Local damage identification method for 32-meter simply-supported box girder

Abstract

The invention discloses a local damage identification method for a 32-meter simply-supported box girder, and relates to the technical field of bridge health monitoring. The method comprises the following steps: establishing a train-rail-bridge multi-body dynamic model, extracting strain response of each key section of a bridge as a data set for algorithm verification, dividing the key sections into different plate element components based on a plate element analysis method, converting train-induced strain signals into wavelet time-frequency diagrams through wavelet transform, and calculating the train-rail-bridge multi-body dynamic model. The sum of wavelet coefficients in the time-frequency graph is used as potential indexes of damage characteristics, an optimal frequency interval is selected by using a particle swarm algorithm, and the sum of the wavelet coefficients in the optimal frequency interval is not sensitive to passing vehicles under different baseline working conditions and is only sensitive to the damage size of the bridge; and the damage is identified and quantified by comparing whether the damage index exceeds the confidence boundary of the baseline working condition, so that the damage identification of different plate element components of the key section is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge health monitoring, and in particular to a method for identifying local damage of a 32-meter simply supported box girder. Background Art

[0002] As an important hub of transportation, the safety of bridges directly affects the stable operation of transportation and economic development. More than 85% of bridges in my country are prestressed concrete simply supported box girder bridges of equal span. With the influence of natural environment, climate change and human factors during long-term service, they gradually show aging and local damage, resulting in a decrease in the bearing capacity of bridges and forming potential safety hazards. This poses a severe challenge to the safety and durability of bridges. Once damaged, it will cause serious economic losses and adverse social impacts. As my country's bridges gradually enter the concentrated maintenance period, especially a large number of 32-meter simply supported box girders enter the maintenance stage. Therefore, it is urgent to innovate detection, evaluation and maintenance technologies to ensure the safe and stable operation of bridges.

[0003] Traditional bridge damage detection methods mainly rely on regular manual inspections and static tests. Such methods not only consume a lot of manpower and material resources, but also the detection results are easily affected by human subjective factors, making it difficult to achieve real-time online monitoring of bridges. In addition, traditional methods have weak early recognition capabilities for local small-scale damage and have lags. With the development of sensor technology, data processing technology, and health monitoring theory, bridge damage identification methods based on vehicle-induced load response have gradually become a research hotspot. Through the dynamic response of bridge structures such as strain and vibration under train loads, the time-frequency characteristics of the structure can be extracted to effectively identify the location and extent of local damage. Therefore, in order to ensure the healthy operation of in-service bridges, long-term monitoring of the health of 32-meter simply supported box girders is required.

[0004] Due to long-term operation and unpredictability of the external environment, bridges are easily affected by various factors, resulting in potential damage and structural failure. In order to ensure the safety and reliability of bridge operation, it is particularly important to accurately identify the damage of the bridge control section. Therefore, for the local damage identification of the 32-meter simply supported box girder, the monitoring data of the bridge is obtained through sensors. Based on advanced data processing and analysis technology, combined with effective feature extraction methods, a damage identification model is established to automatically identify the damage of the control section, which is of great practical significance for improving the efficiency and accuracy of bridge maintenance, and for timely repair and ensuring the safety of bridge operation. Summary of the invention

[0005] In order to solve the problem that traditional damage identification methods have weak early identification capabilities for local damage and cannot be applied to 32-meter dual-lane box standard girders with obvious bending-torsion coupling effects, the present invention provides a 32-meter simply supported box girder local damage identification method, which converts vehicle-induced strain signals into wavelet time-frequency diagrams through continuous wavelet transform, uses the wavelet coefficients and in the time-frequency diagram as potential indicators of damage characteristics, and uses a particle swarm algorithm to select the optimal frequency interval. The wavelet coefficients and in the optimal frequency interval are insensitive to the passing of vehicles under different baseline conditions but only sensitive to the damage size of the bridge. The damage is identified and quantified by comparing whether the damage index exceeds the confidence boundary of the baseline condition, thereby realizing damage identification of different plate elements in key sections.

[0006] To achieve the above object, the present invention adopts the following technical solution: a method for identifying local damage of a 32-meter simply supported box girder, comprising the following steps:

[0007] Step 1: Establish a multi-body dynamic model of train-track-bridge to simulate the train passing through a 32-meter simply supported box girder. Based on the plate element analysis method, the key sections of the bridge are divided into different plate elements to identify and quantify their damage. By simulating different vehicle passing conditions of the bridge under baseline conditions and damage conditions, the strain responses of the measuring points at different plate elements of each key section of the bridge under different conditions are extracted as the data set for algorithm verification.

[0008] Step 2: Based on the continuous wavelet transform theory, the vehicle-induced strain signal is transformed into a two-dimensional time-frequency diagram in the time-frequency domain, and the sum of the wavelet coefficients in the two-dimensional time-frequency diagram is used for damage identification;

[0009] Step 3: The wavelet coefficients and sum in the two-dimensional time-frequency diagram are used as potential indicators of damage characteristics. The objective function is established by minimizing the standard deviation of the wavelet coefficients and sum in the selected frequency interval of the baseline conditions considering different speeds, axle loads and track spectra, and minimizing the standard deviation of the wavelet coefficients and sum in the selected frequency interval of the same damage condition, and the particle swarm algorithm is used to select the optimal frequency interval;

[0010] Step 4: The wavelet coefficients of the selected optimal frequency interval are not sensitive to the passing of vehicles under different baseline conditions but only sensitive to the size of bridge damage. The damage is identified and quantified by comparing whether the damage eigenvalue exceeds the confidence boundary of the baseline condition. When the damage eigenvalue is equal to or greater than the confidence boundary, the feature can be regarded as an outlier and damage is considered to have occurred.

[0011] Furthermore, in the step one, a multi-body dynamic model of the train-track-bridge is established using the bridge design drawings and the actual parameters of the train, so that the train passes through the bridge at different speeds for numerical simulation; for the baseline condition, different train weights are set in the multi-body dynamic model, and different track irregularity spectra are considered; for the damage condition, the train passes through the bridge at different speeds, and under each speed condition, three key sections of the bridge, L / 4, L / 2 and 3L / 4, are selected for simulation. When simulating the damage condition of a certain key section, each key section is divided into six plate element components according to the plate element analysis method, including the bottom plate, the left web plate, the right web plate, the top plate, the left track plate and the right track plate, and the damage condition of each component of each key section is simulated as four different degrees of stiffness reduction.

[0012] Furthermore, when extracting the strain response, four measuring points are taken on each critical section, including two points where the bottom plate intersects with the centerline of the web and two points where the two inner track lines intersect with the top plate.

[0013] Furthermore, in step 2, the continuous wavelet transform is defined as follows:

[0014]

[0015] In the formula, x(t) represents the signal to be analyzed, ψ(t) represents the wavelet function, * represents complex conjugate, a and b represent the scaling factor and translation factor of the wavelet function respectively.

[0016] Furthermore, in the step 2, for different speeds, it is necessary to keep the length and height of the two-dimensional time-frequency diagram obtained by continuous wavelet transform consistent, convert the time axis of the two-dimensional time-frequency diagram into the travel distance of the train head, make the train travel the same distance at different speeds, linearly interpolate the time domain signal at each speed so that a data sampling point is obtained every 0.1 meter traveled, thereby obtaining time domain signals of the same signal length when traveling the same distance at different speeds, and then obtain two-dimensional time-frequency diagrams of the same size at different speeds through continuous wavelet transform.

[0017] Furthermore, the step three specifically includes:

[0018] 3.1. When using the particle swarm algorithm to select the optimal frequency interval, it is expected that the variance between the data of the baseline condition in the two-dimensional time-frequency diagram is minimized, and the variance between the data of the same damage condition of the same plate element is minimized. The objective function is established as follows:

[0019]

[0020] In the formula, k represents the number of damage categories of the same type in the damage condition, represents the normalized vector of the wavelet coefficients and the baseline condition of the selected P frequency intervals, represents the normalized vector of the wavelet coefficients and a certain type of damage condition of the selected P frequency intervals, var() represents the function for finding the variance, and λ is the adjustment coefficient between the two sub-items in the objective function;

[0021] 3.2. In order to amplify the damage characteristics, the wavelet coefficients and are processed as follows to obtain the damage characteristic value DF, which is defined as follows:

[0022] DF i =Swc i -mean(Swc baseline ) (3)

[0023] Where Swc represents the wavelet coefficient and value of the selected frequency interval, i represents the i-th working condition, mean() represents the function of finding the mean value, and Swc baseline Wavelet coefficients and values ​​for the selected frequency interval representing the baseline condition.

[0024] Furthermore, the step 4 specifically includes:

[0025] 4.1. Assuming that DF obeys normal distribution, the confidence boundary CB of the damage characteristic of the baseline condition is calculated based on the Gaussian inverse cumulative distribution function, which is defined as follows:

[0026] CB=invF(1-α) (4)

[0027] in,

[0028]

[0029] Where μ, σ, and α represent the mean, standard deviation, and significance level of the baseline eigenvector, respectively.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention aims at the characteristics of the eccentric load distribution of two-lane train loads with a 32-meter standard beam mainly used in my country, converts the vehicle-induced strain signal into a wavelet time-frequency diagram through continuous wavelet transform, takes the wavelet coefficients and in the time-frequency diagram as potential indicators of damage characteristics, and uses a particle swarm algorithm to select the optimal frequency interval. The wavelet coefficients and in the optimal frequency interval are insensitive to the passing of vehicles under different baseline conditions but are only sensitive to the damage size of the bridge. The damage is identified and quantified by comparing whether the damage index exceeds the confidence boundary of the baseline condition, thereby realizing damage identification of different plate elements in key sections. The present invention can be used as a component of the real-time early warning subsystem for bridge health monitoring, thereby more comprehensively evaluating the structural performance and health status of the bridge, improving the automation, intelligence, accuracy and robustness of the intelligent identification of the real-time early warning subsystem for bridge health monitoring, and providing a solution for the establishment of a real-time early warning subsystem for online damage identification of bridge health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a rendering of a multi-body dynamics model of a train-track-bridge in an embodiment;

[0032] Figure 2 Schematic diagram of the arrangement of sensor measuring points of a bridge key section of a multi-body dynamics model in an embodiment;

[0033] Figure 3 is a combined schematic diagram of a baseline working condition and a damage working condition of a multi-body dynamics model in an embodiment;

[0034] Figure 4 is a schematic diagram of dividing a key section of a bridge into plate elements in an embodiment;

[0035] Figure 5 is a curve showing the variation of the standard deviation of the sum of wavelet coefficients under the baseline condition and the damage condition with the adjustment coefficient in the embodiment;

[0036] Figure 6 is a curve showing the variation of the standard deviation of the sum of wavelet coefficients under the baseline working condition and the damaged working condition with the number of optimal frequency intervals in the embodiment;

[0037] Figure 7 is the DF value of the wavelet coefficient of the measuring point P3b in the embodiment in the 15 optimal frequency intervals;

[0038] Figure 8 It is the damage identification result of the L / 4 key section when the measuring points P1b-P9b in the embodiment are used for damage detection, wherein part (a) is the bottom plate and part (b) is the left web plate;

[0039] Fig. 9 It is the damage identification result of the L / 4 key section when the measuring points P10b-P18b in the embodiment are used for damage detection, wherein part (a) is the bottom plate and part (b) is the right web plate;

[0040] Fig.10 It is the damage identification result of the L / 4 key section when the measuring points P1t-P9t in the embodiment are used for damage detection, wherein part (a) is the top plate and part (b) is the left track plate. DETAILED DESCRIPTION

[0041] The technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0042] A local damage identification method for a 32-meter simply supported box girder comprises the following steps:

[0043] Step 1: Establish a multi-body dynamic model of train-track-bridge to simulate the train passing through a 32-meter simply supported box girder. Based on the plate element analysis method, the key sections of the bridge are divided into different plate elements to identify and quantify their damage. By simulating different passing conditions of the bridge under baseline conditions (healthy) and damaged conditions, the strain responses of the measuring points at different plate elements of each key section of the bridge under different conditions are extracted as the data set for algorithm verification, including:

[0044] 1.1. Using the bridge design drawings and actual train parameters, a multi-body dynamic model of train-track-bridge is established to simulate the train passing through the bridge at different speeds. For the baseline condition, different train weights are set in the multi-body dynamic model, and different track irregularity spectra are considered;

[0045] 1.2. For the damage condition, the train passes through the bridge at different speeds. Under each speed condition, three key sections of the bridge, L / 4, L / 2 and 3L / 4, are selected for simulation. When simulating the damage condition of a key section, each key section is divided into six plate element components according to the plate element analysis method, including the bottom plate, left web plate, right web plate, top plate, left track plate and right track plate. The damage condition of each component of each key section is simulated as four different degrees (5%, 10%, 20%, 30%) of stiffness reduction. When extracting the strain response, 4 measuring points are taken on each key section, including two points where the bottom plate and the center line of the web intersect and two points where the two inner track lines intersect with the top plate.

[0046] Step 2: Based on the continuous wavelet transform theory, the vehicle-induced strain signal is transformed into a two-dimensional time-frequency diagram in the time-frequency domain, which specifically includes:

[0047] 2.1. Continuous wavelet transform (CWT) obtains the time-frequency information of a signal by convolving the signal with a wavelet function that can be scaled and translated. It is defined as follows:

[0048]

[0049] In the formula, x(t) represents the signal to be analyzed, ψ(t) represents the wavelet function, * represents complex conjugate, a and b represent the scaling factor and translation factor of the wavelet function respectively.

[0050] CWT transforms the wavelet function ψ(t) by the scaling factor a and the translation factor b, then performs a dot product with the analyzed signal x(t) and integrates it over time to obtain the analysis result on a two-dimensional time-frequency diagram;

[0051] 2.2. For different speeds, the length and height of the two-dimensional time-frequency graph obtained by continuous wavelet transform need to be kept consistent. Therefore, the time axis of the two-dimensional time-frequency graph is converted into the travel distance of the train head, so that the train travels the same distance at different speeds. The time domain signal at each speed is linearly interpolated to obtain a data sampling point every 0.1 meter, so that the time domain signal of the same signal length is obtained when the same distance is traveled at different speeds. Then, it is transformed through continuous wavelet transform to obtain the two-dimensional time-frequency graph of the same size at different speeds, and the sum of the wavelet coefficients in the two-dimensional time-frequency graph is further used for damage identification;

[0052] Step 3: The wavelet coefficients and in the two-dimensional time-frequency diagram are used as potential indicators of damage characteristics. The objective function is established by minimizing the standard deviation of the wavelet coefficients and in the selected frequency interval of the baseline conditions considering different speeds, axle loads and track spectra, and minimizing the standard deviation of the wavelet coefficients and in the selected frequency interval of the same damage condition. The particle swarm algorithm is used to select the optimal frequency interval, which includes:

[0053] 3.1. When using the particle swarm algorithm to select the optimal frequency interval, it is expected that the variance between the baseline condition data in the two-dimensional time-frequency diagram is minimized, and the variance between the data of the same damage condition of the same plate element component is minimized. The following objective function is established:

[0054]

[0055] In the formula, k represents the number of damage categories of the same type in the damage condition, represents the normalized vector of the wavelet coefficients and the baseline condition of the selected P frequency intervals, represents the normalized vector of the wavelet coefficients and a certain type of damage condition of the selected P frequency intervals, var() represents the function of finding the variance, and λ is the adjustment coefficient between the two sub-items in the objective function.

[0056] By selecting an appropriate λ in the objective function, a balance between the two sub-items in the objective function is achieved;

[0057] 3.2. Particle swarm optimization algorithm belongs to a class of algorithms for sensor placement optimization problems. The problem of frequency interval selection in this study is similar to the problem of sensor optimization selection. Therefore, the particle swarm optimization algorithm is used to obtain the optimal selection scheme for the required number of frequency intervals. The wavelet coefficients and under the baseline condition show certain random fluctuations due to the influence of different speeds, axle weights and track spectra. In order to amplify the damage characteristics, the wavelet coefficients and are processed as follows to obtain the damage characteristic value (DF), which is defined as follows:

[0058] DF i =Swc i -mean(Swc baseline ) (3)

[0059] Where Swc represents the wavelet coefficient and value of the selected frequency interval, i represents the i-th working condition, mean() represents the function of finding the mean value, and Swc baseline The wavelet coefficients and values ​​representing the selected frequency interval of the baseline condition;

[0060] Step 4: The wavelet coefficients of the selected optimal frequency interval are not sensitive to the passing of vehicles under different baseline conditions but only sensitive to the size of bridge damage. The damage is identified and quantified by comparing whether the damage characteristic value (DF) exceeds the confidence boundary (CB) of the baseline condition, including:

[0061] 4.1. Assuming that DF follows a normal distribution, outlier analysis can be performed based on the statistical threshold. Under this assumption, the confidence bound (CB) of the damage characteristic of the baseline condition is calculated based on the Gaussian inverse cumulative distribution function, which is defined as follows:

[0062] CB=invF(1-α) (4)

[0063] in,

[0064]

[0065] Where μ, σ, and α represent the mean, standard deviation, and significance level of the baseline eigenvector, respectively.

[0066] Therefore, when DF is equal to or greater than CB, the feature can be considered an outlier and damage is considered to have occurred.

[0067] Example

[0068] This is illustrated using field dynamic load test data combined with a simulation data set of a multi-body dynamics model of a train-track-bridge.

[0069] Since it is impossible to simulate the damage condition of the bridge through field tests (it is impossible to forcibly destroy the serving bridge), the established train-track-bridge multi-body dynamic model was modified according to the dynamic load test data of the 32-meter simply supported box girder. The power spectrum analysis was performed based on the acceleration data measured by the field dynamic load test, and the first-order natural frequency was 5.86Hz. After the model was modified, the first-order frequency was 5.81Hz, with a relative error of 0.85%. It can be considered that the static and dynamic performance of the multi-body dynamic model can well represent the actual bridge; the multi-body dynamic model was used to perform numerical simulations of the baseline (healthy) working condition and the damaged working condition to verify the effectiveness of the proposed method. Since the finite element model has been modified, after verifying the effectiveness of the method, it can be directly applied to the early warning system for health monitoring and damage identification of the 32-meter simply supported box girder on site.

[0070] S1. The train passes through the studied box girder at different speeds in the multi-body dynamics model of train-track-bridge. The effect diagram of the multi-body dynamics model is shown in Figure 1 As shown in the figure, the layout of the key measuring points of the corresponding bridge key sections is shown in the figure. Figure 2 shown.

[0071] For the baseline condition, three different train speeds and four different train weights were set in the model, and eight different track irregularity spectra were considered. Figure 3 UIC good and UIC bad in represent low interference spectrum and high interference spectrum, respectively, which are widely used in European high-speed railways and ordinary railways. In summary, a total of 3×4×8=96 different baseline conditions are simulated. Figure 3 As shown, for the damage condition, it is assumed that the train passes through the box girder at three different speeds. Under each speed condition, three key sections, L / 4, L / 2 and 3L / 4, are used to simulate the damage condition. When simulating the damage condition of a certain key section, each key section is divided into six components according to the plate element analysis method. The damage of different components of the key section corresponds to different damage conditions. In addition, the damage conditions of each component of the key section are simulated as four different stiffness reductions. In summary, a total of 3×3×6×4=216 damage conditions were simulated. At the same time, extraction Figure 2 The strain responses of the measuring points P1-P18 in the middle box girder are used as the data set for subsequent damage identification, and the sampling frequency is set to 1000 Hz in the numerical simulation. Figure 4 As shown in the figure, based on the plate element analysis method, the key section of the box girder is divided into 6 different components, namely the bottom plate, left web plate, right web plate, top plate, left track plate and right track plate.

[0072] S2. The data set of baseline working condition and damage working condition of the numerical simulation of the multi-body dynamic model is taken as an example. The analysis result of the P3b measuring point of the L / 4 section is taken as an example. For different speeds of the train, the time axis of the two-dimensional time-frequency diagram is converted into the travel distance of the train head, so that the train travels the same distance at different speeds. The time domain signal at each speed is linearly interpolated so that a data sampling point is obtained every 0.1 meter. In this way, the time domain signal of the same signal length can be obtained when the same distance is traveled at different speeds. Then, the wavelet time-frequency diagram of the same size for signals at different speeds can be obtained by continuous wavelet transform. The sum of the wavelet coefficients in the wavelet time-frequency diagram is further used for damage identification. In this study, the sampling frequency of the time domain signal is 1000Hz, the decomposition scale of the continuous wavelet transform is 128, and the size of the transformed time-frequency diagram is 512×128.

[0073] S3, such as Figure 5As shown in the figure, a series of λ values ​​are set for the objective function in equation (2), and the coordinate points between the scales of the two readings on the horizontal axis are all equally divided points. For the baseline condition, the value of the standard deviation first tends to be stable and then increases with the increase of the adjustment coefficient λ. For the damage condition, the value of the standard deviation first decreases and then tends to be stable with the increase of the adjustment coefficient λ. It is expected to select λ where the standard deviations of the baseline condition and the damage condition are both in a stable stage and close to the minimum value, so the value of λ is set to 0.4. Figure 6 As shown in the figure, it can be seen that for both the baseline condition and the damage condition, the standard deviation value first decreases, then tends to be stable, and finally increases with the increase of the number of optimal frequency intervals. Therefore, in this study, without affecting the accuracy, a smaller number of frequency intervals is preferentially selected to achieve damage identification, and the number of optimal frequency intervals is selected as 15.

[0074] S4, the data set of the baseline working condition based on the numerical simulation of the multi-body dynamics model is combined with equations (4) and (5), and the confidence boundary CB value of each sensor measuring point can be calculated. When the value of λ is set to 0.4 and the number of optimal frequency intervals is set to 15, taking all baseline working conditions and the partial damage working condition of the L / 4 section as an example, for the P3b measuring point, Figure 7 As shown in the figure, for the baseline condition, the change trend of the damage index DF value is very stable, and the change of DF value between different speeds, train weights and different track irregularity spectra is very small. For the damage condition, whether it is the damage to the left web or the bottom plate of the L / 4 section, when the stiffness reduction coefficient of the plate element increases, the damage index DF value increases accordingly and roughly presents a linear change trend. Even when the stiffness reduction coefficient is 5%, the DF value of the damage condition is still greater than the maximum value of DF under the baseline condition, which shows that this method can effectively identify minor damage to plate element components, and at different speeds, the DF values ​​corresponding to the same stiffness reduction coefficient of the same component are very small, which proves the effectiveness and accuracy of the proposed method.

[0075] When a train is passing in the up lane, Figure 8 The figure shows the damage identification results of each member of the L / 4 section when sensors P1b to P9b are used for damage detection. It can be seen from the figure that when the stiffness reduction factor of the member reaches 5%, the damage of the bottom plate and the left web can be detected at point P3b. As the stiffness reduction factor increases, the DF value also increases.

[0076] like Fig. 9 The figure shows the identification results of the member with L / 4 section when sensors P10b-P18b are used for damage detection. Figure 8The difference is that P12b can detect the damage of the right web because it is the sensor point located below the right web. In addition, when points P3b and P12b are used to detect the damage of the bottom plate at the same time, point P3b has a larger DF value. Figure 2 As shown in , this is because when the train passes through the bridge on the up lane, it is closer to point P3b in the lateral direction of the bridge, so point P3b has a more sensitive dynamic response under the bending-torsion coupling effect.

[0077] Fig.10 The damage identification results of some plate elements of the L / 4 section are shown when sensors P1t to P9t are used for damage detection. It can be seen that when the stiffness reduction factor reaches 5%, the damage of the top plate can be detected at the measuring point P3t (L / 4 section). Similarly, as the stiffness reduction factor increases, the DF value also increases. In addition, although the measuring point P3t is not located on the track plate, P3t is close to the left track plate. Therefore, when the stiffness reduction factor is large enough, that is, when the stiffness reduction factor reaches 20%, the damage of the left track plate can be detected at the measuring point P3t. In general, when the identified section is installed with Figure 2 When the four sensors are used, the proposed method can successfully identify, locate and quantify the damage of all six components of the cross section, regardless of whether the train is going up or down through the bridge. This proves the effectiveness and accuracy of the method, as well as its large-scale application in the real-time early warning subsystem for online damage identification of 32-meter simply supported box girder health monitoring.

[0078] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other forms of assembly without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations within the meaning and range of equivalents of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0079] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for identifying local damage of a 32-meter simply supported box girder, characterized in that: The following steps are involved: Step 1: Establish a multi-body dynamic model of train-track-bridge to simulate the train passing through a 32-meter simply supported box girder. Based on the plate element analysis method, the key sections of the bridge are divided into different plate elements to identify and quantify their damage. By simulating different vehicle passing conditions of the bridge under baseline conditions and damage conditions, the strain responses of the measuring points at different plate elements of each key section of the bridge under different conditions are extracted as the data set for algorithm verification. Step 2: Based on the continuous wavelet transform theory, the vehicle-induced strain signal is transformed into a two-dimensional time-frequency diagram in the time-frequency domain, and the sum of the wavelet coefficients in the two-dimensional time-frequency diagram is used for damage identification; Step 3: The wavelet coefficients and sum in the two-dimensional time-frequency diagram are used as potential indicators of damage characteristics. The objective function is established by minimizing the standard deviation of the wavelet coefficients and sum in the selected frequency interval of the baseline conditions considering different speeds, axle loads and track spectra, and minimizing the standard deviation of the wavelet coefficients and sum in the selected frequency interval of the same damage condition, and the particle swarm algorithm is used to select the optimal frequency interval; Step 4: The wavelet coefficients of the selected optimal frequency interval are not sensitive to the passing of vehicles under different baseline conditions but only sensitive to the size of bridge damage. The damage is identified and quantified by comparing whether the damage eigenvalue exceeds the confidence boundary of the baseline condition. When the damage eigenvalue is equal to or greater than the confidence boundary, the feature can be regarded as an outlier and damage is considered to have occurred.

2. A local damage identification method for a 32-meter simply supported box girder according to claim 1, characterized in that: In the step 1, a multi-body dynamics model of train-track-bridge is established using the bridge design drawings and the actual parameters of the train, so that the train passes through the bridge at different speeds for numerical simulation; for the baseline working condition, different train weights are set in the multi-body dynamics model, and different track irregularity spectra are considered; For the damage condition, the train is made to pass through the bridge at different speeds. Under each speed condition, three key sections of the bridge, L / 4, L / 2 and 3L / 4, are selected for simulation. When simulating the damage condition of a certain key section, each key section is divided into six plate element components according to the plate element analysis method, including the bottom plate, left web plate, right web plate, top plate, left track plate and right track plate. The damage condition of each component in each key section is simulated as four different degrees of stiffness reduction.

3. A local damage identification method for a 32-meter simply supported box girder according to claim 2, characterized in that: When extracting the strain response, four measuring points were taken on each key section, including two points where the bottom plate and the center line of the web intersected and two points where the two inner track lines intersected the top plate.

4. The local damage identification method of a 32-meter simply supported box girder according to claim 1 is characterized in that: In the step 2, the continuous wavelet transform is defined as follows: In the formula, x(t) represents the signal to be analyzed, ψ(t) represents the wavelet function, * represents complex conjugate, a and b represent the scaling factor and translation factor of the wavelet function respectively.

5. The local damage identification method of a 32-meter simply supported box girder according to claim 1 is characterized in that: In the step 2, for different speeds, it is necessary to keep the length and height of the two-dimensional time-frequency diagram obtained by continuous wavelet transform consistent, convert the time axis of the two-dimensional time-frequency diagram into the travel distance of the train head, make the train travel the same distance at different speeds, and linearly interpolate the time domain signal at each speed so that a data sampling point is obtained every 0.1 meter traveled, so as to obtain a time domain signal of the same signal length when traveling the same distance at different speeds, and then obtain the two-dimensional time-frequency diagram of the same size at different speeds through continuous wavelet transform.

6. The local damage identification method of a 32-meter simply supported box girder according to claim 1 is characterized by: The step three specifically includes: 3.

1. When using the particle swarm algorithm to select the optimal frequency interval, it is expected that the variance between the data of the baseline condition in the two-dimensional time-frequency diagram is minimized, and the variance between the data of the same damage condition of the same plate element is minimized. The objective function is established as follows: In the formula, k represents the number of damage categories of the same type in the damage condition, represents the normalized vector of the wavelet coefficients and the baseline condition of the selected P frequency intervals, represents the normalized vector of the wavelet coefficients and a certain type of damage condition of the selected P frequency intervals, var() represents the function for finding the variance, and λ is the adjustment coefficient between the two sub-items in the objective function; 3.

2. In order to amplify the damage characteristics, the wavelet coefficients and are processed as follows to obtain the damage characteristic value DF, which is defined as follows: DF i =Swc i -mean(Swc baseline ) (3) Where Swc represents the wavelet coefficient and value of the selected frequency interval, i represents the i-th working condition, mean() represents the function of finding the mean value, and Swc baseline Wavelet coefficients and values ​​for the selected frequency interval representing the baseline condition.

7. A local damage identification method for a 32-meter simply supported box girder according to claim 6, characterized in that: The step 4 specifically includes: 4.

1. Assuming that DF obeys normal distribution, the confidence boundary CB of the damage characteristic of the baseline condition is calculated based on the Gaussian inverse cumulative distribution function, which is defined as follows: CB=invF(1-α) (4) in, Where μ, σ, and α represent the mean, standard deviation, and significance level of the baseline eigenvector, respectively.

Images