Binary decomposition tree-based vibration modal video no-priori adaptive identification method and system

Abstract

The invention discloses a binary decomposition tree-based vibration mode video no-priori adaptive identification method and system, and belongs to the technical field of vibration mode identification. The method comprises the following steps: firstly, acquiring a multi-modal vibration video; constructing a space-time motion matrix of the multi-modal vibration video, reducing the dimension of the space-time motion matrix, and removing redundant information and noise; according to the construction principle of a binary tree, aliasing signals serve as root nodes, vibration modes serve as leaf nodes, each physical mode is obtained through successive separation along a path from the root nodes to the leaf nodes, vibration mode space distribution and displacement time sequence information are obtained from the physical modes to reconstruct a single-mode vibration video, finally, dynamic characteristic parameters such as vibration frequency are extracted, and the dynamic characteristic parameters such as the vibration frequency are extracted. And vibration mode identification is completed. According to the method, accurate recognition of the vibration mode can be realized under the condition that pre-measurement and parameter setting are not needed, and the flexibility and applicability of mode recognition are greatly improved, especially in the face of an unknown structure or a changing environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of vibration mode identification, and particularly relates to a vibration mode video non-prior adaptive identification method and system based on a binary decomposition tree. Background Art

[0002] Vibration is one of the inherent properties of an object. In key fields such as industrial production, civil engineering, and aerospace, complex structures are prone to multi-modal aliased vibrations. These vibration modes have different dynamic characteristics such as frequencies, damping ratios, and vibration modes. Among them, some modes can reflect the health status and operating conditions of the structure itself, and some modes may affect the normal operation of the structure and reduce the equipment life. Therefore, corresponding monitoring or suppression means need to be taken for different vibration modes, and accurate vibration mode identification is a prerequisite and basic means for corresponding processing of the modes.

[0003] Traditional vibration mode identification methods usually rely on contact sensors to capture the vibration signals of the structure. Although these methods can provide accurate data under certain conditions, due to the limited number and discrete spatial distribution of contact sensors, it has an adverse impact on subsequent mode identification. With the progress of image sensors, non-contact video-based vibration measurement methods have developed rapidly and become an attractive alternative. The high-density spatial distribution of measurement points has also stimulated the rise of video-based vibration mode identification methods. However, existing vibration mode video identification technologies still face challenges. Existing methods mainly use vibration characteristics such as natural frequencies, vibration modes, or the number of modes as prior information, which means that users must know or estimate these parameters in advance, which is not always feasible for unknown structures or changing environments. In addition, operators usually need to manually input and adjust prior parameters to adapt to different application scenarios. This method not only reduces the degree of automation but also easily introduces human errors, affecting the accuracy of the final identification results. More critically, the current methods have not formed a reasonable decomposition structure and lack a general framework that can automatically adapt to different conditions. These problems reduce the degree of automation and reliability of mode identification and restrict the popularization and development of this technology in practical applications.

[0004] In order to reduce the dependence of current vibration mode video identification methods on prior information and manual input, the present invention proposes a vibration mode video non-prior adaptive identification method and system based on a binary decomposition tree. This method can use the binary decomposition structure to adaptively identify the modal components of multi-modal aliased vibrations, fully improving the automation and reliability of vibration mode video identification. Summary of the Invention

[0005] The object of the present invention is to provide a vibration mode video non-prior adaptive recognition method and system based on a binary decomposition tree, which is conducive to realizing the adaptive decomposition and recognition of vibration modes without prior information and manual participation.

[0006] The object of the present invention is achieved by the following technical solutions:

[0007] A vibration mode video non-prior adaptive recognition method based on a binary decomposition tree, the specific steps are as follows:

[0008] Step 1: Obtain a multi-modal vibration video through a video acquisition device;

[0009] Step 2: Construct a spatio-temporal motion matrix of the multi-modal vibration video, which simultaneously includes the displacement time series signal of the vibration and the spatial distribution of the vibration mode;

[0010] Step 3: Reduce the dimension of the spatio-temporal motion matrix in Step 2 to remove redundant information and noise;

[0011] Step 4: Taking the reduced spatio-temporal motion matrix of multiple-order modes in Step 3 as the input, construct a binary decomposition tree structure, take the aliased signal as the root node, and the vibration mode as the leaf node, and sequentially separate along the path from the root node to the leaf node to obtain each physical mode, and obtain the spatial distribution of the vibration mode and the displacement time series information therefrom;

[0012] Step 4.1: Referring to the construction principle of a binary tree, select the number of process modes as 2, and initially assume that the aliased vibration signal is composed of these 2 process modes, and each process mode can capture multiple components with specific center frequencies and finite bandwidths;

[0013] Step 4.2: Minimize the total bandwidth around each center frequency by constraining the variational conditions, and iteratively update the process modes until the convergence criterion is met;

[0014] Step 4.3: Perform recursive binary decomposition on the extracted 2 process modes. If there are multiple frequency components, repeat the binary decomposition step until all resulting modes are single-frequency signals; after the decomposition is completed, evaluate and merge the resulting modes to finally obtain an accurate physical mode recognition result;

[0015] Step 5: Reconstruct a single-modal vibration video using the spatial distribution information in the spatio-temporal motion matrix of each vibration mode, and extract the dynamic characteristic parameters of the vibration frequency using the displacement time series signal in the spatio-temporal motion matrix of each vibration mode to complete the vibration mode recognition.

[0016] Further, the spatio-temporal motion matrix construction method described in Step 2 includes an image pyramid method, a Hilbert transform method, and a wavelet transform method.

[0017] Further, the dimensionality reduction method described in step 3 includes principal component analysis method, kernel principal component analysis method, and linear discriminant analysis method.

[0018] Further, in the aliased vibration signal in step 4.1, the vibration is expressed as a linear combination of physical modes:

[0019]

[0020] where, δ(x,y,t) is the vibration signal, δ m,n (x,y,t) is the physical mode, u m (x,y,t) is the process mode, and each process mode can capture multiple components with specific center frequencies and finite bandwidths. M is the number of process modes, M = 2, and N is the number of frequency components within each process mode;

[0021] Further, in step 4.2, on the premise that the sum of the process modes in step 4.1 accurately reconstructs the original vibration signal, variational constraint conditions are constructed to minimize the total bandwidth around each center frequency:

[0022]

[0023] where, ω m,n represents the center frequency corresponding to the nth component of the mth process mode;

[0024] Further, in step 4.3, after obtaining each physical mode, according to the formula:

[0025]

[0026] the spatial vibration mode distribution and the displacement time series signal q m,n (t) are obtained, where the displacement signal is a dynamic characteristic parameter containing the modal natural frequency.

[0027] Further, the single-modal video reconstruction method in step 5 is the inverse transformation of the spatio-temporal motion matrix construction method described in step 2, and the dynamic parameter extraction methods such as frequency include Fourier transform, short-time Fourier transform, and Hilbert transform.

[0028] A vibration mode video non-prior adaptive recognition system based on a binary decomposition tree, according to a vibration mode video non-prior adaptive recognition method based on a binary decomposition tree, the system includes an acquisition module, a construction module, a dimensionality reduction module, a decomposition module, and an identification module;

[0029] The acquisition module: is used to obtain multi-modal vibration videos, including a CCD industrial camera, a stable light source, a camera tripod, and a computer;

[0030] The construction module: used to construct a spatio-temporal motion matrix containing multi-modal vibration displacement time series signals and mode shape spatial distributions;

[0031] The dimensionality reduction module: used to reduce the dimensionality of the spatio-temporal motion matrix, removing redundant information and noise;

[0032] The decomposition module: used to construct a binary decomposition tree structure, successively separating each physical mode from the vibration with mode mixing, and obtaining the mode shape spatial distribution and displacement time series information therefrom;

[0033] The identification module: used to identify mode shapes and frequency dynamic characteristic parameters from the mode shape spatial distribution and displacement time series information respectively, reconstruct a single-modal vibration video, and complete vibration mode identification.

[0034] A computer-readable storage medium, on which a computer program / instructions are stored, and when the computer program / instructions are executed by a processor, the steps of a vibration mode video non-prior adaptive identification method based on a binary decomposition tree are implemented.

[0035] The beneficial effects of the present invention are as follows:

[0036] The present invention optimizes the adaptive ability of the vibration mode video identification method in the case of no prior information and manual input. The algorithm based on the binary decomposition tree can achieve accurate identification of vibration modes without the need for pre-measurement and setting of parameters (such as natural frequency, mode shape or number of modes), greatly improving the flexibility and applicability of mode identification, especially when facing unknown structures or changing environments.

[0037] The present invention enhances the reliability and processing efficiency of the vibration mode video identification method. This method provides a general framework for mode identification that can automatically adapt to different conditions. By reducing manual intervention, it realizes the full-process automation from data acquisition to mode identification, significantly reducing the possibility of human error and greatly improving the efficiency and result reliability of the identification process. Description of the Drawings

[0038] Figure 1 It is a flowchart of a vibration mode video non-prior adaptive identification method based on a binary decomposition tree of the present invention;

[0039] Figure 2 It is a schematic diagram of possible decomposition situations of a part of the binary decomposition tree of the present invention;

[0040] Figure 3 It is a schematic diagram of the vibration mode video non-prior adaptive identification process based on a binary decomposition tree and the identification result of the displacement time series signal in an embodiment of the present invention;

[0041] Figure 4 It is a schematic diagram of the modal mode shape identification result based on a binary decomposition tree in an embodiment of the present invention. Detailed implementation manners

[0042] The present invention will be further described below with reference to the accompanying drawings.

[0043] Embodiment 1:

[0044] The present invention provides a vibration mode video non - priori adaptive recognition method and system based on a binary decomposition tree. As Figure 1 shown, it mainly includes the following steps:

[0045] Step 1: Obtain a multi - modal vibration video through a video acquisition device, including an industrial camera, a stable light source, a camera tripod, and a computer.

[0046] Step 2: Construct a spatio - temporal motion matrix of the multi - modal vibration video, which simultaneously includes the displacement time - series signal and the vibration mode spatial distribution of the vibration; the construction methods of the spatio - temporal motion matrix include the image pyramid method, the Hilbert transform method, the wavelet transform method, etc.;

[0047] In this embodiment, the spatio - temporal motion matrix is constructed using the Gabor wavelet transform; the mathematical expression of the two - dimensional Gabor filter is where λ is the wavelength of the filter, θ is the direction of the filter, ψ is the phase parameter of the filter, γ is the spatial aspect ratio of the filter, and σ is the standard deviation of the Gaussian factor of the filter.

[0048] Step 3: Reduce the dimension of the spatio - temporal motion matrix to remove redundant information and noise; the dimension reduction methods include the principal component analysis method, the kernel principal component analysis method, the linear discriminant analysis method, etc.; in this embodiment, the principal component analysis method is used to reduce the dimension of the spatio - temporal motion matrix.

[0049] Step 4: As Figure 2 shown, without prior information and manual input, each modal component can be successively identified from the multi - modal aliased vibration, with extremely high adaptability and automation level, which is a reliable structure for realizing modal recognition. The vibration mode non - priori adaptive recognition is realized using a binary decomposition tree structure; the specific steps are as follows:

[0050] Step 41: Referring to the construction principle of a binary tree, select the number of process modes as 2, and initially assume that the aliased vibration signal is composed of these two process modes, and each process mode can capture multiple components with specific center frequencies and finite bandwidths;

[0051] For a multi - modal aliased vibration signal, the vibration can be expressed as a linear combination of physical modes where δ(x, y, t) is the vibration signal, δ m,n (x, y, t) is the physical mode, u m(x, y, t) is the process mode, and each process mode can capture multiple components with specific central frequencies and finite bandwidths. M is the number of process modes, M = 2, and N is the number of frequency components within each process mode;

[0052] Step 42: Minimize the total bandwidth around each central frequency by constraining the variational conditions, and iteratively update the process modes until the convergence criterion is met;

[0053] On the premise that the sum of the process modes in Step 41 can accurately reconstruct the original vibration signal, construct variational constraint conditions to minimize the total bandwidth around each central frequency where ω m,n represents the central frequency corresponding to the nth component of the mth process mode;

[0054] Step 43: Perform recursive binary decomposition on the two extracted process modes. If there are multiple frequency components, repeat the binary decomposition step until all resulting modes are single-frequency signals. After decomposition, evaluate and merge the resulting modes to finally obtain an accurate physical mode identification result;

[0055] After obtaining each physical mode, according to the formula obtain the spatial vibration mode distribution and the displacement time series signal q m,n (t), where the displacement signal contains dynamic characteristic parameters such as the modal natural frequency.

[0056] Step 5: Reconstruct the single-mode vibration video using the spatial distribution information in the spatio-temporal motion matrix of each vibration mode, extract dynamic characteristic parameters such as the vibration frequency using the displacement time series signal in the spatio-temporal motion matrix of each vibration mode to complete vibration mode identification; perform inverse transformation on the identified single-mode spatio-temporal motion matrix to obtain the single-mode vibration video, and perform Fourier transform on the displacement signal to obtain the corresponding spectrum and vibration frequency.

[0057] As Figure 3 shown, the vibration mode video non-prior adaptive identification method based on a binary decomposition tree of the present invention accurately identifies the corresponding vibration modes from the aliased vibration with four modal components through three binary tree decompositions, and obtains the corresponding modal displacement signals and spectra; as Figure 4 shown, a vibration mode video non-prior adaptive identification method provided by the present invention accurately identifies the modal vibration modes that are highly consistent with the reference values from the aliased vibration.

[0058] Embodiment 2:

[0059] A vibration mode video non-prior adaptive identification system based on a binary decomposition tree of the present invention, the system includes:

[0060] Acquisition module: used to obtain multi-modal vibration videos;

[0061] Construction module: used to construct a spatio-temporal motion matrix containing multi-modal vibration displacement time series signals and mode shape spatial distributions;

[0062] Dimensionality reduction module: used to reduce the dimensionality of the spatio-temporal motion matrix, removing redundant information and noise;

[0063] Decomposition module: used to construct a binary decomposition tree structure, successively separating each physical mode from the vibration with mode mixing, and obtaining the mode shape spatial distribution and displacement time series information therefrom;

[0064] Identification module: used to identify dynamic characteristic parameters such as mode shapes and frequencies from the mode shape spatial distribution and displacement time series information respectively, reconstruct a single-modal vibration video, and complete vibration mode identification.

[0065] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A non-priori adaptive recognition method for vibration modal video based on binary decomposition tree, characterized by: The specific steps are as follows: Step 1: Obtain multimodal vibration video through a video acquisition device; Step 2: construct a spatiotemporal motion matrix of the multimodal vibration video, which includes both the displacement time series signal and the vibration mode spatial distribution; Step 3: Reduce the dimension of the spatiotemporal motion matrix in step 2 to remove redundant information and noise; Step 4: Take the reduced-dimensional space-time motion matrix of the multi-order modes in step 3 as input, construct a binary decomposition tree structure, take the aliased signal as the root node and the vibration mode as the leaf node, and separate them one by one along the path from the root node to the leaf node to obtain each physical mode, and obtain the spatial distribution of the vibration mode and the displacement time series information from it; Step 4.1: Referring to the construction principle of the binary tree, the number of process modes is selected as 2. It is preliminarily assumed that the aliased vibration signal is composed of these two process modes, and each process mode can capture multiple components with a specific center frequency and limited bandwidth; Step 4.2: Minimize the total bandwidth around each center frequency by constraining the variational conditions, and iteratively update the process mode until the convergence criterion is met; Step 4.3: Perform recursive binary decomposition on the two extracted process modes. If there are multiple frequency components, repeat the binary decomposition step until all result modes are single-frequency signals; After the decomposition is completed, the resulting modes are evaluated and merged to finally obtain accurate physical mode identification results; Step 5: Reconstruct the single-mode vibration video using the spatial distribution information in the spatiotemporal motion matrix of each vibration mode, and extract the dynamic characteristic parameters of the vibration frequency using the displacement timing signal in the spatiotemporal motion matrix of each vibration mode to complete the vibration mode recognition.

2. The method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to claim 1, characterized in that: The spatiotemporal motion matrix construction method described in step 2 includes an image pyramid method, a Hilbert transform method and a wavelet transform method.

3. The method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to claim 1, characterized in that: The dimensionality reduction method described in step 3 includes a principal component analysis method, a kernel principal component analysis method, and a linear discriminant analysis method.

4. The method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to claim 1, characterized in that: In step 4.1, the vibration signal is aliased, and the vibration is represented as a linear combination of physical modes: Among them, δ(x,y,t) is the vibration signal, δ m,n (x, y, t) is the physical mode, u m (x, y, t) is the process mode, each process mode can capture multiple components with a specific center frequency and limited bandwidth, M is the number of process modes, M=2, and N is the number of frequency components in each process mode.

5. The method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to claim 1, characterized in that: The step 4.2 constructs a variational constraint condition to minimize the total bandwidth around each center frequency, under the premise that the process modal sum in step 4.1 accurately reconstructs the original vibration signal: Among them, ω m,n Indicates the center frequency corresponding to the nth component of the mth process mode.

6. The method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to claim 1, characterized in that: After obtaining each physical mode, step 4.3 is performed according to the formula: Obtaining spatial mode shape distribution and displacement timing signal q m,n (t), where the displacement signal is a dynamic characteristic parameter including the modal natural frequency.

7. The method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to claim 1, characterized in that: The single-mode video reconstruction method in step 5 is the inverse transformation of the spatiotemporal motion matrix construction method described in step 2, and the frequency and other dynamic parameter extraction methods include Fourier transform, short-time Fourier transform, and Hilbert transform.

8. A vibration modal video non-priori adaptive recognition system based on binary decomposition tree, characterized by: According to a method for non-a priori adaptive recognition of vibration modal video based on binary decomposition tree according to any one of claims 1 to 7, the system comprises an acquisition module, a construction module, a dimensionality reduction module, a decomposition module and a recognition module; The acquisition module is used to acquire multi-modal vibration videos, including a CCD industrial camera, a stable light source, a camera tripod and a computer; The construction module is used to construct a space-time motion matrix including multi-modal vibration displacement time series signals and vibration mode spatial distribution; The dimension reduction module is used to reduce the dimension of the spatiotemporal motion matrix and remove redundant information and noise; The decomposition module is used to construct a binary decomposition tree structure, to separate each physical mode from the vibration of the modal aliasing, and to obtain the spatial distribution of the vibration mode and the displacement time series information therefrom; The identification module is used to identify vibration mode and frequency dynamic characteristic parameters from vibration mode spatial distribution and displacement time series information respectively, reconstruct single-mode vibration video, and complete vibration mode identification.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images