Bridge deflection prediction and anomaly identification method for eliminating temperature effect hysteresis influence

Through the combination of wavelet decomposition and GRU neural network, the temperature deflection data of bridges was screened and the temperature-temperature deflection model was constructed, which solved the problem of temperature effect lag in bridge structure health monitoring, and realized the accurate prediction and constraint components of bridge temperature deflection. Online identification of service performance improves monitoring accuracy and real-time.

CN118965950AActive Publication Date: 2024-11-15DALIAN UNIV OF TECH +1
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202410921789.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-10
Publication Date
2024-11-15
Estimated Expiration
2044-07-10

AI Technical Summary

Technical Problem

The existing bridge structure health monitoring system cannot effectively eliminate the hysteresis of temperature effects, resulting in low modeling accuracy and affecting the identification of deflection changes caused by deterioration of service performance of bridge constraint components.

Method used

The temperature deflection data of bridges is extracted by wavelet decomposition method, the main temperature variables are screened, and the temperature-temperature deflection relationship model is constructed based on the GRU neural network, combining principal component analysis and EWMA control chart to achieve accurate prediction and abnormality of bridge temperature deflection Identify.

Benefits of technology

The hysteresis effect of temperature effect is eliminated, the accuracy of bridge temperature deflection prediction is improved, and the online identification of service performance of main beam constrained components and real-time monitoring of bridge performance degradation is realized.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965950A_ABST
    Figure CN118965950A_ABST
Patent Text Reader

Abstract

The invention relates to a bridge deflection prediction and anomaly identification method for eliminating temperature effect hysteresis influence, and belongs to the technical field of bridge structure performance monitoring and evaluation, and the method comprises the following steps: (1) preprocessing bridge deflection and temperature monitoring data; (2) constructing a temperature-temperature-induced deflection correlation model based on a GRU neural network; (3) bridge temperature-induced deflection anomaly identification indexes are extracted; (4) setting a bridge temperature-induced deflection anomaly identification control limit; and (5) on-line identification of temperature-induced deflection abnormity of the bridge. According to the method, the temperature-temperature-induced deflection correlation model based on the GRU neural network is established, and accurate prediction of the temperature-induced deflection of the bridge is realized. And secondly, taking a difference value between a model predicted value and a bridge temperature-induced deflection measured value as a bridge deflection anomaly identification index. According to the index, the influence of temperature fluctuation caused by changes of weather, seasons and the like on the deflection of the bridge is eliminated, temperature-induced deflection changes caused by service performance degradation of the bridge constraint component can be accurately reflected, and the method has high engineering application value in the field of bridge temperature-induced deflection anomaly recognition based on monitoring data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of bridge structure performance monitoring and evaluation, and in particular relates to a bridge deflection anomaly identification method that eliminates the hysteresis influence of temperature effect. Background Art

[0002] During the long-term operation of bridges, the coupling of environmental effects and operational loads leads to performance degradation and component damage. Once a destructive accident occurs on a bridge, it will endanger personal safety, and destructive accidents are difficult to repair quickly, affecting vehicle traffic and causing economic losses. Therefore, in order to monitor the service status of bridges during operation, many bridges are equipped with structural health monitoring systems. The mid-span deflection of the bridge can characterize the service performance degradation of the bridge constraint components and is an effective indicator for evaluating the service safety of the bridge.

[0003] In order to analyze and identify the real bridge structure damage during the operation of the bridge, the bridge condition can be evaluated by the structural health monitoring system. In existing studies, the change of mechanical properties caused by local damage or component failure of the bridge is identified by establishing a correlation model between temperature and deflection or beam end displacement. Chen Hui analyzed the correlation between the longitudinal displacement of the beam end and temperature and established a linear correlation model between the two (Correlation Analysis of Longitudinal Displacement and Temperature of Cable-Stayed Bridge). Deng Yang used a polynomial model to model temperature and beam end displacement, and used a mean control chart to identify displacement changes caused by structural damage (Research on the Correlation between End Displacement and Temperature of Long-Span Suspension Bridge and Its Application). With the popularization and application of deep learning methods, researchers have used deep learning methods to establish a correlation model between temperature and displacement. Hu Tieming established a correlation model between temperature and beam end displacement based on the error back propagation (BP) neural network, and issued an early warning for structural damage (Damage Warning of Cable-Stayed Bridge Based on the Correlation between Temperature and Support Displacement). Zhou established a correlation model between the vertical displacement and temperature of the bridge mid-span based on the BP neural network (Modeling Deformation Induced By Thermal Loading Using Long-Term Bridge Monitoring Data). Related studies have shown that there is a hysteresis between temperature and bridge structure response due to factors such as the temperature gradient of the main beam and solar radiation. The temperature-displacement correlation model established by existing research institutes cannot eliminate the hysteresis of the temperature effect, resulting in low modeling accuracy and affecting the identification of deflection changes caused by the degradation of the service performance of bridge restraint components. Summary of the invention

[0004] In order to overcome the problems existing in the background technology, the present invention provides a bridge deflection prediction and abnormality identification method that eliminates the hysteresis effect of temperature effect, which can eliminate the hysteresis effect between the structural temperature and the temperature-induced deflection of the bridge, and realize accurate prediction of the temperature-induced deflection of the bridge, and the predicted value is consistent with the actual monitoring data. In addition, the method of the present invention can also identify the service performance of the main beam constraint component and the abnormality of the temperature-induced deflection of the bridge.

[0005] To achieve the above object, the present invention is implemented through the following technical solutions:

[0006] The bridge deflection prediction method for eliminating the hysteresis effect of temperature effect comprises the following steps:

[0007] S1, wavelet decomposition method is used to extract the temperature-induced deflection data of the bridge and eliminate the high-frequency deflection components caused by vehicle load and wind load.

[0008] S2, screen out the main temperature variables that affect bridge deflection.

[0009] S3, according to the main temperature variables affecting the bridge deflection, a temperature-temperature-induced deflection relationship model is constructed based on the GRU neural network to predict the temperature-induced deflection of the bridge.

[0010] Furthermore, in step S1, the number of layers of wavelet decomposition is determined according to the sampling frequency of the bridge deflection data and the frequency of the low-frequency signal to be separated.

[0011]

[0012] Where, J is the number of wavelet decomposition layers; f n - sampling frequency of bridge deflection data; f m - The cut-off frequency of the bridge temperature-induced deflection data, the atmospheric temperature change cycle is 24 hours, f m Take 1.16×10 -5 Hz.

[0013] Furthermore, in step S2, the bridge temperature data is subjected to dimensionality reduction processing using the principal component analysis method.

[0014] Let t represent a temperature vector collected by the bridge monitoring system sensor, and the temperature variable matrix composed of m temperature measurement points is T = [t1, t2, ..., t m ], the temperature variable matrix is ​​standardized, and the principal component analysis model is performed through the following eigenvalue decomposition:

[0015] TT T =AΛA T (2)

[0016] Λ=diag(λ1,λ2,…,λ m ) (3)

[0017] A=[a1,a2,…,a m ] (4)

[0018] Where Λ represents a diagonal matrix consisting of m eigenvalues; λ i represents the i-th eigenvalue; A represents the standard orthogonal matrix composed of m eigenvectors; diag represents the diagonal matrix; a i represents the i-th eigenvector, i.e., the i-th main direction;

[0019] The eigenvector a of the orthogonal matrix A i The corresponding eigenvalues ​​are arranged in descending order, and the contribution rates of the temperature principal components of each order are accumulated. When the accumulated contribution rate exceeds 90%, the included temperature principal components are used as the main temperature variables of the bridge.

[0020] Furthermore, in step S3, a temperature-temperature-induced deflection correlation model is constructed based on the GRU neural network.

[0021] Furthermore, the temperature-temperature-induced deflection correlation model is constructed based on the GRU neural network. Specifically, the main temperature variables affecting the bridge deflection are screened out in step S2 as the input of the GRU neural network, and the bridge temperature-induced deflection is used as the output of the network, and the model is iteratively trained. When the number of iterations reaches the maximum number of iterations, it means that the model training is completed. When the GRU neural network is working, the data flow is calculated by the following formula:

[0022] z t =sig(V Tz T t +U hz h t-1 +b z ) (5)

[0023] r t =sig(V Tr T t +U hr h t-1 +b r ) (6)

[0024]

[0025] Where, T t Represents the input temperature data; r t Represents the reset gate in the GRU neural network; z t Represents the update gate in the GRU neural network; V Tz 、V Tr and V Th Represent the update gate, reset gate and weight matrix of the candidate layer respectively; U hz, U hr and U hh Respectively represent the connection weight matrix between the current time t-1 and the previous time t; b z 、b r and b h Represent the update gate, reset gate and bias vector of the candidate layer respectively; is the candidate layer state at time t; h t-1 and h t Represent the unit information of the previous moment and the current moment respectively; sig(·) is the sigmoid function, and its range is [0, 1]; tanh(·) is the hyperbolic tangent function, and its range is [-1, 1];

[0026] The GRU neural network passes through the r in each cell nucleus t and z t The control structure performs calculations to realize the forward propagation of data, thereby realizing the prediction of the temperature-induced deflection of the bridge.

[0027] The main beam restraint member service performance degradation identification method determines whether the bridge deflection is abnormal based on the difference between the predicted value of the bridge temperature-induced deflection and the measured value of the bridge temperature-induced deflection. The calculation formula is:

[0028] ε de =d T-pre -d T-mea (9)

[0029] In the formula, ε de It is the identification index of residual error in prediction of temperature-induced deflection; d T-pre is the predicted value of the temperature-induced deflection of the bridge; d T-mea is the measured value of the temperature-induced deflection of the bridge;

[0030] When the bridge deflection data is abnormal, the temperature-induced deflection prediction residual identification index ε de Increase, indicating that the service performance of the restraining component is degraded.

[0031] The method for identifying the service state of a bridge is to calculate the difference ε between the predicted value of the temperature-induced deflection of the bridge and the measured value of the temperature-induced deflection of the bridge. de , with ε de As the residual identification index of temperature-induced deflection prediction, ε de Conduct statistical analysis to identify abnormal temperature-induced deflection of bridges, specifically,

[0032] The EWMA control chart is used to monitor the bridge deflection, and the temperature-induced deflection anomaly caused by bridge damage or material performance degradation is identified based on the distribution of the temperature-induced deflection prediction residual identification index on the control chart.

[0033] The EWMA control chart is expressed as:

[0034]

[0035] In the formula, λ represents the smoothing parameter, 0<λ≤1; represents the mean of the first i indicators; Z i Represents the weighted cumulative sum of the indicators;

[0036] The control limits of the EWMA control chart are expressed as:

[0037]

[0038] In the formula, μ represents the mean value of the indicator under normal conditions; σ represents the standard deviation of the indicator under normal conditions; k represents a constant used to define the width of the control limit when the process is under control; n represents the number of identified indicators; UCL and LCL represent the upper control limit and the lower control limit of the control chart, respectively;

[0039] The index ε de Enter the EWMA control chart, when Z i When the control limit UCL or LCL is exceeded, it indicates that the temperature-induced deflection of the bridge is abnormal.

[0040] Beneficial effects of the present invention:

[0041] 1. The temperature-temperature-induced deflection correlation model based on the GRU neural network established in the present invention can adaptively eliminate the lag effect between the structural temperature and the temperature-induced deflection of the bridge, and the predicted value of the temperature-induced deflection of the bridge is more consistent with the actual monitoring data.

[0042] 2. The method for identifying abnormal temperature-induced deflection of bridges established in the present invention can realize online identification of service performance degradation of main beam constraint components.

[0043] 3. The abnormal identification index of temperature-induced deflection of bridges proposed in the present invention can accurately reflect the changing characteristics of temperature-induced deflection when the performance of the bridge deteriorates, and can realize the abnormal identification of temperature-induced deflection of bridges in combination with the EWMA control chart. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of the method of the present invention;

[0045] Figure 2 A prediction diagram of temperature-induced deflection of a bridge according to an application example of the present invention;

[0046] Figure 3 This is a diagram for identifying abnormal temperature-induced deflection of a bridge according to an application example of the present invention:

[0047] Remark: Figure 3 In the figure, (a) is working condition 1; (b) is working condition 2. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention.

[0049] Example 1

[0050] The bridge deflection prediction method eliminating the hysteresis effect of temperature effect includes the following steps:

[0051] (1) Extraction of bridge temperature-induced deflection data

[0052] Wavelet decomposition is used to extract bridge temperature-induced deflection data. Wavelet decomposition is essentially the division of data in the frequency domain. The number of wavelet decomposition layers needs to be determined based on the sampling frequency of the bridge deflection data and the frequency of the low-frequency signal that needs to be separated.

[0053] It is known that the atmospheric temperature variation cycle is 24h, so the cutoff frequency f of the bridge temperature-induced deflection data to be separated is m 1.16×10 -5 Hz. If the sampling frequency of the bridge deflection data is f n , then the number of wavelet decomposition layers for extracting the temperature-induced deflection of the bridge is:

[0054]

[0055] The db30 wavelet is selected as the basis function, and the number of wavelet decomposition layers is calculated according to formula (1). The high-frequency deflection components caused by vehicle loads and wind loads can be completely eliminated. When the temperature-bridge deflection relationship model is subsequently established, the influence of the deflection components caused by vehicle loads and wind loads on the model prediction accuracy can be effectively eliminated.

[0056] (2) Screening of main temperature variables of bridges

[0057] The principal component analysis method is used to reduce the dimension of the bridge temperature data. Let t represent a temperature vector collected by the bridge monitoring system sensor, then the temperature variable matrix composed of m temperature measurement points is T = [t1, t2, ..., t m Then, the temperature variable matrix is ​​standardized and the principal component analysis model is performed through the following eigenvalue decomposition:

[0058] TT T =AΛA T (2)

[0059] Λ=diag(λ1,λ2,…,λ m ) (3)

[0060] A=[a1,a2,…,a m ] (4)

[0061] Where Λ represents a diagonal matrix consisting of m eigenvalues; λ i represents the i-th eigenvalue; A represents the standard orthogonal matrix composed of m eigenvectors; diag represents the diagonal matrix; a i represents the i-th eigenvector, that is, the i-th principal direction. i According to the corresponding eigenvalues, the contribution rates of the temperature principal components of each order are accumulated. When the accumulated contribution rate exceeds 90%, the included temperature principal components are taken as the main temperature variables of the bridge.

[0062] There is a strong correlation between the temperature data of multiple measuring points in the bridge health monitoring system. If the temperature data of all measuring points are selected, the modeling of bridge temperature and temperature-induced deflection will be extremely complicated and the calculation cost will be huge. At the same time, complex input information may not only fail to improve the modeling accuracy, but also have a negative impact on the modeling accuracy. Therefore, screening out the main temperature variables that affect the bridge deflection can effectively improve the operating efficiency and prediction accuracy of the temperature-temperature-induced deflection correlation model.

[0063] (3) Constructing a temperature-temperature-induced deflection relationship model based on the GRU neural network

[0064] The present invention uses a GRU neural network to construct a temperature-temperature-induced deflection correlation model, takes the main temperature variable screened out in step (2) as the input of the GRU neural network, takes the bridge temperature-induced deflection extracted in step (1) as the output of the network, and performs iterative training on the model. When the maximum number of iterations is reached, it indicates that the model training is completed.

[0065] The GRU neural network has strong learning ability and can adaptively eliminate the lag effect between variables. It is suitable for processing data sequences containing large amounts of data and with correlations between previous and subsequent moments, such as bridge structure temperature and deflection monitoring data.

[0066] The data flow when the GRU neural network is working can be calculated by the following formula:

[0067] z t =sig(V Tz T t +U hz h t-1 +b z ) (5)

[0068] r t =sig(V Tr T t +U hr ht-1 +b r ) (6)

[0069]

[0070] Where, T t Represents the input temperature data; r t Represents the reset gate in the GRU neural network; z t Represents the update gate in the GRU neural network; V Tz 、V Tr and V Th Represent the update gate, reset gate and weight matrix of the candidate layer respectively; U hz , U hr and U hh Respectively represent the connection weight matrix between the current time t-1 and the previous time t; b z 、b r and b h Represent the update gate, reset gate and bias vector of the candidate layer respectively; is the candidate layer state at time t; h t-1 and h t Represent the unit information of the previous moment and the current moment respectively; sig(·) is the sigmoid function, and its range is [0, 1]; tanh(·) is the hyperbolic tangent function, and its range is [-1, 1].

[0071] The input data of the GRU neural network includes not only the input temperature data T t , and also includes information from the previous moment h t-1 After being processed with independent biases and weights, the information h from the previous moment t-1 will be input to the current unit. The GRU neural network passes through the r in each cell nucleus t and z t The control structure is used to perform corresponding operations to realize the forward propagation of data and further realize the prediction of the temperature-induced deflection of the bridge.

[0072] Example 2

[0073] A method for identifying the service performance degradation of bridge restraint components by building a temperature-temperature-induced deflection relationship model based on GRU neural network

[0074] The difference between the predicted value of Example 1 and the measured value of the temperature-induced deflection of the bridge is used as the bridge deflection anomaly identification index. This index eliminates the influence of temperature fluctuations caused by changes in weather and seasons on the bridge deflection, and can accurately reflect the change in temperature-induced deflection caused by the degradation of the service performance of the bridge restraint component. The calculation formula is:

[0075] ε de =dT-pre -d T-mea (9)

[0076] In the formula, ε de It is the identification index of residual error in prediction of temperature-induced deflection; d T-pre is the predicted value of the temperature-induced deflection of the bridge; d T-mea It is the measured value of the temperature-induced deflection of the bridge.

[0077] The vertical deflection at the mid-span of a bridge is a direct reflection of the mechanical properties of the bridge. Temperature and vehicle load are the main factors that cause the vertical deflection of the main beam during the operation stage of the bridge, and the influence of temperature is even greater than that of the vehicle. Under the long-term effects of temperature and vehicle load, the service performance of the restraining components of the bridge may deteriorate, resulting in abnormal changes in the deflection at the mid-span of the bridge. Therefore, the mid-span deflection becomes an important identification indicator for the degradation of restraining components during the operation of the bridge. The method of the present invention can accurately predict the temperature-induced deflection. When the measured value of the temperature-induced deflection of the beam deflects too much from the predicted value, the deflection can be accurately predicted based on ε de Determine the service performance of bridge restraint components. This embodiment can adopt the same method as the following embodiment 3, using the EWMA control chart to monitor the bridge deflection, and identify and judge the service performance of bridge restraint components based on the distribution of identification indicators on the control chart.

[0078] Example 3

[0079] Application of the method of the present invention to the identification of abnormal temperature-induced deflection of bridges

[0080] (1) The temperature-temperature-induced deflection relationship model constructed according to the method of Example 1 is used to predict the temperature-induced deflection of the bridge.

[0081] (2) The difference between the predicted value of the bridge temperature-induced deflection and the measured value of the bridge temperature-induced deflection ε de As an indicator for identifying abnormal bridge deflection.

[0082] (3) Collect measured values ​​and predicted values, and de Statistical analysis was performed to identify abnormal temperature-induced deflection of bridges, as follows:

[0083] When the bridge is in normal condition, ε de Since the influence of the prediction error of the temperature-temperature-induced deflection correlation model will fluctuate within a small range and have a stable mean and variance. When the performance of the bridge deteriorates, the amplitude of the measured data of the temperature-induced deflection of the bridge decreases, resulting in ε de Therefore, for ε deBy performing statistical analysis, the service status of the bridge can be determined. The exponentially weighted moving average (EWMA) control chart has been widely used in the field of bridge health monitoring. It can make full use of the historical indicators of the identification process, and assign greater weights to historical indicators that are closer to the current ones, and assign smaller weights to indicators that are farther away from the current ones, thereby realizing the identification of small deviations in indicators. During the service period of the bridge structure, the deterioration of the bridge performance causes the bridge deflection to change. The closer the identification indicator is to the current one, the more it can reflect the current performance of the bridge. In view of this feature, the present invention uses the EWMA control chart to monitor the bridge deflection, and identifies the temperature-induced deflection anomaly caused by bridge damage or material performance degradation based on the distribution of the identification indicators on the control chart. The EWMA control chart can be expressed as:

[0084]

[0085] In the formula, λ represents the smoothing parameter, and 0<λ≤1; represents the mean of the first i indicators; Z i Represents the weighted cumulative sum of the indicator.

[0086] The control limits of the EWMA control chart can be expressed as:

[0087]

[0088] In the formula, μ represents the mean value of the indicator under normal conditions; σ represents the standard deviation of the indicator under normal conditions; k represents a constant used to define the width of the control limit when the process is under control; n represents the number of identified indicators; UCL and LCL represent the upper control limit and lower control limit of the control chart, respectively.

[0089] The index ε de Enter the EWMA control chart, when Z i When the control limit UCL or LCL is exceeded, it indicates that the temperature-induced deflection of the bridge is abnormal.

[0090] The method of the present invention can also perform online identification of abnormal temperature-induced deflection of bridges:

[0091] In the online monitoring stage, the monitoring data is preprocessed according to steps (1) and (2) in Example 1, and the data is input into the model constructed in step (3) to obtain the bridge temperature-induced deflection anomaly identification index ε de , and finally de Enter the EWMA control chart, when Z i When the control limit is exceeded, it is identified as abnormal temperature-induced deflection of the bridge; i When it is within the control limit, the temperature-induced deflection of the bridge is identified as normal.

[0092] Application Examples

[0093] The method of the present invention is verified using the temperature and mid-span deflection monitoring data of a simply supported beam bridge in service in China.

[0094] First, the mid-span deflection data is used to extract the temperature-induced deflection data of the bridge according to step (1) of Example 1. Then, the main temperature variables of the bridge are screened according to step (2) of Example 1.

[0095] Then, according to step (3), the temperature-temperature-induced deflection correlation model based on the GRU neural network is trained with the training set data. Figure 2 The figure shows the prediction ability of the temperature-temperature-induced deflection correlation model based on the GRU neural network. From the figure, it can be concluded that the predicted values ​​of the model are in good agreement with the measured values ​​and the prediction accuracy is high.

[0096] Extracting the abnormal identification index ε of the temperature-induced deflection of the bridge under normal conditions de , and set the control limits of the EWMA control chart according to Example 2.

[0097] According to the method of Example 3, the online identification of abnormal temperature-induced deflection of bridges is realized. In order to verify the effectiveness of the abnormal temperature-induced deflection identification method of bridges proposed in the present invention, the measured data of temperature-induced deflection of bridges are increased by 2 mm and 4 mm respectively to simulate the service performance degradation conditions of two restraint components. Calculate the identification index ε de , and then ε de Enter the EWMA control chart. When the Z i When the control limit is exceeded, it indicates that the temperature-induced deflection of the bridge is abnormal. Figure 3 This is a diagram for identifying abnormal temperature-induced deflection of a bridge. It can be seen from the figure that the fluctuation range of the identification index under normal conditions is small, while the indexes under the two conditions of degradation of the service performance of the restraining components exceed the alarm line, and the higher the abnormal increment of the temperature-induced deflection, the greater the amplitude exceeding the alarm line. It can be seen that by adopting the abnormal temperature-induced deflection identification method of the bridge proposed in the present invention, it is possible to realize the identification of abnormal increments of temperature-induced deflection of the bridge of more than 2 mm. Moreover, as the service performance of the restraining component deteriorates more seriously, the amplitude of the abnormal index exceeding the alarm line becomes greater. In summary, the abnormal temperature-induced deflection identification method of the bridge proposed in the present invention can realize online real-time monitoring and identification of service performance degradation of the bridge restraining component based on monitoring data, which verifies the feasibility and practicality of the present invention.

[0098] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A bridge deflection prediction method that eliminates the hysteresis effect of temperature effect, characterized in that: The following steps are involved: S1, using wavelet decomposition method to extract bridge temperature-induced deflection data and eliminate high-frequency deflection components caused by vehicle load and wind load; S2, screen out the main temperature variables that affect bridge deflection; S3, according to the main temperature variables affecting the bridge deflection, a temperature-temperature-induced deflection relationship model is constructed based on the GRU neural network to predict the temperature-induced deflection of the bridge.

2. The prediction method according to claim 1, characterized in that: In step S1, the number of layers of wavelet decomposition is determined according to the sampling frequency of the bridge deflection data and the frequency of the low-frequency signal to be separated. Where, J is the number of wavelet decomposition layers; f n - sampling frequency of bridge deflection data; f m - The cut-off frequency of the bridge temperature-induced deflection data, the atmospheric temperature change cycle is 24 hours, f m Take 1.16×10 -5 Hz.

3. The prediction method according to claim 1, characterized in that: In step S2, the bridge temperature data is processed by principal component analysis. Let t represent a temperature vector collected by the bridge monitoring system sensor, and the temperature variable matrix composed of m temperature measurement points is T = [t1, t2, ..., t m ], the temperature variable matrix is ​​standardized, and the principal component analysis model is performed through the following eigenvalue decomposition: TT T =AΛA T (2) Λ=diag(λ1,λ2,…,λ m ) (3) <h2 style=";text-align:left;direction:ltr">A=[a1,a2,…,a<h2 style=";text-align:left;direction:ltr"> m <h2 style=";text-align:left;direction:ltr"> ] (4) Where Λ represents a diagonal matrix consisting of m eigenvalues; λ i represents the i-th eigenvalue; A represents the standard orthogonal matrix composed of m eigenvectors; diag represents the diagonal matrix; a i represents the i-th eigenvector, i.e., the i-th main direction; The eigenvector a of the orthogonal matrix A i The corresponding eigenvalues ​​are arranged in descending order, and the contribution rates of the temperature principal components of each order are accumulated. When the accumulated contribution rate exceeds 90%, the included temperature principal components are used as the main temperature variables of the bridge.

4. The prediction method according to claim 1, characterized in that: In step S3, a temperature-temperature-induced deflection correlation model is constructed based on the GRU neural network.

5. The prediction method according to claim 4, characterized in that: The temperature-temperature-induced deflection correlation model based on the GRU neural network is specifically constructed by taking the main temperature variables affecting the bridge deflection screened out in step S2 as the input of the GRU neural network, and the bridge temperature-induced deflection as the output of the network, and iteratively training the model. When the number of iterations reaches the maximum number of iterations, it means that the model training is completed. When the GRU neural network is working, the data flow is calculated by the following formula: z t =sig(V Tz T t +U hz h t-1 +b z ) (5) r t =sig(V Tr T t +U hr h t-1 +b r ) (6) Where, T t Represents the input temperature data; r t Represents the reset gate in the GRU neural network; z t Represents the update gate in the GRU neural network; V Tz 、V Tr and V Th Represent the update gate, reset gate and weight matrix of the candidate layer respectively; U hz , U hr and U hh Respectively represent the connection weight matrix between the current time t-1 and the previous time t; b z , b r and b h Represent the update gate, reset gate and bias vector of the candidate layer respectively; is the candidate layer state at time t; h t-1 and h t Represent the unit information of the previous moment and the current moment respectively; sig(·) is the sigmoid function, and its range is [0, 1]; tanh(·) is the hyperbolic tangent function, and its range is [-1, 1]; The GRU neural network passes through the r in each cell nucleus t and z t The control structure performs calculations to realize the forward propagation of data, thereby realizing the prediction of the temperature-induced deflection of the bridge.

6. A method for identifying service performance degradation of a main beam restraining member, wherein the predicted value of the temperature-induced deflection of a bridge is obtained by using the prediction method according to any one of claims 1 to 5, characterized in that: Whether the bridge deflection is abnormal is determined based on the difference between the predicted value of the bridge temperature-induced deflection and the measured value of the bridge temperature-induced deflection. The calculation formula is: e de =d T-pre -d T-mea (9) In the formula, ε de It is the identification index of residual error in prediction of temperature-induced deflection; d T-pre is the predicted value of the temperature-induced deflection of the bridge; d T-mea is the measured value of the temperature-induced deflection of the bridge; When the bridge deflection data is abnormal, the temperature-induced deflection prediction residual identification index ε de Increase, indicating that the service performance of the restraining component is degraded.

7. A method for identifying the service status of a bridge, wherein the predicted value of the temperature-induced deflection of the bridge is obtained by using the prediction method according to any one of claims 1 to 5, characterized in that: Calculate the difference between the predicted value of bridge temperature-induced deflection and the measured value of bridge temperature-induced deflection ε de , with ε de As the residual identification index of temperature-induced deflection prediction, ε de Conduct statistical analysis to identify abnormal temperature-induced deflection of bridges, specifically, The EWMA control chart is used to monitor the bridge deflection, and the temperature-induced deflection anomaly caused by bridge damage or material performance degradation is identified based on the distribution of the temperature-induced deflection prediction residual identification index on the control chart. The EWMA control chart is expressed as: In the formula, λ represents the smoothing parameter, 0<λ≤1; represents the mean of the first i indicators; Z i Represents the weighted cumulative sum of the indicators; The control limits of the EWMA control chart are expressed as: In the formula, μ represents the mean value of the indicator under normal conditions; σ represents the standard deviation of the indicator under normal conditions; k represents a constant used to define the width of the control limit when the process is under control; n represents the number of identified indicators; UCL and LCL represent the upper control limit and the lower control limit of the control chart, respectively; The index ε de Enter the EWMA control chart, when Z i When the control limit UCL or LCL is exceeded, it indicates that the temperature-induced deflection of the bridge is abnormal.

Citation Information

Patent Citations

  • Bridge health evaluation method

    CN108629144A

  • Modeling method of digital regression model of temperature-dependent deflection monitoring reference value of main beam of cable-stayed bridge

    CN113836783A

  • Cable-stayed bridge girder abnormal deflection monitoring method based on deep learning model

    CN114037002A

  • Bridge bearing capacity evaluation method based on passing vehicle load and dynamic deflection monitoring

    CN115791033A

  • Bridge monitoring deflection and temperature separation method based on physical constraint neural network

    CN117807854A