A data analysis method and system for the bearing capacity of a long-span continuous rigid frame bridge

Abstract

The present invention provides a method for analyzing the bearing capacity data of a long-span continuous rigid-frame bridge, including: obtaining the long-term monitoring data sequence of the target bridge, performing a load distribution analysis on the long-term monitoring data sequence to extract the load dynamic characteristics of the bridge, performing a deformation correlation analysis on the long-term monitoring data sequence to extract the deformation coupling characteristics of the bridge, and the deformation coupling characteristics are used to characterize the cooperative deformation relationship between the main girder and the pier of the bridge; generating a comprehensive evaluation index based on the load dynamic characteristics and the deformation coupling characteristics, and the comprehensive evaluation index is used to quantify the degradation trend of the bearing performance of the bridge structure during its entire life cycle; determining the bearing capacity state of the bridge in the current and future service stages according to the comparison result between the comprehensive evaluation index and the preset bearing capacity threshold. The present invention can improve the sensitivity of bearing performance degradation identification and the reliability of prediction results.

Description

Technical Field

[0001] This application relates to the field of data processing, and specifically relates to a method and system for analyzing the bearing capacity data of a long-span continuous rigid-frame bridge. Background Art

[0002] The evaluation of bridge bearing capacity is the core to ensure the safe operation of long-span continuous rigid-frame bridges. The key lies in accurately identifying the degradation trend of bearing performance through the dynamic analysis of structural response data. In the existing technology, static monitoring data is usually used to independently analyze the load distribution characteristics or local deformation indexes respectively, and the bearing capacity state is determined based on a preset safety factor threshold. However, this method has defects: single-point data cannot capture the dynamic evolution law of the bridge's entire life cycle, resulting in a lag in the identification of the degradation trend; the separate analysis of load characteristics and deformation characteristics destroys the mechanical correlation of structural responses and is difficult to accurately reflect the cooperative working state of the main girder and pier columns; the static evaluation index lacks a multi-dimensional feature fusion mechanism, which can neither quantify the spatio-temporal distribution characteristics of the remaining bearing capacity nor establish a mathematical correlation model between the degradation process and the remaining life, resulting in a lack of foresight in maintenance decisions and difficulty in preventing the systemic failure risk caused by the accumulation of local damage. Summary of the Invention

[0003] This application provides a method and system for analyzing the bearing capacity data of a long-span continuous rigid-frame bridge.

[0004] According to one aspect of this application, a method for analyzing the bearing capacity data of a long-span continuous rigid-frame bridge is provided, including: obtaining a long-term monitoring data sequence of a target bridge, where the long-term monitoring data sequence includes multiple structural response data sets of the bridge arranged according to monitoring time nodes during the service period; performing load distribution analysis on the long-term monitoring data sequence to extract the load dynamic characteristics of the bridge, where the load dynamic characteristics are used to characterize the load transfer law of each key section of the bridge at different monitoring time nodes; performing deformation correlation analysis on the long-term monitoring data sequence to extract the deformation coupling characteristics of the bridge, where the deformation coupling characteristics are used to characterize the cooperative deformation relationship between the main girder and pier columns of the bridge; generating a comprehensive evaluation index based on the load dynamic characteristics and the deformation coupling characteristics, where the comprehensive evaluation index is used to quantify the bearing performance degradation trend of the bridge structure during the entire life cycle; and determining the bearing capacity state of the bridge in the current and future service stages according to the comparison result between the comprehensive evaluation index and a preset bearing capacity threshold.

[0005] According to another aspect of this application, a data analysis system is provided, including: at least one processor; and a memory communicatively connected to the at least one processor; where the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is caused to implement the above method.

[0006] The bridge bearing capacity data analysis method provided by the present invention can comprehensively quantify the degradation trend of the bearing performance of the bridge structure throughout its life cycle and accurately predict its future state by obtaining the long-term monitoring data sequence of the target bridge arranged according to the monitoring time nodes during its service period, combining the load dynamic characteristics extracted from the load distribution analysis and the deformation coupling characteristics extracted from the deformation correlation analysis, generating a comprehensive evaluation index and comparing it with the preset threshold. The long-term monitoring data sequence completely covers the evolution law of the dynamic behavior during the bridge service process through the correlation mapping of the time dimension and the structural response parameters; the load dynamic characteristics accurately capture the time-varying characteristics of the stress state of the key cross-sections through the spatial interpolation calculation of the load transfer law; the deformation coupling characteristics effectively reveal the degradation mechanism of the overall structural stiffness through the co-deformation relationship modeling between the main girder and the pier column. The comprehensive evaluation index based on the multi-source feature fusion incorporates the dynamic changes of the structural stiffness, stress distribution and residual bearing capacity into a unified evaluation system, and can break through the limitations of the traditional static evaluation method by comparing with the trend of the historical reference value, significantly improving the sensitivity of the bearing performance degradation identification and the reliability of the prediction result. Through the multi-level dynamic comparison of the real-time monitoring data and the preset threshold, not only can the current bearing capacity safety level of the bridge be accurately determined, but also the risk evolution path in the future service stage can be prospectively identified based on the time series prediction model, providing a scientific basis with timeliness and forward-looking for the bridge maintenance decision-making, thus effectively avoiding the sudden structural failure caused by the accumulation of local damages and extending the safe service life of the bridge. Description of the Drawings

[0007] Figure 1 Fig. shows a schematic application scenario diagram of the bearing capacity data analysis method for a long-span continuous rigid-frame bridge according to an embodiment of the present application.

[0008] Figure 2 Fig. shows a flowchart of a bearing capacity data analysis method for a long-span continuous rigid-frame bridge according to an embodiment of the present application.

[0009] Figure 3 Fig. shows a schematic composition diagram of a data analysis system according to an embodiment of the present application. Detailed Embodiments

[0010] Figure 1 Fig. shows a schematic diagram of the application scenario provided according to an embodiment of the present application. The application scenario includes one or more sensing and monitoring devices 101, a data analysis system 120, and one or more networks 110 that couple the one or more sensing and monitoring devices 101 to the data analysis system 120.

[0011] In Figure 1In the configuration shown, the data analysis system 120 may include one or more components that implement the functions performed by the data analysis system 120. These components may include software components, hardware components, or a combination thereof that can be executed by one or more processors. The user operating the operation sensing and monitoring device 101 can sequentially utilize one or more applications to interact with the data analysis system 120 to utilize the services provided by these components. It should be understood that various different system configurations are possible, which may vary depending on the application scenario. Therefore, Figure 1 is an example of a system for implementing the various methods described herein and is not intended to be limiting. The sensing and monitoring device 101 is used to monitor data of the bridge, such as pressure, and send the collected data to the data analysis system 120.

[0012] The data analysis system 120 may include one or more general-purpose computers, dedicated server computers (such as PC (personal computer) servers, UNIX servers, midrange servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and / or combination. In this application scenario, one or more databases 130 may also be included. In certain embodiments, these databases may be used to store data and other information.

[0013] Please refer to Figure 2 , the method for analyzing the bearing capacity of a long-span continuous rigid-frame bridge provided by the embodiments of the present application includes the following steps:

[0014] Step S100: Obtain a long-term monitoring data sequence of the target bridge. The long-term monitoring data sequence includes multiple structural response data sets arranged according to monitoring time nodes during the service life cycle of the bridge.

[0015] The long-term monitoring data sequence is a data sequence obtained by continuously monitoring the target bridge throughout its service life cycle. The structural response data set is a set of a series of response data generated by the bridge due to various load actions at each monitoring time node. These response data may include, but are not limited to, physical quantities such as stress, strain, displacement, and acceleration at key parts of the bridge. For example, at a certain monitoring time node, the structural response data set may include the stress value of the mid-span section of the bridge, the displacement value at the top of the pier column, and the acceleration value of the main girder. The long-term monitoring data sequence can be obtained by installing various sensors on the bridge, such as stress sensors, displacement sensors, acceleration sensors, etc. These sensors collect data at preset time intervals and transmit the collected data to the data storage center for storage and management.

[0016] Step S200: Perform a load distribution analysis on the long-term monitoring data sequence to extract the load dynamic characteristics of the bridge. The load dynamic characteristics are used to characterize the load transfer law of each key section of the bridge under different monitoring time nodes.

[0017] The dynamic characteristics of loads are characteristic quantities that reflect the load transfer law of a bridge at different time nodes. In actual analysis, the load transfer law of each key section of the bridge at different monitoring time nodes will be affected by various factors, such as changes in vehicle loads and environmental factors. For example, during periods with high traffic flow, the loads borne by each key section of the bridge will increase accordingly, and the load transfer law will also change. By analyzing the load distribution of the long-term monitoring data sequence, these dynamic load characteristics can be extracted, providing a basis for subsequent evaluation of the bearing performance of the bridge.

[0018] As an implementation method, in step S200, performing load distribution analysis on the long-term monitoring data sequence to extract the dynamic load characteristics of the bridge may specifically include:

[0019] Step S210: Separating a load distribution data subset from the long-term monitoring data sequence, where the load distribution data subset includes the bending moment monitoring values and shear force monitoring values of the mid-span and support sections of the bridge at each monitoring time node.

[0020] The load distribution data subset is a data subset related to load distribution screened from the long-term monitoring data sequence. The bending moment monitoring value refers to the monitoring data of the magnitude of the bending moment borne by the mid-span and support sections of the bridge at each monitoring time node. The shear force monitoring value refers to the monitoring data of the magnitude of the shear force borne by the corresponding section. In actual operation, the load distribution data subset can be separated from the long-term monitoring data sequence through a data screening algorithm. For example, according to the label information of the data, the data related to the bending moment and shear force of the mid-span and support sections of the bridge can be screened. These bending moment monitoring values and shear force monitoring values can intuitively reflect the stress conditions of each key section of the bridge at different monitoring time nodes.

[0021] Step S220: Decomposing the load distribution data subset into a dead load component sequence and a live load component sequence according to a preset load decomposition level, where the dead load component sequence extracts the long-term trend term through low-pass filtering, and the live load component sequence extracts the short-term fluctuation term through high-pass filtering.

[0022] The preset load decomposition level is a hierarchal structure for load decomposition set in advance. The dead load component sequence is the component sequence related to the dead load in the load distribution data subset. The dead load usually refers to relatively stable loads such as the self-weight of the bridge structure and the weight of auxiliary facilities. The live load component sequence is the component sequence related to the live load. The live load includes loads that vary greatly with time, such as vehicle loads, pedestrian loads, and wind loads. For example, in practical applications, digital filters can be used to implement low-pass filtering and high-pass filtering to decompose the load distribution data subset into a dead load component sequence and a live load component sequence.

[0023] Step S230: dynamically superimpose the dead load component sequence and the live load component sequence using a time-varying weight distribution algorithm to generate an equivalent load spectrum corresponding to each monitoring time node.

[0024] The time-varying weight allocation algorithm is an algorithm that can dynamically adjust the weight according to time changes. The equivalent load spectrum is obtained by dynamically superimposing the constant load component sequence and the live load component sequence. It can comprehensively reflect the spectrum of the load conditions borne by the bridge at each monitoring time node. At different monitoring time nodes, the degree of effect of the constant load and live load on the bridge is different. Therefore, it is necessary to use a time-varying weight allocation algorithm to dynamically adjust the weights of the constant load component sequence and the live load component sequence, and then superimpose them to generate an equivalent load spectrum. For example, in time periods with large traffic flow, the effect of live load is relatively large. At this time, the weight of the live load component sequence can be appropriately increased; in time periods with small traffic flow, the effect of constant load is relatively prominent, and the weight of the constant load component sequence can be increased accordingly.

[0025] As an implementation mode, step S230, using a time-varying weight distribution algorithm to dynamically superimpose the dead load component sequence and the live load component sequence to generate an equivalent load spectrum corresponding to each monitoring time node, may specifically include:

[0026] Step S231: Initialize the dead load weight coefficient of each monitoring time node based on the long-term trend slope of each monitoring time node in the dead load component sequence.

[0027] The long-term trend slope is the slope of the long-term change trend of the constant load component sequence at each monitoring time node. The constant load weight coefficient is a coefficient used to measure the proportion of the constant load in the equivalent load spectrum. At each monitoring time node, the long-term trend slope of the constant load component sequence reflects the changing trend of the constant load. For example, if the long-term trend slope of the constant load component sequence at a certain monitoring time node is large, it means that the constant load changes more dramatically near this time node. At this time, the constant load weight coefficient of this time node can be appropriately increased to more accurately reflect the effect of the constant load on the bridge. By analyzing the long-term trend slope of each monitoring time node in the constant load component sequence, the constant load weight coefficient of the corresponding time node can be initialized.

[0028] Step S232: Initialize the live load weight coefficient of the corresponding time node based on the short-term fluctuation amplitude of each monitoring time node in the live load component sequence.

[0029] The short-term fluctuation amplitude is the magnitude of the short-term fluctuations of the live load component sequence at each monitoring time node. The live load weight coefficient is a coefficient used to measure the proportion of the live load in the equivalent load spectrum. The short-term fluctuation amplitude of the live load reflects the degree of intensity of the change of the live load at that time node. For example, during the traffic peak period, the short-term fluctuation amplitude of the live load may be relatively large, indicating that the change of the live load is more frequent and intense. At this time, the live load weight coefficient at this time node can be appropriately increased to better reflect the impact of the live load on the bridge. According to the short-term fluctuation amplitude of each monitoring time node in the live load component sequence, the live load weight coefficient corresponding to the time node can be initialized.

[0030] Step S233: According to the real-time monitoring data set of the environment where the bridge is located, extract the temperature gradient change rate, humidity cumulative effect factor, and traffic flow impact index to generate an environmental dynamic impact set.

[0031] The real-time monitoring data set is the data set obtained by the real-time monitoring of the environment where the bridge is located. The temperature gradient change rate refers to the temperature change rate generated by the temperature difference in different parts of the bridge structure. The humidity cumulative effect factor is a factor that reflects the cumulative impact of humidity on the bridge structure within a certain period of time. The traffic flow impact index is an index that measures the impact of traffic flow on the bridge structure. The environmental dynamic impact set is a set composed of factors such as the temperature gradient change rate, humidity cumulative effect factor, and traffic flow impact index. For example, the real-time monitoring data set can be obtained by installing temperature sensors, humidity sensors, and traffic flow monitoring devices around the bridge, and then analyzing and processing these data to extract the temperature gradient change rate, humidity cumulative effect factor, and traffic flow impact index to generate the environmental dynamic impact set.

[0032] Step S234: Input the environmental dynamic impact set into the weight correction model to perform non-linear coupling adjustment on the dead load weight coefficient and the live load weight coefficient to obtain the corrected dead load weight and the corrected live load weight at each monitoring time node.

[0033] The weight correction model is a model that can correct the dead load weight coefficient and the live load weight coefficient according to the input environmental dynamic impact set. Non-linear coupling adjustment means considering the non-linear interaction relationship between them when adjusting the dead load weight coefficient and the live load weight coefficient. The corrected dead load weight and the corrected live load weight are the weights obtained after being adjusted by the weight correction model and can more accurately reflect the actual situation. For example, the weight correction model can adopt a neural network model, take the environmental dynamic impact set as the input, and through the learning and calculation of the neural network, perform non-linear coupling adjustment on the dead load weight coefficient and the live load weight coefficient to obtain the corrected dead load weight and the corrected live load weight at each monitoring time node.

[0034] Step S235: Adopt a sliding time window mechanism to perform a moving average process on the corrected dead load weights of the current monitoring time node and the previous preset number of nodes, generating a smoothed dead load weight sequence.

[0035] The sliding time window mechanism processes data by setting a time window of a fixed length and sliding this window on the time series. The preset number is the number of previous nodes preset to participate in the moving average process. The moving average process is a process of averaging the corrected dead load weights of the current monitoring time node and the previous preset number of nodes. The smoothed dead load weight sequence is obtained through the moving average process and can make the dead load weights smoother and more stable.

[0036] Step S236: Synchronously perform an exponential decay process on the corrected live load weights of the current monitoring time node and the previous preset number of nodes, generating a decaying live load weight sequence.

[0037] The exponential decay process is a method of performing decay calculations on data, and its decay degree decreases exponentially over time. The decaying live load weight sequence is the live load weight sequence obtained through the exponential decay process. In practical applications, since the short-term impact of live loads is relatively large, as time goes by, the influence of the live load weights of previous nodes on the current node will gradually decrease. Therefore, adopting the exponential decay process can more reasonably consider the time decay characteristics of live load weights. For example, for the corrected live load weights of the current monitoring time node and the previous preset number of nodes, calculate according to the exponential decay formula to obtain the decaying live load weight sequence.

[0038] Step S237: Multiply the data of the corresponding time nodes in the smoothed dead load weight sequence and the dead load component sequence according to the weights to obtain the dead load weighted component.

[0039] The dead load weighted component is the component obtained by multiplying the data of the corresponding time nodes in the smoothed dead load weight sequence and the dead load component sequence according to the weights. At each monitoring time node, multiply the smoothed dead load weight of this node by the data of the corresponding node in the dead load component sequence to obtain the dead load weighted component of this node. For example, at the nth monitoring time node, multiply the weight of the nth node in the smoothed dead load weight sequence by the data of the nth node in the dead load component sequence to obtain the dead load weighted component of the nth node.

[0040] Step S238: Multiply the data of the corresponding time nodes in the decaying live load weight sequence and the live load component sequence according to the weights, obtaining the live load weighted component.

[0041] The weighted component of the live load is obtained by multiplying the attenuated live load weight sequence by the data at the corresponding time nodes in the live load component sequence according to the weights. At each monitoring time node, multiplying the attenuated live load weight at that node by the data at the corresponding node in the live load component sequence gives the weighted component of the live load at that node. For example, at the m-th monitoring time node, multiplying the weight of the m-th node in the attenuated live load weight sequence by the data of the m-th node in the live load component sequence yields the weighted component of the live load at the m-th node.

[0042] Step S239: Perform vector superposition on the weighted dead load component and the weighted live load component at the same monitoring time node to generate an initial equivalent load spectrum.

[0043] Vector superposition is an operation of adding the weighted dead load component and the weighted live load component at the same monitoring time node. The initial equivalent load spectrum is obtained through vector superposition and is a spectral line that preliminarily reflects the load conditions borne by the bridge at each monitoring time node. At each monitoring time node, adding the weighted dead load component and the weighted live load component at that node gives the initial equivalent load value at that node, and so on, to generate the initial equivalent load spectrum.

[0044] Step S2310: Invoke the load distribution verification model to perform spatio-temporal alignment and comparison between the initial equivalent load spectrum and the typical load patterns of bridges of the same type in the historical load distribution feature library.

[0045] The historical load distribution feature library is a database storing the historical load distribution features of bridges of the same type. The typical load pattern refers to a representative load distribution pattern in the historical load distribution feature library. Spatio-temporal alignment and comparison mean matching and comparing the initial equivalent load spectrum and the typical load pattern in the time and space dimensions. For example, the load distribution verification model can adopt a pattern recognition algorithm to perform spatio-temporal alignment between the initial equivalent load spectrum and the typical load patterns of bridges of the same type in the historical load distribution feature library, and compare their similarities and differences in time and space. Through this comparison, it can be judged whether the initial equivalent load spectrum is reasonable and whether it conforms to the load distribution law of bridges of the same type.

[0046] Step S2311: According to the deviation of the load peak position and the difference in distribution uniformity in the comparison results, inversely adjust the coupling adjustment parameters of the weight correction model.

[0047] The deviation of the load peak position refers to the difference in the load peak position between the initial equivalent load spectrum and the typical load pattern. The difference in distribution uniformity refers to the difference in the degree of load distribution uniformity between the two. The coupling adjustment parameter is a parameter used in the weight correction model to adjust the non-linear coupling relationship between the dead load weight coefficient and the live load weight coefficient. According to the comparison results, if the deviation of the load peak position is large or the difference in distribution uniformity is obvious, it indicates that there is a large difference between the initial equivalent load spectrum and the typical load pattern, and it is necessary to make a reverse adjustment to the coupling adjustment parameter of the weight correction model. For example, if the load peak position of the initial equivalent load spectrum is earlier than that of the typical load pattern and the distribution uniformity is poor, the coupling adjustment parameter of the weight correction model can be appropriately adjusted to make the weight distribution of the dead load and the live load more reasonable, so that the initial equivalent load spectrum is closer to the typical load pattern.

[0048] Step S2312: Re-execute the weight correction and subsequent superposition steps using the adjusted weight correction model until the similarity between the spatio-temporal distribution pattern of the initial equivalent load spectrum and the historical typical load pattern reaches a preset threshold.

[0049] The preset threshold is a standard set in advance to measure the similarity between the initial equivalent load spectrum and the historical typical load pattern. After making a reverse adjustment to the coupling adjustment parameter of the weight correction model, it is necessary to re-execute the weight correction and subsequent superposition steps, that is, starting from step S234, recalculate the corrected dead load weight and the corrected live load weight, generate a smooth dead load weight sequence and a decaying live load weight sequence, calculate the dead load weighted component and the live load weighted component, and perform vector superposition to generate a new initial equivalent load spectrum. Then, align the spatio-temporal comparison of the new initial equivalent load spectrum with the historical typical load pattern again and calculate their similarity. Repeat this process until the similarity between the spatio-temporal distribution pattern of the initial equivalent load spectrum and the historical typical load pattern reaches the preset threshold.

[0050] Step S2313: Mark the initial equivalent load spectrum that passes the verification as the final equivalent load spectrum and associate it with the structural response data set corresponding to the monitoring time node.

[0051] The final equivalent load spectrum is a spectrum line that has passed the verification and can accurately reflect the load conditions borne by the bridge at each monitoring time node. After marking the initial equivalent load spectrum that passes the verification as the final equivalent load spectrum, it is necessary to associate it with the structural response data set corresponding to the monitoring time node. In this way, the corresponding relationship between the load and the structural response can be established, providing more comprehensive data support for subsequent analysis and evaluation. For example, at the i-th monitoring time node, associate the final equivalent load spectrum of this node with the structural response data set of this node, so that when analyzing the structural response of this node, the load conditions borne by this node can be considered simultaneously.

[0052] Step S240: Call the pre-trained load transfer model to perform spatial interpolation calculation on the equivalent load spectrum, and obtain the load distribution cloud map within the entire span of the bridge.

[0053] The pre-trained load transfer model is a model that has been pre-trained and can transfer and calculate loads. Spatial interpolation calculation refers to the process of calculating load data at other positions based on the known load data at a finite number of points. The load distribution cloud map is a graph representing the load distribution within the entire span of the bridge in the form of a cloud map. In practical applications, the equivalent load spectrum usually only contains load data at some monitoring positions. In order to obtain the load distribution within the entire span of the bridge, it is necessary to call the pre-trained load transfer model to perform spatial interpolation calculation on the equivalent load spectrum.

[0054] As an implementation, in step S240, calling the pre-trained load transfer model to perform spatial interpolation calculation on the equivalent load spectrum and obtain the load distribution cloud map within the entire span of the bridge may specifically include:

[0055] Step S241: Divide the load data at each monitoring time node in the equivalent load spectrum into several load action sub-domains along the longitudinal axis of the bridge. Each sub-domain includes the starting position coordinates, ending position coordinates, and the average load intensity within the domain.

[0056] The load action sub-domain is a sub-region obtained by dividing the load data in the equivalent load spectrum along the longitudinal axis of the bridge. The starting position coordinates and ending position coordinates respectively represent the starting and ending positions of each load action sub-domain on the longitudinal axis of the bridge. The average load intensity within the domain refers to the average load magnitude within each load action sub-domain. For example, at a certain monitoring time node, the load data in the equivalent load spectrum is divided into multiple load action sub-domains along the longitudinal axis of the bridge. The starting position coordinates and ending position coordinates of each sub-domain can be obtained through measurement and calculation, and the average load intensity within the domain can be obtained by averaging the load data within the sub-domain.

[0057] Step S242: According to the topological relationship of the bridge structure, extract the geometric connection characteristics between each load action sub-domain and the adjacent unmonitored area, and generate a load transfer path network.

[0058] The topological relationship of the bridge structure refers to the connection relationship and spatial layout among various components of the bridge. The geometric connection characteristic refers to the geometric connection mode and characteristics between each load acting sub-domain and the adjacent unmonitored area. The load transfer path network is a network composed of the load transfer paths between each load acting sub-domain and the adjacent unmonitored area. In practical applications, based on the topological relationship of the bridge structure, the geometric connection characteristics between each load acting sub-domain and the adjacent unmonitored area can be analyzed. For example, it can be determined whether they are connected through the beam body, node connection or other means. Then, based on these geometric connection characteristics, the load transfer path network is generated to clarify the load transfer path in the bridge structure. For example, in a continuous rigid frame bridge, by analyzing the topological relationship of the bridge structure, the beam connection mode between each span can be determined, thereby generating the load transfer path network and providing a basis for subsequent load transfer calculation.

[0059] Step S243: Use the load intensity of the load acting sub-domain as the input node feature and the geometric parameters of the load transfer path network as the edge feature to construct the load spatial propagation graph structure.

[0060] The input node feature refers to the parameter used as the node feature in the graph structure. Here, the load intensity of the load acting sub-domain is used as the input node feature. The edge feature refers to the parameter used as the edge feature in the graph structure. Here, the geometric parameters of the load transfer path network are used as the edge feature. The load spatial propagation graph structure is a structure that represents the propagation of loads in the bridge space using a graph structure. When constructing the load spatial propagation graph structure, each load acting sub-domain is regarded as a node, and the feature of the node is the load intensity of the sub-domain; the load transfer paths between each load acting sub-domain are regarded as edges, and the feature of the edge is the geometric parameter of the load transfer path network. For example, in a simple bridge model, there are three load acting sub-domains A, B, and C. A, B, and C are regarded as nodes, and the features of the nodes are their load intensities respectively. The load transfer paths between A and B, and between B and C are regarded as edges, and the features of the edges are the corresponding geometric parameters, thus constructing the load spatial propagation graph structure.

[0061] Step S244: Invoke the graph convolution module in the pre-trained load transfer model to perform multi-scale feature aggregation on the load spatial propagation graph structure and output the load intensity prediction value of each unmonitored area.

[0062] The graph convolution module can perform convolution operations on the graph structure and extract features from the graph structure. Multi-scale feature aggregation refers to the process of aggregating and fusing features in the graph structure at different scales. An unmonitored area refers to an area where load data is not directly monitored in the equivalent load spectrum. The load intensity prediction value is obtained by processing the load space propagation graph structure through the graph convolution module, which is a prediction result of the load intensity in the unmonitored area. For example, the pre-trained load transfer model can adopt a model based on a graph neural network, in which the graph convolution module can perform multi-scale feature aggregation on the load space propagation graph structure, and output the load intensity prediction value of each unmonitored area by learning the features of the nodes and edges in the graph structure.

[0063] Step S245: dynamically adjust the neighborhood sampling radius of the graph convolution module according to the gradient change direction between the measured load intensity and the predicted load intensity of the adjacent load action subdomain.

[0064] The gradient change direction refers to the change direction between the measured load intensity and the predicted load intensity of the adjacent load action subdomain. The neighborhood sampling radius is the range radius of the graph convolution module when performing the convolution operation. In practical applications, there may be a difference between the measured load intensity and the predicted load intensity of the adjacent load action subdomain. According to the gradient change direction of this difference, the neighborhood sampling radius of the graph convolution module can be dynamically adjusted. For example, if the gradient change between the measured load intensity and the predicted load intensity of the adjacent load action subdomain is large, it means that the load change is more drastic. At this time, the neighborhood sampling radius of the graph convolution module can be appropriately increased to obtain more adjacent node information and improve the accuracy of the prediction; conversely, if the gradient change is small, the neighborhood sampling radius can be appropriately reduced.

[0065] Step S246: splicing the predicted load intensity output by the adjusted graph convolution module with the measured load intensity according to the spatial position to form a preliminary load distribution surface.

[0066] The preliminary load distribution surface is obtained by splicing the predicted load intensity output by the adjusted graph convolution module with the measured load intensity according to the spatial position. At each spatial position, the predicted load intensity at that position is combined with the measured load intensity to form the load intensity value at that position. By analogy, a preliminary load distribution surface is formed over the entire bridge space. For example, at a certain cross-sectional position of the bridge, the predicted load intensity at that position is spliced ​​with the measured load intensity to obtain the final load intensity value at that position. The preliminary load distribution surface is formed by performing the same operation at each position of the bridge.

[0067] Step S247: Based on the stiffness distribution characteristics of the bridge main beam cross section, anisotropic smoothing correction is performed on the preliminary load distribution surface to eliminate inter-span mutation noise.

[0068] The stiffness distribution characteristics of the bridge main girder section refer to the stiffness distribution of the bridge main girder at different positions and in different directions. Anisotropic smoothing correction refers to the process of smoothing the preliminary load distribution surface considering the stiffness distribution characteristics of the bridge main girder section. The inter-span abrupt noise refers to the suddenly changing noise in the preliminary load distribution surface at the inter-span position. In practical applications, the section stiffness of the bridge main girder may vary at different positions and in different directions, and this difference will affect the load distribution. Therefore, based on the stiffness distribution characteristics of the bridge main girder section, performing anisotropic smoothing correction on the preliminary load distribution surface can make the load distribution more reasonable and eliminate the inter-span abrupt noise. For example, a method based on finite element analysis can be used to smooth the preliminary load distribution surface according to the stiffness distribution characteristics of the bridge main girder section, so that the load distribution is more continuous and smooth at the inter-span position.

[0069] Step S248: Project the corrected load distribution surface into the bridge three-dimensional coordinate system, and generate a discretized load density lattice according to a preset resolution.

[0070] The bridge three-dimensional coordinate system is a coordinate system used to describe the spatial position of the bridge. The preset resolution is the precision of generating the discretized load density lattice set in advance. The discretized load density lattice is a lattice obtained by discretizing the corrected load distribution surface in the bridge three-dimensional coordinate system according to the preset resolution. After projecting the corrected load distribution surface into the bridge three-dimensional coordinate system, it is discretized according to the preset resolution, converting the surface into a series of discrete points, each point corresponding to a load density value, forming a discretized load density lattice. For example, the preset resolution can be set to 1 meter. In the bridge three-dimensional coordinate system, a point is taken every 1 meter, and the load density value of this point is calculated to form a discretized load density lattice.

[0071] Step S249: According to the spatial correlation between adjacent points in the load density lattice, use an adaptive interpolation kernel function to fill the blank areas between points to generate a continuous load distribution cloud map.

[0072] Spatial correlation refers to the degree of spatial association of load density values between adjacent points in the load density lattice. The adaptive interpolation kernel function is an interpolation function that can automatically adjust the interpolation parameters according to the spatial correlation of adjacent points. The blank area refers to the unfilled area between adjacent points in the discretized load density lattice. The continuous load distribution contour map is obtained by filling the blank area between points using the adaptive interpolation kernel function, and it can continuously represent the load distribution of the entire bridge span. In practical applications, according to the spatial correlation of adjacent points in the load density lattice, a suitable adaptive interpolation kernel function is selected to perform interpolation calculations on the blank area between adjacent points, fill the blank area, and generate a continuous load distribution contour map. For example, a Gaussian interpolation kernel function can be used to automatically adjust the interpolation parameters according to the spatial distance between adjacent points and the difference in load density values, fill the blank area between points, and generate a continuous load distribution contour map.

[0073] Step S2410: Compare the morphological similarity of the load distribution contour map with the typical load patterns in the historical load database, and extract the peak area contour and valley area boundary in the contour map.

[0074] The historical load database is a database that stores historical load data and typical load patterns. Morphological similarity comparison refers to the process of comparing and analyzing the morphology of the load distribution contour map and the typical load patterns in the historical load database. The peak area contour refers to the contour of the area with a higher load intensity in the load distribution contour map. The valley area boundary refers to the boundary of the area with a lower load intensity in the load distribution contour map. In practical applications, comparing the generated load distribution contour map with the typical load patterns in the historical load database for morphological similarity can determine whether the load distribution contour map conforms to historical laws. Through the comparison, the peak area contour and valley area boundary in the contour map are extracted, providing important information for subsequent analysis and evaluation. For example, an image matching algorithm can be used to compare the load distribution contour map with the typical load patterns in the historical load database and extract the peak area contour and valley area boundary.

[0075] Step S2411: According to the morphological differences in the comparison results, reversely optimize the feature aggregation weight coefficients of the graph convolution module.

[0076] The morphological difference refers to the difference in the shape between the load distribution contour map and the typical load patterns in the historical load database. The feature aggregation weight coefficient is the weight coefficient used to aggregate features in the graph convolution module. According to the comparison result, if there is a morphological difference between the load distribution contour map and the typical load pattern, it indicates that the feature aggregation effect of the graph convolution module may not be ideal, and it is necessary to reversely optimize the feature aggregation weight coefficient of the graph convolution module. For example, if the position of the peak region in the load distribution contour map is inconsistent with the typical load pattern, the feature aggregation weight coefficient of the graph convolution module can be adjusted so that the graph convolution module pays more attention to the features related to the peak region when aggregating features, thereby improving the accuracy of load prediction.

[0077] Step S2412: Re - execute the load intensity prediction and subsequent correction steps using the optimized load transfer model until the matching degree of the contour features of the load distribution contour map and the historical typical pattern exceeds the set threshold.

[0078] The set threshold is a pre - set standard for measuring the matching degree between the contour features of the load distribution contour map and the historical typical pattern. After reversely optimizing the feature aggregation weight coefficient of the graph convolution module, re - execute the load intensity prediction and subsequent correction steps using the optimized load transfer model, that is, starting from step S244, re - perform multi - scale feature aggregation on the load space propagation graph structure, output the load intensity prediction values of each unmonitored area, and then perform subsequent adjustment, splicing, correction and other steps to generate a new load distribution contour map. Again, perform a morphological similarity comparison between the new load distribution contour map and the historical typical pattern, and calculate their matching degree. Repeat this process until the matching degree of the contour features of the load distribution contour map and the historical typical pattern exceeds the set threshold. For example, the set threshold can be set to 90%. When the matching degree of the contour features of the new load distribution contour map and the historical typical pattern reaches 90%, it is considered that the load distribution contour map meets the requirements.

[0079] Step S2413: Spatially and temporally align and store the finally verified load distribution contour map with the structural response data set at the corresponding monitoring time node, and mark the coordinate range and intensity level of the key load concentration areas.

[0080] The finally verified load distribution nephogram is the load distribution nephogram that meets the requirements after multiple optimizations and verifications. Spatiotemporal alignment storage refers to the process of matching and storing the finally verified load distribution nephogram with the structural response data set at the corresponding monitoring time nodes in terms of time and space. The key load concentration area refers to the area in the load distribution nephogram where the load intensity is relatively high and has a greater impact on the bridge structure. The coordinate range refers to the position range of the key load concentration area in the three-dimensional coordinate system of the bridge. The intensity level refers to the level of the load intensity in the key load concentration area. When performing spatiotemporal alignment storage on the finally verified load distribution nephogram and the structural response data set at the corresponding monitoring time nodes, it is necessary to ensure their consistency in time and space for subsequent analysis and research. At the same time, annotating the coordinate range and intensity level of the key load concentration area can provide important references for the maintenance and management of the bridge. For example, when storing, the load distribution nephogram and the structural response data set can be associated according to the monitoring time nodes, and the coordinate range and intensity level of the key load concentration area can be annotated on the load distribution nephogram.

[0081] As an implementation manner, the method provided by the embodiment of the present invention further includes a verification process of the load transfer model, which may specifically include:

[0082] Step S2401: Obtain the measured three-dimensional displacement field data of the target bridge under the standard load test condition, where the measured three-dimensional displacement field data includes the spatial coordinate change amounts of the mid-span of the main girder and the pier top section.

[0083] The standard load test condition refers to the condition set during the load test of the bridge. The measured three-dimensional displacement field data is the three-dimensional displacement field data of the target bridge obtained through actual measurement under the standard load test condition. The mid-span of the main girder refers to the middle position of the main girder of the bridge. The pier top section refers to the section at the top of the pier. The spatial coordinate change amount refers to the change amount of the spatial coordinates of the mid-span of the main girder and the pier top section relative to the initial state under the standard load test condition. In actual operation, high-precision measuring devices such as total stations and laser rangefinders can be used to measure the three-dimensional displacement field of the target bridge under the standard load test condition, obtain the spatial coordinate change amounts of the mid-span of the main girder and the pier top section, and obtain the measured three-dimensional displacement field data.

[0084] Step S2402: Input the load parameters of the standard load test condition into the pre-established parametric finite element model, and generate theoretical displacement field distribution data after performing multi-scale mesh division.

[0085] A parametric finite element model is a finite element model established in a parametric way, which can conveniently simulate and analyze different structural parameters and load conditions. Multi-scale mesh generation refers to the process of dividing the finite element model with meshes of different scales according to different parts and characteristics of the structure. The theoretical displacement field distribution data is obtained by simulating and calculating with the parametric finite element model, which is the theoretical displacement field distribution of the target bridge under the standard load test conditions. In practical applications, the load parameters of the standard load test conditions, such as load magnitude, load distribution, etc., are input into the pre-established parametric finite element model, and then the model is subjected to multi-scale mesh generation to improve the calculation accuracy and efficiency. Finally, the model is calculated by finite element analysis software to generate the theoretical displacement field distribution data.

[0086] Step S2403: Extract the absolute error vector of the measured vertical displacement at the mid-span of the main girder in the measured three-dimensional displacement field data and the theoretical value at the corresponding position in the theoretical displacement field distribution data, and at the same time calculate the relative error distribution pattern of the measured horizontal displacement at the pier top and the theoretical value.

[0087] The absolute error vector refers to the difference vector between the measured vertical displacement at the mid-span of the main girder in the measured three-dimensional displacement field data and the theoretical value at the corresponding position in the theoretical displacement field distribution data. The relative error distribution pattern refers to the spatial distribution of the relative error between the measured horizontal displacement at the pier top and the theoretical value. In actual operation, the measured vertical displacement at the mid-span of the main girder is extracted from the measured three-dimensional displacement field data, the theoretical value at the corresponding position is extracted from the theoretical displacement field distribution data, and the difference between them is calculated to obtain the absolute error vector. At the same time, the relative error between the measured horizontal displacement at the pier top and the theoretical value is calculated, and its spatial distribution is analyzed to obtain the relative error distribution pattern.

[0088] Step S2404: Identify the correction priority of the main girder stiffness parameters according to the direction angle of the absolute error vector, and preferentially adjust the elastic modulus adjustment factor orthogonal to the direction of the maximum absolute error.

[0089] The direction angle refers to the direction angle of the absolute error vector in space. The main girder stiffness parameter is a parameter describing the stiffness characteristics of the main girder, such as elastic modulus, etc. The elastic modulus adjustment factor is a factor used to adjust the elastic modulus of the main girder. In practical applications, according to the direction angle of the absolute error vector, the stiffness error situation of the main girder in different directions can be judged. Preferentially adjusting the elastic modulus adjustment factor orthogonal to the direction of the maximum absolute error can more effectively reduce the error. For example, if the direction angle of the absolute error vector indicates that the error is the largest in a certain direction, then preferentially adjust the elastic modulus adjustment factor orthogonal to this direction, so that the stiffness of the main girder in this direction is more in line with the actual situation.

[0090] Step S2405: Synchronously correct the boundary stiffness proportionality coefficient of the pier foundation constraint condition based on the spatial correlation of the relative error distribution pattern.

[0091] Spatial correlation refers to the degree of association of the relative error distribution pattern in space. The pier foundation constraint condition refers to the constraint relationship between the pier foundation and the foundation. The boundary stiffness proportionality coefficient is a coefficient that describes the boundary stiffness of the pier foundation constraint condition. In practical applications, the spatial correlation of the relative error distribution pattern reflects the force conditions and deformation differences of the pier foundation at different positions. Based on this spatial correlation, synchronously correcting the boundary stiffness proportionality coefficient of the pier foundation constraint condition can make the constraint condition of the pier foundation more reasonable and improve the accuracy of the model. For example, if the relative error distribution pattern shows a large relative error in a certain area, it indicates that the constraint condition of the pier foundation in this area may need to be adjusted. At this time, the boundary stiffness proportionality coefficient of this area can be corrected accordingly.

[0092] Step S2406: Re-import the corrected elastic modulus adjustment factor and boundary stiffness proportionality coefficient into the parametric finite element model, and iteratively perform the load-displacement response calculation until the norm of the absolute error vector and the relative error rate of all measurement points are lower than the preset convergence threshold.

[0093] The preset convergence threshold is a standard set in advance to measure the convergence degree of the calculation results. After re-importing the corrected elastic modulus adjustment factor and boundary stiffness proportionality coefficient into the parametric finite element model, the load-displacement response calculation is performed again to obtain new theoretical displacement field distribution data. Then, the absolute error vector and relative error rate between the measured values and the new theoretical values of each measurement point in the three-dimensional displacement field measured data are extracted again. Repeat this process, continuously correcting the elastic modulus adjustment factor and boundary stiffness proportionality coefficient until the norm of the absolute error vector and the relative error rate of all measurement points are lower than the preset convergence threshold.

[0094] Step S2407: Extract the mid-span moment distribution nephogram and the pier bottom shear force gradient curve from the output of the finally converged finite element model to generate a load transfer characteristic verification fingerprint.

[0095] The mid-span moment distribution nephogram is a nephogram representing the moment distribution at the mid-span position of the bridge main girder. The pier bottom shear force gradient curve is a curve describing the change of the shear force gradient at the bottom of the pier. The load transfer characteristic verification fingerprint is composed of the mid-span moment distribution nephogram and the pier bottom shear force gradient curve, and is used to verify the load transfer characteristics. In the output of the finally converged finite element model, extract the mid-span moment distribution nephogram and the pier bottom shear force gradient curve, and combine them to generate a load transfer characteristic verification fingerprint. This verification fingerprint can reflect the transfer characteristics of the bridge under the action of load and is used for subsequent verification and comparison.

[0096] Step S2408: Perform dynamic spatial registration on the load transfer feature verification fingerprint and the real-time monitored load distribution cloud map, and calculate the morphological similarity index of the two in the gradient change region.

[0097] Dynamic spatial registration refers to the process of matching and aligning the load transfer feature verification fingerprint and the real-time monitored load distribution cloud map in space. The gradient change region refers to the region with a large gradient change in the load transfer feature verification fingerprint and the real-time monitored load distribution cloud map. The morphological similarity index is an index to measure the morphological similarity degree of the two in the gradient change region. In practical applications, a spatial registration algorithm is used to perform dynamic spatial registration on the load transfer feature verification fingerprint and the real-time monitored load distribution cloud map to align them in space. Then, analyze the morphological features of the two in the gradient change region and calculate the morphological similarity index. For example, an image matching algorithm can be used to calculate the morphological similarity index of the two in the gradient change region to judge their similarity degree.

[0098] Step S2409: When the morphological similarity index is lower than the historical baseline level for three consecutive monitoring periods, trigger the parameter recalibration instruction of the load transfer model.

[0099] The historical baseline level refers to the average level or reference level of the morphological similarity index in the historical monitoring data. The parameter recalibration instruction is an instruction used to trigger the re-calibration of the load transfer model parameters. During the actual monitoring process, continuously calculate the morphological similarity index of the load transfer feature verification fingerprint and the real-time monitored load distribution cloud map in the gradient change region. If this index is lower than the historical baseline level for three consecutive monitoring periods, it indicates that the parameters of the load transfer model may deviate and need to be re-calibrated. At this time, trigger the parameter recalibration instruction of the load transfer model and start the parameter recalibration process.

[0100] Step S24010: According to the recalibration instruction, call the latest measured data of the three-dimensional displacement field, re-execute the correction process of the elastic modulus adjustment factor and the boundary stiffness ratio coefficient, and update the material constitutive relationship of the finite element model.

[0101] The latest measured data of the three-dimensional displacement field refers to the measured data of the three-dimensional displacement field of the target bridge under the standard load test conditions obtained by the latest monitoring. The material constitutive relationship is the relationship that describes the mechanical properties of materials. After receiving the parameter recalibration instruction, the latest measured data of the three-dimensional displacement field is called, and according to the process of steps S2403 - S2406, the correction process of the elastic modulus adjustment factor and the boundary stiffness ratio coefficient is re-executed. By correcting these parameters, the material constitutive relationship of the finite element model is updated, making the finite element model more accurately reflect the actual mechanical properties of the bridge. For example, according to the latest measured data of the three-dimensional displacement field, the absolute error vector and the relative error distribution pattern are recalculated, the elastic modulus adjustment factor and the boundary stiffness ratio coefficient are adjusted, and the constitutive relationships such as the elastic modulus of the material and the boundary constraint conditions in the finite element model are updated.

[0102] Step S24011: Inject the theoretical displacement field distribution data output by the updated finite element model into the training data set of the load transfer model in reverse, and replace the outdated response features in the historical samples.

[0103] The theoretical displacement field distribution data is the theoretical displacement field distribution of the target bridge under the standard load test conditions calculated by the updated finite element model. The training data set is the data set used to train the load transfer model. The outdated response features refer to the response features in the historical samples that can no longer accurately reflect the actual situation of the bridge. After updating the material constitutive relationship of the finite element model, the updated theoretical displacement field distribution data is calculated through the finite element analysis software. Injecting these data into the training data set of the load transfer model in reverse and replacing the outdated response features in the historical samples make the training data set more accurately reflect the actual situation of the bridge. For example, replacing the corresponding outdated data in the training data set with the displacement values, stress values, etc. in the updated theoretical displacement field distribution data to improve the quality of the training data set.

[0104] Step S24012: Perform online retraining of the load transfer model based on the injected training data set, optimize the weight allocation strategy in the spatial interpolation calculation, and form an adaptive verification closed loop for the load distribution analysis module.

[0105] Online retraining refers to the process of real-time training of a model based on new data during the operation of the model. The weight assignment strategy refers to the strategy of assigning weights to different data points in spatial interpolation calculation. The adaptive verification closed-loop refers to a closed-loop system that enables the load distribution analysis module to automatically adapt to changes in the actual situation of the bridge through continuous verification and adjustment. After injecting the updated theoretical displacement field distribution data into the training dataset, online retraining of the load transfer model is performed based on the injected training dataset. Through retraining, the weight assignment strategy in spatial interpolation calculation is optimized, enabling the load transfer model to perform spatial interpolation calculation more accurately. At the same time, by continuously repeating the above verification and adjustment process, an adaptive verification closed-loop of the load distribution analysis module is formed to ensure the accuracy and reliability of load distribution analysis. For example, during the retraining process, deep learning algorithms are used to train the load transfer model, adjust the parameters of the model, optimize the weight assignment strategy, and improve the performance of the model.

[0106] Step S300: Conduct deformation correlation analysis on the long-term monitoring data sequence, extract the deformation coupling characteristics of the bridge, and the deformation coupling characteristics are used to characterize the co-deformation relationship between the main girder and the pier column of the bridge.

[0107] Deformation correlation analysis refers to the in-depth analysis of the data related to bridge deformation in the long-term monitoring data sequence to understand the correlation relationship of deformation between different parts of the bridge. The deformation coupling characteristic is a characteristic quantity reflecting the co-deformation relationship between the main girder and the pier column of the bridge. In an actual bridge structure, the main girder and the pier column will deform under the action of force, and their deformations are not independent of each other, but there is a co-deformation relationship. By conducting deformation correlation analysis on the long-term monitoring data sequence, this deformation coupling characteristic can be extracted, providing an important basis for evaluating the structural performance and safety of the bridge.

[0108] As an implementation method, in step S300, conducting deformation correlation analysis on the long-term monitoring data sequence and extracting the deformation coupling characteristics of the bridge may specifically include:

[0109] Step S310: Extract the main girder deflection monitoring sequence and the pier column inclination monitoring sequence from the long-term monitoring data sequence, and establish a time-domain synchronous correspondence relationship between the main girder deflection monitoring sequence and the pier column inclination monitoring sequence.

[0110] The main girder deflection monitoring sequence refers to the time - series data extracted from the long - term monitoring data sequence, which reflects the change of the deflection of the bridge main girder. The pier inclination monitoring sequence refers to the time - series data that reflects the change of the inclination of the bridge pier. The time - domain synchronization correspondence relationship means that in the time dimension, the main girder deflection monitoring sequence and the pier inclination monitoring sequence are matched and corresponding, so that the data at the same time point can be correlated with each other. In actual operation, through data screening and processing, the main girder deflection monitoring sequence and the pier inclination monitoring sequence are extracted from the long - term monitoring data sequence. Then, according to the monitoring time nodes, the time - domain synchronization correspondence relationship between them is established.

[0111] Step S320: Extract the local deformation waveform from the main girder deflection monitoring sequence through the time - window sliding algorithm, and extract the phase offset in the pier inclination monitoring sequence of the corresponding time window.

[0112] The time - window sliding algorithm is an algorithm that performs sliding - window operations on time - series data. By setting a time window with a fixed length, the window is slid on the time series for data processing. The local deformation waveform refers to the deformation waveform within a local time period extracted from the main girder deflection monitoring sequence through the time - window sliding algorithm. The phase offset refers to the phase - offset situation in the pier inclination monitoring sequence of the corresponding time window relative to the main girder deflection monitoring sequence. In practical applications, the time - window sliding algorithm is used to slide the time window on the main girder deflection monitoring sequence to extract the local deformation waveform within each time window. At the same time, in the pier inclination monitoring sequence of the corresponding time window, the phase relationship with the main girder deflection monitoring sequence is analyzed to extract the phase offset. For example, if the time - window length is set to 10 monitoring time nodes, the window is slid on the main girder deflection monitoring sequence to extract the local deformation waveform within each window; at the same time, in the pier inclination monitoring sequence of the corresponding window, the phase difference with the main girder deflection monitoring sequence is calculated to obtain the phase offset.

[0113] Step S330: Construct the main - girder - pier displacement transfer function, input the local deformation waveform into the input end of the transfer function, and use the phase offset as the feedback correction term.

[0114] The main girder - pier displacement transfer function is a function used to describe the transfer relationship between the displacement of the main girder and the displacement of the pier. When constructing this function, the local deformation waveform is used as the input to reflect the displacement change of the main girder; the phase offset is used as the feedback correction term to correct the output of the transfer function, so that the transfer function can more accurately reflect the co - deformation relationship between the main girder and the pier. For example, the main girder - pier displacement transfer function can be in the form of a linear function or a non - linear function, which is selected and constructed according to the actual situation. The local deformation waveform is input into the input end of the transfer function, and the predicted displacement of the pier is obtained through the calculation of the function. At the same time, the phase offset is used as the feedback correction term to correct the predicted displacement.

[0115] Step S340: Use the iterative approximation algorithm to optimize the parameter combination of the main girder - pier displacement transfer function until the error between the predicted pier displacement curve at the output end and the measured data is lower than the set tolerance.

[0116] The iterative approximation algorithm is an algorithm that gradually approaches the optimal solution through continuous iterative calculations. The parameter combination refers to the value combination of each parameter in the main girder - pier displacement transfer function. The set tolerance is the pre - set allowable range for measuring the error between the predicted pier displacement curve and the measured data. In practical applications, the iterative approximation algorithm is used to continuously adjust the parameter combination of the main girder - pier displacement transfer function and calculate the predicted pier displacement curve at the output end. Then, the predicted pier displacement curve is compared with the measured data to calculate the error. If the error is higher than the set tolerance, the parameter combination is continued to be adjusted, and the above process is repeated until the error is lower than the set tolerance. For example, the iterative approximation algorithm can use the Newton iterative method or the gradient descent method, etc., to optimize the parameter combination through continuous iterative calculations and improve the accuracy of the transfer function.

[0117] As an implementation, in step S340, the optimization process of the main girder - pier displacement transfer function includes:

[0118] Step S341: Extract the stiffness coupling term coefficient and the damping correction term weight from the initial parameter combination of the transfer function.

[0119] The stiffness coupling term coefficient is the coefficient in the main girder - pier displacement transfer function used to describe the stiffness coupling relationship between the main girder and the pier. The damping correction term weight is the weight used to correct the damping term in the transfer function. In the initial parameter combination of the transfer function, there are multiple parameters, and the stiffness coupling term coefficient and the damping correction term weight are extracted from them for subsequent optimization and adjustment. For example, in a simple main girder - pier displacement transfer function, the parameter combination may include the stiffness coupling term coefficient, the damping correction term weight, the displacement amplification coefficient, etc., and the stiffness coupling term coefficient and the damping correction term weight are extracted from this initial parameter combination.

[0120] Step S342: Call the main girder deflection monitoring sequence and pier inclination monitoring sequence in the historical displacement dataset to generate a displacement transfer verification sample set.

[0121] The historical displacement dataset is a dataset that stores the main girder deflection monitoring sequence and pier inclination monitoring sequence in the historical monitoring data. The displacement transfer verification sample set is a sample set generated by calling the main girder deflection monitoring sequence and pier inclination monitoring sequence in the historical displacement dataset and is used to verify the main girder - pier displacement transfer function. In practical applications, the main girder deflection monitoring sequence and pier inclination monitoring sequence for different time periods are extracted from the historical displacement dataset and combined together to generate the displacement transfer verification sample set.

[0122] Step S343: Input the verification sample set into the transfer function for forward calculation and output the predicted pier displacement curve.

[0123] Forward calculation refers to the process of inputting the main girder deflection monitoring sequence in the verification sample set into the main girder - pier displacement transfer function and performing calculations according to the calculation rules of the function to obtain the predicted pier displacement curve. After inputting the verification sample set into the transfer function, the transfer function calculates and outputs the predicted pier displacement curve based on the input main girder deflection monitoring sequence and the current parameter combination. For example, input the main girder deflection monitoring sequence for a certain time period in the verification sample set into the transfer function, and obtain the predicted pier displacement curve for this time period through the calculation of the function.

[0124] Step S344: Extract the gradient difference distribution characteristics between the predicted curve and the measured curve to generate a parameter correction direction vector.

[0125] The gradient difference distribution characteristic refers to the difference distribution of the gradient changes between the predicted pier displacement curve and the measured pier displacement curve. The parameter correction direction vector is generated based on the gradient difference distribution characteristic and is a vector used to indicate the parameter adjustment direction. In practical applications, the predicted pier displacement curve is compared with the measured pier displacement curve, the differences in their gradient changes are analyzed, and the gradient difference distribution characteristics are extracted. Then, based on the gradient difference distribution characteristics, a parameter correction direction vector is generated to determine the adjustment direction of the spatial weight distribution of the stiffness coupling term coefficient and the time decay factor of the damping correction term weight. For example, if the gradient change of the predicted curve is significantly different from the measured curve in some areas, it indicates that the parameters of the transfer function may need to be adjusted in these areas, and a parameter correction direction vector is generated according to the difference situation.

[0126] Step S345: Adjust the spatial weight distribution of the stiffness coupling term coefficient and the time decay factor of the damping correction term weight according to the correction direction vector.

[0127] The spatial weight distribution refers to the weight allocation of the stiffness coupling term coefficient in space. The time decay factor refers to the decay coefficient of the damping correction term weight varying with time. According to the parameter correction direction vector, adjust the spatial weight distribution of the stiffness coupling term coefficient and the time decay factor of the damping correction term weight. For example, if the parameter correction direction vector indicates that the stiffness coupling term coefficient in a certain area needs to be increased, then adjust the spatial weight distribution of that area accordingly; if it indicates that the time decay characteristic of the damping correction term weight needs to be adjusted, then adjust the time decay factor. Through this adjustment, the displacement transfer function between the main girder and the pier can more accurately reflect the co-deformation relationship between the main girder and the pier.

[0128] Step S346: Input the adjusted parameter combination into the digital twin verification interface and perform time-frequency characteristic matching with the dynamically monitored pier response data.

[0129] The digital twin verification interface is an interface used to verify the parameter combination of the displacement transfer function between the main girder and the pier. The dynamically monitored pier response data refers to the response data of the pier under dynamic action obtained by real-time monitoring, such as acceleration, displacement, etc. Time-frequency characteristic matching refers to the process of inputting the adjusted parameter combination into the digital twin verification interface and matching and comparing the output of the transfer function with the dynamically monitored pier response data in the time and frequency domains. In practical applications, input the adjusted parameter combination into the digital twin verification interface, calculate the predicted dynamic response data of the pier by calling the transfer function through the interface. Then, perform time-frequency characteristic matching on the predicted dynamic response data and the dynamically monitored pier response data, and analyze their similarities and differences in time and frequency. For example, adopt time-frequency analysis methods, such as wavelet transform, to perform time-frequency analysis on the predicted dynamic response data and the real-time monitoring data and compare their time-frequency characteristics.

[0130] Step S347: When the deviation of the low-frequency component energy ratio in the matching result exceeds the preset threshold, trigger the anisotropic compensation mechanism of the stiffness coupling term coefficient.

[0131] The deviation of the low-frequency component energy ratio refers to the difference in the low-frequency component energy ratio between the predicted dynamic response data and the real-time monitoring data. The preset threshold is the allowable range preset to measure the deviation of the low-frequency component energy ratio. The anisotropic compensation mechanism of the stiffness coupling term coefficient is a mechanism used to compensate for the differences in the stiffness coupling term coefficient in different directions. During the time-frequency characteristic matching process, if the deviation of the low-frequency component energy ratio in the matching result exceeds the preset threshold, it indicates that there may be a large error in the low-frequency response of the transfer function, and it is necessary to trigger the anisotropic compensation mechanism of the stiffness coupling term coefficient.

[0132] Step S348: Re - execute the displacement transfer calculation using the compensated parameter combination until the phase - lag feature of the predicted curve matches the measured data time - domain waveform with a satisfactory degree.

[0133] The phase - lag feature refers to the lag in phase between the predicted pier displacement curve and the measured pier displacement curve. The time - domain waveform matching degree refers to the similarity degree of the waveforms of the predicted curve and the measured data in the time domain. After triggering the anisotropy compensation mechanism of the stiffness coupling term coefficient, re - execute the displacement transfer calculation using the compensated parameter combination to obtain a new predicted pier displacement curve. Then, compare the new predicted curve with the measured data and analyze their phase - lag features and time - domain waveform matching degrees. If the matching degree is not satisfactory, continue to adjust the parameter combination and repeat the above process until the phase - lag feature of the predicted curve matches the measured data time - domain waveform with a satisfactory degree.

[0134] Step S349: Associate the finally verified optimal parameter combination with the quantitative expression of the deformation coupling feature to generate a transfer - function feature update instruction.

[0135] The optimal parameter combination is the parameter combination that, after multiple optimizations and verifications, minimizes the error between the predicted curve output by the main - girder - pier displacement transfer function and the measured data. The quantitative expression of the deformation coupling feature is to represent the deformation coupling feature between the bridge main girder and the pier using numerical values or mathematical expressions. The transfer - function feature update instruction is an instruction used to update the features of the main - girder - pier displacement transfer function. After obtaining the finally verified optimal parameter combination, associate it with the quantitative expression of the deformation coupling feature so that the deformation coupling feature can more accurately reflect the co - deformation relationship between the main girder and the pier. At the same time, generate a transfer - function feature update instruction to update the characteristic parameters of the transfer function and ensure the accuracy and effectiveness of the transfer function. For example, associate parameters such as the stiffness coupling term coefficient and the damping correction term weight in the optimal parameter combination with the quantitative expression of the deformation coupling feature to generate a transfer - function feature update instruction.

[0136] Step S3410: Synchronously adjust the reference threshold for displacement anomaly determination in the bridge health monitoring system based on the update instruction.

[0137] The bridge health monitoring system is a system used to monitor the health status of the bridge in real - time. The reference threshold for displacement anomaly determination is the standard threshold used to judge whether the bridge displacement is abnormal. Based on the transfer - function feature update instruction, synchronously adjust the reference threshold for displacement anomaly determination in the bridge health monitoring system. Because after the parameters of the main - girder - pier displacement transfer function are updated, the displacement response of the bridge may change, and the original reference threshold for displacement anomaly determination may no longer be applicable. Therefore, it is necessary to adjust the reference threshold according to the updated transfer - function features so that the bridge health monitoring system can more accurately judge the displacement anomaly of the bridge.

[0138] Step S400: Generate a comprehensive evaluation index based on the load dynamic characteristics and deformation coupling characteristics. The comprehensive evaluation index is used to quantify the degradation trend of the bearing capacity of the bridge structure during its entire life cycle.

[0139] The load dynamic characteristics are extracted by analyzing the load distribution of the long-term monitoring data sequence, and are characteristic quantities reflecting the load transfer law of the bridge at different monitoring time nodes. The deformation coupling characteristics are extracted by analyzing the deformation correlation of the long-term monitoring data sequence, and are characteristic quantities characterizing the co-deformation relationship between the main girder and the pier column of the bridge. The comprehensive evaluation index is generated by comprehensively analyzing the load dynamic characteristics and deformation coupling characteristics, and is an index used to quantify the degradation trend of the bearing capacity of the bridge structure during its entire life cycle. In practical applications, the bearing capacity of the bridge is jointly affected by the load and the deformation. Therefore, it is necessary to combine the load dynamic characteristics and the deformation coupling characteristics for analysis. By generating the comprehensive evaluation index, the degradation trend of the bearing capacity of the bridge structure during its entire life cycle can be evaluated more comprehensively and accurately, providing a scientific basis for the maintenance and management of the bridge.

[0140] As an implementation method, in step S400, generating a comprehensive evaluation index based on the load dynamic characteristics and deformation coupling characteristics may specifically include:

[0141] Step S410: Establish a load-deformation joint analysis model. The input end of the model receives the equivalent load spectrum parameters in the load dynamic characteristics and the main girder-pier column displacement transfer function parameters in the deformation coupling characteristics.

[0142] The load-deformation joint analysis model is a model for comprehensively analyzing the load dynamic characteristics and deformation coupling characteristics. The equivalent load spectrum parameters are important parameters in the load dynamic characteristics, reflecting the equivalent load conditions borne by the bridge at each monitoring time node. The main girder-pier column displacement transfer function parameters are key parameters in the deformation coupling characteristics, describing the displacement transfer relationship between the main girder and the pier column. When establishing the load-deformation joint analysis model, the equivalent load spectrum parameters and the main girder-pier column displacement transfer function parameters are used as the inputs of the model, enabling the model to consider the effects of both the load and the deformation on the bridge structure at the same time. For example, the load-deformation joint analysis model can adopt a neural network model, taking the equivalent load spectrum parameters and the main girder-pier column displacement transfer function parameters as the input nodes of the input layer, and analyzing the combined action of the load and the deformation through the learning and calculation of the neural network.

[0143] Step S420: Set multiple levels of evaluation nodes in the joint analysis model. Among them, the first-level node is used to calculate the structural stiffness degradation rate, the second-level node is used to calculate the stress redistribution coefficient, and the third-level node is used to calculate the residual bearing reserve.

[0144] The multi-level evaluation nodes are the evaluation nodes at different levels in the load-deformation joint analysis model, and each node has different evaluation functions. The structural stiffness degradation rate refers to the degree of stiffness degradation of the bridge structure over time during its service life. The first-level node analyzes the input equivalent load spectrum parameters and the displacement transfer function parameters of the main girder-pier column, and calculates the structural stiffness degradation rate. The stress redistribution coefficient is a coefficient that reflects the stress redistribution of the bridge structure during the stress process. The second-level node calculates the stress redistribution coefficient based on the calculation results of the first-level node and the input parameters. The residual bearing reserve refers to the additional load that the bridge structure can still bear in the current state. The third-level node combines the calculation results of the previous two levels of nodes and calculates the residual bearing reserve. For example, in the joint analysis model, the first-level node can calculate the structural stiffness degradation rate by comparing the stiffness values of the bridge in the initial state and the current state; the second-level node can calculate the stress redistribution coefficient according to the deformation of the structure and the mechanical properties of the material; the third-level node can calculate the residual bearing reserve according to the current bearing capacity and the design bearing capacity of the structure.

[0145] Step S430: Input the output values of the three-level evaluation nodes into the weighted fusion module to obtain the fused comprehensive evaluation value. The weight coefficients of the weighted fusion module are dynamically adjusted according to the bridge design parameters.

[0146] The weighted fusion module is a module used to fuse the output values of the three-level evaluation nodes. The comprehensive evaluation value is the value obtained by fusing the output values of the three-level evaluation nodes through the weighted fusion module, which comprehensively reflects the bearing performance of the bridge structure. The weight coefficient is the coefficient used to adjust the weights of the output values of each evaluation node in the weighted fusion module, and these weight coefficients are dynamically adjusted according to the bridge design parameters. Different bridge design parameters, such as the type, span, and material of the bridge, will affect the importance of the output values of each evaluation node to the comprehensive evaluation value. For example, for long-span bridges, the structural stiffness degradation rate may have a greater impact on the comprehensive evaluation value, and at this time, the weight coefficient of the output value of the first-level node can be appropriately increased. In practical applications, according to the bridge design parameters, the weight coefficients of the weighted fusion module are determined, and the output values of the three-level evaluation nodes are weighted and summed according to the weights to obtain the fused comprehensive evaluation value.

[0147] Step S440: Convert the fused comprehensive evaluation value into a standardized bearing capacity index through a non-linear mapping function.

[0148] The standardized bearing capacity index is obtained after being transformed by a non - linear mapping function, and it is an index used to uniformly measure the bearing performance of bridges. The fused comprehensive evaluation value may have different dimensions and value ranges. For the convenience of comparison and evaluation, it is necessary to transform it into a standardized bearing capacity index through a non - linear mapping function. For example, the non - linear mapping function can adopt forms such as logarithmic functions and exponential functions to map the fused comprehensive evaluation value to a value range, such as between 0 - 100, to obtain the standardized bearing capacity index.

[0149] Step S450: Compare the bearing capacity index with the historical reference value in terms of trend to generate a comprehensive evaluation index including the current state level and the prediction of the remaining life.

[0150] The historical reference value refers to the reference value of the bearing capacity index of the bridge at different past service stages, which can reflect the normal change trend of the bridge's bearing performance. Trend comparison is the process of comparing the current bearing capacity index with the historical reference value and analyzing its change trend. The current state level is the level divided according to the comparison result of the bearing capacity index and the historical reference value to represent the current bearing performance state of the bridge. The prediction of the remaining life is the prediction of the remaining available service time of the bridge based on the change trend of the bearing capacity index. By comparing the bearing capacity index with the historical reference value in terms of trend, it can be judged whether the bearing performance of the bridge is in a normal state, a declining state or an abnormal state, so as to determine the current state level. At the same time, according to the decline rate and trend of the bearing capacity index, the remaining life of the bridge can be predicted. For example, if the current bearing capacity index is lower than the historical reference value and the decline trend is obvious, it indicates that the bearing performance of the bridge is declining and may be in a risk warning state, and it is necessary to further analyze its remaining life. Finally, the information of the current state level and the prediction of the remaining life are integrated together to generate a comprehensive evaluation index containing this information.

[0151] As an implementation method, step S450, comparing the bearing capacity index with the historical reference value in terms of trend to generate a comprehensive evaluation index including the current state level and the prediction of the remaining life, may specifically include:

[0152] Step S451: Extract the set of historical reference values of the bearing capacity index from the bridge service history database. The set of historical reference values includes the mean value, the fluctuation range and the characteristic parameters of the degradation rate of the bearing capacity index at different service stages.

[0153] The bridge service history database is a database that stores various data of the bridge during its entire service life, including historical data of the bearing capacity index. The historical benchmark value set is extracted from the bridge service history database and contains the mean value, fluctuation range, and degradation rate characteristic parameters of the bearing capacity index at different service stages. The mean value of the bearing capacity index reflects the average load-bearing performance of the bridge at a certain service stage, the fluctuation range indicates the change range of the bearing capacity index at this stage, and the degradation rate characteristic parameter describes the degradation speed of the bearing capacity index over time. For example, by querying and analyzing the bridge service history database, the bearing capacity indexes of the bridge in different years are extracted, and the mean value, fluctuation range, and degradation rate characteristic parameters of each year are calculated to form the historical benchmark value set.

[0154] Step S452: Arrange the bearing capacity index sequences continuously obtained within the current monitoring period in chronological order, and use the trend decomposition algorithm to separate the long-term degradation trend term and the short-term environmental disturbance term.

[0155] The current monitoring period refers to the time period currently under monitoring. The bearing capacity index sequence is a sequence composed of the bearing capacity indexes continuously obtained within the current monitoring period. The trend decomposition algorithm is an algorithm used to decompose a time series into different components. The long-term degradation trend term refers to the long-term downward trend of the bearing capacity index over time, which reflects the influence of factors such as the aging and damage of the bridge structure on the load-bearing performance. The short-term environmental disturbance term refers to the fluctuating influence on the bearing capacity index due to short-term environmental factors, such as climate change and temporary loads. In practical applications, the trend decomposition algorithm, such as seasonal decomposition and wavelet decomposition, is used to decompose the bearing capacity index sequence within the current monitoring period into the long-term degradation trend term and the short-term environmental disturbance term. For example, through the seasonal decomposition algorithm, the bearing capacity index sequence is decomposed into a trend term, a seasonal term, and a residual term, where the trend term is the long-term degradation trend term, and the seasonal term and the residual term can be combined into the short-term environmental disturbance term.

[0156] Step S453: Dynamically time-warp and match the long-term degradation trend term with the benchmark trend curve of the corresponding service stage in the historical benchmark value set, and calculate the trend offset and curvature difference degree.

[0157] Dynamic time warping matching is an algorithm used to compare the similarity of two time series. It can dynamically adjust the two sequences on the time axis to achieve the highest matching degree. The baseline trend curve is the change trend curve of the bearing capacity index corresponding to the service stage in the set of historical baseline values. The trend offset is the degree of vertical offset between the long-term degradation trend term and the baseline trend curve, which reflects the difference between the current bridge bearing performance degradation and the historical situation. The curvature difference degree refers to the difference in curve curvature between the two, which can reflect the change in the bearing performance degradation rate. In practical applications, the dynamic time warping matching algorithm is adopted to match the long-term degradation trend term with the baseline trend curve, and calculate the trend offset and the curvature difference degree. For example, through the dynamic time warping algorithm, the best match between the long-term degradation trend term and the baseline trend curve on the time axis is found, and the vertical distance at each time point is calculated to obtain the trend offset; at the same time, the curvature changes of the two curves are analyzed to calculate the curvature difference degree.

[0158] Step S454: Based on the direction and magnitude of the trend offset, determine whether the stiffness degradation mode of the bridge structure in the current monitoring period belongs to the uniform settlement type, local damage type or overall fatigue type.

[0159] The direction and magnitude of the trend offset reflect the relative position and difference degree between the long-term degradation trend term and the baseline trend curve. The uniform settlement type stiffness degradation mode means that the bridge structure undergoes stiffness degradation uniformly as a whole, similar to the situation of uniform settlement. The local damage type stiffness degradation mode means that a certain local part of the bridge structure is damaged, resulting in a significant decrease in the stiffness of that part. The overall fatigue type stiffness degradation mode means that due to long-term load-bearing, the bridge structure shows overall fatigue damage, resulting in a gradual decrease in stiffness. According to the direction and magnitude of the trend offset, the stiffness degradation mode of the bridge structure can be judged. For example, if the direction of the trend offset is negative and the magnitude is large, and the curve is relatively smooth, it may belong to the overall fatigue type stiffness degradation mode; if the trend offset suddenly increases in a certain local area, it may belong to the local damage type stiffness degradation mode; if the trend offset is relatively uniform throughout the time series, it may belong to the uniform settlement type stiffness degradation mode.

[0160] Step S455: According to the stiffness degradation mode, match the preset remaining life prediction rule base, select the corresponding life prediction model and load its time window length adjustment parameter.

[0161] The preset remaining life prediction rule base is established in advance and is a database containing the remaining life prediction rules corresponding to different stiffness degradation modes. The life prediction model is a model used to predict the remaining life of a bridge. Different stiffness degradation modes may require different life prediction models. The time window length adjustment parameter is a parameter used to adjust the time window length of the life prediction model, and it can adjust the prediction time range of the prediction model according to the actual situation. After determining the stiffness degradation mode of the bridge structure, match it in the preset remaining life prediction rule base and select the corresponding life prediction model. At the same time, load the time window length adjustment parameter of this model to adapt to different prediction requirements.

[0162] Step S456: Input the end slope of the long-term degradation trend term, the curvature difference degree, and the average degradation rate in the historical reference value set into the life prediction model, and iteratively deduce the decay trajectory of the bearing capacity index in multiple future service cycles.

[0163] The end slope of the long-term degradation trend term reflects the degradation speed of the current bearing capacity index, the curvature difference degree reflects the change of the degradation speed, and the average degradation rate in the historical reference value set provides a historical degradation reference. The life prediction model iteratively deduces the decay trajectory of the bearing capacity index in multiple future service cycles by receiving these parameters. In practical applications, take the end slope of the long-term degradation trend term, the curvature difference degree, and the average degradation rate as inputs, substitute them into the calculation formula of the life prediction model, and perform iterative calculations. Each iterative calculation obtains a new bearing capacity index, and so on, to obtain the decay trajectory of the bearing capacity index in multiple future service cycles. For example, the life prediction model can adopt a linear regression model or a non-linear regression model, calculate the bearing capacity index of each future service cycle according to the input parameters, and form a decay trajectory.

[0164] Step S457: During the deduction process, monitor the trajectory similarity between the decay trajectory and the failure case trajectories of similar bridges in the historical reference value set in real time, and dynamically adjust the time window length to optimize the prediction step size.

[0165] The failure case trajectories of similar bridges refer to the change trajectories of the bearing capacity index when bridges with similar structures, usage environments, etc. as the target bridge fail. The trajectory similarity refers to the degree of similarity between the decay trajectory and the failure case trajectories of similar bridges. The time window length is the time range used for prediction in the life prediction model, and the prediction step size refers to the time interval of each prediction. During the process of deducing the decay trajectory of the bearing capacity index in multiple future service cycles, monitor the trajectory similarity between the decay trajectory and the failure case trajectories of similar bridges in the historical reference value set in real time. If the similarity is high, it indicates that the target bridge may have a high failure risk. At this time, the time window length can be dynamically adjusted to shorten the prediction step size in order to more accurately predict the remaining life of the bridge.

[0166] Step S458: When the decay trajectory first touches the preset critical failure threshold, record the corresponding service cycle number as the remaining life prediction value.

[0167] The preset critical failure threshold is a pre-set bearing capacity index threshold used to determine whether the bridge has reached the failure state. When the deduced bearing capacity index decay trajectory first touches the preset critical failure threshold, it indicates that the bridge may fail at this moment. Record the corresponding service cycle number at this time and use it as the remaining life prediction value. For example, the preset critical failure threshold is 20. During the deduction process, when the bearing capacity index decay trajectory first drops to 20, if the corresponding service cycle number is recorded as 5 years, then the remaining life prediction value is 5 years.

[0168] Step S459: Divide the state level into normal operation level, risk warning level or emergency intervention level according to the coverage ratio of the bearing capacity index fluctuation range in the current monitoring period to the historical reference value fluctuation range.

[0169] The bearing capacity index fluctuation range in the current monitoring period refers to the range between the maximum and minimum values of the bearing capacity index in the current monitoring period. The historical reference value fluctuation range is the fluctuation range of the bearing capacity index corresponding to the service stage in the historical reference value set. The coverage ratio refers to the overlapping degree of the bearing capacity index fluctuation range in the current monitoring period and the historical reference value fluctuation range. According to the size of the coverage ratio, the state level of the bridge can be divided. If the coverage ratio is high, it means that the current fluctuation situation of the bearing capacity index is similar to the historical situation, and the bridge is in the normal operation level; if the coverage ratio is moderate, but the fluctuation range has a tendency to exceed the historical range, it means that the bridge may have certain risks and is in the risk warning level; if the coverage ratio is low and the fluctuation range greatly exceeds the historical range, it means that the bridge may face an emergency and is in the emergency intervention level.

[0170] Step S4510: Encode the remaining life prediction value, state level and stiffness degradation mode into a structured evaluation vector, and embed the time stamp and spatial location identifier.

[0171] The structured evaluation vector is a vector formed by encoding information such as the remaining life prediction value, state level and stiffness degradation mode. The time stamp is used to record the time of evaluation, and the spatial location identifier is used to indicate the specific location of the bridge. After obtaining the remaining life prediction value, state level and stiffness degradation mode, encode them to form a structured evaluation vector. At the same time, embed the time stamp and spatial location identifier so that the evaluation result has time and spatial information. For example, represent the remaining life prediction value with a numerical value, represent the state level with a classification code, and represent the stiffness degradation mode with another classification code. Combine these codes into a vector, and then add the time stamp and spatial location identifier to form a structured evaluation vector.

[0172] Step S4511: Retrieve through similarity search with the historical evaluation case base to correct the short-term evaluation bias caused by environmental disturbance terms.

[0173] The historical evaluation case base is a database that stores the case information of bridge evaluations in history. Similarity search refers to the process of finding cases similar to the current evaluation vector in the historical evaluation case base. Environmental disturbance terms are separated in step S452. Due to the fluctuating impact of short-term environmental factors on the bearing capacity index, they may cause short-term evaluation biases. By performing a similarity search of the current evaluation vector with the historical evaluation case base, finding similar historical cases, analyzing the impact and correction methods of environmental disturbance terms in these cases, the current evaluation results are corrected to eliminate the short-term evaluation bias caused by environmental disturbance terms. For example, find a case with a high similarity to the current evaluation vector in the historical evaluation case base, and find that the environmental disturbance term in this case causes the evaluation result to be too high. By analyzing its correction method, the current evaluation result is adjusted accordingly.

[0174] Step S4512: Convert the corrected evaluation vector into a comprehensive evaluation index that includes a visualized degradation path diagram, maintenance priority suggestions, and a list of key monitoring areas.

[0175] The visualized degradation path diagram is a chart that graphically shows the degradation process of the bridge's bearing capacity, which can intuitively reflect the changing trend of the bridge's bearing capacity in the future for a period of time. Maintenance priority suggestions are suggestions for ranking the priorities of bridge maintenance work based on the evaluation results. The list of key monitoring areas is a list that lists the areas of the bridge that need to be monitored with priority, and these areas may be areas prone to problems or areas where damage has already occurred. Convert the corrected evaluation vector to generate a comprehensive evaluation index that includes a visualized degradation path diagram, maintenance priority suggestions, and a list of key monitoring areas. For example, draw a visualized degradation path diagram based on the remaining life prediction value and the decay trajectory of the bearing capacity index in the evaluation vector; formulate maintenance priority suggestions based on the state level and stiffness degradation mode; determine the list of key monitoring areas based on the damage conditions found during the evaluation process.

[0176] Step S4513: Dynamically associate the comprehensive evaluation index with the real-time monitoring data in the bridge digital twin model to trigger the model self-update mechanism to synchronously reflect the latest structural state.

[0177] The digital twin model of a bridge is a virtual model obtained by digitally modeling the physical entity of the bridge, which can reflect the structural state of the bridge in real time. The comprehensive evaluation index is dynamically associated with the real-time monitoring data in the digital twin model of the bridge, so that the comprehensive evaluation index can be combined with the real-time monitoring data. When the comprehensive evaluation index changes, the model self-update mechanism is triggered to update the digital twin model of the bridge to synchronously reflect the latest structural state of the bridge. For example, comprehensive evaluation indexes such as the visualized degradation path map, maintenance priority suggestions, and list of key monitoring areas are associated with real-time displacement, stress and other monitoring data in the digital twin model of the bridge. When the comprehensive evaluation index is updated, the relevant parameters and visual display in the digital twin model of the bridge are automatically updated, so that the model can accurately reflect the latest situation of the bridge.

[0178] Step S500: Determine the bearing capacity state of the bridge in the current and future service stages according to the comparison result between the comprehensive evaluation index and the preset bearing capacity threshold.

[0179] The preset bearing capacity threshold is a standard threshold set in advance for dividing the bearing capacity state of the bridge. By comparing the comprehensive evaluation index with the preset bearing capacity threshold, the bearing capacity state of the bridge in the current and future service stages can be judged. The comprehensive evaluation index comprehensively reflects the bearing performance of the bridge, including information such as the current state level and remaining life prediction. The preset bearing capacity threshold can be set according to factors such as the design requirements and safety standards of the bridge. For example, according to the comparison result, it can be judged whether the bridge is in a safe service state, a state where traffic needs to be restricted, or a state where it needs to be urgently closed. At the same time, combined with the remaining life prediction and the future bearing capacity index decay trajectory in the comprehensive evaluation index, the change trend of the bearing capacity state of the bridge in the future service stage can also be predicted.

[0180] As an implementation method, step S500, determining the bearing capacity state of the bridge in the current and future service stages according to the comparison result between the comprehensive evaluation index and the preset bearing capacity threshold, may specifically include:

[0181] Step S510: Compare the bearing capacity index in the comprehensive evaluation index with the preset bearing capacity threshold level by level. The preset bearing capacity threshold includes the normal use state threshold, the restricted traffic state threshold, and the emergency closure state threshold.

[0182] The bearing capacity index is an important part of the comprehensive evaluation index, which reflects the current bearing performance of the bridge. The preset bearing capacity thresholds include the normal use state threshold, the restricted access state threshold, and the emergency closure state threshold, which are set according to factors such as the design standards and safety requirements of the bridge. The normal use state threshold is the lower limit of the bearing capacity index when the bridge can be normally accessed and used. When the bearing capacity index is higher than this threshold, the bridge is in the normal use state. The restricted access state threshold is when the bearing capacity index is lower than this threshold, restrictions need to be imposed on the access of the bridge, such as restricting the load and speed of vehicles. The emergency closure state threshold is when the bearing capacity index is lower than this threshold, the bridge needs to be urgently closed to prohibit the passage of vehicles and pedestrians to ensure safety. In actual operation, the bearing capacity index in the comprehensive evaluation index is compared step by step with these preset bearing capacity thresholds to judge the current bearing state of the bridge. For example, first compare the bearing capacity index with the normal use state threshold. If it is higher than this threshold, the bridge is in the normal use state; if it is lower than the normal use state threshold, continue to compare it with the restricted access state threshold, and so on.

[0183] Step S520: Determine the current bearing capacity state according to the interval membership relationship between the bearing capacity index and the preset bearing capacity threshold. The current bearing capacity state includes the safe service state, the restricted access preparation state, and the emergency closure trigger state.

[0184] The interval membership relationship refers to the relationship that the bearing capacity index falls within different intervals divided by the preset bearing capacity threshold. According to the interval membership relationship between the bearing capacity index and the preset bearing capacity threshold, the current bearing capacity state of the bridge can be determined. If the bearing capacity index is higher than the normal use state threshold, the bridge is in the safe service state, and at this time the bridge can be normally accessed and used. If the bearing capacity index is lower than the normal use state threshold but higher than the restricted access state threshold, the bridge is in the restricted access preparation state, and it is necessary to closely monitor the state of the bridge and may take some preventive restricted access measures. If the bearing capacity index is lower than the restricted access state threshold, the bridge is in the emergency closure trigger state, and it is necessary to immediately close the bridge to prohibit the passage of vehicles and pedestrians.

[0185] Step S530: Input the historical bearing capacity index sequence of the comprehensive evaluation index into the time series prediction model, and output the future bearing capacity index prediction sequence for the next three monitoring periods.

[0186] The historical bearing capacity index sequence is a sequence composed of the bearing capacity indexes of the bridge recorded in the comprehensive evaluation index over a past period of time. The time series prediction model is a model used to predict time series data. It can predict future data values according to the changing trend of historical data. Inputting the historical bearing capacity index sequence into the time series prediction model, through the analysis and learning of historical data, the model outputs the future bearing capacity index prediction sequence for the next three monitoring cycles. For example, the time series prediction model can adopt the autoregressive integrated moving average model (ARIMA), long short-term memory network (LSTM), etc. Taking the historical bearing capacity index sequence as the input, through the calculation of the model, the predicted values of the bearing capacity indexes for the next three monitoring cycles are obtained, forming the future bearing capacity index prediction sequence.

[0187] Step S540: Compare the future bearing capacity index prediction sequence with the preset bearing capacity threshold value cycle by cycle, and identify the future monitoring cycle when the future bearing capacity index prediction sequence is first lower than the threshold level corresponding to the current monitoring cycle.

[0188] Comparing cycle by cycle means the process of comparing each predicted value in the future bearing capacity index prediction sequence with the preset bearing capacity threshold value one by one. The threshold level corresponding to the current monitoring cycle refers to the preset bearing capacity threshold level corresponding to the bridge state of the current monitoring cycle, such as the threshold value in the normal use state, the threshold value in the restricted passage state, or the threshold value in the emergency closure state. By comparing the future bearing capacity index prediction sequence with the preset bearing capacity threshold value cycle by cycle, find the future monitoring cycle when the future bearing capacity index prediction sequence is first lower than the threshold level corresponding to the current monitoring cycle. For example, when the bridge is in a safe service state in the current monitoring cycle, corresponding to the threshold value in the normal use state, compare each predicted value in the future bearing capacity index prediction sequence with the threshold value in the normal use state, and find the future monitoring cycle corresponding to the first predicted value lower than this threshold value.

[0189] Step S550: Determine the degradation trend of the bearing capacity state in the future service stage according to the time interval and the index decline gradient of the future monitoring cycle when it is first lower than the current threshold level, and generate a bearing capacity state decision report including the current bearing capacity state classification and the future bearing capacity state warning level.

[0190] The time interval refers to the time difference between the future monitoring cycle when it first drops below the current threshold level and the current monitoring cycle. The exponential decline gradient refers to the decline rate of the bearing capacity index in the process of the future bearing capacity index prediction sequence approaching and dropping below the current threshold level. Based on the time interval and the exponential decline gradient, the degradation trend of the bearing capacity state in the future service stage can be determined. If the time interval is short and the exponential decline gradient is large, it indicates that the bearing capacity state of the bridge will degrade rapidly and face a high risk in the future; if the time interval is long and the exponential decline gradient is small, it indicates that the degradation of the bearing capacity state of the bridge is relatively slow. According to the current bearing capacity state classification (such as safe service state, restricted access preparation state, emergency closure trigger state) and the future bearing capacity state degradation trend, a future bearing capacity state warning level is generated, such as low-risk warning, medium-risk warning, high-risk warning, etc. Finally, the current bearing capacity state classification and the future bearing capacity state warning level are integrated to generate a bearing capacity state decision report, providing a decision-making basis for the management and maintenance of the bridge.

[0191] As an implementation manner, the training process of the time series prediction model includes:

[0192] Step S5301: Extract the bearing capacity index time series from the historical monitoring data, and intercept continuous degradation feature segments according to a preset cycle length.

[0193] The historical monitoring data is the data accumulated from the long-term monitoring of the bridge, which contains the historical records of the bearing capacity index. The bearing capacity index time series is a sequence composed of the bearing capacity indexes extracted from the historical monitoring data and arranged in chronological order. The preset cycle length is the time length preset for intercepting continuous degradation feature segments. The continuous degradation feature segment refers to the time period during which the bearing capacity index shows a continuous degradation trend within a certain period. In actual operation, the bearing capacity index data is screened from the historical monitoring data to form the bearing capacity index time series. Then, according to the preset cycle length, continuous degradation feature segments are intercepted in the time series.

[0194] Step S5302: Use an adaptive decomposition algorithm to decompose each feature segment into multiple intrinsic mode components and a trend remainder.

[0195] The adaptive decomposition algorithm is an algorithm that can automatically decompose according to the characteristics of the data itself. The intrinsic mode components refer to the components with different frequencies and amplitudes obtained by decomposing the feature segments through the adaptive decomposition algorithm, and each intrinsic mode component represents a variation pattern in the data. The trend residue is the part remaining after decomposition that represents the long-term trend of the data. In practical applications, the adaptive decomposition algorithm, such as empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD), etc., is used to decompose each continuous degradation feature segment into multiple intrinsic mode components and a trend residue. For example, using the empirical mode decomposition algorithm to decompose a continuous degradation feature segment, multiple intrinsic mode components are obtained, each component having different frequencies and amplitudes, reflecting the variation of the bearing capacity index on different time scales, and at the same time a trend residue is obtained, representing the long-term degradation trend of this feature segment.

[0196] Step S5303: Extract the time-frequency domain energy distribution characteristics for each intrinsic mode component and establish the associated mapping relationship with the external environmental variables.

[0197] The time-frequency domain energy distribution characteristics refer to the energy distribution of the intrinsic mode component in the time and frequency domains, which can reflect the energy concentration degree and variation law of the intrinsic mode component. The external environmental variables refer to the external factors that may affect the bridge bearing capacity index, such as temperature, humidity, traffic flow, etc. For each intrinsic mode component, time-frequency analysis methods, such as wavelet transform, short-time Fourier transform, etc., are used to extract its energy distribution characteristics in the time-frequency domain. Then, analyze the relationship between these characteristics and the external environmental variables and establish the associated mapping relationship. For example, by performing wavelet transform on a certain intrinsic mode component, its energy distribution characteristics in the time-frequency domain are obtained, and then analyze the correlation between this characteristic and external environmental variables such as temperature and traffic flow, and establish the associated mapping relationship to consider the influence of external environmental factors during prediction.

[0198] Step S5304: Configure independent autoregressive predictors for each mode component according to the associated mapping relationship and load the covariance matrix of the corresponding environmental variables.

[0199] An autoregressive predictor is a predictor based on an autoregressive model that can predict future data according to historical data. The covariance matrix is a matrix used to describe the correlation between external environmental variables. According to the association mapping relationship established in step S5303, an independent autoregressive predictor is configured for each intrinsic mode component. Each autoregressive predictor can make predictions based on the historical data of that mode component and the associated external environmental variables. At the same time, the covariance matrix of the corresponding environmental variables is loaded to consider the correlation between external environmental variables during the prediction process. For example, for an intrinsic mode component, according to its association mapping relationship with temperature and traffic flow, an autoregressive predictor is configured, and the covariance matrices of temperature and traffic flow are loaded. This predictor can predict the future value of this mode component based on the historical data of this mode component, the current values of temperature and traffic flow, and their correlation.

[0200] Step S5305: Input the trend remainder into the non-linear extended predictor and generate a long-term degradation baseline in combination with the bridge material aging parameters.

[0201] A non-linear extended predictor is a predictor that can handle non-linear relationships and can further predict the trend remainder. The bridge material aging parameters are parameters that describe the aging degree of bridge materials during use, such as the carbonation depth of concrete and the corrosion rate of steel. The trend remainder is input into the non-linear extended predictor and combined with the bridge material aging parameters to generate a long-term degradation baseline. The long-term degradation baseline reflects the degradation trend of the bridge bearing capacity index in the long term and takes into account the influence of factors such as bridge material aging. For example, the non-linear extended predictor can adopt a neural network model. Taking the trend remainder and the bridge material aging parameters as inputs, through the learning and calculation of the model, a long-term degradation baseline is generated, which can be used to predict the change trend of the bridge bearing capacity index over a long period of time.

[0202] Step S5306: Call the parameter coordinator to perform time-domain alignment and energy ratio calibration on the outputs of each predictor to generate a fused prediction result.

[0203] The parameter coordinator is a module used to coordinate the outputs of each predictor. It can perform time-domain alignment and energy ratio calibration on the outputs of each predictor. Time-domain alignment means aligning the outputs of each predictor in the time dimension so that they have the same starting time and time interval. Energy ratio calibration means performing weighted summation on the outputs of each predictor according to the energy ratios of each intrinsic mode component and the trend remainder to obtain a more accurate fused prediction result. In practical applications, the parameter coordinator is called to perform time-domain alignment on the prediction outputs of each autoregressive predictor for the intrinsic mode components and the prediction output of the non-linear extended predictor for the trend remainder, and then perform weighted summation according to the energy ratios of each component to generate a fused prediction result.

[0204] Step S5307: Compare the morphological similarity between the fusion result and the look-ahead window of the measured data, and extract the spatial distribution hotspots of the prediction error.

[0205] The look-ahead window refers to a data window of a future period selected from the measured data for comparison with the fusion prediction result. The morphological similarity comparison refers to comparing the shape and trend between the fusion prediction result and the look-ahead window of the measured data to analyze their similarity degree. The spatial distribution hotspots of the prediction error refer to the regions or time periods with relatively large errors in the prediction error. By comparing the morphological similarity between the fusion prediction result and the look-ahead window of the measured data, and calculating similarity metrics between the two, such as the correlation coefficient, mean square error, etc., the accuracy of the prediction is evaluated. At the same time, analyze the distribution of the prediction error in time and space, and extract the spatial distribution hotspots of the prediction error. For example, select the data of the next 10 monitoring cycles in the measured data as the look-ahead window, compare the fusion prediction result with this look-ahead window, calculate the correlation coefficient. If the correlation coefficient is low, it indicates that the prediction accuracy is poor. Further analyze the distribution of the prediction error within these 10 monitoring cycles to find the time periods with relatively large errors as the spatial distribution hotspots of the prediction error.

[0206] Step S5308: Adjust the weight allocation strategy of each predictor according to the hotspot distribution, and focus on strengthening the prediction logic of the modal components associated with the hotspot regions.

[0207] The weight allocation strategy refers to the way of weight allocation when performing weighted summation on the outputs of each predictor in the parameter coordinator. The hotspot region refers to the region or time period corresponding to the spatial distribution hotspots of the prediction error. According to the spatial distribution hotspots of the prediction error, adjust the weight allocation strategy of each predictor, and increase the weight of the prediction logic of the modal components associated with the hotspot regions. Because the prediction error in the hotspot regions is relatively large, it indicates that the prediction of the modal components corresponding to these regions may be inaccurate, and it is necessary to focus on strengthening their prediction logic. For example, if the spatial distribution hotspots of the prediction error are concentrated in a certain time period, and this time period is mainly related to a certain intrinsic modal component, then increase the weight of the autoregressive predictor corresponding to this intrinsic modal component, and at the same time optimize the parameters and model structure of this predictor to improve its prediction accuracy.

[0208] Step S5309: Connect the optimized predictor combination to the real-time data stream interface to establish a dynamic parameter exchange channel between the prediction model and the monitoring system.

[0209] The real-time data stream interface is an interface for receiving real-time monitoring data. The dynamic parameter exchange channel refers to the channel through which the prediction model and the monitoring system can exchange parameters in real time. Connect the optimized predictor combination to the real-time data stream interface so that the prediction model can obtain the latest data of the monitoring system in real time. At the same time, establish a dynamic parameter exchange channel between the prediction model and the monitoring system. During the prediction process, if the parameters of the monitoring system change, such as changes in external environmental variables, updates of bridge structure parameters, etc., these changes can be transmitted to the prediction model in real time through the dynamic parameter exchange channel, and the prediction model can adjust the prediction results in a timely manner according to these changes. For example, when the monitoring system detects a sudden increase in temperature, it transmits the temperature change information to the prediction model through the dynamic parameter exchange channel, and the prediction model can adjust the prediction results according to the correlation mapping relationship between temperature and each modal component.

[0210] Step S5310: Based on the real-time feedback data in the exchange channel, perform online fine-tuning of the predictor parameters to maintain the model prediction accuracy.

[0211] The real-time feedback data refers to the data obtained from the monitoring system in real time through the dynamic parameter exchange channel. Online fine-tuning refers to the process of adjusting the parameters of the predictor in real time according to the real-time feedback data during the operation of the model. Based on the real-time feedback data in the exchange channel, perform online fine-tuning of the predictor parameters to maintain the prediction accuracy of the model. Because the actual situation of the bridge is constantly changing, the real-time feedback data can reflect these changes. By performing online fine-tuning of the predictor parameters, the prediction model can better adapt to these changes and improve the prediction accuracy. For example, when the real-time feedback data shows that the change trend of a certain intrinsic modal component deviates from the prediction result, fine-tune the parameters of the autoregressive predictor corresponding to this modal component according to the deviation situation, such as adjusting the autoregressive coefficient, etc., to improve the prediction accuracy.

[0212] Please refer to Figure 3, is a structural block diagram of the data analysis system 120 of the present application. The data analysis system 120 includes a computing unit 1001, which can execute various appropriate actions and processes according to the computer program stored in the ROM (i.e., read-only memory) 1002 or the computer program loaded from the storage unit 1008 into the RAM (i.e., random access memory) 1003. In the RAM 1003, various programs and data required for the operation of the data analysis system 120 can also be stored. The computing unit 1001, the ROM 1002, and the RAM 1003 are connected to each other through a bus 1004. The input / output (I / O) interface 1005 is also connected to the bus 1004. Multiple components in the data analysis system 120 are connected to the I / O interface 1005, including: an input unit 1006, an output unit 1007, a storage unit 1008, and a communication unit 1009. The input unit 1006 can be any type of device that can input information into the data analysis system 120. The output unit 1007 can be any type of device that can present information. The storage unit 1008 can include, but is not limited to, magnetic disks and optical disks.

Claims

1. A data analysis method for the bearing capacity of a long-span continuous rigid frame bridge, characterized in that, Including: Obtain the long-term monitoring data sequence of the target bridge, where the long-term monitoring data sequence includes multiple structural response data sets arranged according to monitoring time nodes during the service life cycle of the bridge; Separate the load distribution data subset from the long-term monitoring data sequence. The load distribution data subset contains the bending moment monitoring values and shear force monitoring values at the mid-span and support sections of the bridge under each monitoring time node. Decompose the load distribution data subset into a dead load component sequence and a live load component sequence according to a preset load decomposition level. Among them, the dead load component sequence extracts the long-term trend term through low-pass filtering, and the live load component sequence extracts the short-term fluctuation term through high-pass filtering. Use the time-varying weight allocation algorithm to dynamically superimpose the dead load component sequence and the live load component sequence to generate the equivalent load spectrum corresponding to each monitoring time node. Call the pre-trained load transfer model to perform spatial interpolation calculation on the equivalent load spectrum to obtain the load distribution cloud map of the entire bridge span. Take the geometric parameters and mechanical parameters of the maximum gradient change region in the load distribution cloud map as the load dynamic characteristics, and the load dynamic characteristics are used to characterize the load transfer law of each key section of the bridge under different monitoring time nodes; Extract the main girder deflection monitoring sequence and the pier inclination monitoring sequence from the long-term monitoring data sequence, and establish the time-domain synchronous correspondence relationship between the main girder deflection monitoring sequence and the pier inclination monitoring sequence. Extract the local deformation waveform on the main girder deflection monitoring sequence through the time window sliding algorithm, and extract the phase offset in the pier inclination monitoring sequence of the corresponding time window. Construct a main girder-pier displacement transfer function, input the local deformation waveform into the input end of the transfer function, and use the phase offset as the feedback correction term. Use the iterative approximation algorithm to optimize the parameter combination of the main girder-pier displacement transfer function until the error between the predicted pier displacement curve at the output end and the measured data is lower than the set tolerance. Take the characteristic parameter set of the optimized main girder-pier displacement transfer function as the deformation coupling characteristic, and the deformation coupling characteristic is used to characterize the cooperative deformation relationship between the main girder and the pier of the bridge; Generate a comprehensive evaluation index based on the load dynamic characteristics and the deformation coupling characteristics, and the comprehensive evaluation index is used to quantify the degradation trend of the bearing capacity of the bridge structure during the whole life cycle; Determine the bearing capacity state of the bridge in the current and future service stages according to the comparison result between the comprehensive evaluation index and the preset bearing capacity threshold.

2. The method according to claim 1, wherein The generating of the comprehensive evaluation index based on the load dynamic characteristics and the deformation coupling characteristics includes: Establish a load-deformation joint analysis model, and the input end of the model receives the equivalent load spectrum parameters in the load dynamic characteristics and the main girder-pier displacement transfer function parameters in the deformation coupling characteristics; Set multiple evaluation nodes in the joint analysis model. Among them, the first-level node is used to calculate the structural stiffness degradation rate, the second-level node is used to calculate the stress redistribution coefficient, and the third-level node is used to calculate the residual bearing reserve; Input the output value of the three - level evaluation node into the weighted fusion module to obtain the fused comprehensive evaluation value, where the weight coefficients of the weighted fusion module are dynamically adjusted according to the bridge design parameters; Convert the fused comprehensive evaluation value into a standardized bearing capacity index through a non - linear mapping function; Compare the trend of the bearing capacity index with the historical reference value to generate a comprehensive evaluation index including the current state level and the prediction of the remaining life; 3. The method according to claim 2, characterized in that Determine the bearing capacity state of the bridge in the current and future service stages according to the comparison result between the comprehensive evaluation index and the preset bearing capacity threshold, including: Compare the bearing capacity index in the comprehensive evaluation index with the preset bearing capacity thresholds step by step. The preset bearing capacity thresholds include the normal use state threshold, the restricted passage state threshold, and the emergency closure state threshold; Determine the current bearing capacity state according to the interval membership relationship between the bearing capacity index and the preset bearing capacity threshold. The current bearing capacity state includes the safe service state, the restricted passage preparation state, and the emergency closure trigger state; Input the historical bearing capacity index sequence of the comprehensive evaluation index into the time - series prediction model to output the future bearing capacity index prediction sequence for the next three monitoring periods; Compare the future bearing capacity index prediction sequence with the preset bearing capacity thresholds period by period to identify the future monitoring period when the future bearing capacity index prediction sequence is first lower than the threshold level corresponding to the current monitoring period; Determine the degradation trend of the bearing capacity state in the future service stage according to the time interval and the index decline gradient of the future monitoring period when it is first lower than the current threshold level, and generate a bearing capacity state decision report including the classification of the current bearing capacity state and the early warning level of the future bearing capacity state; 4. The method according to claim 1, wherein The optimization process of the main girder - pier displacement transfer function includes: Extract the stiffness coupling term coefficient and the damping correction term weight from the initial parameter combination of the transfer function; Call the main girder deflection monitoring sequence and the pier inclination monitoring sequence in the historical displacement dataset to generate a displacement transfer verification sample set; Input the verification sample set into the transfer function for forward calculation to output the predicted pier displacement curve; Extract the gradient difference distribution characteristics between the predicted curve and the measured curve to generate a parameter correction direction vector; Adjust the spatial weight distribution of the stiffness coupling term coefficient and the time decay factor of the damping correction term weight according to the correction direction vector; Input the adjusted parameter combination into the digital twin verification interface to perform time - frequency characteristic matching with the real - time monitored pier dynamic response data; When the deviation of the energy ratio of the low - frequency component in the matching result exceeds the preset threshold, trigger the anisotropic compensation mechanism of the stiffness coupling term coefficient; Re - execute the displacement transfer calculation with the compensated parameter combination until the phase - lag characteristic of the predicted curve is in line with the measured data time - domain waveform; Associate the finally verified optimal parameter combination with the quantitative expression of the deformation coupling characteristic to generate a transfer function characteristic update instruction; Synchronously adjust the reference threshold for displacement anomaly determination in the bridge health monitoring system based on the update instruction; 5. The method according to claim 3, characterized in that, The training method of the time - series prediction model includes: Extract the bearing capacity index time series from historical monitoring data and intercept continuous degradation feature segments according to a preset cycle length; Use an adaptive decomposition algorithm to decompose each feature segment into multiple intrinsic mode components and a trend residue; Extract time-frequency domain energy distribution features for each intrinsic mode component and establish an associated mapping relationship with external environmental variables; Configure an independent autoregressive predictor for each mode component according to the associated mapping relationship and load the covariance matrix of the corresponding environmental variables; Input the trend residue into a non-linear extended predictor and generate a long-term degradation baseline in combination with bridge material aging parameters; Call a parameter coordinator to perform time-domain alignment and energy ratio calibration on the outputs of each predictor to generate a fused prediction result; Compare the morphological similarity between the fused result and the forward-looking window of the measured data to extract the spatial distribution hotspots of the prediction error; Adjust the weight allocation strategy of each predictor according to the hotspot distribution, and focus on strengthening the prediction logic of the mode components associated with the hotspot area; Connect the optimized predictor combination to the real-time data stream interface to establish a dynamic parameter exchange channel between the prediction model and the monitoring system; Based on the real-time feedback data in the exchange channel, perform online fine-tuning of the predictor parameters to maintain the prediction accuracy of the model.

6. The method according to claim 1, characterized in that The verification method of the load transfer model includes: Obtain the measured three-dimensional displacement field data of the target bridge under the standard load test condition, and the measured three-dimensional displacement field data includes the spatial coordinate changes of the mid-span of the main girder and the pier top section; Input the load parameters of the standard load test condition into a pre-established parametric finite element model, and generate theoretical displacement field distribution data after performing multi-scale mesh division; Extract the absolute error vector between the measured vertical displacement value at the mid-span of the main girder in the measured three-dimensional displacement field data and the theoretical value at the corresponding position in the theoretical displacement field distribution data, and calculate the relative error distribution pattern between the measured horizontal displacement value and the theoretical value at the pier top at the same time; Identify the correction priority of the main girder stiffness parameters according to the direction angle of the absolute error vector, and preferentially adjust the elastic modulus adjustment factor orthogonal to the direction of the maximum absolute error; Based on the spatial correlation of the relative error distribution pattern, synchronously correct the boundary stiffness ratio coefficient of the pier column foundation constraint condition; Re-import the corrected elastic modulus adjustment factor and boundary stiffness ratio coefficient into the parametric finite element model, and iteratively perform load-displacement response calculations until the norm of the absolute error vector and the relative error rate of all measurement points are lower than the preset convergence threshold; Extract the mid-span bending moment distribution cloud map and the pier bottom shear force gradient curve from the output of the finally converged finite element model to generate a load transfer feature verification fingerprint; Perform dynamic spatial registration on the load transfer feature verification fingerprint and the real-time monitored load distribution cloud map, and calculate the morphological similarity index of the two in the gradient change area; When the morphological similarity index is lower than the historical baseline level for three consecutive monitoring periods, trigger a parameter recalibration instruction for the load transfer model; Call the latest measured three-dimensional displacement field data according to the recalibration instruction, re-execute the correction process of the elastic modulus adjustment factor and the boundary stiffness ratio coefficient, and update the material constitutive relationship of the finite element model; Inject the theoretical displacement field distribution data output by the updated finite element model into the training dataset of the load transfer model in reverse to replace the outdated response features in the historical samples; Perform online retraining of the load transfer model based on the injected training dataset, optimize the weight allocation strategy in the spatial interpolation calculation, and form an adaptive verification closed loop for the load distribution analysis module.

7. The method according to claim 1, characterized in that The dynamic superposition of the dead load component sequence and the live load component sequence using the time-varying weight allocation algorithm to generate the equivalent load spectrum corresponding to each monitoring time node includes: Initialize the dead load weight coefficient corresponding to each monitoring time node based on the long-term trend slope of each monitoring time node in the dead load component sequence; Initialize the live load weight coefficient corresponding to each monitoring time node based on the short-term fluctuation amplitude of each monitoring time node in the live load component sequence; Extract the temperature gradient change rate, humidity cumulative effect factor, and traffic flow impact index from the real-time monitoring data set of the bridge environment to generate an environmental dynamic impact set; Input the environmental dynamic impact set into the weight correction model to perform non-linear coupling adjustment on the dead load weight coefficient and the live load weight coefficient to obtain the corrected dead load weight and the corrected live load weight at each monitoring time node; Adopt a sliding time window mechanism to perform a moving average process on the corrected dead load weights of the current monitoring time node and the previous preset number of nodes to generate a smoothed dead load weight sequence; Synchronously perform an exponential decay process on the corrected live load weights of the current monitoring time node and the previous preset number of nodes to generate a decaying live load weight sequence; Multiply the smoothed dead load weight sequence by the data at the corresponding time node in the dead load component sequence according to the weight to obtain the dead load weighted component; Multiply the decaying live load weight sequence by the data at the corresponding time node in the live load component sequence according to the weight to obtain the live load weighted component; Perform vector superposition on the dead load weighted component and the live load weighted component at the same monitoring time node to generate an initial equivalent load spectrum; Call the load distribution verification model to perform spatio-temporal alignment comparison between the initial equivalent load spectrum and the typical load patterns of bridges of the same type in the historical load distribution feature library; According to the load peak position deviation and distribution uniformity difference in the comparison result, reversely adjust the coupling adjustment parameters of the weight correction model; Re-execute the weight correction and subsequent superposition steps using the adjusted weight correction model until the spatio-temporal distribution form of the initial equivalent load spectrum reaches a preset threshold of similarity with the historical typical load pattern; Mark the initial equivalent load spectrum that passes the verification as the final equivalent load spectrum and associate it with the structural response data set corresponding to the monitoring time node.

8. A data analysis system, characterized in that, Including: At least one processor; And a memory communicatively connected to the at least one processor; Wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method according to any one of claims 1 to 7.

Images