A method for setting risk threshold of beam bridge cluster structures based on multi-point strain detection dataset

Abstract

The present invention discloses a method for setting a risk threshold of a beam bridge cluster structure under a multi-measurement point strain detection data set. The method uses a K-medoids algorithm based on Gaussian similarity to divide the categories of individual bridges in the cluster, uses a five-point cubic smoothing analysis method to reconstruct the monitoring data of the individual bridge structure under the reference state in the cluster, constructs the initial value of the risk threshold of two measuring points, introduces the multi-measurement point strain detection data set of the bridge structure under the reference state, uses an entropy weight-Bayesian variable weight fusion algorithm to correct the initial value of the risk threshold of two measuring points, uses a leave-one-out cross-validation method to establish a bridge cluster structure risk threshold, divides the beam bridge cluster structure into multi-level warning levels, and realizes effective warning of bridge structure risks. The method is suitable for risk assessment and monitoring of bridge structures during the operation period. The present invention solves the problem of inaccurate structural risk warning caused by environmental factors and structural stiffness attenuation in the operation period.

Description

Technical Field

[0001] The present invention belongs to the field of bridge structure disease monitoring, and relates to a method for setting a risk threshold value of a beam bridge cluster structure, and specifically to a method for setting a risk threshold value of a beam bridge cluster structure under a multi-measurement point strain detection data set. Background Art

[0002] As an important part of transportation infrastructure, bridges have an increasingly greater impact on people. Daily travel is inseparable from the connection of bridges. At present, my country has more than 1 million highway bridges and railway bridges, making it the world's largest bridge country. Among them, more than 790,000 bridges are small and medium-span bridges. At the same time, most of these bridges are beam bridges with relatively simple structures. At the same time, with the continuous development of China's urbanization process, small and medium-span beam bridges are developing in the direction of clustering and similarity. Once the bridge structure is built and opened to traffic, it will inevitably be subjected to various loads, resulting in a decrease in the resistance of the bridge structure and damage to the structure, which may even endanger people's lives in serious cases.

[0003] Structural health monitoring (SHM), as the main means of monitoring changes in bridge structure status, is gradually becoming an important part of bridge structure operation. However, in the actual operation and use stage of the bridge structure health monitoring system, the collected data will inevitably be affected by various environmental factors, which will have a great impact on the results identified by the monitoring data. Therefore, effectively removing the impact of environmental factors in the monitoring data is an important part of building a bridge structure early warning system; at the same time, the existing bridge structure risk thresholds are mostly set directly by the bridge structure health monitoring data, and some indicators that can directly show the state of the bridge structure are not fully utilized, making it difficult to achieve effective early warning of bridge structure risks. From the perspective of structural damage detection, according to the state of the bridge structure, correcting the structural risk threshold through some technical means can undoubtedly achieve effective early warning of bridge structure risks. Summary of the invention

[0004] In view of the deficiencies in the risk warning of the existing bridge structures, the present invention provides a method for setting a risk threshold of a beam bridge cluster structure under a multi-measurement point strain detection data set to achieve effective early warning of bridge structure risks.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] A method for setting a risk threshold of a beam bridge cluster structure under a multi-measurement point strain detection data set comprises the following steps:

[0007] Step 1: Based on the structural monitoring data of the beam bridge cluster, the Gaussian similarity-based K-medoids algorithm is used to classify the individual bridges in the cluster. The specific steps are as follows:

[0008] Step 1: Based on the monitoring data of the beam bridge cluster structure, a monitoring data set of the natural frequency of the beam bridge cluster structure is constructed:

[0009]

[0010] In the formula, (·) T is the transpose of the matrix; m f is the total number of sampling points for monitoring data; k f is a sampling point of the monitoring data, k f ∈(1,2,…,m f ), For the kth f A sample set of natural frequency monitoring data;

[0011] Step 12: For the monitoring data set of the natural frequency of the beam bridge cluster structure constructed in step 11, select m cluster center points, calculate the Gaussian similarity between the remaining objects and the cluster center points, and realize the preliminary classification of the monitoring data set categories:

[0012]

[0013] In the formula, C is the set of cluster centers; m is the number of cluster centers; c m is the mth cluster center point; For the kth f The Gaussian similarity between the sample and the jth cluster center; σ is The variance of For the kth f The Euclidean distance between the sample and the jth cluster center;

[0014] Step 13: Calculate the Gaussian similarity between the remaining objects and the cluster center points calculated in steps 1 and 2, and calculate the similarity of all objects in the monitoring data set and class cluster C represents object c j The sum of absolute errors is:

[0015]

[0016] Step 14: According to the calculation results in step 13, randomly select sample data of non-cluster center points to replace the center points for iterative calculation. When the iteration meets the convergence condition, stop the iteration. The convergence condition is:

[0017]

[0018] Where U is the total difference value; s is the number of iterations; E s and E s-1are the sum of the absolute errors of the sth and s-1th iterations respectively; when the iterations converge, the clustering of the entire sample set is completed;

[0019] Step 2: For any single bridge structure in the different bridge categories that have been divided, use the five-point cubic smoothing analysis method to reconstruct the monitoring data of the single bridge structure under the reference state. The specific steps are as follows:

[0020] Step 21: Based on the beam bridge cluster structure monitoring data in step 1, construct a long-term strain monitoring data set of the same type of single bridges in the cluster:

[0021] Y=[ε 1 ,ε 2 ,…ε i ,…ε m′ ]

[0022] Where i is any measuring point of bridge structure strain, i∈(1,2,…,m′); m′ is the number of measuring points; ε i is the vector composed of long-term monitoring data of any strain measuring point i;

[0023] Step 22: For the long-term strain monitoring data set of the same type of single bridges in the cluster constructed in step 21, five-point cubic smoothing analysis is used to reduce the impact of noise and random vehicle loads on the strain data:

[0024]

[0025] In the formula, is the strain data value at the kth monitoring time point of any strain measuring point i after smoothing;

[0026] Step 23: Based on the processing results in step 22, the strain data set matrix of any single bridge in the cluster is obtained:

[0027]

[0028] In the formula, It is the vector composed of strain data of all measuring points at any k-th monitoring time point after the original data preprocessing;

[0029] Step 3: According to the monitoring data of the single bridge structure reconstructed in step 2, establish the early warning feature vector, use the maximum entropy method to fit the probability density function of the early warning feature vector, and construct the initial value of the risk threshold of the two measuring points of the single bridge structure. The specific steps are as follows:

[0030] Step 31: Based on the monitoring data of the single bridge structure under the reference state reconstructed in step 2, establish the early warning feature vector:

[0031] θ(i)=subpace(S0i ,S 01 )

[0032] Where θ(i) is the strain characteristic angle; S 0i and S 01 They are the monitoring data of the bridge beam structure at the 0i measuring point and the 01 measuring point under the reference state;

[0033] Step 32: For the warning feature vector established in step 31, use the maximum entropy method to fit the probability density function of the warning feature vector:

[0034]

[0035] Where f(θ(i)) is the maximum entropy probability density function of the warning feature vector θ(i); N is the center distance order of the selected warning feature vector θ(i); a z is the unknown coefficient of f(θ(i));

[0036] Step 33: According to the probability density function obtained in step 32, determine the initial value of the risk threshold of the single bridge structure:

[0037]

[0038] In the formula, F(·) is the cumulative probability distribution function; α is the confidence level; when the total number of warning characteristic values ​​below a certain probability density accounts for 1-α of the total number of confidence levels, the probability density at this time is set as the initial value of the risk threshold of the two measuring points 0i and 01

[0039] Step 4: Introduce the strain detection data set of multiple measuring points of the bridge structure under the reference state, use the entropy weight-Bayesian variable weight fusion algorithm to construct the overall degradation coefficient of the bridge structure, and calculate the effective stiffness of the bridge structure based on the design stiffness of the bridge structure. The specific steps are as follows:

[0040] Step 41: Calculate the weights of the components at key parts of the bridge based on the strain detection data set of multiple measuring points at key parts of the bridge structure:

[0041]

[0042] Where Y vh The result of standardization of the original data of strain detection at section v when the test section is h; X vh is the unprocessed original data; X v is the vector of all v sections; P vh is the proportion of the sample value of section v when the test section is h; M is the number of test sections; e v is the entropy value of the vth section; g v is the coefficient of variation of the vth section; wv is the weight of the key components of the bridge;

[0043] Step 42: Based on the weights of the key components of the bridge calculated in step 41, a prior distribution considering the bearing capacity detection and assessment information is obtained using the Bayesian estimation method:

[0044]

[0045] In the formula, ψ B is the overall degradation coefficient; ξ B,v is the degradation coefficient of section v; w M is the contribution weight of the components in the key parts of the bridge; is the prior distribution function; f v (ξ B,v |ψ B ) is the maximum likelihood function; M is the number of detection sections; Q is a regularization constant;

[0046] Step 43: Based on the prior distribution obtained in step 42, under the square loss function, the overall degradation coefficient ψ is obtained B Bayesian estimation of for:

[0047]

[0048] Step 44: Based on the overall degradation coefficient ψ obtained in step 43 B Bayesian estimation of Calculate the effective stiffness of the bridge structure:

[0049]

[0050] In the formula, B′ is the effective stiffness of the bridge structure; B is the stiffness of the bridge structure in a healthy state;

[0051] Step 5: Based on the effective stiffness of the bridge structure calculated in step 4, establish a dynamic evolution finite element model of the bridge structure, calculate the risk threshold correction coefficient, and build a two-point risk threshold based on the initial value of the risk threshold of the two-point single bridge structure established in step 3. The specific steps are as follows:

[0052] Step 51: According to the effective stiffness of the bridge structure calculated in step 4, establish the dynamic evolution finite element model of the bridge structure and calculate the risk threshold correction coefficient:

[0053]

[0054] In the formula, ξ fis the threshold correction coefficient of different types of structural responses after stiffness reduction, f represents various types of result responses such as acceleration, displacement and strain; ε′ is the various types of structural responses under the design load after stiffness correction; ε is the various types of structural responses under the design load under normal circumstances;

[0055] Step 52: Based on the initial value of the risk threshold of the two measuring points of the single bridge structure established in step 3 and the risk threshold correction coefficient calculated in step 51, construct the risk threshold of the two measuring points:

[0056]

[0057] Where P l o is the risk threshold of two measuring points: o and l;

[0058] Step 6: Based on the risk threshold of the two measuring points established in step 5, the leave-one-out cross-validation method is used to set the risk threshold of the beam bridge cluster structure and divide the beam bridge cluster structure into multiple levels of warning levels. The specific steps are as follows:

[0059] Step 61: Based on the strain data set matrix of any single bridge in the cluster reconstructed in step 2 and the two-point risk threshold construction method in step 5, establish the decision warning index of the l1th measuring point with the oth measuring point as the reference in the state to be diagnosed:

[0060]

[0061] In the formula, is the risk threshold of two measuring points of the l1th measuring point with the oth measuring point as the reference;

[0062] Step 62: Based on the decision-making early warning indicators constructed in step 61, the leave-one-out cross-validation method is adopted to select any measuring point o among the strain measuring points as the reference measuring point for cross-validation, and establish a set of decision-making early warning indicators for other measuring points of the same type of bridges in the cluster:

[0063]

[0064] Where η is the set of decision warning indicators for the same type of bridge structures in the cluster under the decision-making state; η o is a set of decision warning indicators in the decision-making state with the oth sensor data as reference; The data of the oth sensor in the decision-making state is used as the reference for the decision warning indicator of the l1th sensor; n1 is the number of sensors of the same type on the same type of bridges in the cluster;

[0065] Step 63: Based on the decision-making warning indicator set established in step 62, establish the risk threshold of the beam bridge cluster structure:

[0066]

[0067] In the formula, is the structural risk threshold of the beam bridge cluster, which is the median of the 95% confidence probability of the bridge structure decision warning indicator set in the decision-making state; is the decision-making and early warning indicator set of the damage diagnosis factors of all measuring points, It is the decision warning indicator of the l1th sensor data in the decision-to-be state; is the early warning index factor of the l1th sensor data; if the warning index of the l1th sensor data satisfies Right now If the monitoring data of the sensor position exceeds the bridge structure threshold, a structural warning is issued; otherwise, the structural state at the measuring point is considered normal;

[0068] Step 64: Based on the risk threshold of the beam bridge cluster structure established in step 63, establish the warning level standard for the cross-validation results of the bridge structures within the cluster:

[0069]

[0070] In the formula, w f is the proportion of the number of overrun measurement points of the same type to all measurement points in the beam bridge cluster; num(overrun) is the number of overrun measurement points of the same type in the cluster obtained by cross-validation; num(total) is the total number of measurement points of the same type in the cluster; according to w f The calculation results are used to divide the warning levels of the same type of bridge structures in the cluster.

[0071] Compared with the prior art, the present invention has the following advantages:

[0072] 1. The present invention utilizes the K-medoids algorithm based on Gaussian similarity to classify the individual bridges in the cluster, adopts the five-point cubic smoothing analysis method to reconstruct the monitoring data of the individual bridge structures in the cluster under the reference state, constructs the initial value of the risk threshold of two measuring points, introduces the strain detection data set of multiple measuring points of the bridge structure under the reference state, adopts the entropy weight-Bayesian variable weight fusion algorithm, corrects the initial value of the risk threshold of two measuring points, adopts the leave-one-out cross-validation method, establishes the risk threshold of the beam bridge cluster structure, divides the beam bridge cluster structure into multi-level warning levels, realizes the effective warning of the bridge structure risk, and is suitable for the risk assessment and monitoring of the bridge structure during the operation period.

[0073] 2. The present invention extracts the structural damage information contained in the historical inspection data of the bridge structure, and corrects the initial value of the risk threshold of the two measuring points established according to the bridge structure monitoring data, so as to achieve the purpose of effective early warning of bridge structure risks, thereby providing effective guarantee for the safe operation and maintenance of the bridge structure during the operation period, and solving the problem of inaccurate structural risk warning caused by environmental factors and structural stiffness attenuation of the bridge structure during the operation period. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A flowchart of the method for constructing early warning limits of beam bridge cluster structure monitoring data based on multi-measurement point strain detection data sets;

[0075] Figure 2 Figure 2 is the structural diagram of the selected beam bridge cluster.

[0076] Figure 3 This is the measurement point location one of the single bridge.

[0077] Figure 4 This is the second measuring point location of the single bridge.

[0078] Figure 5 Schematic diagram of historical strain monitoring data of a single bridge structure.

[0079] Figure 6 This is a schematic diagram of the probability density fitting results of the warning feature vector of a single bridge.

[0080] Figure 7 Schematic diagram of the historical multi-point strain detection data set of a single bridge structure.

[0081] Figure 8 Schematic diagram of concrete damage at measuring point No. 14.

[0082] Fig. 9 It is a finite element model of a single bridge structure.

[0083] Fig.10 This is the bridge structure damage warning result. DETAILED DESCRIPTION

[0084] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0085] The present invention provides a method for setting a risk threshold of a beam bridge cluster structure under a multi-measurement point strain detection data set. The method extracts the structural state information contained in the multi-measurement point strain detection data set of the bridge structure, corrects the initial value of the two-measurement point risk threshold constructed based on the bridge structure monitoring data, establishes a risk threshold of the beam bridge cluster structure, and divides the beam bridge cluster structure into multiple levels of early warning levels, thereby achieving effective early warning of bridge structure risks. The existing setting of risk thresholds for bridge structures often follows the following idea: statistical analysis is performed using bridge structure health monitoring data, and values ​​that meet a certain confidence level are selected as risk thresholds, thereby providing early warnings for bridge structure risks. However, bridge structure health monitoring data are affected by various environmental factors, and directly setting risk thresholds often makes it difficult to fully utilize indicators that can directly reflect the state of the bridge structure. Therefore, the present invention adopts a five-point cubic smoothing analysis method and a maximum entropy method to construct an initial value of the bridge structure risk threshold, and introduces a multi-measurement point strain detection data set of the bridge structure under a reference state, corrects the initial value of the risk threshold, establishes a bridge structure risk threshold, achieves effective early warning of bridge structure risks, and ensures the safe operation of the bridge structure. Taking the historical monitoring and detection data of the beam bridge cluster as an example, as Figure 1 As shown, the specific steps include:

[0086] Step 1: Based on the structural monitoring data of the beam bridge cluster, the Gaussian similarity-based K-medoids algorithm is used to classify the individual bridges in the cluster. The specific steps are as follows:

[0087] Step 1: Based on the monitoring data of the beam bridge cluster structure, a monitoring data set of the natural frequency of the beam bridge cluster structure is constructed:

[0088]

[0089] In the formula, (·) T is the transpose of the matrix; m f is the total number of sampling points for monitoring data; k f is a sampling point of the monitoring data, k f ∈(1,2,…,m f ), For the kth f A sample set of natural frequency monitoring data;

[0090] Step 12: For the monitoring data set of the natural frequency of the beam bridge cluster structure constructed in step 11, select m cluster center points, calculate the Gaussian similarity between the remaining objects and the cluster center points, and realize the preliminary classification of the monitoring data set categories:

[0091]

[0092] In the formula, C is the cluster center set; c m is the mth cluster center point; For the kth f The Gaussian similarity between the sample and the jth cluster center; σ is The variance of For the kth f The Euclidean distance between the sample and the jth cluster center;

[0093] Step 13: Calculate the Gaussian similarity between the remaining objects and the cluster center points calculated in steps 1 and 2, and calculate the similarity of all objects in the monitoring data set and class cluster C represents object c j The sum of absolute errors is:

[0094]

[0095] Step 14: According to the calculation results in step 13, randomly select sample data of non-cluster center points to replace the center points for iterative calculation. When the iteration meets the convergence condition, stop the iteration. The convergence condition is:

[0096]

[0097] Where U is the total difference value; s is the number of iterations; E s and E s-1 are the sum of the absolute errors of the sth and s-1th iterations respectively; when the iterations converge, the clustering of the entire sample set is completed;

[0098] Step 2: For any single bridge structure in the different bridge categories that have been divided, use the five-point cubic smoothing analysis method to reconstruct the monitoring data of the single bridge structure under the reference state. The specific steps are as follows:

[0099] Step 21: Based on the beam bridge cluster structure monitoring data in step 1, construct a long-term strain monitoring data set of the same type of single bridges in the cluster:

[0100] Y=[ε 1 ,ε 2 ,…ε i ,…ε m′ ]

[0101] Where i is any measuring point of bridge structure strain, i∈(1,2,…,m′); m′ is the number of measuring points; ε i is the vector composed of long-term monitoring data of any strain measuring point i;

[0102] Step 22: For the long-term strain monitoring data set of the same type of single bridges in the cluster constructed in step 21, five-point cubic smoothing analysis is used to reduce the impact of noise and random vehicle loads on the strain data:

[0103]

[0104] In the formula, is the strain data value at the kth monitoring time point of any strain measuring point i after smoothing;

[0105] Step 23: Based on the processing results in step 22, the strain data set matrix of any single bridge in the cluster is obtained:

[0106]

[0107] In the formula, It is the vector composed of strain data of all measuring points at any k-th monitoring time point after the original data preprocessing;

[0108] Step 3: According to the monitoring data of the single bridge structure reconstructed in step 2, establish the early warning feature vector, use the maximum entropy method to fit the probability density function of the early warning feature vector, and construct the initial value of the risk threshold of the two measuring points of the single bridge structure. The specific steps are as follows:

[0109] Step 31: Based on the monitoring data of the single bridge structure under the reference state reconstructed in step 2, establish the early warning feature vector:

[0110] θ(i)=subpace(S 0i ,S 01 )

[0111] Where θ(i) is the strain characteristic angle; S 0i and S 01 They are the monitoring data of the bridge beam structure at the 0i measuring point and the 01 measuring point under the reference state;

[0112] Step 32: For the warning feature vector established in step 31, use the maximum entropy method to fit the probability density function of the warning feature vector:

[0113]

[0114] Where f(θ(i)) is the maximum entropy probability density function of the warning feature vector θ(i); N is the center distance order of the selected warning feature vector θ(i); a z is the unknown coefficient of f(θ(i));

[0115] Step 33: According to the probability density function obtained in step 32, determine the initial value of the risk threshold of the single bridge structure:

[0116]

[0117] In the formula, F(·) is the cumulative probability distribution function; α is the confidence level; when the total number of warning characteristic values ​​below a certain probability density accounts for 1-α of the total number of confidence levels, the probability density at this time is set as the initial value of the risk threshold of the two measuring points 0i and 01

[0118] Step 4: Introduce the strain detection data set of multiple measuring points of the bridge structure under the reference state, use the entropy weight-Bayesian variable weight fusion algorithm to construct the overall degradation coefficient of the bridge structure, and calculate the effective stiffness of the bridge structure based on the design stiffness of the bridge structure. The specific steps are as follows:

[0119] Step 41: Calculate the weights of the components at key parts of the bridge based on the strain detection data set of multiple measuring points at key parts of the bridge structure:

[0120]

[0121] Where Y vh The result of standardization of the original data of strain detection at section v when the test section is h; X vh is the unprocessed original data; X v is the vector of all v sections; P vh is the proportion of the sample value of section v when the test section is h; M is the number of test sections; e v is the entropy value of the vth section; g v is the coefficient of variation of the vth section; w v is the weight of the key components of the bridge;

[0122] Step 42: Based on the weights of the key components of the bridge calculated in step 41, a prior distribution considering the bearing capacity detection and assessment information is obtained using the Bayesian estimation method:

[0123]

[0124] In the formula, ψ B is the overall degradation coefficient; ξ B,v is the degradation coefficient of section v; w M is the contribution weight of the components in the key parts of the bridge; is the prior distribution function; f v (ξ B,v |ψ B ) is the maximum likelihood function; M is the number of detection sections; Q is a regularization constant;

[0125] Step 43: Based on the prior distribution obtained in step 42, under the square loss function, the overall degradation coefficient ψ is obtained B Bayesian estimation of for:

[0126]

[0127] Step 44: Based on the overall degradation coefficient ψ obtained in step 43 B Bayesian estimation of Calculate the effective stiffness of the bridge structure:

[0128]

[0129] In the formula, B′ is the effective stiffness of the bridge structure; B is the stiffness of the bridge structure in a healthy state;

[0130] Step 5: Based on the effective stiffness of the bridge structure calculated in step 4, establish a dynamic evolution finite element model of the bridge structure, calculate the risk threshold correction coefficient, and build a two-point risk threshold based on the initial value of the risk threshold of the two-point single bridge structure established in step 3. The specific steps are as follows:

[0131] Step 51: According to the effective stiffness of the bridge structure calculated in step 4, establish the dynamic evolution finite element model of the bridge structure and calculate the risk threshold correction coefficient:

[0132]

[0133] In the formula, ξ f is the threshold correction coefficient of different types of structural responses after stiffness reduction, f represents various types of result responses such as acceleration, displacement and strain; ε′ is the various types of structural responses under the design load after stiffness correction; ε is the various types of structural responses under the design load under normal circumstances;

[0134] Step 52: Based on the initial value of the risk threshold of the two measuring points of the single bridge structure established in step 3 and the risk threshold correction coefficient calculated in step 51, construct the risk threshold of the two measuring points:

[0135]

[0136] Where P l o is the risk threshold of two measuring points: o and l;

[0137] Step 6: Based on the risk threshold of the two measuring points established in step 5, the leave-one-out cross-validation method is used to set the risk threshold of the beam bridge cluster structure and divide the beam bridge cluster structure into multiple levels of warning levels. The specific steps are as follows:

[0138] Step 61: Based on the strain data set matrix of any single bridge in the cluster reconstructed in step 2 and the two-point risk threshold construction method in step 5, establish the decision warning index of the l1th measuring point with the oth measuring point as the reference in the state to be diagnosed:

[0139]

[0140] In the formula, is the risk threshold of two measuring points of the l1th measuring point with the oth measuring point as the reference;

[0141] Step 62: Based on the decision-making early warning indicators constructed in step 61, the leave-one-out cross-validation method is adopted to select any measuring point o among the strain measuring points as the reference measuring point for cross-validation, and establish a set of decision-making early warning indicators for other measuring points of the same type of bridges in the cluster:

[0142]

[0143] Where η is the set of decision warning indicators for the same type of bridge structures in the cluster under the decision-making state; η o is a set of decision warning indicators in the decision-making state with the oth sensor data as reference; The data of the oth sensor in the decision-making state is used as the reference for the decision warning indicator of the l1th sensor; n1 is the number of sensors of the same type on the same type of bridges in the cluster;

[0144] Step 63: Based on the decision-making warning indicator set established in step 62, establish the risk threshold of the beam bridge cluster structure:

[0145]

[0146] In the formula, is the structural risk threshold of the beam bridge cluster, which is the median of the 95% confidence probability of the bridge structure decision warning indicator set in the decision-making state; is the decision-making and early warning indicator set of the damage diagnosis factors of all measuring points, It is the decision warning indicator of the l1th sensor data in the decision-to-be state; is the early warning index factor of the l1th sensor data; if the warning index of the l1th sensor data satisfies Right now If the monitoring data of the sensor position exceeds the bridge structure threshold, a structural warning is issued; otherwise, the structural state at the measuring point is considered normal;

[0147] Step 64: Based on the risk threshold of the beam bridge cluster structure established in step 63, establish the warning level standard for the cross-validation results of the bridge structures within the cluster:

[0148]

[0149] In the formula, w fis the proportion of the number of overrun measurement points of the same type to all measurement points in the beam bridge cluster; num(overrun) is the number of overrun measurement points of the same type in the cluster obtained by cross-validation; num(total) is the total number of measurement points of the same type in the cluster; according to w f The calculation results are used to divide the warning levels of the same type of bridge structures in the cluster.

[0150] Girder bridge cluster structure Figure 2 As shown in Figure 2, the measurement points of the single bridge are located as follows: Figure 3 and Figure 4 As shown in Figure 2, the historical strain monitoring data of the single bridge structure is as follows: Figure 5 As shown in the figure, the probability density fitting results of the single bridge warning feature vector are as follows: Figure 6 As shown in the figure, the historical multi-point strain detection dataset of a single bridge structure is as follows: Figure 7 As shown in the figure, the concrete damage at measuring point 14 is as follows Figure 8 As shown in Figure 2, the finite element model of the single bridge structure is as follows: Fig. 9 As shown in Figure 2, the bridge structure damage warning results are as follows: Fig.10 When measuring points 10, 11, and 35 are used as reference points, it is identified that the early warning index factor of measuring point 14 exceeds the limit value. Measuring point 14 is set as the reference point, and it is found that the early warning index factors of the remaining measuring points all exceed the limit value, that is, the structural state at measuring point 14 is significantly different from that at other measuring points, and the structural state at measuring point 14 can be effectively identified.

Claims

1. A method for setting risk threshold of beam bridge cluster structure under multi-measurement point strain detection data set, characterized in that The method comprises the following steps: Step 1: Based on the structural monitoring data of the beam bridge cluster, the Gaussian similarity-based K-medoids algorithm is used to classify the individual bridges in the cluster; Step 2: For any single bridge structure in the different bridge categories that have been divided, use the five-point cubic smoothing analysis method to reconstruct the monitoring data of the single bridge structure under the reference state; Step 3: Based on the monitoring data of the single bridge structure reconstructed in step 2, establish the early warning feature vector, use the maximum entropy method to fit the probability density function of the early warning feature vector, and construct the initial value of the risk threshold of the two measuring points of the single bridge structure; Step 4: Introduce the strain detection data set of multiple measuring points of the bridge structure under the reference state, use the entropy weight-Bayesian variable weight fusion algorithm to construct the overall degradation coefficient of the bridge structure, and calculate the effective stiffness of the bridge structure based on the design stiffness of the bridge structure; Step 5: Based on the effective stiffness of the bridge structure calculated in step 4, a dynamic evolution finite element model of the bridge structure is established, and the risk threshold correction coefficient is calculated. Based on the initial value of the risk threshold of the two measuring points of the single bridge structure established in step 3, a two-measuring-point risk threshold is constructed; Step 6: Based on the risk threshold of the two measuring points established in step 5, the leave-one-out cross-validation method is used to set the risk threshold of the beam bridge cluster structure and divide the beam bridge cluster structure into multiple levels of warning levels.

2. The method for setting the risk threshold of beam bridge cluster structure under the multi-measurement point strain detection data set according to claim 1 is characterized in that The specific steps of step one are as follows: Step 1: Based on the monitoring data of the beam bridge cluster structure, a monitoring data set of the natural frequency of the beam bridge cluster structure is constructed: In the formula, (·) T is the transpose of the matrix; m f is the total number of sampling points for monitoring data; k f is a sampling point of the monitoring data, k f ∈(1,2,…,m f ), For the kth f A sample set of natural frequency monitoring data; Step 12: For the monitoring data set of the natural frequency of the beam bridge cluster structure constructed in step 11, select m cluster center points, calculate the Gaussian similarity between the remaining objects and the cluster center points, and realize the preliminary classification of the monitoring data set categories: In the formula, C is the cluster center set; c m is the mth cluster center point; For the kth f The Gaussian similarity between the sample and the jth cluster center; σ is The variance of For the kth f The Euclidean distance between the sample and the jth cluster center; Step 13: Calculate the Gaussian similarity between the remaining objects and the cluster center points calculated in steps 1 and 2, and calculate the similarity of all objects in the monitoring data set and class cluster C represents object c j The sum of absolute errors is: Step 14: According to the calculation results in step 13, randomly select sample data of non-cluster center points to replace the center points for iterative calculation. When the iteration meets the convergence condition, stop the iteration. The convergence condition is: Where U is the total difference value; s is the number of iterations; E s and E s-1 are the sum of the absolute errors of the sth and s-1th iterations respectively; when the iterations converge, the clustering division of the entire sample set is completed.

3. The method for setting the risk threshold of beam bridge cluster structure under the multi-measurement point strain detection data set according to claim 1 is characterized in that The specific steps of step 2 are as follows: Step 21: Based on the beam bridge cluster structure monitoring data in step 1, construct a long-term strain monitoring data set of the same type of single bridges in the cluster: Y=[e 1 ,he 2 ,…eh i ,…eh m′ ] Where i is any measuring point of bridge structure strain, i∈(1,2,…,m′); m′ is the number of measuring points; ε i is the vector composed of long-term monitoring data of any strain measuring point i; Step 22: For the long-term strain monitoring data set of the same type of single bridges in the cluster constructed in step 21, five-point cubic smoothing analysis is used to reduce the impact of noise and random vehicle loads on the strain data: In the formula, is the strain data value at the kth monitoring time point of any strain measuring point i after smoothing; Step 23: Based on the processing results in step 22, the strain data set matrix of any single bridge in the cluster is obtained: In the formula, It is a vector composed of strain data of all measuring points at any kth monitoring time point after the original data is preprocessed.

4. The method for setting the risk threshold of beam bridge cluster structure under the multi-measurement point strain detection data set according to claim 1 is characterized in that The specific steps of step three are as follows: Step 31: Based on the monitoring data of the single bridge structure under the reference state reconstructed in step 2, establish the early warning feature vector: θ(i)=subpace(S 0i ,S 01 ) Where, θ(i) is the strain characteristic angle; S 0i and S 01 They are the monitoring data of the bridge beam structure at the 0i measuring point and the 01 measuring point under the reference state; Step 32: For the warning feature vector established in step 31, use the maximum entropy method to fit the probability density function of the warning feature vector: Where f(θ(i)) is the maximum entropy probability density function of the warning feature vector θ(i); N is the center distance order of the selected warning feature vector θ(i); a z is the unknown coefficient of f(θ(i)); Step 33: According to the probability density function obtained in step 32, determine the initial value of the risk threshold of the single bridge structure: In the formula, F(·) is the cumulative probability distribution function; α is the confidence level; when the total number of warning characteristic values ​​below a certain probability density accounts for 1-α of the total number of confidence levels, the probability density at this time is set as the initial value of the risk threshold of the two measuring points 0i and 01 5. The method for setting the risk threshold of beam bridge cluster structure under the multi-measurement point strain detection data set according to claim 1 is characterized in that The specific steps of step 4 are as follows: Step 41: Calculate the weights of the components at the key parts of the bridge based on the strain detection data set of multiple measuring points at the key parts of the bridge structure: Where Y vh The result of standardization of the original data of strain detection at section v when the test section is h; X vh is the unprocessed original data; X v is the vector of all v sections; P vh is the proportion of the sample value of section v when the test section is h; M is the number of test sections; e v is the entropy value of the vth section; g v is the coefficient of difference of the vth section; w v is the weight of the key components of the bridge; Step 42: Based on the weights of the key components of the bridge calculated in step 41, a prior distribution considering the bearing capacity detection and assessment information is obtained using the Bayesian estimation method: In the formula, ψ B is the overall degradation coefficient; ξ B,v is the degradation coefficient of section v; w M is the contribution weight of the components in the key parts of the bridge; is the prior distribution function; f v (ξ B,v |ψ B ) is the maximum likelihood function; M is the number of detection sections; Q is a regularization constant; Step 43: Based on the prior distribution obtained in step 42, under the square loss function, the overall degradation coefficient ψ is obtained B Bayesian estimation of for: Step 44: Based on the overall degradation coefficient ψ obtained in step 43 B Bayesian estimation of Calculate the effective stiffness of the bridge structure: Where B′ is the effective stiffness of the bridge structure; B is the stiffness of the bridge structure in a healthy state.

6. The method for setting the risk threshold of beam bridge cluster structure under the multi-measurement point strain detection data set according to claim 1 is characterized in that The specific steps of step five are as follows: Step 51: According to the effective stiffness of the bridge structure calculated in step 4, establish the dynamic evolution finite element model of the bridge structure and calculate the risk threshold correction coefficient: In the formula, ξ f is the threshold correction coefficient of different types of structural responses after stiffness reduction, and f represents various types of result responses of acceleration, displacement and strain; ε′ is the various structural responses under the design load after stiffness correction; ε is the various structural responses under design loads in normal conditions; Step 52: Based on the initial value of the risk threshold of the two measuring points of the single bridge structure established in step 3 and the risk threshold correction coefficient calculated in step 51, construct the risk threshold of the two measuring points: Where P l o is the risk threshold of the two measuring points o and l, is the initial value of the risk threshold of the two measuring points o and l.

7. The method for setting the risk threshold of beam bridge cluster structure under the multi-measurement point strain detection data set according to claim 3 is characterized in that The specific steps of step six are as follows: Step 61: Based on the strain data set matrix of any single bridge in the cluster reconstructed in step 2 and the risk threshold of the two measuring points in step 5, establish the decision warning index of the l1th measuring point with the oth measuring point as the reference in the state to be diagnosed: In the formula, is the risk threshold of two measuring points of the l1th measuring point with the oth measuring point as the reference; Step 62: Based on the decision-making early warning indicators constructed in step 61, the leave-one-out cross-validation method is adopted to select any measuring point o among the strain measuring points as the reference measuring point for cross-validation, and establish a set of decision-making early warning indicators for other measuring points of the same type of bridges in the cluster: Where η is the set of decision warning indicators for the same type of bridge structures in the cluster under the decision-making state; η o is a set of decision warning indicators in the decision-making state with the oth sensor data as reference; is the decision warning indicator of the l1th sensor with the data of the oth sensor as reference in the decision-making state; n1 is the number of sensors of the same type on the same type of bridges in the cluster; Step 63: Based on the decision-making warning indicator set established in step 62, establish the risk threshold of the beam bridge cluster structure: In the formula, is the structural risk threshold of the beam bridge cluster, which is the median of the 95% confidence probability of the bridge structure decision warning indicator set in the decision-making state; is the decision-making and early warning indicator set of the damage diagnosis factors of all measuring points, It is the decision warning indicator of the l1th sensor data in the decision-to-be state; is the early warning index factor of the l1th sensor data; if the decision early warning index of the l1th sensor data satisfies Right now If the monitoring data of the sensor position exceeds the bridge structure threshold, a structural warning is issued; otherwise, the structural state at the measuring point is considered normal; Step 64: Based on the risk threshold of the beam bridge cluster structure established in step 63, establish the warning level standard for the cross-validation results of the bridge structures within the cluster: In the formula, w f The ratio of the number of over-limit measuring points of the same type to all measuring points in the beam bridge cluster; num(overrun) is the number of overrun points of the same type in the cluster obtained by cross-validation; num(total) is the total number of points of the same type in the cluster; according to w f The calculation results are used to divide the warning levels of the same type of bridge structures in the cluster.

Images