A method and device for video analysis of vibration modalities of a combination electrical appliance

Abstract

The application provides a kind of based on combination electric appliance vibration modal video analysis method and device, it is related to electric power system maintenance field, comprising: obtaining the high frame rate video data of gas insulated combination electric appliance in dynamic operation process;Pixel level displacement analysis is carried out to high frame rate video data, and the three-dimensional displacement field data of each pixel point in high frame rate video data is obtained;Three-dimensional displacement field data and pre-constructed mechanical modal analysis model are used, and the modal parameter of gas insulated combination electric appliance in dynamic operation process is obtained;Modal parameter is input into pre-constructed deep learning hybrid model, and the fault type and position of gas insulated combination electric appliance in dynamic operation process are obtained.The application can diagnose non-contact for the mechanical abnormal state such as loosening, stress unevenness, structure displacement and geological subsidence of gas insulated combination electric appliance in high-voltage power grid.

Description

Technical Field

[0001] This application relates to the field of power equipment maintenance, specifically a method and device for video-based vibration modal analysis of combined electrical equipment. Background Technology

[0002] Gas-insulated switchgear (GIS) is widely used in ultra-high voltage and extra-high voltage power grids due to its small footprint and excellent insulation performance. However, its enclosed structure makes mechanical failures difficult to monitor directly. According to the failure statistics report of this type of equipment, mechanical failures account for a high proportion, mainly manifested as follows: (1) Loose connection: The pre-tightening force of flange bolts decreases, resulting in abnormal vibration amplitude (for example, when operating a GIS circuit breaker, the vibration acceleration at the loose flange can reach 3 to 5 times the normal state); (2) Uneven stress: Installation deviation or temperature deformation causes local stress concentration (when the stress at the busbar weld exceeds a certain value, fatigue cracks are likely to occur); (3) Structural displacement: When the circuit breaker and the busbar are relatively displaced, the contact resistance of the contacts will increase; (4) Geological collapse: When the foundation settles unevenly, the overall structure tilts, causing sealing failure.

[0003] The existing methods for detecting the aforementioned mechanical faults have the following problems: (1) Contact monitoring method: This method uses vibration sensors and strain gauges. However, it must be directly installed on the surface of the equipment through a rigid connection. It is prone to signal distortion and measurement errors in high voltage or strong electromagnetic interference environments. In addition, it can only acquire data from a single point and cannot capture the overall dynamic characteristics of the equipment, such as spatial vibration distribution or global modal changes.

[0004] (2) Signal analysis methods: existing analysis methods such as Fourier transform and wavelet analysis are used. However, they mainly deal with one-dimensional time series signals and can only extract frequency or time-frequency features. They are difficult to effectively correlate the spatial distribution features of structural mechanical modes and cannot handle the positional changes or local deformation effects of vibration modes in three-dimensional structures.

[0005] (3) Data-driven models: such as traditional machine learning. They rely heavily on manual feature engineering, requiring experts to manually design and select features, which increases complexity and subjectivity; when faced with complex mode distortions, their recognition accuracy is insufficient, especially in nonlinear vibration or abnormal deformation scenarios, and the model's generalization ability is weak.

[0006] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section. Summary of the Invention

[0007] To address the problems in the existing technology, this application provides a video-based method and device for vibration modal analysis of combined electrical equipment, which can perform non-contact diagnosis of mechanical anomalies such as loosening, uneven stress, structural displacement and geological collapse of gas-insulated combined electrical equipment in high-voltage power grids.

[0008] To solve the above-mentioned technical problems, this application provides the following technical solution: In a first aspect, this application provides a method for video-based vibration modal analysis of combined electrical equipment, including: Acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation; Pixel-level displacement analysis is performed on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data; Using the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, the modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained; wherein, the mechanical modal analysis model is constructed based on the constraints of discretized physical equations; The modal parameters are input into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

[0009] Further, the step of performing pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data includes: The original image sequence corresponding to the high frame rate video data is preprocessed to obtain an enhanced image sequence; Gaussian pyramid hierarchical matching is performed on the enhanced image sequence to obtain a pixel-level displacement initial field; Subpixel displacement calculation is performed on the initial pixel-level displacement field to obtain the subpixel displacement field; The three-dimensional displacement field data of each pixel is generated based on the binocular camera system parameters and the sub-pixel displacement field.

[0010] Furthermore, the modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained by utilizing the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, including: The three-dimensional displacement field data is divided into displacement time history windows to obtain normalized displacement time history segments; After inputting the normalized displacement time history segment into the elastic body dynamics equation, the elastic body dynamics equation is subjected to finite difference processing to obtain discretized physical equation constraints. The three-dimensional displacement field data and the constraints of the discretized physical equations are input into the mechanical modal analysis model to obtain the modal parameters.

[0011] Furthermore, the step of pre-constructing the mechanical modal analysis model includes: The training dataset, constructed from historical data of three-dimensional displacement field and historical modal parameters, is input into the neural network; The network parameters of the neural network are iteratively optimized based on a composite loss function that includes constraints from discretized physical equations. The neural network containing the iteratively optimized network parameters is determined as the mechanical modal analysis model.

[0012] Furthermore, the step of inputting the modal parameters into a pre-constructed deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation includes: The modal parameters are filtered for invalid modal parameters based on a preset modal parameter threshold to obtain valid modal parameters; The effective modal parameters are input into the deep learning hybrid model to obtain the fault type and location; wherein, the deep learning hybrid model includes a three-dimensional deep learning network composed of a convolutional neural network and a long short-term memory network.

[0013] Furthermore, the step of constructing the deep learning hybrid model includes: The historical modality parameters are input into a three-dimensional convolutional neural network layer of the neural network to obtain the spatial features of the historical modality parameters; The historical modality parameters are input into the long short-term memory network layer of the neural network, and the time characteristics of the historical modality parameters are described. The spatial features and the temporal features are fused to obtain spatiotemporal features; Using the spatiotemporal features, corresponding fault types and locations, the three-dimensional deep learning network composed of convolutional neural networks and long short-term memory networks is trained end-to-end to obtain the deep learning hybrid model.

[0014] Secondly, this application provides a video-based vibration modal analysis device for combined electrical equipment, comprising: The video data acquisition unit is used to acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation. The three-dimensional displacement analysis unit is used to perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data. The mechanical modal analysis unit is used to input the three-dimensional displacement field data into a pre-constructed mechanical modal analysis model to obtain the modal parameters of the gas-insulated combined electrical equipment during dynamic operation; wherein, the mechanical modal analysis model is constructed based on discretized physical equation constraints; The fault diagnosis unit is used to input the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

[0015] Thirdly, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the video-based vibration modal analysis method for combined electrical equipment.

[0016] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the video-based vibration modal analysis method for combined electrical equipment.

[0017] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the video-based vibration modal analysis method for combined electrical equipment.

[0018] To address the problems in existing technologies, this application provides a video-based method and apparatus for vibration modal analysis of combined electrical equipment. This method uses a high-speed camera (frame rate ≥ 2000fps) to capture micro-vibration images of gas-insulated combined electrical equipment under excitations such as switching operations and overvoltage. It generates normal / abnormal state samples (including four types of faults such as loosening and uneven stress) through physical simulation experiments and labels the true values ​​of modal parameters. It calculates the three-dimensional displacement field based on an improved digital image correlation (DIC) method. It uses a physical information neural network (PINN) to fuse displacement data with the elastic body dynamics equations to solve for modal parameters such as frequency, mode shape, and damping ratio. Finally, it achieves joint diagnosis of fault type and location through a 3D CNN-LSTM multi-task model. The method provided in this application is the first to achieve non-contact, full-field monitoring of the mechanical state of gas-insulated combined electrical equipment. Compared with traditional methods, it can provide early warning of potential faults and significantly improve the reliability of power grid operation and maintenance. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the vibration modal video analysis method for combined electrical equipment in the embodiments of this application; Figure 2 This is a flowchart illustrating the three-dimensional displacement field data obtained in the embodiments of this application; Figure 3 This is a flowchart illustrating the process of obtaining modal parameters in the embodiments of this application; Figure 4 This is a flowchart illustrating the construction of the mechanical modal analysis model in an embodiment of this application; Figure 5 This is a flowchart illustrating the fault types and locations obtained in the embodiments of this application; Figure 6 This is a flowchart illustrating the construction of a deep learning hybrid model in an embodiment of this application; Figure 7 This is a structural diagram of the vibration modal video analysis device for combined electrical equipment in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of the electronic device in the embodiments of this application; Figure 9 This is a schematic diagram of the overall process of video analysis of vibration modes of combined electrical equipment in the embodiments of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0022] The information collected in the technical solution of this application is information and data authorized by the user or fully authorized by all parties. The collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with the relevant laws, regulations and standards of the relevant countries and regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation portals are provided for users to choose to authorize or refuse.

[0023] Provide users with corresponding operation entry points, allowing them to choose to agree to or reject the automated decision results; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0024] In one embodiment, see Figure 1 In order to enable non-contact diagnosis of mechanical anomalies such as loosening, uneven stress, structural displacement, and geological collapse in gas-insulated combined electrical equipment in high-voltage power grids, this application provides a video-based vibration modal analysis method for combined electrical equipment, including: S101: Acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation; S102: Perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data; S103: Using the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, the modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained; wherein, the mechanical modal analysis model is constructed based on the constraints of discretized physical equations; S104: Input the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

[0025] Understandably, see Figure 9 This application's embodiments, based on high-speed image acquisition and pixel-level displacement analysis, perform mechanical modal reconstruction of equipment to achieve multimodal recognition. Through non-contact full-field monitoring and a physical information fusion algorithm, it achieves high-precision diagnosis of mechanical anomalies. First, high-speed image acquisition technology captures the dynamic operation of the machinery, acquiring high-frame-rate video data. Then, pixel-level displacement analysis is performed on the high-frame-rate video data to accurately calculate the positional change of each pixel and identify micro-deformations. Based on this, a mechanical modal reconstruction algorithm reconstructs the vibration modes and stress distribution of the mechanical system. Finally, a comprehensive judgment is made by combining multimodal recognition methods. The entire process relies on non-contact full-field monitoring, avoiding interference with equipment operation and covering the entire monitoring area. Furthermore, it applies a physical information fusion algorithm to deeply integrate image data with the mechanical physical model, improving diagnostic accuracy and reliability, thereby achieving early warning and accurate identification of mechanical anomalies.

[0026] The method provided in this application includes the following steps: S1. Multi-source excitation image acquisition: 2 to 4 high-speed cameras (frame rate 2000 to 10000fps, resolution 1920×1080 to 4096×2160) are used to capture images of key areas of GIS (circuit breaker arc-extinguishing chamber, disconnector contact, busbar flange) simultaneously with binoculars; simulated excitation conditions include: switch operation (opening and closing), overvoltage (lightning / operational impact), and vibration (0.1 to 10Hz low-frequency excitation).

[0027] S2. Label Database Construction: Four types of abnormal samples are generated through physical simulation.

[0028] (1) Mechanical loosening state: Adjust the preload of the flange, gas chamber connecting bolts or internal connecting parts of the circuit breaker disc spring, such as reducing it to 50%~70% of the standard value; (2) Uneven stress in independent air chambers: A lateral force (above 5000N) is applied to one of the air chambers, or a height difference is created by using a bracket (the angle with the horizontal plane does not exceed 2°). (3) Structural displacement of equipment casing: control axial / radial deviation of 1~5mm and angular deflection of 0.1°~0.5°; (4) Geological collapse: simulate unilateral settlement of 1~5mm or bilateral differential settlement of 1~3mm; The labeled data includes: three-dimensional displacement time history (sampling frequency 10~50kHz), modal parameters (true values ​​from laser vibrometer), fault type and location coordinates.

[0029] S3. Perform pixel-level displacement analysis on high frame rate video data.

[0030] S4. Perform the mechanical mode reconstruction process, including: using Physics-Informed Neural Networks (PINN) to determine the modal parameters of the gas-insulated combined electrical equipment during dynamic operation.

[0031] S5. Abnormal state identification, including: using 3D-CNN-LSTM to determine the fault type and location of gas-insulated combined electrical equipment during dynamic operation.

[0032] In step S2, the physical simulation experiment must meet the following requirements: 50 to 100 sets of samples are generated for each abnormal state; the true values ​​are collected synchronously using a laser vibrometer (accuracy ±1μm) and a strain gauge (±1με); data augmentation is performed using random cropping (80% of the region), time scale transformation (0.8 to 1.2 times speed), Gaussian noise (SNR 30 to 40dB), and the sample augmentation ratio is 1:5.

[0033] In practical implementation, this application's embodiments achieve high-precision identification of abnormal mechanical states in GIS through high-speed image full-field monitoring, physical information fusion modeling, and tag-based recognition. The feasibility of this method is ensured by steps such as distortion correction, sub-pixel calculation of the attention mechanism, and discretization of the physical information equation. Finally, experimental data verified that the 3D displacement field RMSE ≤ 0.05mm, modal frequency error ≤ 0.4Hz, and fault identification accuracy ≥ 96.2%, significantly outperforming existing technologies. The specific experimental process is described below: (1) The parameters of the GIS sample and equipment are as follows.

[0034] Test sample: 220kV three-phase common box type GIS (total length 10m, including circuit breaker, disconnect switch and busbar).

[0035] High-speed cameras: 2 units, frame rate 5000fps, resolution 1920×1080, lens focal length 50mm.

[0036] True value equipment (physical simulation experiment): laser vibrometer (Polytec PDV-100, ±1μm), strain gauge (±1με).

[0037] High-speed image acquisition module: contains 2 to 4 high-speed cameras using global shutter CMOS imaging mode, which should be synchronously triggered with the excitation source in physical simulation experiments.

[0038] Computing module: GPU computing power ≥ 8 TFLOPS, single-frame displacement analysis latency ≤ 20ms.

[0039] Database module: Stores image sequences, displacement fields, modal parameters and annotation data, with a capacity of ≥10TB and a write speed of ≥500MB / s.

[0040] Visualization terminal: Real-time display of displacement cloud map, vibration mode animation and fault diagnosis results, refresh rate ≥25fps.

[0041] (2) The key experimental results of steps S3 and S4 are as follows.

[0042] Table 1: Comparison of S3 Image Preprocessing Effects

[0043] Table 2: S4 Modal Parameter Identification Results (Normal State)

[0044] Table 3: Key Thresholds and Processing Logic

[0045] As described above, the vibration modal video analysis method for combined electrical equipment provided in this application can capture micro-vibration images of gas-insulated combined electrical equipment under excitations such as switching operations and overvoltage using a high-speed camera (frame rate ≥ 2000fps); generate normal / abnormal state samples (including four types of faults such as loosening and uneven stress) through physical simulation experiments, and label the true values ​​of modal parameters; calculate the three-dimensional displacement field based on the improved digital image correlation (DIC) method; use a physical information neural network (PINN) to fuse displacement data with the elastic body dynamics equations to solve for modal parameters such as frequency, mode shape, and damping ratio; and achieve joint diagnosis of fault type and location through a 3D CNN-LSTM multi-task model. The method provided in this application is the first to achieve non-contact full-field monitoring of the mechanical state of gas-insulated combined electrical equipment. Compared with traditional methods, it can provide early warning of potential faults and significantly improve the reliability of power grid operation and maintenance.

[0046] In one embodiment, see Figure 2 Pixel-level displacement analysis is performed on the high frame rate video data to obtain three-dimensional displacement field data of each pixel in the high frame rate video data, including: S201: Perform image preprocessing on the original image sequence corresponding to the high frame rate video data to obtain an enhanced image sequence; S202: Perform Gaussian pyramid hierarchical matching on the enhanced image sequence to obtain the pixel-level displacement initial field; S203: Perform sub-pixel displacement calculation on the pixel-level displacement initial value field to obtain the sub-pixel displacement field; S204: Generate three-dimensional displacement field data for each pixel based on the binocular camera system parameters and the sub-pixel displacement field.

[0047] It is understandable that steps S201 to S204 correspond to step S3 mentioned above.

[0048] S3, the pixel-level displacement analysis process is as follows.

[0049] S3-1, Image Preprocessing Subprocess: First, the raw image sequence acquired by the binocular high-speed camera is processed: distortion correction is performed by calibrating the imaging equipment parameters, specifically using Zhang's calibration method, based on the following formula to correct the radial / tangential distortion of the raw image:

[0050] in, x , y These are the normalized, distortion-free x and y coordinates, respectively. ; , These are the horizontal and vertical coordinates of the distortion space normalized from the original image; , , Radial distortion coefficient; , is the tangential radial distortion coefficient.

[0051] Second, adjust the parameters dynamically according to the noise level and perform bilateral filtering: if the gray standard deviation σg>30, set the kernel size of the bilateral filtering model to 7×7, the spatial domain standard deviation σs=2.0, and the value domain standard deviation σr=40; otherwise, the kernel size is 3×3, the spatial domain standard deviation σs=1.0, and the value domain standard deviation σr=20. Third, the image enhancement algorithm Contrast Limited Adaptive Histogram Equalization (CLAHE) is used, with a clip limit of 2.0~4.0, and the information entropy H of the enhanced image is required to be ≥6.5. Finally, the enhanced image output is directly used as the input parameter for the subsequent Gaussian pyramid hierarchical matching process in step S3-2, to ensure that the hierarchical matching process is performed on a distortion-free, high-contrast image.

[0052] S3-2, Gaussian pyramid hierarchical matching: First, the enhanced image sequence (containing GIS equipment surface texture features) preprocessed by S3-1 is processed. The pyramid parameters are set as follows:

[0053] Second, the pixel-level displacement field is calculated through the three-layer progressive matching process described above. The initial displacement estimate obtained from the top layer is scaled and used as the search starting point for the middle layer. After refinement by the middle layer matching, it is scaled again to the bottom layer as the initial position for precise matching, obtaining the displacement vector of each tracking point. After optimization by quadratic surface interpolation, the displacement measurement accuracy reaches 0.1~0.5 pixels. Finally, the pixel-level initial displacement field (up, vp) and the zero-mean normalized cross-correlation (ZNCC) coefficient matrix are output for screening reliable matching points.

[0054] Finally, the output pixel-level displacement initial field and ZNCC coefficient matrix are used as input parameters for the S3-3 sub-pixel displacement calculation process, providing initial values ​​and weights for sub-pixel-level optimization.

[0055] S3-3, Subpixel displacement calculation (using attention mechanism): First, the low-precision regions (edges and blurred texture regions) in the initial pixel displacement field are processed by ROI weight allocation; where edge pixels (extracted by Sobel operator) have a weight of w=1.2 and non-edge regions have a weight of w=1.0.

[0056] Second, surface fitting is performed on the 3×3 pixel neighborhood of the ZNCC peak, as shown in the following formula:

[0057] in, This refers to the normalized cross-correlation values, which represent the similarity measure of template matching. u,v These represent the horizontal and vertical offsets centered on the initial pixel-level displacement, respectively; a, b, c, d, e, and f are the quadratic surface fitting coefficients. Subpixel displacement is solved using partial derivatives, with an accuracy requirement of 0.01~0.05 pixels. Anomaly point masks are used to mark unreliable points with ZNCC < 0.85, which can be subsequently corrected by interpolation. Finally, the output subpixel displacement field and anomaly mask are used as inputs to the S3-4 3D displacement field construction and verification process, and the 3D coordinates are calculated in combination with binocular parallax.

[0058] S3-4. Construction and Verification of Three-Dimensional Displacement Field: First, input the subpixel displacement field output by S3-3, and the parameters of the stereo camera system; where the baseline distance is b, the focal length is f, and the parallax is d; use the stereo triangulation formula to obtain the subpixel displacement fields Z of the left and right views: (b=0.5~2m, f=50~100mm).

[0059] Second, dynamic verification was performed using a checkerboard calibration plate, requiring the root mean square error of the three-dimensional displacement to be ≤0.05mm.

[0060] Third, obtain three-dimensional displacement field data ( u x , u y , u z (Spatial resolution and temporal resolution) After the accuracy verification reaches RMSE≤0.05mm, proceed to S4 process; otherwise, return to S3-1 for reprocessing. Finally, the output three-dimensional displacement field data is directly used as the input parameters for the S4-1 PINN model preprocessing process, realizing the transformation from image pixels to physical quantities.

[0061] As can be seen from the above description, the vibration modal video analysis method for combined electrical equipment provided in this application can perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data.

[0062] In one embodiment, see Figure 3 The process of obtaining modal parameters of the gas-insulated combined electrical equipment during dynamic operation using the three-dimensional displacement field data and a pre-constructed mechanical modal analysis model includes: S301: Divide the three-dimensional displacement field data into displacement time history windows to obtain normalized displacement time history segments; S302: After inputting the normalized displacement time history segment into the elastic body dynamics equation, the elastic body dynamics equation is subjected to finite difference processing to obtain discretized physical equation constraints. S303: Input the three-dimensional displacement field data and the constraints of the discretized physical equations into the mechanical modal analysis model to obtain the modal parameters.

[0063] It is understandable that steps S201 to S204 correspond to step S4 mentioned above.

[0064] S4. The mechanical mode reconstruction process is as follows.

[0065] S4-1, PINN model input preprocessing: First, input the three-dimensional displacement field data output by S3-4 ( , , The displacement time history window is divided as follows: length T = 1024 frames (preferably 51.2ms for 20kHz sampling), overlap rate 50%; Second, normalize the displacement time history segments:

[0066] in, u This refers to the original displacement value, i.e., the aforementioned three-dimensional displacement field data segment; This refers to the minimum value of the original displacement; This refers to the maximum value of the original displacement; This refers to the normalized displacement value. It covers three vibration cycles, marks valid segments, and deletes segments from periods when the excitation is unstable. Finally, the normalized displacement time history segment output is used as the input for the constraint discretization of the S4-2 physical equations to ensure that the training data of the PINN model satisfies the time continuity and physical dimension consistency.

[0067] S4-2, Constraint Discretization of Physical Equations: First, the displacement time history output by S4.1 is normalized, and the elastic body dynamics equations are adopted. If the purpose is to measure the density of materials used in GIS equipment, the default setting can be density. The elasticity matrix C, measured in kg / m³, needs to be obtained beforehand through materials mechanics experiments. The equations are discretized using the finite difference method. Second derivative in space:

[0068] Where h refers to the pixel pitch, which is taken as 0.07mm here; In terms of spatial location Displacement value at; In position The displacement values ​​at the adjacent points on the left and right sides; Second derivative of time: Refers to the time interval, here taken as ; In The displacement value at time; In The displacement values ​​one time interval before and after the given moment.

[0069] Thus, the discretized physical constraints, the spatial second derivative, and the temporal second derivative are obtained.

[0070] Second, obtain the PINN model training dataset, including the data loss term ( ) and physical loss item ( ),in, The mean square error of the original displacement data and the network-predicted displacement is composed of the original displacement data and the network-predicted displacement, ensuring that the network output closely matches the actual observation. The network solution is forced to satisfy the basic physical laws by the mean square error of the residuals of the discretized elastic dynamics equations. Finally, the output discretized physical constraints and training dataset are directly input into the PINN model in the S4-3 modal parameter extraction and post-processing process for physical regularization solution of modal parameters.

[0071] S4-3. Modal parameter extraction and post-processing: Process the raw modal parameters (including noise interference and invalid modes) output by the PINN model: 1) Output: First 5 modal parameters (frequency) , vibration shape Damping ratio The first three frequency ranges must satisfy the following: , , ; 2) Invalid mode handling: If the frequency is out of range or MAC < 0.9 (you can choose to compare with the reference mode of the laser vibrometer), reselect the input window and increase the PINN physical loss weight. Up to 0.2.

[0072] In S4, the training parameters of the PINN model are: number of hidden layer neurons 512→256→128→64→32, using the Adam optimizer (learning rate 0.001, decaying by 10% every 100 rounds), batch size 64, and training rounds 2000.

[0073] As can be seen from the above description, the vibration modal video analysis method for combined electrical equipment provided in this application can obtain the modal parameters of the gas-insulated combined electrical equipment during dynamic operation by utilizing the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model.

[0074] In one embodiment, see Figure 4 The step of pre-constructing the mechanical modal analysis model includes: S401: Input the training dataset constructed from the historical data of the three-dimensional displacement field and the historical modal parameters into the neural network; S402: Iteratively optimize the network parameters of the neural network based on a composite loss function that includes constraints from the discretized physical equations; S403: The neural network containing the iteratively optimized network parameters is determined as the mechanical modal analysis model.

[0075] Understandably, the core steps in constructing a mechanical modal analysis model lie in the deep integration of physical laws and data-driven methods. First, it is necessary to integrate historical data of the three-dimensional displacement field and historical modal parameters to construct a high-quality training dataset, which is then input into the neural network architecture to provide fundamental data support for model learning. Subsequently, a composite loss function containing constraints from discretized physical equations is introduced. This function not only includes data fitting terms to measure the deviation between the model's predictions and historical data but also incorporates physical equation constraint terms to ensure that the model output strictly adheres to fundamental physical laws such as mechanical equilibrium and energy conservation. Through optimization algorithms such as gradient descent, the parameters of the neural network, such as weights and biases, are iteratively optimized. Under the dual guidance of data-driven and physical constraints, the model parameters are gradually adjusted to minimize the composite loss function. This process enables the model to capture complex features in the data while maintaining physical consistency. Finally, when the parameter optimization reaches convergence, the neural network is determined as a mechanical modal analysis model for prediction, achieving efficient and physically reliable mechanical analysis.

[0076] As can be seen from the above description, the vibration modal video analysis method for combined electrical equipment provided in this application can pre-construct the mechanical modal analysis model.

[0077] In one embodiment, see Figure 5 The step of inputting the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation includes: S501: The modal parameters are filtered for invalid modal parameters according to the preset modal parameter threshold to obtain valid modal parameters; S502: Input the effective modal parameters into the deep learning hybrid model to obtain the fault type and location; wherein, the deep learning hybrid model includes a three-dimensional deep learning network composed of a convolutional neural network and a long short-term memory network.

[0078] In one embodiment, see Figure 6 The steps for constructing the deep learning hybrid model include: S601: Input the historical modality parameters into the three-dimensional convolutional neural network layer of the neural network to obtain the spatial features of the historical modality parameters; S602: Input the historical modality parameters into the long short-term memory network layer of the neural network, wherein the historical modality parameters have temporal characteristics; S603: Perform spatiotemporal feature fusion on the spatial features and the temporal features to obtain spatiotemporal features; S604: Using the spatiotemporal features, the corresponding fault type and location, perform end-to-end training on the three-dimensional deep learning network composed of a convolutional neural network and a long short-term memory network to obtain the deep learning hybrid model.

[0079] It is understandable that steps S501 to S502 and steps S601 to S604 correspond to step S5.

[0080] S5. Abnormal State Identification: Employing a 3D CNN-LSTM multi-task model: S5-1 Input Data Preprocessing: First, the time-domain sequence of modal parameters output from S4-3 is received. This sequence contains the frequencies, mode shapes, and damping ratios of the first five dominant modes. To construct the input tensor for the model, the modal parameters are structurally reorganized: the mode shapes (two-dimensional spatial distribution) at each time step are stacked along the modal order dimension to form a three-dimensional tensor (modal order, spatial height, spatial width). At the same time, the frequencies and damping ratios at the corresponding time steps are associated with this tensor as auxiliary feature vectors.

[0081] Second, the modal parameter sequence is normalized and aligned in the time dimension. The sequence length is adjustable, ranging from 100 to 500 frames, to accommodate observations of different durations. Simultaneously, scalar parameters such as frequency and damping ratio are standardized to bring them to similar numerical magnitudes.

[0082] Finally, a structured, spatiotemporally aligned sequence of modal parameter tensors is output as input to the deep learning hybrid model in S5-2, used to learn the mapping relationship between fault modes and spatiotemporal features. S5-2, 3D CNN-LSTM multi-task model: First, a deep learning hybrid model (3D CNN-LSTM) consisting of a 3D convolutional neural network and a long short-term memory network is constructed. The 3D CNN module obtains the spatial features of historical modality parameters; the LSTM module obtains the temporal features of historical modality parameters. Then, a two-stage transfer learning strategy was employed to train the hybrid model. The first stage involved pre-training the modal feature extractor: using large-scale historical data, the fully trained PINN model in S4 was used as a foundation, its core parameters were frozen, and it was used as a dedicated physical feature extractor, whose output consisted of modal features containing physical information. The second stage involved end-to-end fine-tuning of the multi-task network: the pre-trained modal feature extractor was concatenated with a 3D CNN-LSTM network to form a complete diagnostic model. Using a historical dataset containing accurate fault types and location labels, the entire network was fine-tuned end-to-end using a multi-task loss function (jointly optimizing classification cross-entropy loss and localization smoothing L1 loss).

[0083] S5-3, Model Output Results: The model can receive real-time modal parameter sequences and output two key results in parallel: first, the classification probability of fault types (normal, mechanical loosening, uneven stress in independent air chambers, structural displacement of equipment casing, geological collapse); second, the three-dimensional spatial coordinate prediction of the fault location, with a positioning accuracy requirement of no more than 5mm.

[0084] As can be seen from the above description, the vibration modal video analysis method for combined electrical equipment provided in this application can construct the deep learning hybrid model and input the modal parameters into the pre-constructed deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation.

[0085] Based on the same inventive concept, this application also provides a device for video analysis of vibration modes of combined electrical equipment, which can be used to implement the method described in the above embodiments, as described in the following embodiments. Since the principle of solving the problem based on the video analysis of vibration modes of combined electrical equipment is similar to that of the video analysis method for vibration modes of combined electrical equipment, the implementation of the device can refer to the implementation of the method based on software performance benchmark determination, and will not be repeated. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0086] In one embodiment, see Figure 7 In order to enable non-contact diagnosis of mechanical anomalies such as loosening, uneven stress, structural displacement, and geological collapse in gas-insulated combined electrical equipment in high-voltage power grids, this application provides a video-based vibration modal analysis device for combined electrical equipment, comprising: The video data acquisition unit 701 is used to acquire high frame rate video data of the gas-insulated combined electrical equipment during dynamic operation. The three-dimensional displacement analysis unit 702 is used to perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data. The mechanical modal analysis unit 703 is used to input the three-dimensional displacement field data into a pre-constructed mechanical modal analysis model to obtain the modal parameters of the gas-insulated combined electrical equipment during dynamic operation; wherein, the mechanical modal analysis model is constructed based on discretized physical equation constraints; The fault diagnosis unit is used to input the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

[0087] From a hardware perspective, this is to enable non-contact diagnosis of mechanical anomalies such as loosening, uneven stress, structural displacement, and geological collapse in gas-insulated combined electrical equipment in high-voltage power grids.

[0088] This application provides an embodiment of an electronic device for implementing all or part of the aforementioned method for video-based vibration modal analysis of combined electrical equipment. The electronic device specifically includes the following components: The system comprises a processor, a memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the device for video analysis of vibration modes of combined electrical equipment and core business systems, user terminals, and related databases; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the video analysis method for vibration modes of combined electrical equipment and the embodiments of the device for video analysis of vibration modes of combined electrical equipment, the contents of which are incorporated herein by reference, and repeated details will not be described again.

[0089] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0090] In practical applications, some parts of the vibration modal video analysis method for combined electrical equipment can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific choice depends on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0091] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0092] Figure 8 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 8 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 8 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0093] In one embodiment, the video-based vibration modal analysis method for combined electrical equipment can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control: S101: Acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation; S102: Perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data; S103: Using the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, the modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained; wherein, the mechanical modal analysis model is constructed based on the constraints of discretized physical equations; S104: Input the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

[0094] As described above, the vibration modal video analysis method for combined electrical equipment provided in this application can capture micro-vibration images of gas-insulated combined electrical equipment under excitations such as switching operations and overvoltage using a high-speed camera (frame rate ≥ 2000fps); generate normal / abnormal state samples (including four types of faults such as loosening and uneven stress) through physical simulation experiments, and label the true values ​​of modal parameters; calculate the three-dimensional displacement field based on the improved digital image correlation (DIC) method; use a physical information neural network (PINN) to fuse displacement data with the elastic body dynamics equations to solve for modal parameters such as frequency, mode shape, and damping ratio; and achieve joint diagnosis of fault type and location through a 3D CNN-LSTM multi-task model. The method provided in this application is the first to achieve non-contact full-field monitoring of the mechanical state of gas-insulated combined electrical equipment. Compared with traditional methods, it can provide early warning of potential faults and significantly improve the reliability of power grid operation and maintenance.

[0095] In another embodiment, the combined electrical equipment vibration modal video analysis device can be configured separately from the central processing unit 9100. For example, the data composite transmission device can be configured as a chip connected to the central processing unit 9100 based on the combined electrical equipment vibration modal video analysis device, and the function of the combined electrical equipment vibration modal video analysis method can be realized through the control of the central processing unit.

[0096] like Figure 8As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 8 All components shown; in addition, the electronic device 9600 may also include Figure 8 For components not shown, please refer to existing technologies.

[0097] like Figure 8 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0098] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0099] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0100] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0101] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0102] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0103] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is also coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored sound via the speaker 9131.

[0104] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the video-based vibration modal analysis method for combined electrical equipment, where the execution subject is a server or client, as described in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the video-based vibration modal analysis method for combined electrical equipment, where the execution subject is a server or client, as described in the above embodiments. For example, when the processor executes the computer program, it implements the following steps: S101: Acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation; S102: Perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data; S103: Using the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, the modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained; wherein, the mechanical modal analysis model is constructed based on the constraints of discretized physical equations; S104: Input the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

[0105] As described above, the vibration modal video analysis method for combined electrical equipment provided in this application can capture micro-vibration images of gas-insulated combined electrical equipment under excitations such as switching operations and overvoltage using a high-speed camera (frame rate ≥ 2000fps); generate normal / abnormal state samples (including four types of faults such as loosening and uneven stress) through physical simulation experiments, and label the true values ​​of modal parameters; calculate the three-dimensional displacement field based on the improved digital image correlation (DIC) method; use a physical information neural network (PINN) to fuse displacement data with the elastic body dynamics equations to solve for modal parameters such as frequency, mode shape, and damping ratio; and achieve joint diagnosis of fault type and location through a 3D CNN-LSTM multi-task model. The method provided in this application is the first to achieve non-contact full-field monitoring of the mechanical state of gas-insulated combined electrical equipment. Compared with traditional methods, it can provide early warning of potential faults and significantly improve the reliability of power grid operation and maintenance.

[0106] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for video-based vibration modal analysis of combined electrical equipment, characterized in that, include: Acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation; Pixel-level displacement analysis is performed on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data; Using the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, the modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained; wherein, the mechanical modal analysis model is constructed based on the constraints of discretized physical equations; The modal parameters are input into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

2. The method for video-based vibration modal analysis of combined electrical equipment according to claim 1, characterized in that, The step of performing pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data includes: The original image sequence corresponding to the high frame rate video data is preprocessed to obtain an enhanced image sequence; Gaussian pyramid hierarchical matching is performed on the enhanced image sequence to obtain a pixel-level displacement initial field; Subpixel displacement calculation is performed on the initial pixel-level displacement field to obtain the subpixel displacement field; The three-dimensional displacement field data of each pixel is generated based on the binocular camera system parameters and the sub-pixel displacement field.

3. The method for video-based vibration modal analysis of combined electrical equipment according to claim 1, characterized in that, The modal parameters of the gas-insulated combined electrical equipment during dynamic operation are obtained by utilizing the three-dimensional displacement field data and the pre-constructed mechanical modal analysis model, including: The three-dimensional displacement field data is divided into displacement time history windows to obtain normalized displacement time history segments; After inputting the normalized displacement time history segment into the elastic body dynamics equation, the elastic body dynamics equation is subjected to finite difference processing to obtain discretized physical equation constraints. The three-dimensional displacement field data and the constraints of the discretized physical equations are input into the mechanical modal analysis model to obtain the modal parameters.

4. The method for video-based vibration modal analysis of combined electrical equipment according to claim 3, characterized in that, The steps for pre-constructing the mechanical modal analysis model include: The training dataset, constructed from historical data of three-dimensional displacement field and historical modal parameters, is input into the neural network; The network parameters of the neural network are iteratively optimized based on a composite loss function that includes constraints from discretized physical equations. The neural network containing the iteratively optimized network parameters is determined as the mechanical modal analysis model.

5. The method for video-based vibration modal analysis of combined electrical equipment according to claim 1, characterized in that, The modal parameters are input into a pre-built deep learning hybrid model to obtain the fault types and locations of the gas-insulated combined electrical equipment during dynamic operation, including: The modal parameters are filtered for invalid modal parameters based on a preset modal parameter threshold to obtain valid modal parameters; The effective modal parameters are input into the deep learning hybrid model to obtain the fault type and location; wherein, the deep learning hybrid model includes a three-dimensional deep learning network composed of a convolutional neural network and a long short-term memory network.

6. The method for video-based vibration modal analysis of combined electrical equipment according to claim 5, characterized in that, The steps for constructing the deep learning hybrid model include: The historical modality parameters are input into a three-dimensional convolutional neural network layer of the neural network to obtain the spatial features of the historical modality parameters; The historical modality parameters are input into the long short-term memory network layer of the neural network to obtain the temporal characteristics of the historical modality parameters; The spatial features and the temporal features are fused to obtain spatiotemporal features; Using the spatiotemporal features, corresponding fault types and locations, the three-dimensional deep learning network composed of convolutional neural networks and long short-term memory networks is trained end-to-end to obtain the deep learning hybrid model.

7. A device for video-based vibration modal analysis of combined electrical equipment, characterized in that, include: The video data acquisition unit is used to acquire high frame rate video data of gas-insulated combined electrical equipment during dynamic operation. The three-dimensional displacement analysis unit is used to perform pixel-level displacement analysis on the high frame rate video data to obtain the three-dimensional displacement field data of each pixel in the high frame rate video data. The mechanical modal analysis unit is used to input the three-dimensional displacement field data into a pre-constructed mechanical modal analysis model to obtain the modal parameters of the gas-insulated combined electrical equipment during dynamic operation; wherein, the mechanical modal analysis model is constructed based on discretized physical equation constraints; The fault diagnosis unit is used to input the modal parameters into a pre-built deep learning hybrid model to obtain the fault type and location of the gas-insulated combined electrical equipment during dynamic operation; wherein, the deep learning hybrid model is constructed based on the time-domain sequence corresponding to the historical modal parameters.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method for video analysis of vibration modes of combined electrical equipment as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method for video analysis of vibration modes of combined electrical equipment as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the method for video analysis of vibration modes of combined electrical equipment as described in any one of claims 1 to 6.

Images