A method and system for intelligent evaluation and abnormality identification of cable-stayed cable operation period behavior

Abstract

The application discloses a method and system for intelligent evaluation and abnormality identification of cable-stayed cable operation period behavior, and the method comprises the following steps: completing a multi-source data acquisition system construction; completing data preprocessing and establishing a data time axis unified template; extracting and fusing features of the data, reorganizing the data according to typical performance dimensions of the cable-stayed cable and based on theme types; completing standardization and structuralization processing of all sensing data; constructing a dynamic input vector containing time and space correlation characteristics of multi-source heterogeneous data by using A / B type core indicators, mapping nonlinear characteristics by using a self-adaptive optimized kernel function, and improving parameter optimization efficiency by using a cross-validation and Bayesian optimization fusion mode to complete model training; completing cable-stayed cable state evaluation and output based on the trained model, and binding the model with a bridge model to automatically complete visual disease labeling. The application provides a scientific basis for cable-stayed cable whole life cycle management.

Description

Technical Field

[0001] This invention belongs to the field of bridge cable-stayed bridge monitoring technology, and more specifically, relates to a method and system for intelligent assessment and anomaly identification of the operational performance of cable-stayed bridges. Background Technology

[0002] As a key load-bearing structure of cable-stayed bridges, the performance of stay cables directly determines the safe operation and service life of the bridge. During long-term service, stay cables continuously endure complex effects such as vehicle loads, temperature changes, and wind and rain erosion, inevitably leading to problems such as cable tension fluctuations, wire corrosion, sheath damage, and anchorage zone failure. If these problems are not detected and addressed in a timely manner, they may gradually deteriorate and trigger a chain reaction. At best, this can lead to a decline in bridge performance; at worst, it can cause structural collapse, threatening life and property. Furthermore, emergency repairs and maintenance can disrupt traffic, resulting in significant economic losses and social impact.

[0003] Currently, there are many practical challenges in the condition management of cable-stayed bridges. At the data acquisition level, monitoring methods are fragmented: while automated sensors can acquire dynamic data such as cable force and vibration in real time, they struggle to capture structural defects such as sheath appearance and internal wire breaks; manual inspections, while recording visible damage, are limited by long cycles, strong subjectivity, and limited coverage; specialized inspections, although highly accurate, often target local components and fail to reflect the overall condition. This decentralized acquisition model results in fragmented data, making it impossible to form a complete condition profile.

[0004] In the data processing and evaluation stage, existing methods focus more on the independent analysis of single indicators, lacking in-depth integration of multi-source data. This makes it difficult to uncover the correlation patterns between different parameters, leading to a one-sided judgment on the true state of the cable-stayed bridge. At the same time, anomaly identification often relies on experience-based judgment or simple models, which are poorly adaptable to nonlinear characteristics in complex environments, easily resulting in missed or false positives, and failing to meet the needs of accurate early warning.

[0005] These current circumstances have rendered the full lifecycle management of cable-stayed bridges reactive, making it impossible to anticipate potential risks or formulate scientific maintenance strategies. Therefore, building a technological system capable of comprehensive perception, accurate assessment, and timely early warning is of great significance for improving cable-stayed bridge management, ensuring bridge safety, and reducing operation and maintenance costs. It is also a key issue that urgently needs to be addressed in the field of bridge engineering. Summary of the Invention

[0006] This invention aims to address the problems of scattered and insufficiently integrated data, as well as the strong subjectivity and low accuracy of assessment methods in the monitoring and evaluation of cable-stayed bridges during their operational phase. By constructing a multi-source data acquisition system, it achieves the fusion and structured processing of monitoring and detection data. Combined with intelligent model training, it improves the accuracy of anomaly identification and the reliability of condition assessment, providing a scientific basis for the full life-cycle management of cable-stayed bridges, supporting intelligent maintenance decisions, and ensuring the safe operation of bridges.

[0007] To address the aforementioned deficiencies or improvement needs of existing technologies, as a first aspect of this invention, the present invention provides a method for intelligent assessment and anomaly identification of the operational performance of cable-stayed bridges, comprising:

[0008] S1. Complete the construction of a multi-source data acquisition system, including: monitoring data collected in real time by automated sensing equipment, routine testing data collected periodically by manual methods, and special testing data for acquiring the status of local or key components based on high-precision instruments;

[0009] S2. Complete data preprocessing and establish a unified template for the data timeline; extract and fuse features from the data, and reorganize the data based on the typical performance dimensions of the cable-stayed bridge and the theme type;

[0010] S3. Standardize and structure all perceived data; construct a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators; map nonlinear features through adaptively optimized kernel functions; and improve parameter optimization efficiency by using a fusion of cross-validation and Bayesian optimization to complete model training.

[0011] S4. Based on the trained model, complete the state assessment and output of the cable-stayed bridge and bind it with the bridge BIM model to automatically complete the visual annotation of defects.

[0012] Furthermore, the monitoring data mentioned in S1 includes cable tension data, dynamic response data collected by acceleration and vibration sensors, load and time history records of passing vehicles, environmental parameters including temperature, humidity, wind speed, wind direction and rainfall collected by environmental sensors, and overall response background data through full bridge displacement and deflection monitoring.

[0013] The routine inspection data includes visual and structural defect data such as sheath appearance, wire breakage rate, anchorage damage, coating, linearity, and water seepage, as well as internal information data including grout solidification and rust-preventive oil condition.

[0014] The specific testing data includes performance evaluation data of vibration damping devices, non-destructive testing data of anchors and cables, and status data of local or key components.

[0015] Furthermore, the unified template for the data timeline in S2 is specifically as follows:

[0016] Let the multi-source dataset be... ,in For the first The data source is a type of data whose original time series is... The corresponding data value is ,in, For the first The first class of data A timestamp;

[0017] The unified timeline template is defined as a standard time series:

[0018] ,

[0019] in, The starting time is the earliest timestamp among all data sources; To ensure a uniform time interval; The length of the time axis satisfies That is, the time range that covers all data.

[0020] For each data source Through interpolation function Mapping the original data to a standard timeline yields an aligned data sequence. :

[0021] ,

[0022] Ultimately, an integrated dataset with a unified timeline is formed. :

[0023] ,

[0024] Consistency of multi-source heterogeneous data in the time dimension was achieved through time scale standardization and interpolation mapping.

[0025] Furthermore, the topic types in S2 include PE sheath appearance identification topic, anchorage zone status identification topic, cable force status assessment topic, remaining bearing capacity calculation topic, fatigue life assessment topic, and vibration reduction performance assessment topic.

[0026] The PE sheath appearance identification theme focuses on the integrity and apparent defects of the PE sheath; the anchorage zone status identification theme is used to determine aspects including the sealing performance of the anchorage zone, anchor head protective layer detachment, and wire breakage; the cable force status assessment theme is used for anomaly detection in stay cables; the remaining bearing capacity calculation theme is used to quantify the bearing margin of stay cables; the fatigue life assessment theme is based on fatigue theory for life prediction; and the vibration reduction performance assessment theme evaluates the effectiveness of the vibration reduction device based on characteristics including additional damping of stay cables, amplitude control effect, and mode shape distribution.

[0027] Furthermore, the standardization and structuring processes in S3 specifically include:

[0028] Standardize and structure the indicators according to the format of "indicator object - indicator name - importance - indicator definition - format example - indicator source";

[0029] The indicator objects are determined based on the structural composition and performance dimensions of the stay cables; the indicator names are determined for the corresponding indicator objects, combined with data characteristics and evaluation requirements; the importance is divided into categories A, B, C, and D according to the degree of influence of the indicator on the stay cable status evaluation, corresponding to key indicators, important indicators, secondary important indicators, and general indicators, respectively; the indicator definitions are determined based on their physical meaning, measurement standards, and evaluation objectives; the format examples are a unified data format presentation method for different types of indicators; the indicator sources are the data collection methods for the corresponding indicators, used to distinguish from monitoring data, routine testing data, or special testing data.

[0030] Furthermore, the specific process in S3 of constructing a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators is as follows:

[0031] Establish a set of core indicators (A / B categories) in the integrated inspection and supervision indicator system. for:

[0032] ,

[0033] in, For the first Item;

[0034] Introducing the spatiotemporal correlation feature matrix ,in Indicators and The spatiotemporal correlation coefficient is calculated using mutual information or covariance:

[0035] ,

[0036] in, For the first Class data at time The value of ; For the first Class data at time The value of ; This is a time lag parameter, reflecting dynamic correlation; It is a variance function; It is the covariance function;

[0037] Then the dynamic input vector For the integration of indicators and spatiotemporal characteristics:

[0038] ,

[0039] in, It is the identity matrix; For correlation weight coefficients, control correction terms The intensity of the impact.

[0040] Furthermore, the specific process of mapping nonlinear features through an adaptively optimized kernel function in S3 is as follows:

[0041] The traditional RBF kernel function is:

[0042] ,

[0043] in, For kernel function output; The input sample represents the two classes of data to be compared. The sensitivity parameter of the kernel function;

[0044] Introducing adaptive coefficients Related to the importance of the indicator:

[0045] ,

[0046] in, The result of the adaptive kernel function calculation; Let be the input sample, with dimension . ,correspond One monitoring indicator; For the first Weight coefficients for each dimension; For the first Sensitivity coefficients in each dimension; For the sample The Each dimension component;

[0047] The precise mapping of nonlinear features is achieved through optimization using maximum likelihood estimation, where the th... Sensitivity parameters in each dimension Follows a gamma distribution. The shape parameter of the gamma distribution. It is a rate parameter.

[0048] Furthermore, the specific process of improving parameter optimization efficiency by fusing cross-validation and Bayesian optimization in S3 is as follows:

[0049] Let the SVM model parameter set be... , The penalty coefficient is the objective function, which is the expected classification accuracy.

[0050] ,

[0051] in, Use parameters for the model Accuracy at that time; It is the number of partitions in the dataset, that is... Cross-validation; For the first Folded dataset; For parameters No. Datasets The accuracy rate;

[0052] Alternative grid search is performed using Bayesian optimization based on Gaussian process priors. Iterative parameter update:

[0053] ,

[0054] in, It is a Gaussian process; Let be the mean function of a Gaussian process; The kernel function for a Gaussian process; This represents two different sets of model parameter configurations; For the function to be improved; Let represent the model parameters obtained through optimization in the (t+1)th iteration.

[0055] As a second aspect of the present invention, the present invention provides an intelligent assessment and anomaly identification system for the operational performance of cable-stayed bridges, comprising:

[0056] The multi-source data acquisition unit is used to complete the construction of a multi-source data acquisition system, including: monitoring data collected in real time by automated sensing devices, routine detection data collected periodically by manual methods, and special detection data for acquiring the status of local or key components based on high-precision instruments.

[0057] The data fusion and reorganization unit is used to complete data preprocessing and establish a unified template for the data timeline; it performs feature extraction and fusion on the data, reorganizes the data based on the typical performance dimensions of the cable-stayed bridge and the theme type;

[0058] The model training optimization unit is used to standardize and structure all the perceived data; it constructs a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators, maps nonlinear features through an adaptively optimized kernel function, and improves the efficiency of parameter optimization by using a combination of cross-validation and Bayesian optimization to complete model training.

[0059] The condition assessment annotation unit is used to complete the condition assessment and output of the stay cable based on the trained model, and is bound to the bridge BIM model to automatically complete the visual annotation of defects.

[0060] As a third aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor of any step of the described method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge.

[0061] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0062] 1. The intelligent assessment and anomaly identification method for the operational performance of cable-stayed bridges of this invention achieves comprehensive perception through the construction of a multi-source data acquisition system, providing a solid data foundation for subsequent assessment. This system integrates three major categories of data: monitoring data, routine testing data, and specialized testing data. It covers real-time monitoring information such as cable tension, vibration, and environmental parameters; periodic manual testing data such as sheath appearance and wire breakage rate; and specialized data such as vibration damping device performance and internal non-destructive testing data. The collaborative acquisition of multi-dimensional data overcomes the limitations of a single data source, ensuring comprehensive capture of various state characteristics of cable-stayed bridges during operation, providing rich and diverse original information support for accurate performance assessment and anomaly identification.

[0063] 2. The intelligent assessment and anomaly identification method for the operational performance of cable-stayed bridges of this invention improves assessment reliability through data fusion reconstruction and standardization. After preprocessing, a unified timeline template is established, and the data is reorganized according to six major themes, including PE sheath appearance and anchorage zone status. Simultaneously, an integrated inspection and monitoring indicator system is constructed to achieve data standardization and structuring. This processing method eliminates the heterogeneity of multi-source data, forming an organic whole from scattered data. The complementary verification of data from different themes significantly improves the systematicness and accuracy of the dimensional analysis of cable-stayed bridge performance, laying a high-quality data foundation for subsequent model training and status assessment.

[0064] 3. The intelligent assessment and anomaly identification method for the operational performance of cable-stayed bridges of this invention achieves accurate identification and judgment through SVM model training and state assessment mechanisms. Input vectors are constructed using core indicators, and the trained model can automatically identify anomalies such as cable force deviation and vibration reduction performance degradation, classifying state levels based on the quantitative values ​​of the indicators. This mechanism fully utilizes the features resulting from the fusion of multi-source data, outputting clear state levels through a standardized model. This not only achieves automated identification of cable-stayed bridge anomalies but also provides clear decision-making basis for operation and maintenance, effectively supporting intelligent management throughout the entire lifecycle of cable-stayed bridges. Attached Figure Description

[0065] Figure 1 This is a flowchart of the intelligent assessment and anomaly identification method for the operational performance of cable-stayed bridges according to an embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram of a typical abnormal event of a bridge suspension cable according to an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of typical defects in bridge suspension cables according to an embodiment of the present invention;

[0068] Figure 4 This is a schematic diagram of the defect identification results of the PE sheath of the cable-stayed bridge according to an embodiment of the present invention;

[0069] Figure 5 This is a schematic diagram of the defect identification results of the PE sheath of the cable-stayed bridge according to an embodiment of the present invention;

[0070] Figure 6 This is a schematic diagram of a spiral wire falling during a cable-stayed bridge inspection based on a drone, according to an embodiment of the present invention.

[0071] Figure 7 This is a system unit diagram of an embodiment of the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0073] Example 1

[0074] Please refer to Figure 1 This embodiment 1 provides a method for intelligent assessment and anomaly identification of the operational performance of cable-stayed bridges, including:

[0075] S1. Complete the construction of a multi-source data acquisition system, including: monitoring data collected in real time by automated sensing equipment, routine testing data collected periodically by manual methods, and special testing data for acquiring the status of local or key components based on high-precision instruments;

[0076] S2. Complete data preprocessing and establish a unified template for the data timeline; extract and fuse features from the data, and reorganize the data based on the typical performance dimensions of the cable-stayed bridge and the theme type;

[0077] S3. Standardize and structure all perceived data; construct a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators; map nonlinear features through adaptively optimized kernel functions; and improve parameter optimization efficiency by using a fusion of cross-validation and Bayesian optimization to complete model training.

[0078] S4. Based on the trained model, complete the state assessment and output of the cable-stayed bridge and bind it with the bridge BIM model to automatically complete the visual annotation of defects.

[0079] This embodiment 1 further elaborates on the above steps.

[0080] (1) Multi-source data acquisition

[0081] Please refer to Figure 2 As the core load-bearing component of a cable-stayed bridge, the performance of the stay cables during their operational period directly affects the bridge's safety and durability. Long-term exposure to loads, environmental factors, and other influences can easily lead to various problems, making comprehensive and accurate data collection fundamental to understanding their condition. Therefore, a multi-source data acquisition system needs to be constructed to achieve a holistic understanding of the stay cable's condition.

[0082] The system is formed by the coordinated collection of three types of data: monitoring data, routine testing data, and specialized testing data. In a preferred embodiment, the monitoring data includes cable tension data, dynamic response data collected by acceleration and vibration sensors, load and time records of passing vehicles, environmental parameters including temperature, humidity, wind speed, wind direction, and rainfall collected by environmental sensors, and overall response background data through full bridge displacement and deflection monitoring.

[0083] The routine inspection data includes visual and structural defect data such as sheath appearance, wire breakage rate, anchorage damage, coating, linearity, and water seepage, as well as internal information data including grout solidification and rust-preventive oil condition.

[0084] The specific testing data includes performance evaluation data of vibration damping devices, non-destructive testing data of anchors and cables, and status data of local or key components.

[0085] The three types of data complement each other, ranging from real-time dynamics and periodic details to key in-depth exploration levels, and together form a complete data collection system, providing comprehensive and systematic raw information support for subsequent intelligent assessment and anomaly identification.

[0086] (2) Data fusion and reorganization

[0087] In the data processing stage, the temporal heterogeneity of multi-source data is first addressed by preprocessing and establishing a unified timeline template. Since different types of data have varying collection frequencies and time records, a standard time series covering the entire data range is constructed by determining the earliest start time and setting unified time intervals according to data type (e.g., seconds for monitoring data and days for routine detection data). Then, interpolation is used to map various types of data to this timeline, forming an integrated dataset.

[0088] In a preferred embodiment, the unified template for the data timeline is as follows:

[0089] Let the multi-source dataset be... ,in For the first The data source is a type of data whose original time series is... The corresponding data value is ,in, For the first The first class of data A timestamp;

[0090] The unified timeline template is defined as a standard time series:

[0091] ,

[0092] in, The starting time is the earliest timestamp among all data sources; To standardize the time interval, it is determined based on the data collection frequency; for example, monitoring data is taken in seconds, while routine detection data is taken in days. The length of the time axis satisfies That is, the time range that covers all data.

[0093] For each data source Through interpolation function Mapping the original data to a standard timeline yields an aligned data sequence. :

[0094] ,

[0095] Ultimately, an integrated dataset with a unified timeline is formed. :

[0096] ,

[0097] Consistency of multi-source heterogeneous data in the time dimension was achieved through time scale standardization and interpolation mapping.

[0098] This operation eliminates inconsistencies in the time dimension, lays the foundation for correlation analysis of multi-source data, and ensures that subsequent evaluations can uncover dynamic correlations between data based on a unified time scale.

[0099] Please refer to Figures 3-6 Subsequently, feature extraction and fusion are performed, and the data is reorganized by theme. The feature extraction stage extracts key information from multi-source data—dynamic features such as cable force fluctuation amplitude and vibration frequency are extracted from monitoring data, and static features such as sheath damage area and anchorage corrosion degree are extracted from inspection data. These dispersed features are then correlated and integrated using fusion technology to form a comprehensive feature set that fully reflects the state of the stay cables. Based on this, the data is reorganized according to themes including PE sheath appearance identification, anchorage condition identification, cable force condition assessment, remaining load-bearing capacity calculation, fatigue life assessment, and vibration reduction performance assessment, forming a structured system around the core performance dimensions of the stay cables.

[0100] In a preferred embodiment, please refer to Figure 4 as well as Figure 5 The PE sheath appearance identification theme focuses on the integrity and apparent defects of the PE sheath; the anchorage zone status identification theme is used to determine the sealing performance of the anchorage zone, anchor head protective layer detachment, and wire breakage; the cable force status assessment theme is used for anomaly detection of stay cables; the remaining bearing capacity calculation theme is used to quantify the bearing margin of stay cables; the fatigue life assessment theme is based on fatigue theory for life prediction; and the vibration reduction performance assessment theme evaluates the effectiveness of the vibration reduction device based on characteristics including the additional damping of stay cables, amplitude control effect, and mode shape distribution.

[0101] This processing method not only preserves the key features of the data, but also enhances the complementarity of different types of data through thematic integration, allowing subsequent performance evaluation and anomaly identification to accurately focus on each key performance indicator, thereby improving the systematicness and accuracy of the evaluation.

[0102] (3) Model training optimization

[0103] After completing data preprocessing, timeline unification, and thematic reorganization, the data needs to be standardized and structured to achieve standardized integration. Following the framework of "indicator object - name - importance - definition - format - source," the attributes and classifications of various indicators are clarified (A / B categories are core indicators). This not only unifies the data format but also highlights key information through importance classification, providing a clear and consistent input foundation for subsequent model training and avoiding the impact of chaotic data formats or ambiguous core information on evaluation accuracy.

[0104] In a preferred embodiment, the standardization and structuring processes specifically involve:

[0105] Standardize and structure the indicators according to the format of "indicator object - indicator name - importance - indicator definition - format example - indicator source";

[0106] The indicator objects are determined based on the structural composition and performance dimensions of the stay cables; the indicator names are determined for the corresponding indicator objects, combined with data characteristics and evaluation requirements; the importance is divided into categories A, B, C, and D according to the degree of influence of the indicator on the stay cable status evaluation, corresponding to key indicators, important indicators, secondary important indicators, and general indicators, respectively; the indicator definitions are determined based on their physical meaning, measurement standards, and evaluation objectives; the format examples are a unified data format presentation method for different types of indicators; the indicator sources are the data collection methods for the corresponding indicators, used to distinguish from monitoring data, routine testing data, or special testing data.

[0107] The study covers no fewer than 61 indicators, encompassing dimensions such as cable acceleration response, modal characteristics, cable stress, sheath condition, anchorage damage, fatigue performance, and environmental impact. Please refer to Table 1, which lists some indicators and provides examples of some data as follows:

[0108] ,

[0109] Based on standardized A / B class core indicators, a dynamic input vector with spatiotemporal correlation features is constructed. By calculating the correlation coefficients of different indicators in the time dimension, the original indicators are fused with spatiotemporal correlation features, so that the input vector not only contains basic data but also reflects the dynamic interaction relationships between indicators. This processing overcomes the limitations of single data, allowing the model to capture the collaborative change patterns of multi-source data and provide richer information support for accurate evaluation.

[0110] In a preferred embodiment, the specific process of constructing a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data based on A / B class core indicators is as follows:

[0111] Establish a set of core indicators (A / B categories) in the integrated inspection and supervision indicator system. for:

[0112] ,

[0113] in, For the first Item;

[0114] Introducing the spatiotemporal correlation feature matrix ,in Indicators and The spatiotemporal correlation coefficient is calculated using mutual information or covariance:

[0115] ,

[0116] in, For the first Class data at time The value of ; For the first Class data at time The value of ; This is a time lag parameter, reflecting dynamic correlation; It is a variance function; It is the covariance function;

[0117] Then the dynamic input vector For the integration of indicators and spatiotemporal characteristics:

[0118] ,

[0119] in, It is the identity matrix; For correlation weight coefficients, control correction terms The intensity of the impact.

[0120] To uncover complex nonlinear relationships in the data, an adaptively optimized kernel function is used for feature mapping. By introducing sensitivity and weight coefficients related to the importance of indicators, the kernel function can adjust its measurement of data differences based on the actual impact of different indicators. Furthermore, probability distribution optimization is used to improve mapping accuracy.

[0121] In a preferred embodiment, the specific process of mapping nonlinear features using an adaptively optimized kernel function is as follows:

[0122] The traditional RBF kernel function is:

[0123] ,

[0124] in, For kernel function output; The input sample represents the two classes of data to be compared. The sensitivity parameter of the kernel function;

[0125] Introducing adaptive coefficients Related to the importance of the indicator:

[0126] ,

[0127] in, The result of the adaptive kernel function calculation; Let be the input sample, with dimension . ,correspond One monitoring indicator; For the first Weight coefficients for each dimension; For the first Sensitivity coefficients in each dimension; For the sample The Each dimension component;

[0128] The precise mapping of nonlinear features is achieved through optimization using maximum likelihood estimation, where the th... Sensitivity parameters in each dimension Follows a gamma distribution. The shape parameter of the gamma distribution. It is a rate parameter.

[0129] During the model training phase, cross-validation and Bayesian optimization are combined to improve parameter optimization efficiency. Cross-validation across multiple datasets ensures the model's generalization ability, while probabilistic modeling using Bayesian optimization quickly locates the optimal parameters, replacing inefficient grid search. This combination guarantees model stability while significantly reducing the computational cost of parameter tuning, enabling the trained model to more efficiently and reliably support subsequent behavior evaluation and anomaly identification.

[0130] In a preferred embodiment, the specific process of improving parameter optimization efficiency by fusing cross-validation and Bayesian optimization is as follows:

[0131] Let the SVM model parameter set be... , The penalty coefficient is the objective function, which is the expected classification accuracy.

[0132] ,

[0133] in, Use parameters for the model Accuracy at that time; It is the number of partitions in the dataset, that is... Cross-validation; For the first Folded dataset; For parameters No. Datasets The accuracy rate;

[0134] Alternative grid search is performed using Bayesian optimization based on Gaussian process priors. Iterative parameter update:

[0135] ,

[0136] in, It is a Gaussian process; Let be the mean function of a Gaussian process; The kernel function for a Gaussian process; This represents two different sets of model parameter configurations; For the function to be improved; Let represent the model parameters obtained through optimization in the (t+1)th iteration.

[0137] Overall, this series of processes builds upon the previous data integration work, and by standardizing data formats, strengthening correlation features, adapting to nonlinear relationships, and optimizing training efficiency, it constructs a complete transformation path from data to model, providing accurate and efficient technical support for the performance evaluation of cable-stayed bridges during operation.

[0138] (4) Status assessment labeling

[0139] After completing model training and ensuring its stability and accuracy, based on the recognition results output by the SVM model and combined with the quantified values ​​of each parameter in the previously constructed indicator system, this invention further constructs a multi-dimensional evaluation model for the service status of cable-stayed bridges. This model comprehensively considers multiple dimensions such as structural performance, degree of damage, and safety risks, classifying the cable-stayed bridge status into four levels: Level I represents the normal state, where all performance indicators of the cable-stayed bridge are within the design limits, the structure is intact, and no special intervention is required; Level II is a slightly degraded state, characterized by minor fluctuations or deterioration trends in some non-critical indicators, but without affecting the overall load-bearing capacity, and it is recommended to increase the monitoring frequency to monitor its development; Level III is a moderately damaged state, where key components show identifiable damage (such as localized wire corrosion, minor sheath damage, etc.), which has already had a certain impact on structural performance, requiring the development of a maintenance plan for timely handling; Level IV is a severely damaged state, with major defects that endanger structural safety (such as excessive number of broken wires, risk of anchor failure, etc.), requiring immediate activation of the emergency response mechanism to avoid safety accidents.

[0140] The assessment results output has diverse presentation and application capabilities. On the one hand, the results can be visualized on the digital monitoring platform in the form of charts, heat maps, etc., intuitively showing the status distribution and changing trends of various parts of the cable stays; on the other hand, the system can automatically integrate assessment data and generate lifecycle management reports that include status analysis, risk warnings, and maintenance recommendations, providing decision support for the operation and maintenance team.

[0141] Most importantly, this method can be deeply integrated with bridge BIM models, automatically linking assessed damage information (such as damage location, extent, and grade) to corresponding components in the BIM model via a data interface, enabling three-dimensional visualization and annotation of the damage. The annotation information includes details such as damage type, discovery time, and development trend, facilitating managers to intuitively locate problems in the virtual model and formulate targeted maintenance plans.

[0142] Furthermore, this method has good versatility. Its core logic and evaluation framework can be transferred to other cable-stayed structures (such as the suspension cables of suspension bridges and the stiffening cables of arch bridges). By simply adjusting the parameter weights and evaluation thresholds in the index system according to the mechanical characteristics and damage modes of different cable structures, accurate evaluation of various cable-stayed structures can be achieved.

[0143] This process builds upon the previous model training and parameter optimization work, and achieves a leap from data identification to state determination through a multi-dimensional evaluation model. It also enhances the practicality and management efficiency of the results by leveraging visualization and BIM integration technologies. At the same time, its universal design expands the application boundaries of the technology and provides a systematic solution for the safe operation and maintenance of various cable structures.

[0144] Example 2

[0145] Please refer to Figure 7 This embodiment 2 provides a smart assessment and anomaly identification system for the operational performance of cable-stayed bridges, including:

[0146] The multi-source data acquisition unit is used to complete the construction of a multi-source data acquisition system, including: monitoring data collected in real time by automated sensing devices, routine detection data collected periodically by manual methods, and special detection data for acquiring the status of local or key components based on high-precision instruments.

[0147] The data fusion and reorganization unit is used to complete data preprocessing and establish a unified template for the data timeline; it performs feature extraction and fusion on the data, reorganizes the data based on the typical performance dimensions of the cable-stayed bridge and the theme type;

[0148] The model training optimization unit is used to standardize and structure all the perceived data; it constructs a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators, maps nonlinear features through an adaptively optimized kernel function, and improves the efficiency of parameter optimization by using a combination of cross-validation and Bayesian optimization to complete model training.

[0149] The condition assessment annotation unit is used to complete the condition assessment and output of the stay cable based on the trained model, and is bound to the bridge BIM model to automatically complete the visual annotation of defects.

[0150] Example 3

[0151] This embodiment 3 also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement any step of a method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge.

[0152] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0153] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0154] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent assessment and anomaly identification of the operational performance of cable-stayed bridges, characterized in that, include: S1. Complete the construction of a multi-source data acquisition system, including: monitoring data collected in real time by automated sensing equipment, routine testing data collected periodically by manual methods, and special testing data for acquiring the status of local or key components based on high-precision instruments; S2. Complete data preprocessing and establish a unified template for the data timeline; extract and fuse features from the data, and reorganize the data based on the typical performance dimensions of the cable-stayed bridge and the theme type; S3. Standardize and structure all perceived data; construct a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators; map nonlinear features through adaptively optimized kernel functions; and improve parameter optimization efficiency by using a fusion of cross-validation and Bayesian optimization to complete model training. S4. Based on the trained model, complete the state assessment and output of the stay cables, and bind them with the bridge BIM model to automatically complete the visual annotation of defects; The specific process of constructing a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators in S3 is as follows: Establish a set of core indicators (A / B categories) in the integrated inspection and supervision indicator system. for: in, For the first Item; Introducing the spatiotemporal correlation feature matrix ,in Indicators and The spatiotemporal correlation coefficient is calculated using mutual information or covariance: in, For the first Class data at time The value of ; For the first Class data at time The value of ; This is a time lag parameter, reflecting dynamic correlation; It is a variance function; It is the covariance function; Then the dynamic input vector For the integration of indicators and spatiotemporal characteristics: in, It is the identity matrix; For correlation weight coefficients, control correction terms The intensity of the impact; The process of mapping nonlinear features through adaptively optimized kernel functions specifically involves introducing sensitivity coefficients and weight coefficients related to the importance of indicators, allowing the kernel function to adjust the way it measures data differences according to the actual impact of different indicators, and then improving the mapping accuracy through probability distribution optimization.

2. The method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge according to claim 1, characterized in that, The monitoring data mentioned in S1 includes cable tension data, dynamic response data collected by acceleration and vibration sensors, load and time history records of passing vehicles, environmental parameters including temperature, humidity, wind speed, wind direction and rainfall collected by environmental sensors, and overall response background data through full bridge displacement and deflection monitoring. The routine inspection data includes visual and structural defect data such as sheath appearance, wire breakage rate, anchorage damage, coating, linearity, and water seepage, as well as internal information data including grout solidification and rust-preventive oil condition. The specific testing data includes performance evaluation data of vibration damping devices, non-destructive testing data of anchors and cables, and status data of local or key components.

3. The method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge according to claim 1, characterized in that, The unified template for the data time axis in S2 is as follows: Let the multi-source dataset be... ,in For the first The data source is a type of data whose original time series is... The corresponding data value is ,in, For the first The first class of data A timestamp; The unified timeline template is defined as a standard time series: in, The starting time is the earliest timestamp among all data sources; To ensure a uniform time interval; The length of the time axis satisfies That is, the time range that covers all data; For each data source Through interpolation function Mapping the original data to a standard timeline yields an aligned data sequence. : Ultimately, an integrated dataset with a unified timeline is formed. : Consistency of multi-source heterogeneous data in the time dimension was achieved through time scale standardization and interpolation mapping.

4. The method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge according to claim 1, characterized in that, The topic types in S2 include PE sheath appearance identification topic, anchorage zone status identification topic, cable force status assessment topic, remaining bearing capacity calculation topic, fatigue life assessment topic, and vibration reduction performance assessment topic. The PE sheath appearance identification theme focuses on the integrity and apparent defects of the PE sheath; the anchorage zone status identification theme is used to determine aspects including the sealing performance of the anchorage zone, anchor head protective layer detachment, and wire breakage; the cable force status assessment theme is used for anomaly detection in stay cables; the remaining bearing capacity calculation theme is used to quantify the bearing margin of stay cables; the fatigue life assessment theme is based on fatigue theory for life prediction; and the vibration reduction performance assessment theme evaluates the effectiveness of the vibration reduction device based on characteristics including additional damping of stay cables, amplitude control effect, and mode shape distribution.

5. The method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge according to claim 1, characterized in that, The standardization and structuring processes in S3 specifically refer to: Standardize and structure the indicators according to the format of "indicator object - indicator name - importance - indicator definition - format example - indicator source"; The indicator objects are determined based on the structural composition and performance dimensions of the stay cables; the indicator names are determined for the corresponding indicator objects, combined with data characteristics and evaluation requirements; the importance is divided into categories A, B, C, and D according to the degree of influence of the indicator on the stay cable status evaluation, corresponding to key indicators, important indicators, secondary important indicators, and general indicators, respectively; the indicator definitions are determined based on their physical meaning, measurement standards, and evaluation objectives; the format examples are a unified data format presentation method for different types of indicators; the indicator sources are the data collection methods for the corresponding indicators, used to distinguish from monitoring data, routine testing data, or special testing data.

6. The method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge according to claim 1, characterized in that, The specific process of mapping nonlinear features through an adaptively optimized kernel function in S3 is as follows: The traditional RBF kernel function is: in, This is the output of the kernel function; The input sample represents the two classes of data to be compared. The sensitivity parameter of the kernel function; Introducing adaptive coefficients Related to the importance of the indicator: in, The result of the adaptive kernel function calculation; Let be the input sample, with dimension . ,correspond One monitoring indicator; For the first Weight coefficients for each dimension; For the first Sensitivity coefficients in each dimension; For the sample The Each dimension component; The precise mapping of nonlinear features is achieved through optimization using maximum likelihood estimation, where the th... Sensitivity parameters in each dimension Follows a gamma distribution. The shape parameter of the gamma distribution. It is a rate parameter.

7. The method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge according to claim 1, characterized in that, The specific process of improving parameter optimization efficiency by fusing cross-validation and Bayesian optimization in S3 is as follows: Let the SVM model parameter set be... , The penalty coefficient is the objective function, which is the expected classification accuracy. in, Parameters for the model Accuracy at that time; It is the number of partitions in the dataset, that is... Cross-validation; For the first Folded dataset; For parameters No. Datasets The accuracy rate; Alternative grid search is performed using Bayesian optimization based on Gaussian process priors. Iterative parameter update: in, It is a Gaussian process; Let be the mean function of a Gaussian process; Let be the kernel function of the Gaussian process; This represents two different sets of model parameter configurations; For the function to be improved; Let represent the model parameters obtained through optimization in the (t+1)th iteration.

8. A smart system for evaluating and identifying the operational performance of cable-stayed bridges, characterized in that, include: The multi-source data acquisition unit is used to complete the construction of a multi-source data acquisition system, including: monitoring data collected in real time by automated sensing devices, routine detection data collected periodically by manual methods, and special detection data for acquiring the status of local or key components based on high-precision instruments. The data fusion and reorganization unit is used to complete data preprocessing and establish a unified template for the data timeline; it performs feature extraction and fusion on the data, reorganizes the data based on the typical performance dimensions of the cable-stayed bridge and the theme type; The model training optimization unit is used to standardize and structure all the perceived data; it constructs a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators, maps nonlinear features through an adaptively optimized kernel function, and improves the efficiency of parameter optimization by using a combination of cross-validation and Bayesian optimization to complete model training. The condition assessment annotation unit is used to complete the condition assessment and output of the stay cables based on the trained model, and is bound to the bridge BIM model to automatically complete the visual annotation of defects; The specific process of constructing a dynamic input vector containing spatiotemporal correlation features of multi-source heterogeneous data using A / B class core indicators in the model training and optimization unit is as follows: Establish a set of core indicators (A / B categories) in the integrated inspection and supervision indicator system. for: in, For the first Item; Introducing the spatiotemporal correlation feature matrix ,in Indicators and The spatiotemporal correlation coefficient is calculated using mutual information or covariance: in, For the first Class data at time The value of ; For the first Class data at time The value of ; This is a time lag parameter, reflecting dynamic correlation; It is a variance function; It is the covariance function; Then the dynamic input vector For the integration of indicators and spatiotemporal characteristics: in, It is the identity matrix; For correlation weight coefficients, control correction terms The intensity of the impact; The process of mapping nonlinear features through adaptively optimized kernel functions specifically involves introducing sensitivity coefficients and weight coefficients related to the importance of indicators, allowing the kernel function to adjust the way it measures data differences according to the actual impact of different indicators, and then improving the mapping accuracy through probability distribution optimization.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor as described in any one of claims 1-7: a method for intelligent assessment and anomaly identification of the operational performance of a cable-stayed bridge.

Images