Method and system for intelligently identifying abnormity of expansion constraint device of large-span bridge

Abstract

The invention discloses an intelligent abnormity identification method and system for a large-span bridge expansion constraint device, and belongs to the technical field of intelligent expansion joint detection. The method comprises the steps that the beam bridge structure temperature and the beam bridge longitudinal displacement representing the state of the beam bridge constraint device are acquired; eliminating the longitudinal displacement noise of the beam bridge by using a wavelet separation method, and obtaining dynamic and static time sequence fusion data of the beam bridge structure temperature and the longitudinal displacement of the beam bridge based on a temperature-displacement long-term change rule; and identifying the dynamic and static time sequence fusion data by using a constructed Adaboost data anomaly identification model and judging the abnormal state of the expansion joint restraint device of the large-span steel box girder bridge. According to the method and the system for intelligently identifying the abnormity of the large-span bridge expansion constraint device, whether the response data of the bridge structure exceeds the limit or is abnormal can be effectively monitored, so that the service state of the expansion constraint device can be accurately identified.

Description

Technical Field

[0001] This invention belongs to the field of intelligent detection technology for expansion joints in beam bridges, specifically relating to an intelligent identification method and system for abnormal expansion restraint devices in long-span bridges. Background Technology

[0002] Expansion joints are crucial constraint structures for regulating the longitudinal displacement of bridges, affecting overall bridge safety and traffic stability. However, damage to them is common during service due to complex natural environmental conditions. Currently, the condition assessment of bridge expansion joints primarily relies on visual inspection, while the low-frequency, large-amplitude periodic displacement temperature generated by expansion joints is mainly caused by temperature loads, which is particularly significant for cross-river bridges with large displacement characteristics.

[0003] With the development of artificial intelligence methods such as neural networks, machine learning has provided new approaches for identifying anomalies in bridge structural conditions. However, longitudinal displacement studies based on long-term monitoring data mainly focus on cable-stayed bridges, with limited research on bridge components. Furthermore, current expansion joint assessment methods primarily rely on correlation analysis to determine the relationship between data points, resulting in relatively singular assessment indicators that cannot quantify model errors and lack effective long-term data fusion methods. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for intelligent identification of anomalies in expansion joint restraint devices for long-span bridges. This method and system can effectively monitor whether the response data of the bridge structure exceeds the limits or has anomalies, thereby accurately identifying the service status of the expansion joint restraint device.

[0005] To achieve the above objectives, the present invention is implemented using the following technical solution: In a first aspect, the present invention provides a method for intelligent identification of anomalies in the expansion joint restraint device of a long-span bridge, comprising: Obtain the structural temperature and longitudinal displacement of the beam bridge that characterize the state of the beam bridge restraint device; Wavelet separation method was used to remove longitudinal displacement noise of the beam bridge, and dynamic and static time series fusion data of beam bridge structure temperature and longitudinal displacement were obtained based on the long-term variation law of temperature-displacement. The constructed Adaboost data anomaly identification model is used to identify and determine the abnormal state of the expansion joint restraint device of the long-span steel box girder bridge by using the dynamic and static time-series fusion data.

[0006] Furthermore, several temperature sensor measuring points are arranged on at least one pier section and at least one mid-span section of the main bridge for both directions. At four different locations on the upper surface, lower surface, top plate, and bottom plate of the box girder at the same pier section and mid-span section, one temperature sensor measuring point is arranged at each of these locations. The average temperature of several temperature sensor measuring points is selected as the temperature of the beam bridge structure. ; In the formula T i Let be the temperature of the i-th sensor; m is the number of temperature sensors.

[0007] Furthermore, a longitudinal displacement sensor measuring point was arranged at the upstream and downstream supports of the two end piers on the main bridge. The cable-stayed displacement gauge was used, with one end fixed to the upper part of the crossbeam under the main tower and the other end passing over the expansion joint. The data of the measuring point with the largest displacement change amplitude at the upstream and downstream supports of the two end piers on the main bridge were selected as the longitudinal displacement of the beam bridge.

[0008] Furthermore, based on the number of temperature sensors and the measured temperatures, combined with the top area of ​​the main bridge and the bottom area of ​​the beams and their respective proportions, the average temperature of the bridge structure is calculated using the area weighting method.

[0009] Furthermore, the temperature-displacement feature fusion method based on time-series signals includes the following steps: Multi-resolution analysis techniques based on discrete wavelet decomposition decompose long-period fluctuations caused by temperature in displacement into the local features of the signal in both time and frequency domains, while eliminating short-period fluctuations caused by noise and vehicle load. The structural temperature measurement value at the current monitoring time t is used as the instantaneous static characteristic value of the temperature. Considering the time lag effect of temperature-displacement, the structural temperature at time t-1 is taken as the dynamic characteristic value of temperature time series. The longitudinal displacement measurement value of the beam at the current monitoring time t is used as the static characteristic value of the displacement. The trained LSTM network model is used to extract instantaneous static and dynamic displacement features from the real-time measured longitudinal displacement features of the bridge. The instantaneous static and dynamic characteristics of temperature and displacement of the beam bridge structure at a certain moment are used to construct a comprehensive information time series for that moment using a multi-dimensional feature serial fusion method.

[0010] Furthermore, the training and construction method of the Adaboost data anomaly detection model includes the following steps: Anomalies in the displacement data were labeled using an expert experience method. 70% of the samples were used to construct the training set, and the remaining 30% were used as the test set. Using the processed integrated information time series, combined with the BP neural network classification algorithm, multiple weak classifiers are generated. Reduce the weight of samples that have been accurately classified by the weak classifier, and increase the weight of samples that have been misclassified by the weak classifier. The Adaboost algorithm uses a weighted voting mechanism to integrate multiple weak classifiers into a strong classifier to predict the class label of test samples.

[0011] Secondly, this invention provides an intelligent identification system for abnormal expansion and contraction restraint devices of long-span bridges, comprising the following modules: The data acquisition module is used to acquire the structural temperature and longitudinal displacement of the beam bridge, which characterize the state of the beam bridge restraint device. The time-series fusion module is used to remove longitudinal displacement noise of the beam bridge using wavelet separation method, and to obtain dynamic and static time-series fusion data of beam bridge structure temperature and longitudinal displacement. The anomaly identification module is used to identify and judge the abnormal state of the expansion joint restraint device of the long-span steel box girder bridge by using the constructed Adaboost data anomaly identification model.

[0012] Furthermore, several temperature sensor measuring points are arranged on at least one pier section and at least one mid-span section of the main bridge for both directions. At four different locations on the upper surface, lower surface, top plate, and bottom plate of the box girder at the same pier section and mid-span section, one temperature sensor measuring point is arranged at each of these locations. The average temperature of several temperature sensor measuring points is selected as the temperature of the beam bridge structure. ; In the formula T i Let m be the temperature of the i-th sensor; m is the number of temperature sensors. A longitudinal displacement sensor measuring point was set up at the upstream and downstream supports of the two end piers on the main bridge. The cable displacement meter was used, with one end fixed to the upper part of the crossbeam under the main tower and the other end passing over the expansion joint. The data of the measuring point with the largest displacement change amplitude at the upstream and downstream supports of the two end piers on the main bridge were selected as the longitudinal displacement of the beam bridge.

[0013] Thirdly, the present invention provides a computer-readable storage medium that, when the instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to perform the intelligent identification method for abnormality of the expansion joint constraint device of a long-span bridge as described in any one of the first aspects.

[0014] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: The intelligent identification method and system for anomalies in long-span bridge expansion joint constraint devices provided by this invention utilizes long-term monitoring data of the expansion joint constraint device for data fusion, which is divided into monitoring data processing, time-series feature fusion, and anomaly identification. First, multi-sensor data fusion is performed in the data processing part. Then, multi-dimensional features are extracted based on beam temperature and bridge longitudinal displacement data for serial feature fusion. Efficient machine learning classification algorithms are used for training and comparison. Finally, the longitudinal displacement at this moment is verified based on the fused model to determine whether there is a real anomaly, thereby judging the real state of the expansion joint constraint device and improving the effectiveness and practicality of anomaly identification. Attached Figure Description

[0015] Figure 1 A flowchart of an intelligent identification method for anomalies in a long-span bridge expansion joint constraint device provided in an embodiment of the present invention.

[0016] Figure 2 This is a schematic diagram of a sensor arrangement provided in an embodiment of the present invention.

[0017] Figure 3 Temperature diagram of a mid-span sensor structure provided in an embodiment of the present invention.

[0018] Figure 4 This is a longitudinal displacement diagram of a beam end provided in an embodiment of the present invention.

[0019] Figure 5 This is a schematic diagram of temperature-displacement consistency provided in an embodiment of the present invention.

[0020] Figure 6 This is a diagram of an LSTM structure provided in an embodiment of the present invention.

[0021] Figure 7 This is a time series diagram provided as an embodiment of the present invention.

[0022] Figure 8 This is a schematic diagram of an AdaBoost classification process provided in an embodiment of the present invention.

[0023] Figure 9 This is a schematic diagram of a beam end expansion joint provided in an embodiment of the present invention. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention. Example

[0025] like Figure 1 As shown, this embodiment of the invention provides a method for intelligent identification of anomalies in the expansion joint constraint device of a long-span bridge, including: Obtain the structural temperature and longitudinal displacement of the beam bridge that characterize the state of the beam bridge restraint device; Wavelet separation method was used to remove longitudinal displacement noise of the beam bridge, and dynamic-static time-series fusion data of beam bridge structure temperature and longitudinal displacement were obtained based on the long-term variation law of temperature-displacement. The constructed Adaboost data anomaly identification model is used to identify and determine the abnormal state of the expansion joint restraint device of the long-span steel box girder bridge by using the dynamic and static time-series fusion data.

[0026] In this embodiment, the Chongqi Bridge is taken as a specific example. It is an important part of the G40 National Expressway. The main bridge uses 102+4×185+102m steel box girders, and the approach bridge is equipped with 6×50m span external prestressed precast segmental continuous box girders, which is currently the largest span of this type of bridge in China.

[0027] The substructure employs rectangular solid piers, integral abutments, and steel pipe pile foundations, and is also equipped with auxiliary facilities such as TMD dampers, seismic spherical steel bearings, and modular expansion joints. Currently, the Chongqi Bridge's health monitoring system comprehensively monitors over twenty items, including beam temperature and humidity, main beam displacement, cable force, and damper displacement. The longitudinal displacement and structural temperature sensors are arranged as follows: Figure 2 .

[0028] One longitudinal displacement sensor was installed at each of the upstream and downstream supports of piers 1 and 7 of the main steel box girder bridge, totaling four longitudinal displacement measurement points at the beam ends. A draw-wire displacement gauge was used. Simultaneously, considering factors such as structural type, span length, cross-section, component dimensions, and sunlight conditions, structural temperature sensors were installed at the cross-sections of piers 4 (both upstream and downstream), the mid-span section of the fourth span (both upstream and downstream), pier 7 (both upstream and downstream), pier 3 (upstream), and the mid-span section of the third span (upstream). These sensors were deployed in conjunction with temperature compensation measurement points for strain monitoring, totaling 64 temperature sensor points. Specific experimental data from the temperature sensor monitoring are shown in Table 1 below.

[0029] Table 1 Temperature and longitudinal displacement data with a sampling frequency of 1Hz were selected from the main steel box girder bridge to characterize the state of the bridge's restraint devices. Eight temperature sensors were installed at the mid-span sections of piers 4 and 5 for monitoring, with measuring points located on the upper and lower surfaces and top and bottom plates of the same section of the box girder. Therefore, four mid-span sensors were selected to represent the temperature changes at different locations within a single section. Due to the strong correlation between these sensors, the average temperature was selected for subsequent data fusion.

[0030] ; In the formula, Ti is the temperature of the i-th sensor; m is the number of temperature sensors.

[0031] like Figure 3 As shown, a period of temperature change was selected, and the effective temperature of the beam was calculated based on a weighted average of the top and bottom of the beam, ultimately yielding the average temperature of the structure. In this embodiment, based on the number of temperature sensors and the measured temperatures, combined with the area of ​​the main bridge top and bottom of the beam and their respective proportions, the area-weighted method was used to calculate the average temperature of the bridge structure. Specifically, longitudinal displacement sensors at the north end of the bridge were installed on both the upstream and downstream sides of the bridge. One end of the sensor was fixed to the upper part of the crossbeam below the main tower, and the other end crossed the expansion joint. The longitudinal displacement change at the north beam end is shown in the figure. Figure 4As shown, the downstream support data exhibits the largest variation and is used to represent the longitudinal displacement of the structure.

[0032] In this embodiment, the longitudinal displacement of the expansion joint can be decomposed into low-frequency components caused by temperature loads and high-frequency components caused by vehicle and wind loads. Considering the multi-resolution analysis characteristics of the discrete wavelet decomposition method, which can characterize the local features of the signal in both the time and frequency domains, the wavelet separation method is used to decompose the long-period fluctuations caused by temperature in the displacement and to remove the short-period fluctuations caused by noise and vehicle loads. The wavelet transform is based on the Fourier transform. First, the mother wavelet y(t) is scaled and translated to obtain a wavelet sequence. For the continuous case: ; In the formula a The scaling factor; b This is the translation factor.

[0033] Then any function f ( t The continuous wavelet transform of ) is: ;

[0034] The signal under study was analyzed using multi-resolution analysis. x ( n The signal is transformed into approximate and detail terms. The transformation formula for the original signal is shown below:

[0035] ;

[0036] In the formula x It is a time series; M The number of decomposition levels; a M,k These are approximate coefficients; φ M,k It is a scaling function; d j,k For the coefficients of the j-th resolution; ψ j,k It is a wavelet function.

[0037] Discrete wavelet reconstruction is the inverse process of decomposition, used to recover the original signal. It involves processing and recombining the signal approximation coefficients and detail coefficients. In practical applications, it has been found that the bd8 wavelet has a better separation effect in the discrete wavelet decomposition method. At the same time, different decomposition levels will cause differences in the results of temperature-induced displacement of expansion joints. Therefore, the decomposition level M selected in the process of extracting temperature-induced displacement of long-span bridges is 9.

[0038] Based on waveform separation analysis of monitoring data, structural temperature and displacement data corresponding to the same time period are extracted. The time consistency results of temperature and displacement are as follows: Figure 5As shown, the trends of the overall bridge structural temperature and beam end displacement over time are generally consistent, but a time lag effect exists. Therefore, the measured structural temperature at monitoring time t is [value missing]. y t At the same time, the structural temperature at time t-1 also needs to be considered. y t-1 The influence of temperature at time t is considered, therefore the temperature characteristic at time t is set. T t for[ y t y t-1 ] .

[0039] In this embodiment, when the monitored longitudinal displacement of the bridge exceeds the allowable design value, it indicates that the restraint device may have been damaged. Furthermore, to more accurately assess the structural safety of the bridge, it is also necessary to analyze the trend of displacement data over time, such as abnormal fluctuations during displacement changes. Such discrepancies with the trend may also indicate structural damage.

[0040] Therefore, the longitudinal displacement characteristics of bridges can be divided into instantaneous and static displacement characteristics. ν ts and displacement dynamic characteristics ν td The sensor measurement at time t represents the instantaneous static characteristics of the displacement. ν ts， Then the bridge displacement characteristics at time t ν t for[ ν ts ν td The dynamic characteristics of displacement represent the periodic pattern of longitudinal deformation of a bridge structure over time, typically exhibiting nonlinear characteristics, and can be extracted using an LSTM model. LSTM is a variant of recurrent neural networks, widely used for processing time series data and identifying discontinuous pattern features. Its network structure is as follows: Figure 6 As shown.

[0041] In the LSTM model, historical displacement data of the bridge is used as input, and the displacement measurement at time t+1 is used as the target ground truth and label ground truth during training. The ultimately trained LSTM network can extract the feature matrix during the process, capturing the dynamic displacement features of the time series. ν tdThis is transformed into a 1×n dimension matrix. In the dynamic feature extraction of bridge displacement, the training and test sets are divided in an 80% and 20% ratio, respectively. The gradient threshold is set to 1, and the initial learning rate is specified as 0.01. The model is trained for 100 epochs. To prevent gradient explosion, the learning rate is reduced by multiplying by a factor of 0.2 after the 25th epoch. After training, the LSTM model has a loss function value of 0.0007 and a root mean square error (RMSE) of 0.0154.

[0042] In this embodiment, since the dynamic and static features have different dimensions and are interdependent, it is necessary to merge different feature data into a single feature vector at the underlying level. This helps to completely preserve the information of temperature and longitudinal displacement and reduces computational complexity. Feature fusion can be categorized into three types based on the model structure: serial, parallel, and hybrid. In the serial strategy, the model has only one branch, requiring feature extraction, selection, and transformation before feature fusion is performed according to a specific merging order, using the output of one model as the input of the next. For measurement time t, a new feature Z is constructed using the serial fusion method. t ,in Z t =[ y t y t-1 ν ts ν td The integrated information time series at a certain moment after the fusion of all features is defined as a 1×13 dimensional vector, and its fused time series features are as follows: Figure 7 As shown.

[0043] In this embodiment, a machine learning classification algorithm is used to determine whether there are any anomalies in the longitudinal displacement of a long-span steel box girder bridge in real time. The AdaBoost algorithm is an ensemble learning enhancement method suitable for classifying multi-dimensional fused data and exhibits good accuracy across various classification problems. For different time-series samples, multiple weak classifiers are generated by the basic learning algorithm, and then the weight distribution is adjusted. The process involves reducing the weight of samples that have been accurately classified while increasing the weight of samples misclassified by the weak classifiers. Samples with larger weights receive more attention in subsequent iterations. Finally, AdaBoost uses a weighted voting mechanism to integrate these weak classifiers into a strong classifier to predict the class label of the test sample, such as... Figure 8 As shown.

[0044] 70% of the monitored longitudinal displacement data samples were used to construct the training set, and the remaining 30% were used as the test set. Anomalies in the displacement data were labeled using expert experience. A backpropagation (BP) neural network was used as a weak classifier to analyze the accuracy of AdaBoost. After feature processing and fusion, the results were input into the AdaBoost model, Gaussian Kernel SVM, and logistic regression model for training. The results of validation using the test set are shown in Table 2. The balanced accuracy of the trained models were 98.9%, 96.6%, and 91.3%, respectively, and the macro-F1 scores were 0.985, 0.956, and 0.911, respectively. It can be observed that the latter two did not outperform AdaBoost in classification performance. This may be because AdaBoost is essentially an ensemble learner, capable of better evaluating all feature variables and improving the final classification performance by integrating multiple weak learners.

[0045] Table 2: Performance Comparison of Different Algorithms In this embodiment, after training the AdaBoost model for the expansion joint device, the monitoring data can be fused and analyzed to compare with the results of manual inspections. Validation using data from the second half of 2022 to the first half of 2023 revealed persistent anomalies in the data of the bridge's critical restraint devices from April to June 2023. This necessitates attention to whether specific components of the critical restraint devices are clogged with debris, have uneven gaps, or exhibit wear, requiring targeted inspection and repair. According to the actual bridge inspection and comprehensive inspection reports, no abnormalities were found in the expansion joints during the December 2022 bridge inspection. However, an inspection in July 2023 revealed blockages in all four expansion joints, and slight rust was found on the surface of the carriageway steel after wear of the anti-rust layer. No other obvious defects were found, fully validating the effectiveness and practicality of the restraint device anomaly identification model.

[0046] This invention provides an intelligent identification method for anomalies in expansion joint constraint devices of long-span bridges. It utilizes longitudinal displacement and structural temperature data of long-span steel box girder bridges to perform data-level and feature-level fusion. Finally, an AdaBoost state diagnosis model based on a weak classifier of a neural network, combined with long-term historical data, is used to identify abnormal states of the expansion joint constraint devices in long-span steel box girder bridges. On one hand, dynamic and static multi-dimensional feature indicators of temperature and longitudinal displacement are proposed. The trends of overall structural temperature and beam end longitudinal displacement over time show a good correlation and a lag effect. On the other hand, an AdaBoost data anomaly identification model is constructed, and the accuracy of different types of machine learning algorithms is compared. The model's average accuracy reaches 98.9%, indicating that the ensemble algorithm can effectively monitor whether there are real anomalies in the multi-dimensional fused data. Finally, the method is validated by combining on-site bridge inspection data, realizing the fusion of monitoring and inspection data to determine whether there are problems with the bridge expansion joint devices. Future research can combine this with fault prediction to help bridge maintenance personnel take targeted intervention measures in advance.

[0047] Secondly, this invention provides an intelligent identification system for abnormal expansion and contraction restraint devices of long-span bridges, comprising the following modules: The data acquisition module is used to acquire the structural temperature and longitudinal displacement of the beam bridge, which characterize the state of the beam bridge restraint device. The time-series fusion module is used to remove longitudinal displacement noise of the beam bridge using wavelet separation method, and to obtain dynamic and static time-series fusion data of beam bridge structure temperature and longitudinal displacement. The anomaly identification module is used to identify and judge the abnormal state of the expansion joint restraint device of the long-span steel box girder bridge by using the constructed Adaboost data anomaly identification model.

[0048] Several temperature sensor measuring points are arranged on at least one pier section and at least one mid-span section of the main bridge for both directions. At four different locations on the upper surface, lower surface, top plate, and bottom plate of the box girder at the same pier section and mid-span section, one temperature sensor measuring point is arranged at each of these locations. The average temperature of several temperature sensor measuring points is selected as the temperature of the beam bridge structure. ; In the formula T i Let be the temperature of the i-th sensor; m is the number of temperature sensors.

[0049] A longitudinal displacement sensor measuring point was set up at the upstream and downstream supports of the two end piers on the main bridge. The cable displacement meter was used, with one end fixed to the upper part of the crossbeam under the main tower and the other end passing over the expansion joint. The data of the measuring point with the largest displacement change amplitude at the upstream and downstream supports of the two end piers on the main bridge were selected as the longitudinal displacement of the beam bridge.

[0050] Thirdly, embodiments of the present invention provide a computer-readable storage medium that, when the instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to perform the above-described intelligent identification method for abnormalities in the expansion joint constraint device of a long-span bridge.

[0051] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent identification of anomalies in expansion joint restraint devices of long-span bridges, characterized in that: include: Obtain the structural temperature and longitudinal displacement of the beam bridge to characterize the state of the beam bridge's restraint devices. Wavelet separation method was used to remove longitudinal displacement noise of the beam bridge, and dynamic and static time series fusion data of beam bridge structure temperature and longitudinal displacement were obtained based on the long-term variation law of temperature-displacement. The constructed Adaboost data anomaly identification model is used to identify and determine the abnormal state of the expansion joint restraint device of the long-span steel box girder bridge by using the dynamic and static time-series fusion data.

2. The intelligent identification method for abnormalities in the expansion joint restraint device of a long-span bridge according to claim 1, characterized in that, Several temperature sensor measuring points are arranged on at least one pier section and at least one mid-span section of the main bridge for both directions. At four different locations on the upper surface, lower surface, top plate, and bottom plate of the box girder at the same pier section and mid-span section, one temperature sensor measuring point is arranged at each of these locations. The average temperature of several temperature sensor measuring points is selected as the temperature of the beam bridge structure. ； In the formula T i Let be the temperature of the i-th sensor; m is the number of temperature sensors.

3. The intelligent identification method for abnormalities in the expansion joint restraint device of a long-span bridge according to claim 2, characterized in that, A longitudinal displacement sensor measuring point was set up at the upstream and downstream supports of the two end piers on the main bridge. The cable displacement meter was used, with one end fixed to the upper part of the crossbeam under the main tower and the other end passing over the expansion joint. The data of the measuring point with the largest displacement change amplitude at the upstream and downstream supports of the two end piers on the main bridge were selected as the longitudinal displacement of the beam bridge.

4. The intelligent identification method for abnormalities in the expansion joint restraint device of a long-span bridge according to claim 3, characterized in that, Based on the number of temperature sensors and the measured temperatures, combined with the top area of ​​the main bridge and the bottom area of ​​the beams and their proportions, the average temperature of the bridge structure is calculated using the area weighting method.

5. The intelligent identification method for abnormalities in the expansion joint restraint device of a long-span bridge according to claim 1, 2, or 3, characterized in that, The temperature-position movement static feature fusion method based on time-series signals includes the following steps: Multi-resolution analysis techniques based on discrete wavelet decomposition decompose long-period fluctuations caused by temperature in displacement into the local features of the signal in both time and frequency domains, while eliminating short-period fluctuations caused by noise and vehicle load. The structural temperature measurement value at the current monitoring time t is used to represent the instantaneous static temperature characteristic. Considering the time lag effect of temperature-displacement, the structural temperature at time t-1 is taken as the temperature time-series dynamic characteristic. The longitudinal displacement measurement value of the beam at the current monitoring time t is used as the instantaneous static characteristic of the displacement. The trained LSTM network model is used to extract dynamic displacement features from the longitudinal displacement features of the bridge measured in real time. The instantaneous static and dynamic characteristics of temperature and displacement of the beam bridge structure at a certain moment are used to construct a comprehensive information time series for that moment using a multi-dimensional feature serial fusion method.

6. The intelligent identification method for abnormalities in the expansion joint restraint device of a long-span bridge according to claim 5, characterized in that, The training and construction method of the Adaboost data anomaly detection model includes the following steps: Anomalies in the displacement data were labeled using an expert experience method. 70% of the samples were used to construct the training set, and the remaining 30% were used as the test set. Using the processed integrated information time series, combined with the BP neural network classification algorithm, multiple weak classifiers are generated. Reduce the weight of samples that have been accurately classified by the weak classifier, and increase the weight of samples that have been misclassified by the weak classifier. The Adaboost algorithm uses a weighted voting mechanism to integrate multiple weak classifiers into a strong classifier to predict the class label of test samples.

7. An intelligent identification system for abnormalities in the expansion joint restraint device of a long-span bridge, characterized in that: Includes the following modules: The data acquisition module is used to acquire the structural temperature and longitudinal displacement of the beam bridge, which characterize the state of the beam bridge restraint device. The time-series fusion module is used to remove longitudinal displacement noise of the beam bridge using wavelet separation method, and to obtain dynamic and static time-series fusion data of beam bridge structure temperature and longitudinal displacement. The anomaly identification module is used to identify and judge the abnormal state of the expansion joint restraint device of the long-span steel box girder bridge by using the constructed Adaboost data anomaly identification model.

8. The intelligent identification system for abnormalities in the expansion joint restraint device of a long-span bridge according to claim 7, characterized in that, Several temperature sensor measuring points are arranged on at least one pier section and at least one mid-span section of the main bridge for both directions. At four different locations on the upper surface, lower surface, top plate, and bottom plate of the box girder at the same pier section and mid-span section, one temperature sensor measuring point is arranged at each of these locations. The average temperature of several temperature sensor measuring points is selected as the temperature of the beam bridge structure. ； In the formula T i Let m be the temperature of the i-th sensor; m is the number of temperature sensors. A longitudinal displacement sensor measuring point was set up at the upstream and downstream supports of the two end piers on the main bridge. The cable displacement meter was used, with one end fixed to the upper part of the crossbeam under the main tower and the other end passing over the expansion joint. The data of the measuring point with the largest displacement change amplitude at the upstream and downstream supports of the two end piers on the main bridge were selected as the longitudinal displacement of the beam bridge.

9. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform the intelligent identification method for abnormality of the long-span bridge expansion and contraction restraint device as described in any one of claims 1 to 6.

Images