specifically, a weighted time-cumulative average is performed on the numerical magnitude features and topological morphological features over the entire time span to construct a sequential likelihood function based on the time-averaged distance. This involves introducing a morphological weight adjustment factor. For the entire time period (total) The numerical magnitude feature distance calculated for each frame within a frame. Distance from topological features Perform linear weighted fusion; then perform full-time fusion. The weighted distances of the frames are accumulated and summed, then divided by the total number of frames. A time-dimensional averaging process is performed; finally, a sequential likelihood function based on the time-averaged distance is constructed in negative exponential form. .

[0089] Sequential likelihood function based on time-averaged distance The calculation formula is as follows:

[0090] In the formula, In the first Failure modes Full-time observation data appears under certain conditions. The sequential likelihood; For the first The first failure mode (corresponding to the first failure mode in the prior sample library) (samples) This represents the total number of frames (sampling points) observed throughout the entire time period. For the first The numerical magnitude characteristic distance at each frame time; For the first The distance of topological features at frame time; The shape weight adjustment factor (in this example, the value is [value]). ).

[0091] S5. Based on the posterior probability and each of the physical failure sub-regions, lock the target physical failure sub-region with the highest aggregation confidence in the first-level identification.

[0092] Step S5 includes the following sub-steps: D1. Extract multiple candidate failure modes with the highest posterior probability across the entire bridge range to form a candidate set.

[0093] Specifically, a candidate failure mode set is formed by extracting multiple candidate failure modes with the highest posterior probabilities across the entire bridge. This involves: after calculating the updated posterior probability distribution of all potential failure modes across the entire bridge using Bayesian inference in step S4, all failure modes are sorted in descending order of their posterior probabilities; the top K high-probability terms are set as a screening threshold, and the top K failure modes are selected. This set of K failure modes is defined as the candidate failure mode set. .

[0094] D2. For each physical failure sub-region, the sum of the probabilities of failure modes falling within the candidate set is calculated and used as the aggregate score for that physical failure sub-region.

[0095] Specifically, for each physical failure sub-region, the sum of the probabilities of failure modes falling within the candidate set is calculated and used as the aggregate score for that physical failure sub-region. This is achieved by iterating through each physical failure sub-region divided in step S3. ( ); for the current physical failure sub-region Find those that simultaneously belong to this sub-region And included in the candidate failure mode set All intersecting failure modes are identified; the normalized posterior probabilities of these intersecting failure modes are summed to calculate the physical failure sub-region. aggregate score .

[0096] Physical Failure Sub-region Aggregate Score The calculation formula is as follows:

[0097] In the formula, For the first Physical failure sub-regions The aggregate score, Ranked by global posterior probability The set of candidate failure modes for a given bit (in this example) (Value is 5) To belong to the A set of failure modes for each physical failure sub-region For the first Types of failure modes For the first Failure modes in full-time observation data Normalized posterior probability under given conditions.

[0098] D3. Select the physical failure sub-region with the highest aggregation score as the target physical failure sub-region.

[0099] Specifically, the physical failure sub-region with the highest aggregation score is selected as the target physical failure sub-region. This involves comparing the aggregation scores of all physical failure sub-regions. Search and extract the physical failure sub-regions that maximize the aggregation score, and lock them as the final target physical failure sub-regions. .

[0100] Target physical failure sub-region The locking formula is as follows:

[0101] In the formula, For the locked target physically failed sub-region, For the first Aggregate scores of each physically failed sub-region To find the sub-region index that maximizes the expression within parentheses. The function.

[0102] S6. Within the locked target physical failure sub-region, perform a second-level identification, renormalize the posterior probability of each specific failure component within the sub-region, and output the target failure component with the highest probability.

[0103] Step S6 includes the following sub-steps: E1. Taking the target physical failure sub-region as a condition, perform conditional probability normalization calculation on the posterior probability of all failure modes within the sub-region, so that the sum of the conditional posterior probabilities of each failure mode within the sub-region is 1.

[0104] Among them, the target physical failure sub-region is identified and locked at the first level. As a condition, the failure space is strictly converged from the global domain to the target physical failure sub-region. The initial normalized posterior probabilities of each failure mode within the target physical failure sub-region are obtained after the full-time update. Based on the definition of conditional probability, the initial normalized posterior probability of each failure mode within the target physical failure sub-region is divided by the sum of the initial normalized posterior probabilities of all failure modes within the target physical failure sub-region. This achieves dynamic redistribution of sample weights and normalization of conditional probability calculation within the sub-space. The corrected formula for calculating the conditional posterior probability is as follows:

[0105] In the formula, To lock the target physical failure sub-region Under the condition of the first Failure modes The corrected conditional posterior probability; For the first Failure modes in full-time observation data Globally normalized posterior probability under given conditions; This is the first level of identification to lock the target physically failed sub-region; To belong to the target physical failure sub-region The internal first Types of failure mode samples; For the target physical failure sub-region The sum of the globally normalized posterior probabilities of all failure modes.

[0106] E2. Based on the normalized conditional posterior probability, the failure component with the highest probability value is determined as the final target failure component.

[0107] Among them, traversing the target physical failure sub-region Conditional posterior probabilities of all internal failure modes after normalization calculation Within the target sub-region, numerical comparisons are performed to search for and pinpoint the failure mode that maximizes the conditional posterior probability globally. The specific cable component number corresponding to this failure mode is extracted and output as the final target failure component. The determination formula is as follows:

[0108] In the formula, The final output should specify the target failed component (e.g., the specific sling number). For the first Failure modes In the target physical failure sub-region The corrected conditional posterior probability under the given conditions; To seek sub-regions that cause physical failure of the target A function of the failure mode / component index corresponding to the maximum value of the posterior probability of internal conditions.

[0109] This example provides a dynamic identification method for the main failure modes of a through-arch bridge. First, real-time monitoring signals of the bridge structure under service conditions are acquired, and a pre-generated prior sample library of structural failure modes is called. Then, an adaptive time-frequency decomposition of the real-time monitoring signals is performed using a heuristic swarm intelligence filtering algorithm to remove low-frequency environmental drift and high-frequency random noise, extracting the main trend term with concentrated energy and significant mean as a steady-state feature component, and reconstructing a high signal-to-noise ratio observation vector. Next, based on the reconstructed observation vector, a spatial correlation matrix is ​​calculated, and block diagonalized distribution features are extracted to divide the potential failure space of the entire bridge into multiple [variables / modes]. The scheme identifies physical failure sub-regions exhibiting coordinated stress behavior. It then calculates the absolute numerical deviation and cosine similarity between measured and predicted vectors over the entire timescale, constructing a weighted time-averaged sequential likelihood function that integrates numerical magnitude and topological features. This function, combined with a prior sample library, updates the posterior probability of each failure mode using Bayesian inference. Finally, a two-stage hierarchical identification process is performed. The first stage identifies the target physical failure sub-region with the highest aggregation confidence by statistically summing the probabilities falling within the candidate set. The second stage re-normalizes the conditional posterior probability within this target physical failure sub-region and outputs the target failed component with the highest probability. This scheme offers the following advantages: 1. By employing a heuristic swarm intelligence filtering algorithm to perform adaptive iterative decomposition on the original monitoring signal and combining it with the energy-mean comprehensive index to screen the main trend term, it can automatically remove low-frequency drift caused by solar radiation temperature difference, vehicle impact and high-frequency noise from sensors. This effectively avoids the numerical drift and likelihood distortion problems that are easily generated when the traditional Euclidean distance likelihood function is directly calculated based on strong noise non-stationary signals, and significantly improves the signal-to-noise ratio of the input observation vector and the restoration of the true state of structural stress.

[0110] 2. By constructing a weighted sequential likelihood function that integrates numerical absolute deviation and cosine similarity distance, the geometric and topological features are captured by utilizing the invariance of cosine similarity to the overall numerical scaling and translation of vectors. Combined with the cumulative averaging effect of full-time evidence, this overcomes the defect that the magnitude of residual vectors is difficult to distinguish under strong noise background by a single Euclidean distance. It can effectively suppress the likelihood misleading caused by high-amplitude non-Gaussian abrupt changes and local noise interference, and significantly widen the posterior probability gradient gap between the target component and the highly correlated interference term.

[0111] 3. By calculating the block diagonalization feature of the spatial correlation matrix, the full bridge is divided into physical failure sub-regions. A two-stage hierarchical inference mechanism of "first-level macroscopic aggregation localization and second-level conditional normalization precise localization within sub-regions" is adopted to converge the high-dimensional global search to the target physical cluster. This breaks the computational bottleneck caused by the traditional blind global search, completely eliminates the probability confusion and residual interference of neighboring components in non-target regions, and achieves high-confidence, high-robust dynamic locking and accurate identification of the main failure mode under complex and noisy environments.

[0112] For those consistent with the above, please refer to Figure 6 , Figure 6 This application provides a schematic diagram of the structure of a dynamic identification system for the main failure modes of a through-arch bridge. (See attached diagram.) Figure 6 As shown, the system includes: The first module 1 is used to acquire real-time monitoring signals of the bridge structure in service and call a pre-generated prior sample library of structural failure modes.

[0113] The second module 2 is used to perform adaptive time-frequency decomposition on the real-time monitoring signal, extract steady-state feature components that reflect the actual stress state of the structure, and reconstruct a high signal-to-noise ratio observation vector.

[0114] The third module 3 is used to calculate the correlation spectrum of the structural response in the spatial distribution based on the reconstructed observation vector, and to divide the potential failure space of the full bridge into multiple physical failure sub-regions according to the correlation spectrum.

[0115] The fourth module 4 is used to construct a sequential likelihood function that integrates numerical magnitude features and topological morphology features. Combined with the prior sample library, the posterior probability of each failure mode is updated through Bayesian inference.

[0116] The fifth module 5 is used to identify the target physical failure sub-region with the highest aggregate confidence in the first-level identification based on the posterior probability.

[0117] The sixth module 6 is used to perform a second-level identification within the locked target physical failure sub-region, renormalize the posterior probability of each specific failure component within the sub-region, and output the target failure component with the highest probability.

[0118] For examples consistent with the above embodiments, please refer to... Figure 7 , Figure 7 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application, such as... Figure 7 As shown, it includes a processor, an input device, an output device, and a memory, which are interconnected. The memory is used to store a computer program, which includes program instructions. The processor is configured to call the program instructions. The program includes instructions for performing the following steps. The system acquires real-time monitoring signals of the bridge structure under service conditions and calls upon a pre-generated prior sample library of structural failure modes.

[0119] The real-time monitoring signal is subjected to adaptive time-frequency decomposition to extract steady-state feature components that reflect the actual stress state of the structure, and a high signal-to-noise ratio observation vector is reconstructed.

[0120] Based on the reconstructed observation vectors, the correlation spectrum of the structural response in spatial distribution is calculated, and the potential failure space of the full bridge is divided into multiple physical failure sub-regions according to the correlation spectrum.

[0121] A sequential likelihood function integrating numerical magnitude features and topological morphology features is constructed, and the posterior probability of each failure mode is updated by Bayesian inference in combination with the prior sample library.

[0122] Based on the posterior probability and each of the physical failure sub-regions, the target physical failure sub-region with the highest aggregate confidence is identified in the first-level identification.

[0123] Within the locked target physical failure sub-region, a second level of identification is performed, the posterior probability of each specific failure component within the sub-region is renormalized, and the target failure component with the highest probability is output.

[0124] The above mainly describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, the terminal includes the corresponding hardware structure and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0125] This application embodiment can divide the terminal into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0126] This application embodiment also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data interchange, the computer program causing a computer to perform some or all of the steps of any of the dynamic identification methods for main failure modes of under-deck arch bridges as described in the above method embodiments.

[0127] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps of any of the dynamic identification methods for main failure modes of under-deck arch bridges as described in the above method embodiments.

[0128] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0129] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0130] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0131] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0132] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software program module.

[0133] If the integrated unit is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0134] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk, etc.

[0135] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for dynamic identification of main failure modes of a through-arch bridge, characterized in that, include: Acquire real-time monitoring signals of the bridge structure under service conditions and call up a pre-generated prior sample library of structural failure modes; Adaptive time-frequency decomposition is performed on the real-time monitoring signal to extract steady-state feature components that reflect the actual stress state of the structure and reconstruct a high signal-to-noise ratio observation vector; Based on the reconstructed observation vectors, the correlation spectrum of the structural response in the spatial distribution is calculated, and the potential failure space of the full bridge is divided into multiple physical failure sub-regions according to the correlation spectrum. A sequential likelihood function integrating numerical magnitude features and topological morphology features is constructed, and the posterior probability of each failure mode is updated by Bayesian inference in combination with the prior sample library. Based on the posterior probability and each of the physical failure sub-regions, the target physical failure sub-region with the highest aggregation confidence is locked in the first-level identification. Within the locked target physical failure sub-region, a second level of identification is performed, the posterior probability of each specific failure component within the sub-region is renormalized, and the target failure component with the highest probability is output.

2. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 1, characterized in that, The adaptive time-frequency decomposition of the real-time monitoring signal to extract steady-state feature components reflecting the true stress state of the structure includes: The original signal is iteratively decomposed using a heuristic swarm intelligence filtering algorithm; Separate low-frequency environmental drift components from high-frequency random noise components; The main trend term with concentrated energy and significant mean is automatically selected from the intrinsic modal components obtained by decomposition using comprehensive evaluation indicators, and is used as the steady-state characteristic component.

3. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 2, characterized in that, The iterative decomposition of the original signal using a heuristic swarm intelligence filtering algorithm includes: The original monitoring signal is discretized into an input sequence, and the iteration parameters are initialized. Perform energy spectral density analysis on the current input sequence to estimate the dominant frequency characteristics of the current input sequence; Based on the dominant frequency characteristics, the group size parameter and flexibility time interval parameter of the group filter are adaptively calculated; Based on the population size parameter and the flexibility time interval parameter, a population filtering operation is performed on the current input sequence to obtain the filtered output. Calculate the iteration deviation between the current filter output and the previous filter output. If the deviation is greater than a preset threshold, then use the current filter output as a new input sequence. When the iterative deviation converges to within a preset threshold, the current filtered output is extracted as a single oscillation component, and this component is stripped from the original signal. The above process is repeated on the residual signal until the energy of the residual signal is lower than the preset threshold.

4. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 1, characterized in that, The calculation of the correlation map of the structural response in spatial distribution based on the reconstructed observation vectors includes: Calculate the Pearson correlation coefficient between the response vectors of each monitoring point after reconstruction, and construct a spatial correlation matrix; Extract the blocky diagonalized distribution features from the spatial correlation matrix; Based on the blocky diagonalized distribution characteristics, adjacent monitoring points with similar stress-cooperative behavior are aggregated into the same physical failure sub-region.

5. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 1, characterized in that, The construction of the sequential likelihood function that integrates numerical magnitude features and topological morphological features includes: For full-time observation data, the absolute numerical deviation between the predicted value of the prior sample and the measured observation vector at each time point is calculated as a numerical magnitude feature; Calculate the cosine similarity between the two vectors at each time step and convert it into a distance metric as a topological feature; We construct a sequential likelihood function based on the time-averaged distance by performing weighted cumulative averaging of the numerical magnitude features and topological features over the entire time period.

6. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 5, characterized in that, The sequential likelihood function based on the time-averaged distance is as follows: In the formula, For the measured observation vector, For the first Types of failure modes This represents the total number of frames observed throughout the entire time period. For the first The numerical magnitude of the feature distance of the frame, For the first The distance of the topological morphological features of the frame. This is the morphological weight adjustment factor.

7. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 5, characterized in that, The topological morphological features are measured using vector cosine similarity. By calculating the cosine value of the angle between the predicted vector and the measured vector, and mapping it to a distance interval within a preset range, the influence of overall vector scaling and translation on morphological comparison is eliminated.

8. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 1, characterized in that, The step of identifying the target physical failure sub-region with the highest aggregate confidence in the first-level identification based on the posterior probability includes: Extract the top-ranking candidate failure modes with the highest posterior probabilities across the entire bridge and form a candidate set; For each physical failure sub-region, the sum of the probabilities of failure modes falling within the candidate set is calculated and used as the aggregate score for that physical failure sub-region. The physical failure sub-region with the highest aggregation score is selected as the target physical failure sub-region.

9. The method for dynamic identification of main failure modes of a through-arch bridge according to claim 1, characterized in that, The second-level identification within the locked target physical failure sub-region includes: Using the target physical failure sub-region as a condition, the posterior probabilities of all failure modes within the sub-region are calculated using conditional probability normalization, so that the sum of the conditional posterior probabilities of each failure mode within the sub-region is 1. Based on the normalized conditional posterior probability, the failure component with the highest probability value is determined as the final target failure component.

10. A dynamic identification system for the main failure modes of a through-arch bridge, characterized in that, include: The first module is used to acquire real-time monitoring signals of the bridge structure in service and call a pre-generated prior sample library of structural failure modes. The second module is used to perform adaptive time-frequency decomposition on the real-time monitoring signal, extract steady-state feature components that reflect the actual stress state of the structure, and reconstruct a high signal-to-noise ratio observation vector. The third module is used to calculate the correlation spectrum of the structural response in the spatial distribution based on the reconstructed observation vector, and to divide the potential failure space of the full bridge into multiple physical failure sub-regions according to the correlation spectrum. The fourth module is used to construct a sequential likelihood function that integrates numerical magnitude features and topological morphology features. Combined with the prior sample library, the posterior probability of each failure mode is updated through Bayesian inference. The fifth module is used to identify the target physical failure sub-region with the highest aggregate confidence in the first-level identification based on the posterior probability. The sixth module is used to perform a second-level identification within the locked target physical failure sub-region, renormalize the posterior probability of each specific failure component within the sub-region, and output the target failure component with the highest probability.

Images