Bridge damage identification method based on vibration energy entropy

Abstract

The application discloses a bridge damage identification method based on vibration energy entropy, and comprises the following steps: dividing a bridge into multiple units with equal length and setting multiple measuring points; sampling and acquiring velocity response time history vectors of the measuring points; defining vibration energy density, vibration energy entropy and calculation methods; defining damage factors of the units and calculation methods; establishing a finite element model; constructing a sensitivity matrix; calculating the theoretical vibration energy entropy of the bridge by using the finite element model and updating the value of the sensitivity matrix; calculating the actual vibration energy entropy of the bridge; correcting the current value of each unit damage factor after solving the sensitivity equation; and outputting the unit damage factors after iteration to convergence. The identification method overcomes the problems of the prior art, such as the need to set a large number of sensors when identifying bridge damage, complex data processing, low efficiency, insufficient identification accuracy and the like, and realizes rapid, efficient and accurate damage identification of the bridge by fusing the data of each measuring point and then identifying the bridge damage by using the fused data.

Description

Technical Field

[0001] This invention relates to the field of bridge technology, and in particular to a bridge damage identification method based on vibration energy entropy. Background Technology

[0002] During the use of bridges after construction, damage may occur due to various factors such as external forces and aging of building materials. In order to ensure the normal use of bridges, it is necessary to conduct regular inspections of the bridge's condition, determine the location and extent of damage, and carry out timely repairs in accordance with safety standards.

[0003] Using a single-point testing method for bridges can lead to problems such as significant noise interference and poor stability. In practical applications, a common approach is to first randomly deploy a small number of points to identify potentially damaged areas, and then deploy more points within those areas to identify damaged units and their extent. If a large number of testing points are directly deployed on the bridge structure to improve identification accuracy, this would increase the number of required testing equipment and other testing costs. Furthermore, if the number of sensors is very large, the amount of data obtained will also be enormous, making rapid data processing and accurate identification of damage locations a major challenge. Chinese Patent Publication No. CN111506870B, published on March 31, 2023, entitled "A Time-Varying Structural Damage Identification Method Based on Wavelet Transform," discloses a structural damage identification method. This method uses wavelet transform to identify the instantaneous frequency and mode shape of a time-varying structure and calculates the difference between instantaneous mode shapes, which can be used for bridge damage identification. However, the damage identification method provided by this patent requires wavelet transform of the response signals of each measuring point. This not only requires setting up a large number of sensors, but also the high complexity of data processing such as multiple wavelet transforms will affect the identification efficiency, cannot make full use of the correlation characteristics of the data of each measuring point, and the identification accuracy is easily affected by local noise. Summary of the Invention

[0004] This invention overcomes the problems of existing technologies that require a large number of sensors, have complex and inefficient data processing, and lack sufficient accuracy in identifying bridge damage. It provides a bridge damage identification method based on vibration energy entropy. The method divides the bridge into multiple units and sets a corresponding number of measuring points. By defining and calculating energy density, vibration energy entropy, and damage factor, and utilizing the sensitivity of the bridge vibration energy entropy to the damage factor, the data from each measuring point are first fused, and then the bridge damage is identified using the fused data, thus achieving fast, efficient, and accurate damage identification of bridges.

[0005] To achieve the above objectives, the present invention adopts the following solution:

[0006] The bridge damage identification method based on vibration energy entropy includes the following steps:

[0007] S1: Divide the bridge into multiple units of equal length along the bridge length, and set multiple measuring points evenly on the bridge body;

[0008] S2: Sample and obtain the velocity response time history vector of each measuring point at several sampling time points within the velocity response time history;

[0009] S3: Definition of vibration energy density and vibration energy entropy, and their calculation methods:

[0010]

[0011] H(t)=E(t)ln(E(t))

[0012] Where t is the sampling time corresponding to the current sampling time point, p is the number of measurement points, and u j The velocity response time history vector obtained for the j-th measuring point. Let E(t) be the energy density at the j-th measuring point, E(t) be the sum of vibration energy densities at all measuring points, and H(t) be the vibration energy entropy.

[0013] The relative reduction in the bending stiffness of an element is defined as the element's damage factor, which is calculated as follows:

[0014]

[0015] Among them, (EI) i0 and (EI) i These are the bending stiffnesses of the i-th element before and after the damage, respectively.

[0016] S4: Based on the structural characteristics of the bridge, establish a finite element model for calculating the theoretical vibration energy entropy of the bridge corresponding to different damage factors. Based on the sensitivity of the bridge vibration energy entropy to each element damage factor, construct a sensitivity matrix and set initial values ​​for each element damage factor.

[0017] S5: Using the current value of the damage factor of each unit as input, calculate the theoretical vibration energy entropy of the bridge using the finite element model and update the value of the sensitivity matrix;

[0018] S6: Select an unselected sampling time point according to the sampling time sequence, calculate the vibration energy density of each measuring point and the actual vibration energy entropy of the bridge based on the velocity response time history vector obtained from this sampling time point; establish a sensitivity equation based on the difference between the theoretical vibration energy entropy and the actual vibration energy entropy of the bridge and the sensitivity matrix, solve the sensitivity equation to obtain the correction value of the damage factor of each unit, and correct the current value of the damage factor of each unit accordingly.

[0019] S7: Iterate through steps S5-S6 until the values ​​to be corrected for the damage factors of each unit meet the convergence condition. After the iteration is completed, output the values ​​of the damage factors of each unit as the identification results.

[0020] As a preferred method, the sensitivity matrix is ​​constructed in step S4 using the following approach:

[0021] The vibration energy entropy of the bridge is relative to the damage factor α of the i-th unit. i The sensitivity is expressed in vector form:

[0022]

[0023] Where m is the number of sampling time points, The vibration energy entropy of the entire bridge relative to the damage factor α of the i-th unit. i Sensitivity, The vibration energy entropy at the k-th measuring point relative to the damage factor α of the i-th unit. i Sensitivity;

[0024] The sensitivity of the bridge's vibration energy entropy to the damage factor of each element was calculated separately. Combined into matrix form:

[0025]

[0026] Where n is the number of units divided, and matrix S is the constructed sensitivity matrix;

[0027] The sensitivity equation established in step S5 is based on the following sensitivity relationship:

[0028]

[0029] Where ΔH is the difference between the theoretical vibration energy entropy and the actual vibration energy entropy, and Δα is the value to be corrected for the damage factor.

[0030] Preferably, the convergence condition described in step S7 is: or |Δα k | <T2;

[0031] Where, Δα k The value to be corrected for the element damage factor calculated in the k-th iteration is... T1 and T2 are the current values ​​of the unit damage factor before the k-th correction, and both are convergence threshold values.

[0032] Preferably, this identification method further includes the following steps:

[0033] S8: Units with damage factor values ​​higher than the threshold are considered as damage units. A damage localization neural network model is established based on the mapping relationship between the velocity response time history vectors of the three measurement points before and after the damage unit and the specific damage location on the damage unit. The sampling data of previous damage identification and the accurate damage location data after verification and correction are packaged into a dataset and input into the damage localization neural network model for model training.

[0034] S9: Input the velocity response time history vectors of the identified damaged unit and the three measurement points before and after it into the trained damage localization neural network model to obtain the accurate damage location on the identified damaged unit. After verifying the damage location in the field, use the correct data to correct the recognition accuracy of the damage localization neural network model again.

[0035] Preferably, step S8 further includes the following steps:

[0036] For units where it is impossible to obtain three measurement points forward or backward, three measurement points are set separately on each unit. An auxiliary localization neural network model is established based on the mapping relationship between the velocity response time history vector of the three measurement points on the unit and the specific damage location on the unit. The sampling data and identification results of these units as damage units are selected and the damage localization neural network model is trained.

[0037] Step S9 further includes the following steps:

[0038] For the damaged unit identified this time, if it is impossible to obtain three measurement points forward or backward, the velocity response time history vectors of the three measurement points set separately on the damaged unit are input into the trained auxiliary localization neural network model to obtain the accurate damage location on the identified damaged unit. After verifying the damage location in the field, the recognition accuracy of the auxiliary localization neural network model is corrected again with the correct data.

[0039] Preferably, step S1 further includes: running a test train on the bridge, and setting up a measuring point at the front and rear of the train and between the two carriages;

[0040] Step S2 also includes: sampling and obtaining the speed response time history vector of each measuring point on the train at several sampling time points within the speed response time history;

[0041] Step S4 also includes: establishing a finite element model for calculating the theoretical vibration energy entropy of the train corresponding to different damage factors based on the structural characteristics of the bridge and the operating state characteristics of the train; constructing a train sensitivity matrix based on the sensitivity of the train vibration energy entropy to the damage factors of each unit; and creating a second damage factor for each unit and setting an initial value.

[0042] Step S5 also includes: using the current value of the second damage factor of each unit as input, calculating the theoretical vibration energy entropy of the train using the finite element model and updating the value of the train sensitivity matrix;

[0043] Step S6 further includes: selecting an unselected train response data sampling time point according to the sampling time sequence; calculating the vibration energy density of each measuring point on the train and the actual vibration energy entropy of the train based on the speed response time history vector obtained from this sampling time point; establishing a train sensitivity equation based on the difference between the theoretical vibration energy entropy and the actual vibration energy entropy of the train and the train sensitivity matrix; solving the train sensitivity equation to obtain the correction value of the second damage factor of each unit and correcting the current value of the second damage factor of each unit accordingly.

[0044] Step S7 further includes: iterating through steps S5-S6 until the values ​​to be corrected for the second damage factor of each unit all meet the convergence condition. After the iteration is completed, the values ​​of the second damage factor of each unit are output, and the reliability of the damage factor of the potentially damaged units in the identification results is also output. The reliability of the damage factor is calculated as follows:

[0045]

[0046] Where, q x To determine the reliability of the damage factor for the xth potentially damaged unit, α x1 α is the damage factor of this element, calculated based on the energy entropy of bridge vibration. x2 This is the second damage factor of the unit, calculated based on the train vibration energy entropy.

[0047] The present invention has at least the following beneficial effects: (1) By defining and calculating energy density and vibration energy entropy, the required feature quantities can be extracted quickly and accurately from the sampled data, greatly shortening the time required for damage identification; (2) By using the method of integrating multiple measurement point data for bridge damage identification, the data of all affected measurement points are identified in combination with the damage location, resulting in high identification accuracy and less susceptibility to local noise; (3) By using the iterative convergence method, accurate identification results can be obtained without calculating all sampled point data, resulting in high identification efficiency; (4) The train response model and the bridge response model can be combined to further improve the accuracy of identification, and the identification method is more flexible, effectively solving the problems of inconvenient bridge measurement point setting and data deviation; (5) The bridge response model can be assisted by the damage localization neural network model to quickly obtain the location and degree of damage of the damage unit and accurately locate the specific damage location on the damage unit. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the present invention;

[0049] Figure 2This is a schematic diagram of a test bridge structure according to the present invention;

[0050] Figure 3 The result is the velocity response identification at the 5th measuring point under operating condition 1.

[0051] Figure 4 The result is the velocity response identification at the 7th measuring point under operating condition 1.

[0052] Figure 5 The result is the velocity response identification at the 9th measuring point under operating condition 1.

[0053] Figure 6 The result is the velocity response identification at the 11th measuring point under operating condition 1.

[0054] Figure 7 The result is the velocity response identification at the 13th measuring point under operating condition 1.

[0055] Figure 8 The average of the velocity response identification results at measuring points 5, 7, 9, 11, and 13 under operating condition 1;

[0056] Figure 9 The identification result of the identification method provided by the present invention under working condition 1;

[0057] Figure 10 The result of velocity response identification at the 13th measuring point when local noise is increased;

[0058] Figure 11 The recognition result of the recognition method provided by the present invention when local noise is increased;

[0059] Figure 12 The identification results of the identification method provided by the present invention when identifying double-damaged bridges. Detailed Implementation

[0060] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0061] like Figure 1-2 As shown, the bridge damage identification method based on vibration energy entropy provided by the present invention includes the following steps:

[0062] S1: Divide the bridge into multiple units of equal length along its length, and evenly distribute multiple measuring points across the bridge. The specific number of units is determined based on the actual length of the bridge. Generally, the bridge is divided into units of equal length. For areas with a high probability of damage, the length of the units in those areas can be shortened accordingly. The number of measuring points and their distribution on the bridge are determined by considering factors such as the number of units, the complexity of the bridge structure, and the influence of the testing environment on the sampling data. As an example, for instance... Figure 2As shown, the bridge is 32 meters long. The bridge is divided into 16 equal units, and a total of 17 measuring points are set at both ends of the bridge and between the units.

[0063] S2: At several sampling time points within the velocity response time history, the velocity response time history vectors of each measuring point are obtained. The velocity response time history is the time interval during which the bridge structure generates a velocity response when a train passes. The velocity response time history vector is the representation vector of the bridge's velocity response within this time interval. The sampling time points are selected according to the actual response period and response duration of the bridge as an optimal range. The sampling interval is set between 0.01 and 0.05 seconds. A velocity sensor is used for sampling. The main data collected is the vertical velocity of the bridge vibration. Each measuring point is sampled at each sampling time point and combined into a set of sampling data.

[0064] S3: Definition of vibration energy density and vibration energy entropy, and their calculation methods:

[0065]

[0066] H(t)=E(t)ln(E(t))

[0067] Where t is the sampling time corresponding to the current sampling time point, p is the number of measurement points, and u j The velocity response time history vector obtained for the j-th measuring point. Let E(t) be the energy density of the j-th measuring point, E(t) be the sum of the vibration energy densities of all measuring points, and H(t) be the vibration energy entropy. Since the velocity response time history vector is a vector, the physical meaning of directly linearly adding the responses of multiple measuring points is unclear, and the effect of directly using it for damage identification is not good. The vibration energy change value obtained by summing the velocity response after converting it into an energy response is highly correlated with the damage location and degree of the bridge. Because the vibration energy is proportional to the square of the vibration velocity, the square of the velocity response time history vector of the measuring point is defined as the energy density to characterize the vibration energy response of the measuring point. The entropy of the energy density is defined as the vibration energy entropy. Due to the continuity of the bridge structure itself, there is a correlation between the divided units. When the bridge structure is damaged, the response of all measuring points will change, and the vibration energy entropy of the measuring points will also change accordingly. This can more directly and accurately reflect the specific damage location and degree of the bridge.

[0068] The relative reduction in the bending stiffness of an element is defined as the element's damage factor, which is calculated as follows:

[0069]

[0070] Among them, (EI) i0 and (EI) iThese are the bending stiffness of the i-th element before and after damage, respectively. Bending stiffness is an important performance indicator of a bridge, and the degree of decrease in the bending stiffness of a bridge is the main way to characterize bridge damage. The relative reduction value of the bending stiffness of a bridge element is defined as the damage factor of the element, which is used to directly reflect the degree of damage to the bridge at each element location.

[0071] S4: Based on the structural characteristics of the bridge, a finite element model is established to calculate the theoretical vibration energy entropy of the bridge corresponding to different damage factors. A sensitivity matrix is ​​constructed based on the sensitivity of the bridge vibration energy entropy to the damage factors of each element, and initial values ​​are set for the damage factors of each element. The finite element model is a model established using the finite element analysis method, treating the bridge as an assembly of elements. Each element transmits force and is constrained only at the element connection. Based on this model, the correspondence between the damage factors of each element of the bridge and the vibration energy entropy of the bridge is analyzed and calculated. When the damage factor values ​​of each element are known or preset, the theoretical vibration energy entropy of the bridge under this damage state can be quickly calculated based on the correspondence. The degree of change of the bridge vibration energy entropy when the damage factors of each element change is defined as sensitivity. The sensitivity of the overall vibration energy entropy of the bridge to the damage factors of each element is combined into a sensitivity vector. Then, based on multiple sets of different damage factor data, the sensitivity vector is expanded into a sensitivity matrix.

[0072] S5: Using the current value of the damage factor for each element as input, the theoretical vibration energy entropy of the bridge is calculated using the finite element model, and the value of the sensitivity matrix is ​​updated. Since the finite element model has established the correspondence between the damage factor of each element and the vibration energy entropy of the bridge, given the preset or updated value of the damage factor, the theoretical vibration energy entropy of the bridge in this state can be quickly calculated using the finite element model. The value of the theoretical vibration energy entropy calculated by the finite element model is compared with the actual vibration energy entropy calculated from the actual data detected at the measuring points. The difference between the two is converted into a correction value for the damage factor through sensitivity, so that the theoretically calculated damage factor value continuously approaches the actual value until the convergence condition is met. The value of each element in the sensitivity matrix is ​​the sensitivity of the bridge vibration energy entropy to the damage factor. Its value will change due to the correction of the damage factor value, so it must be calculated and updated accordingly in each iteration.

[0073] S6: Select an unselected sampling time point according to the sampling time sequence, calculate the vibration energy density of each measuring point and the actual vibration energy entropy of the bridge based on the velocity response time history vector obtained from this sampling time point; establish a sensitivity equation based on the difference between the theoretical vibration energy entropy and the actual vibration energy entropy of the bridge and the sensitivity matrix, solve the sensitivity equation to obtain the correction value of the damage factor of each unit, and correct the current value of the damage factor of each unit accordingly.

[0074] S7: Iterate through steps S5-S6 until the correction values ​​of the damage factors for each element meet the convergence condition. After the iteration, output the values ​​of the damage factors for each element as the identification result. The iterative process is essentially to continuously correct the calculated element damage factors so that the difference between them and the actual damage factors is reduced to the convergence range. Since the actual value of the damage factors or their numerical range cannot be directly measured, the actual vibration energy entropy is calculated using the actual measured velocity response data, and the theoretical vibration energy entropy is calculated using the theoretical value of the damage factors after the previous correction and the finite element model. The difference between the theoretical vibration energy entropy and the actual vibration energy entropy is converted into the value that the theoretical value of the damage factors needs to be corrected in this iteration through sensitivity. Note that the correction value here is not necessarily the exact difference between the theoretical value and the actual value of the damage factors. The accuracy of this correction is reflected in the calculation of the difference between the theoretical vibration energy entropy and the actual vibration energy entropy using the corrected damage factor value in the next iteration. When the difference between the theoretical energy entropy and the actual energy entropy is extremely small, the difference between the theoretical value and the actual value of the damage factors is negligible. At this time, the iteration ends and the theoretical value of the damage factors at this time is output as the actual value.

[0075] Compared to directly using velocity response time history vectors for superposition or other vector-based calculations, this method defines and calculates energy density and vibration energy entropy, enabling rapid and accurate extraction of necessary feature quantities from sampled data. It transforms vectors into scalars, reducing computational complexity and significantly shortening the time required for damage identification. In terms of data processing, this method integrates data from multiple measurement points for bridge damage identification, combining damage location with data from all affected measurement points. This results in high accuracy and is less susceptible to local noise. Furthermore, this method employs an iterative convergence approach, utilizing data from all measurement points for the calculation of damage factors in each unit. The data is relatively stable and reliable, unlike methods such as direct velocity response vector calculations which are easily affected by fluctuations in single-point data. Under normal circumstances, the number of data sets required within each iteration is small, eliminating the need to calculate data from all sampling points to obtain accurate identification results, thus achieving high identification efficiency.

[0076] In another technical solution, the sensitivity matrix is ​​constructed in step S4 using the following method:

[0077] The vibration energy entropy of the bridge is relative to the damage factor α of the i-th unit. i The sensitivity is expressed in vector form:

[0078]

[0079] Where m is the number of sampling time points, The vibration energy entropy of the entire bridge relative to the damage factor α of the i-th unit. i Sensitivity, The vibration energy entropy at the k-th measuring point relative to the damage factor α of the i-th unit. i Sensitivity;

[0080] The sensitivity of the bridge's vibration energy entropy to the damage factor of each element was calculated separately. Combined into matrix form:

[0081]

[0082] Where n is the number of units divided, and matrix S is the constructed sensitivity matrix;

[0083] The sensitivity equation established in step S5 is based on the following sensitivity relationship:

[0084]

[0085] Where ΔH is the difference between the theoretical vibration energy entropy and the actual vibration energy entropy, and Δα is the value to be corrected for the damage factor.

[0086] The convergence condition mentioned in step S7 is or |Δα k | <T2;

[0087] Where, Δα k The value to be corrected for the element damage factor calculated in the k-th iteration is... T1 and T2 are the current values ​​of the unit damage factor before the k-th correction, and both are convergence threshold values.

[0088] In another technical solution, the identification method further includes the following steps:

[0089] S8: Units with damage factor values ​​exceeding a threshold are considered damaged units. A damage localization neural network model is established based on the mapping relationship between the velocity response time history vectors of three measurement points before and after the damaged unit and the specific damage location on the damaged unit. Previous damage identification sampling data and the accurate damage location data after verification and correction are packaged into a dataset and input into the damage localization neural network model for model training. The data used to train the damage localization neural network model utilizes the velocity response time history vectors of three measurement points before and after the damaged unit. The data from these measurement points contains most of the velocity response information of the damaged unit. Detection data outside these measurement points, after being evenly distributed, are used to determine the specific damage on the unit. Location is not very meaningful. Removing these meaningless data can reduce the amount of training data for the damage localization neural network model, speeding up the training without affecting the training effect. It should be noted that the damage unit and its surrounding measurement point data mentioned here are not necessarily independent. When there are two or more adjacent damage units on the bridge, there will be mutual influence such as data superposition between damage units. Therefore, all damage units and their related surrounding measurement point data in a single test result are placed in a set of data. The input data of the damage localization neural network model includes not only the damage factor value of each unit and the actual damage location after inspection and verification, but also information such as the number of damage units, their distribution location, and the superposition of measurement point data.

[0090] S9: Input the velocity response time history vectors of the identified damaged unit and the three measurement points before and after it into the trained damage localization neural network model to obtain the accurate damage location on the identified damaged unit. After verifying the damage location in the field, use the correct data to correct the recognition accuracy of the damage localization neural network model again.

[0091] Step S8 also includes the following steps:

[0092] For units where it is impossible to obtain three measurement points forward or backward, three measurement points are set separately on each unit. An auxiliary localization neural network model is established based on the mapping relationship between the velocity response time history vector of the three measurement points on the unit and the specific damage location on the unit. The sampling data and identification results of these units as damage units are selected and the damage localization neural network model is trained. For units at the beginning and end of the bridge, the measurement point data has continuity towards the middle of the bridge but not towards the two ends. The construction structure at the two ends is also different from that in the middle section. Considering the special characteristics of the beginning and end of the bridge, a limited number of independent measurement points are set for them. Without increasing the equipment cost too much, the comprehensive detection of the bridge and the accurate identification of the real-time position can be achieved with the help of the auxiliary localization neural network model.

[0093] Step S9 further includes the following steps:

[0094] For the damaged unit identified this time, if it is impossible to obtain three measurement points forward or backward, the velocity response time history vectors of the three measurement points set separately on the damaged unit are input into the trained auxiliary localization neural network model to obtain the accurate damage location on the identified damaged unit. After verifying the damage location in the field, the recognition accuracy of the auxiliary localization neural network model is corrected again with the correct data.

[0095] In another technical solution, step S1 further includes: running a test train on the bridge, setting up a measuring point at the front and rear of the train and between two carriages; based on condition 1, the train is composed of 4 10DOF carriages, maintaining a constant speed of 25m / s.

[0096] Step S2 also includes: sampling and obtaining the speed response time history vector of each measuring point on the train at several sampling time points within the speed response time history;

[0097] Step S4 also includes: establishing a finite element model for calculating the theoretical vibration energy entropy of the train corresponding to different damage factors based on the structural characteristics of the bridge and the operating state characteristics of the train; constructing a train sensitivity matrix based on the sensitivity of the train vibration energy entropy to the damage factors of each unit; and creating a second damage factor for each unit and setting an initial value.

[0098] Step S5 also includes: using the current value of the second damage factor of each unit as input, calculating the theoretical vibration energy entropy of the train using the finite element model and updating the value of the train sensitivity matrix;

[0099] Step S6 further includes: selecting an unselected train response data sampling time point according to the sampling time sequence; calculating the vibration energy density of each measuring point on the train and the actual vibration energy entropy of the train based on the speed response time history vector obtained from this sampling time point; establishing a train sensitivity equation based on the difference between the theoretical vibration energy entropy and the actual vibration energy entropy of the train and the train sensitivity matrix; solving the train sensitivity equation to obtain the correction value of the second damage factor of each unit and correcting the current value of the second damage factor of each unit accordingly.

[0100] Step S7 further includes: iterating through steps S5-S6 until the values ​​to be corrected for the second damage factor of each unit all meet the convergence condition. After the iteration is completed, the values ​​of the second damage factor of each unit are output, and the reliability of the damage factor of the potentially damaged units in the identification results is also output. The reliability of the damage factor is calculated as follows:

[0101]

[0102] Where, q x To determine the reliability of the damage factor for the xth potentially damaged unit, α x1The damage factor of this element is calculated based on the energy entropy of bridge vibration. x2 This is the second damage factor of the unit, calculated based on the train vibration energy entropy. The combination and comparison of the two measurement methods further increases the accuracy of the detection. For units where the identification results differ significantly between the two methods in rare cases, other verification methods can be considered.

[0103] The methods for processing the arrangement of measuring points on the train, the calculation of the train vibration energy entropy, and its correspondence with the damage factors of each unit of the bridge can all be deduced from the methods for processing bridge measuring points and measuring point data, which is obvious to those in the field.

[0104] Based on actual application scenarios, taking some test data as an example, the bridge adopts the following... Figure 2 The design structure shown has a bridge length of 32 meters. The bridge is divided into 16 equal units, and a total of 17 measuring points are set at both ends of the bridge and between the units. Based on this, the design condition 1 is: the damaged unit is the 6th unit, the damage factor is 10%, and the test noise is set to 5%.

[0105] First, the damage factors are directly identified and obtained using the velocity response identification method at each measuring point, such as... Figure 3-7 The results shown are the velocity response identification results for measuring points 5, 7, 9, 11, and 13, respectively. Figure 8 The figure shows the average data of each identification result. From the velocity identification results of measuring points 5, 7, 9, 11, and 13 in the figure, it can be seen that the results obtained by identifying the 6th unit damage using the velocity response of each measuring point individually can identify the damage even when the damage factor is large and the noise is small. However, units at both ends of the bridge, such as the 1st and 16th units, are easily misidentified even when there is no damage. Furthermore, because the bridge structure has a certain degree of symmetry to maintain stability, this symmetry is also reflected in the response to damage. Therefore, the 11th unit, which is symmetrical to the damaged unit about the midpoint of the bridge, also clearly has a misidentification problem. These units with large identified damage factors but no actual damage are called false identification units. The existence of false identification units leads to an underestimation of the damage factor of the identified actual damaged units of the bridge. As can be seen from the figure, the sum of the damage factors of the 6th and 11th units is approximately 10%. As can be seen from the average of the identification results of each measurement point, the identification results have basically reached the level of being able to identify damage. However, the two types of false identification units when using single measurement point identification still have an impact on the identification results. This impact will further increase with the increase of noise. The simple single measurement point velocity response identification and averaging still has limitations in the effect of unit damage identification.

[0106] For condition 1, under the same conditions, the identification results obtained using the bridge damage identification method based on vibration energy entropy provided in this application are as follows: Figure 9 As shown in the figure, the bridge damage identification method based on vibration energy entropy is significantly better than the results of identifying each measuring point using velocity response separately. In the figure, only the adjacent seventh unit has a very small value, except for the sixth damaged unit. The seventh unit with such a value will not affect the judgment of the actual damaged unit, so it cannot be called a false identification unit. This is enough to show the superiority of the bridge damage identification method based on vibration energy entropy provided in this application. By comparison, it can be seen that this method is much better than the identification result obtained by first using the velocity response of a single measuring point to identify the damage and then averaging the identification results of all measuring points.

[0107] After repeated modifications to the test conditions, the test data showed that the bridge damage identification method based on vibration energy entropy required an average of 8 iterations, indicating a fast convergence speed. In contrast, the program execution time of the method using single-point velocity response identification was generally more than five times longer than that of the method based on vibration energy entropy. This is attributed to the fact that the identification method provided in this application can fully utilize the information from each measuring point, and the feature quantities obtained after processing each set of data are relatively stable, which can significantly reduce the impact of local noise on the identification speed and effect. To verify this beneficial effect, in another test, with other conditions unchanged, the local noise was increased to 10% at the 13th measuring point. Damage identification was performed using both the single-point velocity response identification method and the bridge damage identification method based on vibration energy entropy. The identification results are as follows: Figure 10 and Figure 11 As shown, where Figure 10 The identification results of the 13th measurement point in the single-point velocity response identification method clearly show that the damage factor values ​​of the false identification units 10 and 11 are higher than the value of the actual damage unit 6. Figure 11 The identification results using the bridge damage identification method based on vibration energy entropy show that the identification effect is not significantly different from that before increasing local noise, which is sufficient to demonstrate the performance improvement of the identification method provided in this application in combating the influence of local noise.

[0108] In another test condition, condition 2, there are two damaged units, the 5th and 12th units, with damage factors of 12% and 8% respectively. The test noise is set to 3% for both units. Under these conditions, the identification results using the bridge damage identification method based on vibration energy entropy are as follows: Figure 12As shown in the figure, this method successfully identified two damaged units on the bridge. In other tests, it also showed good results in identifying multiple damaged units. Therefore, the identification method provided in this application is applicable to the detection of different numbers of damaged units and is not limited by the number of damaged units.

[0109] It should be noted that although the steps are described in a specific order above, this does not mean that they must be performed in that order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required functionality is achieved. The number of devices and processing scale described herein are for simplification of the invention; applications, modifications, and variations of this invention will be readily apparent to those skilled in the art.

[0110] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A bridge damage identification method based on vibration energy entropy, characterized in that, Includes the following steps: S1: Divide the bridge into multiple units of equal length along the bridge length, and set multiple measuring points evenly on the bridge body; S2: Sample and obtain the velocity response time history vector of each measuring point at several sampling time points within the velocity response time history; S3: Definition of vibration energy density and vibration energy entropy, and their calculation methods: H(t)=E(t)ln(E(t)) Where t is the sampling time corresponding to the current sampling time point, p is the number of measurement points, and u j The velocity response time history vector obtained for the j-th measuring point. Let E(t) be the energy density at the j-th measuring point, E(t) be the sum of vibration energy densities at all measuring points, and H(t) be the vibration energy entropy. The relative reduction in the bending stiffness of an element is defined as the element's damage factor, which is calculated as follows: Among them, (EI) i0 and (EI) i These are the bending stiffnesses of the i-th element before and after the damage, respectively. S4: Based on the structural characteristics of the bridge, establish a finite element model for calculating the theoretical vibration energy entropy of the bridge corresponding to different damage factors. Based on the sensitivity of the bridge vibration energy entropy to each element damage factor, construct a sensitivity matrix and set initial values ​​for each element damage factor. S5: Using the current value of the damage factor of each unit as input, calculate the theoretical vibration energy entropy of the bridge using the finite element model and update the value of the sensitivity matrix; S6: Select an unselected sampling time point according to the sampling time sequence, calculate the vibration energy density of each measuring point and the actual vibration energy entropy of the bridge based on the velocity response time history vector obtained from this sampling time point; establish a sensitivity equation based on the difference between the theoretical vibration energy entropy and the actual vibration energy entropy of the bridge and the sensitivity matrix, solve the sensitivity equation to obtain the correction value of the damage factor of each unit, and correct the current value of the damage factor of each unit accordingly. S7: Iterate through steps S5-S6 until the values ​​to be corrected for the damage factors of each unit meet the convergence condition. After the iteration is completed, output the values ​​of the damage factors of each unit as the identification results.

2. The bridge damage identification method based on vibration energy entropy according to claim 1, characterized in that, The sensitivity matrix is ​​constructed in step S4 using the following method: The vibration energy entropy of the bridge is relative to the damage factor α of the i-th unit. i The sensitivity is expressed in vector form: Where m is the number of sampling time points, The vibration energy entropy of the entire bridge relative to the damage factor α of the i-th unit. i Sensitivity, The vibration energy entropy at the k-th measuring point relative to the damage factor α of the i-th unit. i Sensitivity; The sensitivity of the bridge's vibration energy entropy to the damage factor of each element was calculated separately. Combined into matrix form: Where n is the number of units divided, and matrix S is the constructed sensitivity matrix; The sensitivity equation established in step S5 is based on the following sensitivity relationship: Where ΔH is the difference between the theoretical vibration energy entropy and the actual vibration energy entropy, and Δα is the value to be corrected for the damage factor.

3. The bridge damage identification method based on vibration energy entropy according to claim 1, characterized in that, The convergence condition mentioned in step S7 is: or |Δα k | <T2; Where, Δα k The value to be corrected for the element damage factor calculated in the k-th iteration is... T1 and T2 are the current values ​​of the unit damage factor before the k-th correction, and both are convergence threshold values.

4. The bridge damage identification method based on vibration energy entropy according to claim 1, characterized in that, It also includes the following steps: S8: Units with damage factor values ​​higher than the threshold are considered as damage units. A damage localization neural network model is established based on the mapping relationship between the velocity response time history vectors of the three measurement points before and after the damage unit and the specific damage location on the damage unit. The sampling data of previous damage identification and the accurate damage location data after verification and correction are packaged into a dataset and input into the damage localization neural network model for model training. S9: Input the velocity response time history vectors of the identified damaged unit and the three measurement points before and after it into the trained damage localization neural network model to obtain the accurate damage location on the identified damaged unit. After verifying the damage location in the field, use the correct data to correct the recognition accuracy of the damage localization neural network model again.

5. The bridge damage identification method based on vibration energy entropy according to claim 4, characterized in that, Step S8 further includes the following steps: For units where it is impossible to obtain three measurement points forward or backward, three measurement points are set separately on each unit. An auxiliary localization neural network model is established based on the mapping relationship between the velocity response time history vector of the three measurement points on the unit and the specific damage location on the unit. The sampling data and identification results of these units as damage units are selected and the damage localization neural network model is trained. Step S9 further includes the following steps: For the damaged unit identified this time, if it is impossible to obtain three measurement points forward or backward, the velocity response time history vectors of the three measurement points set separately on the damaged unit are input into the trained auxiliary localization neural network model to obtain the accurate damage location on the identified damaged unit. After verifying the damage location in the field, the recognition accuracy of the auxiliary localization neural network model is corrected again with the correct data.

6. The bridge damage identification method based on vibration energy entropy according to claim 1, characterized in that, Step S1 further includes: running a test train on the bridge, and setting up a measuring point at the front and rear of the train and between the two carriages; Step S2 further includes: sampling and obtaining the speed response time history vector of each measuring point on the train at several sampling time points within the speed response time history; Step S4 also includes: establishing a finite element model for calculating the theoretical vibration energy entropy of the train corresponding to different damage factors based on the structural characteristics of the bridge and the operating state characteristics of the train; constructing a train sensitivity matrix based on the sensitivity of the train vibration energy entropy to the damage factors of each unit; and creating a second damage factor for each unit and setting an initial value. Step S5 also includes: using the current value of the second damage factor of each unit as input, calculating the theoretical vibration energy entropy of the train using the finite element model and updating the value of the train sensitivity matrix; Step S6 further includes: selecting an unselected train response data sampling time point according to the sampling time sequence; calculating the vibration energy density of each measuring point on the train and the actual vibration energy entropy of the train based on the speed response time history vector obtained from this sampling time point; establishing a train sensitivity equation based on the difference between the theoretical vibration energy entropy and the actual vibration energy entropy of the train and the train sensitivity matrix; solving the train sensitivity equation to obtain the correction value of the second damage factor of each unit and correcting the current value of the second damage factor of each unit accordingly. Step S7 further includes: iterating through steps S5-S6 until the values ​​to be corrected for the second damage factor of each unit all meet the convergence condition. After the iteration is completed, the values ​​of the second damage factor of each unit are output, and the reliability of the damage factor of the potentially damaged units in the identification results is also output. The reliability of the damage factor is calculated as follows: Where, q x To determine the reliability of the damage factor for the xth potentially damaged unit, α x1 α is the damage factor of this element, calculated based on the energy entropy of bridge vibration. x2 This is the second damage factor of the unit, calculated based on the train vibration energy entropy.

Images