Fuzzy identification method of structural damage based on vibration modal features and wavelet packet transform

Through the modal feature fusion wavelet packet transformation method of structural vibration signals, fuzzy recognition of structural damage is achieved, and the problems of low identification accuracy and efficiency in the prior art are solved, and different damage levels can be identified and computational efficiency can be improved.

CN115541723BActive Publication Date: 2024-11-22TONGJI UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211055624.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-31
Publication Date
2024-11-22
Estimated Expiration
2042-08-31

AI Technical Summary

Technical Problem

In the identification of structural damage, the change amplitude of structural modal parameters needs to reach more than 10% to ensure reliability and accuracy. In addition, it is difficult to establish a scale reduction model for large-scale engineering structures, and local damage information cannot be directly identified.

Method used

The structural damage fuzzy recognition method based on vibration mode features fusion wavelet packet transformation is adopted. The modal feature parameters are extracted through the structural vibration signal, and fuzzy clustering is performed to identify different damage levels without relying on damage situation presets or modal parameter change amplitude.

Benefits of technology

The identification and evaluation of structural damage is realized, the calculation efficiency and accuracy are improved, and the local damage information can be identified when the overall modal parameters of the structure are not obvious.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115541723B_ABST
    Figure CN115541723B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of structural damage identification in structural health monitoring, and proposes a fuzzy identification method for structural damage based on vibration modal features fused with wavelet packet transform. The method extracts structural modal feature parameters based on structural vibration signals and fuses wavelet packet transform, and then fuzzy clusters the structural response signals under different damage conditions, thereby realizing the identification of structural damage. The present invention performs fuzzy clustering based on the energy distribution of the vibration signal modal frequency band, and achieves dimensionality reduction of vibration data while retaining the structural stiffness characteristics, thereby improving the computational efficiency and accuracy of fuzzy clustering processing. The present invention does not need to rely on damage scenario presets or the amplitude of changes in structural modal parameters. It can identify and divide different damage levels of the structure only based on the actual response of the structure, and has great engineering practice value for structural damage assessment in the field of structural health monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of structural damage identification in structural health monitoring. Background Art

[0002] In structural health monitoring, extracting structural modal features based on vibration signals is one of the important ways to identify structural damage. Modal parameters (such as modal frequency, damping ratio, etc.) are basic mechanical performance indicators of structural stiffness characteristics. Their numerical values ​​can reflect the damage of the structure to a certain extent, such as stiffness degradation. A large number of modal parameter identification methods have been developed based on structural vibration signals, which can be divided into time domain method, frequency domain method and time-frequency domain combination method. Among them, the random subspace method (SSI) in frequency domain identification can obtain modal parameter information based only on the structural output response, and has been widely used in related fields.

[0003] Based on modal parameter identification, a large number of structural damage identification methods have been derived, such as modal analysis, local damage method, non-parametric analysis method and time series method. These methods preset the response characteristics of the structure under different damage scenarios through the structural scale model, and construct a structural damage probability model, thereby realizing the detection and identification of structural damage.

[0004] However, a large number of studies have shown that damage identification based on modal parameters can only guarantee its reliability and accuracy when the variation of the structural modal parameters reaches more than 10%. For actual structures of large-scale projects, on the one hand, it is difficult to establish a scaled model and it is impossible to preset different damage scenarios of the structure. On the other hand, the sensitivity between the overall modal parameters of the structure and local damage may not be obvious, resulting in the inability to directly identify structural damage information through modal analysis. Therefore, the present invention proposes a fuzzy identification method for structural damage based on vibration modal features fused with wavelet packet transform. Summary of the invention

[0005] The purpose of the present invention is to propose a fuzzy identification method for structural damage based on vibration modal features fused with wavelet packet transform. This method extracts structural modal feature parameters based on structural vibration signals and fuses them with wavelet packet transform, and then fuzzy clusters the structural response signals under different damage conditions, thereby realizing the identification of structural damage. The present invention does not need to rely on damage scenario presets or the amplitude of structural modal parameter changes. It can identify and divide different damage levels of the structure based only on the actual response of the structure, and has great engineering practice value for structural damage assessment in the field of structural health monitoring.

[0006] To achieve the purpose of the present invention, the present invention is implemented by the following technical solutions:

[0007] A structural damage fuzzy identification method based on vibration modal features fused with wavelet packet transform comprises the following steps:

[0008] S1. Use sensors to collect acceleration response signals of the structure under different damage scenarios, and cut them into time series of appropriate length according to the sampling frequency {a1}{a2}…{a N}; Obtain N samples of N time periods and provide them to S2 and S3.

[0009] S2. Perform modal frequency identification based on acceleration signals and use random subspace method to identify modal parameters to obtain the first n-order modal frequencies {f n}, provided to S3.

[0010] S3, the collected acceleration signal {a i} is processed by wavelet packet transform, and then the structural modal frequency {f n} corresponding to the energy of the wave packet, we get {Energy node}, provided to S4.

[0011] S4. node} is normalized to obtain the relative modal frequency band energy {Energy rel}, provided to S5.

[0012] S5. Calculate the n-dimensional vector {Energy rel}, as a similarity measure. For vectors x and y, the Euler distance is calculated as follows:

[0013]

[0014] S6. Use the sum of square errors between each type of sample (N types of samples) and the class mean as the clustering criterion and define the criterion function:

[0015]

[0016] in

[0017]

[0018] J e It is a commonly used clustering criterion function, which represents the sum of squares of the total errors generated when N samples are clustered into c classes. Its value depends on the c cluster centers. y is the i-th cluster family Γ i The sample points in m i is the center of the i-th cluster.

[0019] S7, according to the minimum distance principle, assign each sample to the class closest to the representative point, constantly recalculate the class center, adjust each sample category, and finally make the clustering criterion function J eMinimum. (t) times of iteration, the sample point classification principle is

[0020]

[0021] In this way, the clustering results of N acceleration sequences under c different damage scenarios can be achieved. In practical applications, the number of damage types c can be defined in advance according to needs, and usually four damage levels can be taken, namely no damage, slight damage, moderate damage and severe damage.

[0022] S8. Obtain the clustering result of the acceleration response data (c=4). To further find out the degree of damage corresponding to the cluster family, perform modal frequency identification on the sample time series in each family according to the method in step S2. According to the mean value of the first-order modal frequency obtained by each class identification, the order from high to low is corresponding to no damage, slight damage, moderate damage and severe damage.

[0023] In S1, the time length of the vibration signal division needs to include at least 5 complete cycles of the structure.

[0024] In S2, when using the random subspace method for modal parameter identification, in order to effectively eliminate the interference of false modes, the stable spectrum is used to screen the modes. The basic idea is to use the characteristic of the system's eigenvalues ​​being conjugate with each other, and compare the frequency and damping calculation results of adjacent orders in the frequency-order two-dimensional spectrum in turn. If the difference between the two is within the set tolerance range, the two are treated as the same order and are stable points, otherwise they are unstable points.

[0025] In S3, the wavelet packets filter the low-frequency subbands of the input signal into progressively finer equal-width intervals, thereby obtaining a higher resolution in the frequency domain. The bandwidth (in Hz) of the complete wavelet packet down to level j is

[0026] F S is the sampling frequency.

[0027] In S4, the energy of the wave packet data is normalized, and the relative energy distribution information of the structural frequency band is used as the fuzzy clustering feature to eliminate the interference of the amplitude of the time domain vibration signal.

[0028] In S5, the calculation formula of Euler distance is:

[0029]

[0030] In S6, J e It is a commonly used clustering criterion function, which represents the sum of squares of the total errors generated when N samples are clustered into c classes. Its value depends on the c cluster centers.

[0031] Matters not covered by the present invention are applicable to the prior art.

[0032] The beneficial effects of the present invention are as follows: the modal frequency characteristics of the structure are extracted by using S2 modal frequency identification. According to the modal frequency identification result, S3 decomposes the vibration signal wavelet packet to obtain the energy distribution information of the corresponding frequency band of the structure. S5, S6, and S7 perform fuzzy clustering based on the signal modal frequency band energy vector, thereby obtaining the clustering results of the acceleration response under different damage scenarios. Finally, the modal frequency of the acceleration signal in the class sample is identified in S8, and its average value result is used as the corresponding basis for the degree of damage of the class. That is, only through the acceleration response of the structure, the identification and judgment of the overall damage state of the structure are realized.

[0033] The present invention achieves the acquisition of energy distribution information of a specific frequency band (corresponding to the modal frequency) of the vibration signal by fusing modal feature extraction and wavelet packet transformation, and uses it as a high-dimensional stiffness feature of the structure for fuzzy clustering, thus getting rid of the requirement of the variation amplitude of modal parameters in the process of identifying structural damage in the traditional modal analysis method. At the same time, fuzzy clustering is performed according to the energy distribution of the modal frequency band of the vibration signal, which achieves the dimension reduction of the vibration data while retaining the structural stiffness feature, thereby improving the computational efficiency and computational accuracy of the fuzzy clustering process. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is the overall flow chart of the present invention;

[0035] Figure 2 A modal frequency identification stability spectrum according to an embodiment of the present invention;

[0036] Figure 3 FIG. 4 is the absolute energy distribution of the wavelet packet decomposition nodes according to an embodiment of the present invention.

[0037] Figure 4 FIG. 1 is a diagram showing relative energy distribution of wavelet packet decomposition nodes according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. The technical solution of the present invention is described by taking the acceleration response signal of a high-rise building structure under environmental excitation as an example.

[0039] The design concept of the present invention is: extract the structural modal frequency through modal parameter identification, compare the wavelet packet decomposition results of the structural vibration signal, extract the energy distribution of the structural modal frequency band, and perform fuzzy clustering after normalization. The vibration signals of the same category in the clustering results indicate that they have similar modal energy distribution, that is, structural stiffness characteristics, and therefore can correspond to the structural damage level.

[0040] Example

[0041] A fuzzy identification method of structural damage based on vibration modal features fused with wavelet packet transform, such as Figure 1 As shown, the following steps are included:

[0042] S1. Using sensors to collect structural vibration response signals (taking acceleration signals as an example in this embodiment), and dividing them into time periods of appropriate length according to sampling frequency;

[0043] S2. Modal frequency identification is performed based on vibration signals. The random subspace method combined with the stable spectrum method, namely the SSI-COV method, is used to identify modal parameters and obtain the first n-order modal frequencies of the structure {f n} (in this embodiment, n=8), such as Figure 2 .

[0044] S3, using wavelet packet transform (WPT) to decompose the vibration signal and extract the structural modal frequency {f n}The energy of the wave packet corresponding to node},like Figure 3 .

[0045] S4. Normalize the wave packet energy to obtain the relative modal frequency band energy {Energy rel},like Figure 4 .

[0046] S5. Calculate the 8-dimensional vector {Energy rel The Euler distance between two vectors x and y in} is used as a similarity measure:

[0047]

[0048] S6. Use the sum of square errors between each class of samples and the class mean as the clustering criterion and define the criterion function:

[0049]

[0050] in

[0051]

[0052] J e It is a commonly used clustering criterion function, which represents the sum of squares of the total errors generated when N samples are clustered into 4 categories. Its value depends on the 4 cluster centers.

[0053] S7, according to the minimum distance principle, assign each sample to the class closest to the representative point, constantly recalculate the class center, adjust each sample category, and finally make the clustering criterion function J e Minimum.

[0054] S8. Obtain the clustering results of the acceleration response data (c=4). To further find out the damage degree corresponding to each cluster family, perform modal frequency identification on the sample time series in each family according to the method in step S2. According to the mean value of the first-order modal frequency obtained by identification of each class, in descending order, they correspond to no damage, slight damage, moderate damage and severe damage.

[0055] The above embodiments describe the basic principles, main features and technical routes of the present invention, and also show the feasibility and effectiveness of the present invention in realizing structural damage identification. It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive, and do not limit the patent scope of the present invention. Any changes to the present invention directly or indirectly by those skilled in the art according to the technical solution of the present invention are also included in the patent protection scope of the present invention. The scope of protection required by the present invention is defined by the attached claims and their equivalents.

Claims

1. Fuzzy identification method of structural damage based on vibration modal features fused with wavelet packet transform, It is characterized in that The following steps are involved: S1. Use sensors to collect acceleration response signals of the structure under different damage scenarios, and cut them into time series of appropriate length according to the sampling frequency {a 1 }{a 2 }…{a N }; Obtain N samples of N time periods and provide them to S2 and S3; S2. Perform modal frequency identification based on acceleration signals and use random subspace method to identify modal parameters to obtain the first n-order modal frequencies {f n }, provided to S3; S3, the collected acceleration signal {a i } is processed by wavelet packet transform, and then the structural modal frequency {f n } corresponding to the energy of the wave packet, we get {Energy node }, provided to S4; S4. node } is normalized to obtain the relative modal frequency band energy {Energy rel }, provided to S5; S5. Calculate the n-dimensional vector {Energy rel }, as a similarity measure; S6. Use the sum of square errors between each class of samples and the class mean as the clustering criterion and define the criterion function: in J e It is a commonly used clustering criterion function, which represents the sum of squares of the total errors generated when N samples are clustered into c classes. Its value depends on the c cluster centers. y is the i-th cluster family Γ i The sample points in m i is the center of the i-th cluster; S7, according to the minimum distance principle, assign each sample to the class closest to the representative point, constantly recalculate the class center, adjust each sample category, and finally make the clustering criterion function J e Minimum; (t) times of iteration, the sample point classification principle is Thus, the clustering results of N acceleration sequences under c different damage scenarios are achieved; in actual application, the number of damage types c is defined in advance according to the needs, and 4 damage levels are taken, namely no damage, slight damage, moderate damage and severe damage; S8. Obtain the clustering results of the acceleration response data. To further find out the degree of damage corresponding to the cluster family, perform modal frequency identification on the sample time series in each family according to the method in step S2. According to the mean value of the first-order modal frequency obtained by each class identification, the order from high to low is corresponding to no damage, slight damage, moderate damage and severe damage.

2. The structural damage fuzzy identification method based on vibration modal features fused with wavelet packet transform as claimed in claim 1, It is characterized in that In S1, the time length of the vibration signal division needs to include at least 5 complete cycles of the structure.

3. The structural damage fuzzy identification method based on vibration modal features fused with wavelet packet transform as claimed in claim 1, It is characterized in that In S2, when the random subspace method is used for modal parameter identification, in order to effectively eliminate the interference of false modes, the stable spectrum method is used to screen the modes; The basic idea is to use the pairwise conjugation characteristic of the system's eigenvalues ​​to compare the frequency and damping calculation results of adjacent orders in the frequency-order two-dimensional spectrum in turn. If the difference between the two is within the set tolerance range, they will be treated as the same order and are stable points, otherwise they are unstable points.

4. The structural damage fuzzy identification method based on vibration modal features fused with wavelet packet transform as claimed in claim 1, It is characterized in that In S3, the wavelet data packet converts the low-frequency subband filtering of the input signal into gradually finer equal-width intervals, thereby obtaining a higher resolution in the frequency domain; the bandwidth of the complete wavelet packet down to level j is k=0,1,...2 j -1, F S is the sampling frequency, in Hz.

5. The structural damage fuzzy identification method based on vibration modal features fused with wavelet packet transform as claimed in claim 1, It is characterized in that In S4, the energy of the wave packet data is normalized, and the relative energy distribution information of the structural frequency band is used as the fuzzy clustering feature to eliminate the interference of the amplitude of the time domain vibration signal.

6. The structural damage fuzzy identification method based on vibration modal features fused with wavelet packet transform as claimed in claim 1, It is characterized in that In S5, for vector x, y, the calculation formula of its Euler distance is: