Bridge damage real-time identification method considering temperature nonlinear influence

Abstract

The invention discloses a bridge damage real-time identification method considering temperature nonlinear influence, and belongs to the technical field of bridge damage identification, and the method comprises the steps: firstly, carrying out the preprocessing of an abnormal value of bridge vibration monitoring data, and identifying a main beam frequency under the action of environment temperature; secondly, non-linearly extracting a main beam frequency slow feature vector based on index slow feature analysis; then, adaptively searching a first neighbor sample of each frequency slow feature vector and a neighborhood set thereof, estimating a local mean value and a standard deviation of the neighborhood set, and carrying out local standardization on the slow feature vectors; and finally, constructing a robust mahalanobis distance damage index of minimum covariance estimation, carrying out statistical analysis on probability distribution of the damage index, setting a real-time dynamic alarm threshold, and judging whether the bridge is damaged or not according to a threshold judgment criterion. The method is simple in principle and operation process, and the accuracy, robustness and timeliness of a bridge damage identification result can be improved.

Description

Technical Field

[0001] The invention belongs to the technical field of bridge damage identification, and in particular relates to a real-time bridge damage identification method taking into account the nonlinear influence of temperature. Background Art

[0002] Bridges are the key link of transportation infrastructure. Ensuring the service performance of in-service bridges is an important guarantee for the long life of transportation infrastructure. In vibration-based bridge damage identification, the modal frequency that characterizes the structural dynamic characteristics is the most commonly used damage-sensitive feature. However, the temperature of the bridge under long-term service is the most important environmental factor causing frequency changes, and this change may mask the frequency changes caused by structural damage. In order to effectively evaluate the service status of the bridge, real-time and refined elimination of the temperature interference effect in the frequency is a hot issue in the field of bridge damage identification technology.

[0003] At present, the existing bridge damage identification methods mainly include explicit methods based on temperature and frequency correlation models and implicit methods based on multi-order frequency correlation modeling. The explicit method has clear physical meaning, but is limited by the completeness of monitoring data. The implicit method takes temperature as a potential influencing variable of frequency and directly performs correlation analysis on frequency to determine the frequency change caused by damage. Since it does not rely on temperature measurement data, the implicit method has become a research hotspot. However, in actual bridge damage identification, the non-uniform temperature distribution of the main beam and noise interference often lead to a large difference in the fluctuation changes of each order of frequency, making the frequency data statistically nonlinearly correlated and non-Gaussian, thereby reducing the accuracy and real-time performance of the damage identification results of the traditional method. Therefore, it is of great engineering significance to study the real-time identification method of bridge damage considering the nonlinear influence of temperature. Summary of the invention

[0004] The purpose of the present invention is to provide a real-time bridge damage identification method taking into account the nonlinear influence of temperature. The method principle and operation process are simple. When environmental factors are unknown, the nonlinear influence of temperature on the main beam frequency can be considered and eliminated, thereby improving the accuracy, robustness and timeliness of bridge damage identification results.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A real-time bridge damage identification method considering the nonlinear effect of temperature includes the following steps: Step 1: Preprocess the abnormal values ​​of bridge vibration monitoring data and identify the main beam frequency under the effect of ambient temperature; Step 2: Extract the main beam frequency slow feature vector based on the nonlinear exponential slow feature analysis; Step 3: Adaptively search the first nearest neighbor sample and its neighborhood set of each frequency slow feature vector, estimate the local mean and standard deviation of the neighborhood set, and perform local standardization on the slow feature vector; Step 4: Construct a robust Mahalanobis distance damage index with minimum covariance estimation, statistically analyze the probability distribution of the damage index and set a real-time dynamic alarm threshold, and determine whether the bridge is damaged according to the threshold judgment criterion.

[0006] Furthermore, step 1 specifically includes the following process: Step 1.1: Collect the vibration acceleration response data recorded by the bridge health monitoring system, screen for abnormal data types such as jump points, drift, periodic components and missing data, and perform error control preprocessing of the data by median filtering, low-pass filtering, differential filtering and trimming and filling; Step 1.2: Divide the acceleration response data pre-processed in step 1.1 into multiple parts in hours, identify the operational modal parameters of each part of the data, and obtain the main beam frequency data under the effect of ambient temperature. ,in, and Respectively represent the number of frequency samples and the number of variables, and divide the preprocessed frequency data into training set, validation set and test set; Furthermore, step 2 specifically includes the following process: The exponential slow feature analysis is used to transform the original frequency data preprocessed in step 1 into a new space through exponential mapping nonlinearly. , then, the slow frequency features are extracted by solving the following optimization problem: In the formula, represents the first-order covariance matrix and , represents the covariance matrix and , for About Time The first derivative of for The average value along the time axis, is the coefficient matrix, is the number of slow features; The Lagrange multiplier method is used to solve the above optimization problem, and the solution formula is: ,in, is the eigenvalue corresponding to the slow feature; then, the slow eigenvector of the main beam frequency is calculated based on the solved coefficient matrix ; Furthermore, step 3 specifically includes the following process: Step 3.1: Use the similarity measurement method to find the similarity in step 2 in the training data set. Slow feature vector of moment frequency The nearest neighbor sample , and then calculate the nearest neighbor sample With Neighbors Euclidean distance , get the nearest neighbor sample The neighborhood sample subset composed of the first k nearest neighbor samples of ,in, ; Step 3.2: Estimate the k-neighborhood set The local neighborhood mean and local standard deviation of Slow feature vector of moment frequency After standardization, the calculation formula is as follows: In the formula, and are the slow frequency eigenvectors The nearest neighbor vector The mean and standard deviation of the neighborhood set of The slow frequency feature vector is normalized for the local neighborhood; Step 3.3: Loop through steps 3.1 and 3.2 to traverse the frequency slow feature vectors at all moments to obtain the local neighborhood normalized values.

[0007] Furthermore, step 4 specifically includes the following process: Step 4.1: For each frequency slow feature vector in step 3 Perform local neighborhood normalization and construct a robust Mahalanobis distance damage index based on minimum covariance estimation ,in, and Slow frequency features The mean vector and robust covariance matrix of ; Step 4.2: Use the kernel density estimation method to statistically analyze the probability distribution of each damage indicator in step 4.1, and then set the damage threshold; Step 4.3: Calculate the frequency slow feature and its nearest neighbor vector of the online test data set, search for the neighborhood sample subset of the nearest neighbor vector in the training data set, and calculate its local mean and standard deviation. Construct the damage index based on the standardized result, and use the threshold discrimination criterion to determine in real time whether the bridge is damaged.

[0008] Beneficial effects of the present invention: 1. The present invention utilizes the exponential slow feature analysis to nonlinearly extract the slow frequency feature of the main beam under the effect of ambient temperature, which greatly reduces the influence of measurement noise, data multicollinearity, etc. on the practicality of monitoring data.

[0009] 2. The present invention utilizes the nearest neighbor and its neighborhood set of the slow frequency feature of the main beam to complete the local normalization of the data, avoiding the phenomenon of misselection of local sample subsets when directly selecting the first k nearest neighbor samples, and effectively reducing the non-Gaussian distribution and nonlinear correlation of frequency data under temperature interference.

[0010] 3. The present invention can fully consider the nonlinear effect of temperature on the main beam frequency when environmental factors are unknown, which helps to accurately reveal the changes in structural dynamic characteristics induced by internal changes such as sudden structural damage or performance degradation, improves the accuracy, robustness and timeliness of bridge damage identification results, and has strong universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The present invention will be further described below in conjunction with the accompanying drawings.

[0012] Figure 1 is a flow chart of the method of the present invention; Figure 2 The first two frequency slow characteristic time history curves in the embodiment of the present invention; Figure 3 It is a correlation scatter plot of the first two slow frequency features in an embodiment of the present invention; Figure 4 : is the probability density fitting curve of the first two slow frequency features in the embodiment of the present invention; Figure 5 This is a correlation scatter plot of the normalization of the local neighborhood of the first two slow-frequency features in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0014] The present invention provides a real-time identification method for bridge damage considering the nonlinear influence of temperature, which is specifically divided into four steps: "pre-processing abnormal values ​​of bridge vibration monitoring data and identifying the main beam frequency under the effect of ambient temperature", "nonlinearly extracting the main beam frequency slow feature vector based on exponential slow feature analysis", "adaptively searching the first nearest neighbor sample of each frequency slow feature vector and its neighborhood sample set, estimating the local mean and local standard deviation of the neighborhood sample set, and performing local standard on the slow feature vector" and "constructing a robust Mahalanobis distance damage index with minimum covariance estimation, statistically analyzing the probability distribution of the damage index and setting a real-time dynamic alarm threshold, and judging whether the bridge is damaged according to the threshold judgment criterion". The implementation process is as follows: Figure 1 shown.

[0015] Furthermore, the specific method of step 1 is: Step 1.1: Collect the vibration acceleration response data recorded by the bridge health monitoring system, screen for abnormal data types such as jump points, drift, periodic components and missing data, and perform error control preprocessing of the data by median filtering, low-pass filtering, differential filtering and trimming and filling; Step 1.2: Divide the acceleration response data pre-processed in step 1.1 into multiple parts in hours, identify the operational modal parameters of each part of the data, and obtain the main beam frequency data under the effect of ambient temperature. ,in, and Respectively represent the number of frequency samples and the number of variables, and divide the preprocessed frequency data into training set and test set; The specific method of step 2 is: The exponential slow feature analysis is used to transform the original frequency data preprocessed in step 1 into a new space through exponential mapping nonlinearly. , then, the slow frequency features are extracted by solving the following optimization problem: In the formula, represents the first-order covariance matrix and , represents the covariance matrix and , for About Time The first derivative of for The average value along the time axis, is the coefficient matrix, is the number of slow features; The above optimization problem characterizes the exponential slow feature analysis model as a generalized eigenvalue decomposition problem, which is solved by the Lagrange multiplier method. The solution formula is: ,in, is the eigenvalue corresponding to the slow feature; then, the slow eigenvector of the main beam frequency is calculated based on the solved coefficient matrix ; The specific method of step 3 is: Step 3.1: Use the similarity measurement method to find the similarity in step 2 in the training data set. Slow feature vector of moment frequency The nearest neighbor sample , and then calculate the nearest neighbor sample With Neighbors Euclidean distance , get the nearest neighbor sample The neighborhood sample subset composed of the first k nearest neighbor samples of ,in, ; Step 3.2: Estimate the k-neighborhood set The local neighborhood mean and local standard deviation of Slow feature vector of moment frequency After standardization, the calculation formula is as follows: In the formula, and are the slow frequency eigenvectors The nearest neighbor vector The mean and standard deviation of the neighborhood set of The slow frequency feature vector is normalized for the local neighborhood; Step 3.3: Loop through steps 3.1 and 3.2 to traverse the frequency slow feature vectors at all moments to obtain the local neighborhood normalized values.

[0016] The specific method of step 4 is: Step 4.1: For each frequency slow feature vector in step 3 Perform local neighborhood normalization and construct a robust Mahalanobis distance damage index based on minimum covariance estimation ,in, and Slow frequency features The mean vector and robust covariance matrix of ; Step 4.2: Use the kernel density estimation method to statistically analyze the probability distribution of each damage indicator in step 4.1, and then set the damage threshold. As a preferred method, the significance level is set to 0.01; Step 4.3: Calculate the frequency slow feature and its nearest neighbor vector of the online test data set, search for the neighborhood sample subset of the nearest neighbor vector in the training data set, and calculate its local mean and standard deviation. Construct the damage index based on the standardized result, and use the threshold discrimination criterion to determine in real time whether the bridge is damaged.

[0017] The effectiveness of the proposed method was verified using actual monitoring data of a beam bridge for 10 months. The acceleration response data of the main beam in normal service at different ambient temperatures were selected as the training set and the validation set, and the acceleration response collected when the bridge was damaged by foundation tilt, soffit concrete spalling, and prestressed tendon fracture was selected as the test set, with a sampling frequency of 100 Hz.

[0018] (1) According to step 1, the abnormal data such as jump points, drift, periodic components and missing data in the measured vibration acceleration response of the collected bridge are preprocessed, and then the preprocessed acceleration response is used for operational modal parameter identification in a time unit of 1 hour to obtain the first four measured frequency data of the bridge under the action of ambient temperature.

[0019] (2) According to the process described in step 2, extract the first two slow eigenvectors of the main beam frequency, such as Figure 2 As shown. Figure 3 and Figure 4 It can be seen that the eigenvector greatly reduces the spatial dimension of the original frequency data and reduces the interference of measurement noise, but it statistically presents nonlinear correlation and non-Gaussian distribution characteristics, which can easily mask the frequency variability induced by bridge structure damage.

[0020] (3) Calculate the local neighborhood sample subset according to the process described in step 3, and standardize the slow feature vector using the local mean and local standard deviation of the local neighborhood sample subset. Figure 5 It is a correlation scatter plot of the normalization of the local neighborhood of the first two frequency slow features. When the structure is intact, the normalized slow feature vectors are linearly correlated and statistically approximately Gaussian distributed, and are very different from the statistical characteristics of the monitoring samples under the structural damage state. This shows that the present invention can effectively eliminate the potential nonlinear influence of temperature.

[0021] (4) The false alarm rate of the training set and the validation set is about 0.58%, indicating that the bridge is in normal service condition; while the alarm rate of the test set is about 99.13%, indicating that the present invention can timely and effectively capture small abnormal changes in frequency when the bridge is in a damaged state, and give a correct judgment on the health status of the serving bridge in real time.

[0022] The above contents are merely examples and explanations of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.

Claims

1. A real-time bridge damage identification method considering the nonlinear effect of temperature, characterized in that: The steps include: Step 1: Preprocess the abnormal values ​​of bridge vibration monitoring data and identify the main beam frequency under the effect of ambient temperature; Step 2: Extract the main beam frequency slow feature vector based on the nonlinear exponential slow feature analysis; Step 3: Adaptively search the first nearest neighbor sample and its neighborhood set of each frequency slow feature vector, estimate the local mean and standard deviation of the neighborhood set, and perform local standardization on the slow feature vector; Step 4: Construct a robust Mahalanobis distance damage index with minimum covariance estimation, statistically analyze the probability distribution of the damage index and set a real-time dynamic alarm threshold, and determine whether the bridge is damaged according to the threshold judgment criterion.

2. The real-time bridge damage identification method considering the nonlinear effect of temperature according to claim 1 is characterized in that: Step 1 specifically includes the following process: Step 1.1: Collect the vibration acceleration response data recorded by the bridge health monitoring system, screen the abnormal data types, and perform error control preprocessing of median filtering, low-pass filtering, differential filtering and trimming and filling on the data; Step 1.2: Divide the acceleration response data pre-processed in step 1.1 into multiple parts in hours, identify the operational modal parameters of each part of the data, and obtain the main beam frequency data under the effect of ambient temperature. ,in, and They represent the number of frequency samples and the number of variables respectively. The preprocessed frequency data are divided into training set, validation set and test set.

3. The real-time bridge damage identification method considering the nonlinear effect of temperature according to claim 2 is characterized in that: Step 2 specifically includes the following process: The exponential slow feature analysis is used to transform the original frequency data preprocessed in step 1 into a new space through exponential mapping nonlinearly. , then, the slow frequency features are extracted by solving the following optimization problem: In the formula, represents the first-order covariance matrix and , represents the covariance matrix and , for About Time The first derivative of for The average value along the time axis, is the coefficient matrix, is the number of slow features; The Lagrange multiplier method is used to solve the above optimization problem, and the solution formula is: ,in, is the eigenvalue corresponding to the slow feature; then, the slow eigenvector of the main beam frequency is calculated based on the solved coefficient matrix .

4. The real-time bridge damage identification method considering the nonlinear effect of temperature according to claim 1 is characterized in that: Step 3 specifically includes the following process: Step 3.1: Use the similarity measurement method to find the training data set in step 2 Slow feature vector of moment frequency The nearest neighbor sample , and then calculate the nearest neighbor sample With Neighbors Euclidean distance , get the nearest neighbor sample The neighborhood set composed of the first k nearest neighbor samples of ,in, ; Step 3.2: Estimate the k-neighborhood set The local neighborhood mean and local standard deviation of Slow feature vector of moment frequency After standardization, the calculation formula is as follows: In the formula, and are the slow frequency eigenvectors The nearest neighbor vector The mean and standard deviation of the neighborhood set of The slow frequency feature vector is normalized for the local neighborhood; Step 3.3: Loop through steps 3.1 and 3.2 to traverse the frequency slow feature vectors at all moments to obtain the local neighborhood normalized values.

5. The real-time bridge damage identification method considering the nonlinear effect of temperature according to claim 1 is characterized in that: Step 4 specifically includes the following process: Step 4.1: For each frequency slow feature vector in step 3 Perform local neighborhood normalization and construct a robust Mahalanobis distance damage index based on minimum covariance estimation ,in, and Slow frequency features The mean vector and robust covariance matrix of ; Step 4.2: Use the kernel density estimation method to statistically analyze the probability distribution of each damage indicator in step 4.1, and then set the damage threshold; Step 4.3: Calculate the frequency slow feature and its nearest neighbor vector of the online test data set, search for the neighborhood sample subset of the nearest neighbor vector in the training data set, and calculate its local mean and standard deviation. Construct the damage index based on the standardized result, and use the threshold discrimination criterion to determine in real time whether the bridge is damaged.

Images