High-voltage circuit breaker non-disassembly detection spring deformation-modal health state evaluation method

Abstract

The invention discloses a high-voltage circuit breaker non-disassembly detection spring deformation-modal health state evaluation method, and relates to the technical field of image analysis. The method comprises the following steps: step 1, obtaining a spring image sequence when the circuit breaker acts; 2, a displacement time curve of a first turn is obtained through processing of a correlation filtering algorithm, a displacement curve in the stretching direction of the spring is obtained through extraction, and a first intrinsic mode function, a second intrinsic mode function, a third intrinsic mode function and a residual component are obtained through decomposition of a mode decomposition algorithm; performing discrete Fourier transform extraction to obtain a spectrogram of a third intrinsic mode function; 3, when the deviation between the frequency corresponding to the power ratio peak value in the spring spectrogram of the to-be-detected circuit breaker and the frequency corresponding to the power ratio peak value in the spring spectrogram of the normal circuit breaker exceeds a comparison threshold value, determining that a fault occurs; through the steps 1 to 3, the calculation amount is reduced, the tracking speed is increased, the high-voltage circuit breaker spring detection and evaluation efficiency is higher, the method is more scientific, and the technical effect is better.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and particularly to a method for evaluating the health state of a high-voltage circuit breaker's spring deformation-modal without disassembly detection. Background Art

[0002] The prior art measures the motion information of a circuit breaker by installing an angular displacement sensor on the transmission main shaft and calculates the mechanical characteristic parameters of the circuit breaker's operation. Maintenance experience has proved that the operation reliability of a circuit breaker is directly related to the mechanical state of the energy storage spring. In the relevant field, there is no scientific evaluation method for the state of the energy storage spring. General wear, relaxation, creep, and fracture and other potential defects can only be identified by special instruments after disassembly. As the energy storage spring, which is the source of the operating drive energy of the circuit breaker, has a special shape, its energy release structure determines that it is impossible to install a high-precision sensor under the live state. Therefore, it is difficult to detect the defects of the energy storage spring itself in the existing technical conditions.

[0003] The authorized announcement number is CN108921812B, and the name is an intelligent evaluation method for the fatigue state of a circuit breaker spring based on image recognition. It uses a high-speed image sequence and the NCC algorithm to detect and analyze the positions of key moving targets during the spring deformation process, obtains a curve representing the spring fatigue state, and thus obtains a spring fatigue characteristic parameter vector; uses the GA-SALBP model to analyze the obtained spring fatigue characteristic parameter vector, obtains the fatigue state value and stress relaxation condition of the circuit breaker spring, and combines the error energy index of the circuit breaker spring performance evaluation to realize the evaluation of the performance of the tested spring. Due to the large workload of calculating the spring fatigue state curve and analyzing the spring fatigue characteristic parameter vector, the working efficiency of the spring technical parameter analysis and evaluation is low, and the cost performance is poor.

[0004] The application publication number is CN116524006A, and the name is a method, device, and medium for evaluating the state of an energy storage spring. By collecting the sound signal and shooting the image sequence of the energy storage spring, then preprocessing the sound signal and the shot image sequence respectively, and importing the processed sound characteristics and the processed shot image characteristics into the evaluation model. Due to the need to collect and process two types of signals, the workload of calculation and analysis is large, resulting in low working efficiency of the spring technical parameter analysis and evaluation and poor cost performance. Due to the interference of other transmission components in the sound and vibration signals, it is impossible to directly evaluate the performance of the spring only.

[0005] The prior art uses a high-speed camera to obtain an image sequence of the deformation of the energy storage spring of a circuit breaker, takes the first turn of the movable end as the matching target, and proposes a normalized cross-correlation graphic pyramid matching algorithm, abbreviated as NCC-P-E, to track the target and obtain the deformation characteristic parameters of the energy storage spring. Parameters such as spring displacement and speed are extracted in this method, but there is a lack of a reasonable algorithm for further evaluating the spring performance, so the health state of the spring cannot be judged scientifically and reasonably.

[0006] Therefore, the low efficiency and poor effect of the spring detection and evaluation of high-voltage circuit breakers have become technical problems to be solved urgently. Summary of the Invention

[0007] The present invention provides a method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker, which solves the technical problems of low efficiency and poor effect in the spring detection and evaluation of high-voltage circuit breakers.

[0008] To solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0009] A method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker includes the following steps. Step 1: Obtain an image sequence of the spring when the circuit breaker operates; Step 2: Process the image sequence of the spring when the circuit breaker operates through the correlation filtering algorithm DSST to obtain the displacement-time curve of the first turn of the spring, extract the displacement curve of the first turn of the spring in the spring expansion and contraction direction from the displacement-time curve of the first turn of the spring, and decompose it through the empirical mode decomposition algorithm EMD to obtain the first intrinsic mode function, the second intrinsic mode function, the third intrinsic mode function and the residual component, and use the discrete Fourier transform to extract the frequency spectrum diagram of the third intrinsic mode function from the third intrinsic mode function; Step 3: Based on Step 1 and Step 2, obtain the frequency spectrum diagrams of the third intrinsic mode functions of the spring of the circuit breaker to be tested and the spring of the normal circuit breaker respectively. If the deviation between the frequency corresponding to the peak power ratio in the frequency spectrum diagram of the spring of the circuit breaker to be tested and the frequency corresponding to the peak power ratio in the frequency spectrum diagram of the spring of the normal circuit breaker exceeds the set comparison threshold, it is determined as a fault.

[0010] A further technical solution lies in: In the said Step 1, a computer, a camera, an LED lamp and a controller are used to obtain an image sequence of the spring when the circuit breaker operates. The computer is connected to the controller and communicates bidirectionally. The controller is electrically connected to the closing and opening coils of the high-voltage circuit breaker. The computer is electrically connected to the camera through a USB video data cable.

[0011] A further technical solution lies in: The specific division of the said Step 2 includes the following steps,

[0012] Step 201: Process the spring image sequence during the breaker operation through the correlation filtering algorithm DSST to obtain the displacement-time curve X(t) of the first turn of the spring;

[0013] Step 202: Extract from the displacement-time curve X(t) of the first turn of the spring to obtain the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction;

[0014] Step 203: Decompose the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction through the empirical mode decomposition algorithm EMD to obtain the first intrinsic mode function IMF1(t), the second intrinsic mode function IMF2(t), the third intrinsic mode function IMF3(t), and the residual component r3(t);

[0015] Step S204: Use the discrete Fourier transform to extract the frequency spectrum diagram of the third intrinsic mode function IMF3(t) from the third intrinsic mode function IMF3(t).

[0016] A further technical solution lies in that: in the said Step 2, the position filter of the correlation filtering algorithm DSST is a two-dimensional filter and is used for position tracking, the scale filter of the correlation filtering algorithm DSST is a one-dimensional filter and is used for scale estimation, the position filter is represented by Equation (1),

[0017]

[0018] In Equation (1), d represents a d-dimensional feature vector, l represents a dimension of the d-dimensional feature, h is the filter of the Gaussian output response, * is convolution, f is the sample feature, g is the Gaussian function, and λ is the regularization term coefficient;

[0019] Through the discrete Fourier transform, the convolution operation in the time domain in Equation (1) is converted into a point multiplication calculation in the frequency domain for solution, and we get:

[0020]

[0021] In Equation (2), is the conjugate transpose of f in the frequency domain, is the conjugate transpose of g in the frequency domain;

[0022] In a new frame of image, use the position filter trained with the target information of the previous frame to determine the current frame target position, and the output response value of the position filter is:

[0023]

[0024] In Equation (3), y is the filter output response value, is the feature channel of the l-th dimension at time t, and determine the tracking target center position according to the maximum value of the output response.

[0025] A further technical solution lies in that: in the step 2, when performing scale estimation through a scale filter, with the current target position as the center, the previous frame target box is in a non-linear exponential relationship as the candidate scale, and the target sample scale selection principle is:

[0026]

[0027] In formula (4), P and R respectively represent the width and height of the target box in the previous frame image; a is the scale factor; s is the number of scales; n is a constant between;

[0028] Combining the obtained target center position with the optimal target scale to predict the target in the new frame.

[0029] A further technical solution lies in that: in the step 2, before inputting the spring image sequence when the circuit breaker operates into the correlation filtering algorithm DSST, the inter-frame difference method is used to search in the spring image sequence to obtain and remove redundant images.

[0030] A further technical solution lies in that: in the step 2, the gray value of the z-th frame image in the spring image sequence is subtracted from the gray value of the corresponding pixel point in the (z + 1)-th frame image and then the absolute value is taken to obtain a difference image, and then threshold binaryzation is performed. If the white pixel value after binaryzation exceeds the set threshold, then this frame image is determined to be a redundant image.

[0031] A further technical solution lies in that: in the step 2, the modal decomposition algorithm EMD is represented by formula (6),

[0032] y(t) = IMF1(t) + IMF2(t) + IMF3(t) + r3(t) (6)

[0033] In formula (6), y(t) is the displacement curve of the first turn of the spring in the spring expansion and contraction direction, IMF1(t) is the time-domain signal of the first intrinsic mode function IMF, IMF2(t) is the time-domain signal of the second intrinsic mode function IMF, IMF3(t) is the time-domain signal of the third intrinsic mode function IMF, and r3(t) is the time-domain signal of the residual component.

[0034] A further technical solution lies in that: in the step 2, the discrete Fourier transform is represented by formula (7),

[0035] C i (f) = ∫c i (t)e -j2πft dt (7)

[0036] In formula (7), C i (f) is the spectrum of the i-th IMF, and c i (t) is the time-domain signal of the i-th IMF.

[0037] A further technical solution lies in that: in step 3, the set comparison threshold is 3% - 7%.

[0038] The beneficial effects produced by adopting the above technical solution are as follows:

[0039] Through steps 1 to 3, the detection and evaluation efficiency of the high - voltage circuit breaker spring is high and the effect is good, which is reflected in the following aspects.

[0040] By using the displacement curve of the first - turn spring in the spring expansion and contraction direction, the displacement fluctuation in the non - expansion and contraction direction does not need to be considered, which simplifies the calculation amount.

[0041] By adopting the correlation filtering algorithm DSST, the dot - product calculation in the frequency domain is used to replace the complex time - domain convolution operation, greatly reducing the calculation amount, accelerating the tracking speed of the free end of the spring as the deformation target, and thus improving the efficiency.

[0042] Performing discrete Fourier transform on the third intrinsic mode function. Since only one intrinsic mode function is discretely Fourier - transformed, the algorithm is simplified, the calculation amount is reduced, and the efficiency is improved.

[0043] Since the third intrinsic mode function correspondingly reflects the spring damping displacement after removing the influence of most impact forces and longitudinal fluctuations, and can interpret the damping characteristics of the spring after being subjected to impact forces, therefore, detecting and evaluating the health state of the high - voltage circuit breaker spring based on the third intrinsic mode function is more scientific, more reasonable, and has better technical effects.

[0044] Comparing the frequency spectrum diagram of the third intrinsic mode function of the spring of the circuit breaker to be tested with the frequency spectrum diagram of the third intrinsic mode function of the normal circuit breaker spring. When the deviation of the frequency corresponding to the peak power ratio in the frequency spectrum diagram of the spring of the circuit breaker to be tested from the frequency corresponding to the peak power ratio in the frequency spectrum diagram of the normal circuit breaker spring exceeds the set comparison threshold, it is determined as a fault. The method of evaluation and judgment is simple, the calculation amount is reduced, and the efficiency is improved.

[0045] See the description in the specific implementation part for details. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is the flow chart of the present invention;

[0047] Figure 2 is the principle block diagram of the present invention;

[0048] Figure 3 is the flow chart for calculating the displacement time curve of the first turn of the spring based on the correlation filtering algorithm DSST;

[0049] Figure 4a is the response diagram of the position filter;

[0050] Figure 4b is the response diagram of the scale filter;

[0051] Figure 5a is the displacement-time curve diagram of the first turn of the closing spring;

[0052] Figure 5b is the velocity-time curve diagram of the closing spring;

[0053] Figure 6 is the comparison diagram of the displacement-time curves of the normal spring and the faulty spring;

[0054] Figure 7a is the curve diagram of the first eigenmode function of the spring of the normal high-voltage circuit breaker;

[0055] Figure 7b is the curve diagram of the second eigenmode function of the spring of the normal high-voltage circuit breaker;

[0056] Figure 7c is the curve diagram of the third eigenmode function of the spring of the normal high-voltage circuit breaker;

[0057] Figure 8a is the curve diagram of the first eigenmode function of the spring of the old and faulty high-voltage circuit breaker;

[0058] Figure 8b is the curve diagram of the second eigenmode function of the spring of the old and faulty high-voltage circuit breaker;

[0059] Figure 8c is the curve diagram of the third eigenmode function of the spring of the old and faulty high-voltage circuit breaker;

[0060] Figure 9a is the frequency spectrum diagram of the normal spring;

[0061] Figure 9b is the frequency spectrum diagram of the old and faulty spring. Detailed implementation manners

[0062] The circuit breaker spring oscillator is a typical mechanical structure, which consists of key components such as springs, inertial masses and damping devices. During operation, the vibration characteristics and response behaviors of the oscillator are affected by various factors, such as spring stiffness, mass distribution and damping device characteristics, etc. Nonlinear damping is particularly important, especially the damping part of the buffer spring. The spring displacement reflects the law of the change of the vibration amplitude of the oscillator over time, providing valuable information for analyzing its motion law and response characteristics. The potential energy stored in the spring of the operating mechanism provides the initial energy for the operation of the high-voltage circuit breaker. Spring fatigue or even fracture faults have caused many circuit breaker refusal-to-operate accidents. Scientifically evaluating the health state of the spring is the prerequisite for ensuring the reliable operation of the circuit breaker.

[0063] This application uses a high-speed camera to capture the instantaneous release process of the spring when the circuit breaker closes, and uses the discriminative scale space tracker (DSST) algorithm to extract the real-time motion information of the spring target. It solves the problem that the size of the energy storage spring changes rapidly during the stretching process and cannot be accurately tracked. The displacement of the instantaneous release motion of the spring is calculated through the DSST algorithm of the correlation filtering algorithm, and the modal spectrum analysis of the spring displacement is carried out. A reasonable feature is selected to judge the working state of the spring. The experimental results prove that the evaluation method after fusing the preferred signal features by using the damping motion in the free recovery stage of the spring for deformation-modal transformation effectively improves the analysis efficiency and accuracy of the spring health state.

[0064] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and in no way restricts the present application and its application or use. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0065] Many specific details are set forth in the following description in order to provide a thorough understanding of the present application, but the present application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present application. Therefore, the present application is not limited by the specific embodiments disclosed below.

[0066] Embodiment 1:

[0067] As Figure 1 shown, the present invention discloses a method for evaluating the health state of the deformation-modal of the spring for non-disassembly detection of high-voltage circuit breakers, including a preparation step and a detection and evaluation step, which are described in detail as follows.

[0068] The preparation step includes two steps: the first step is to obtain the spring image sequence during the normal operation of the circuit breaker in advance, and the second step is to extract the spectrogram of the third intrinsic mode function IMF3(t) from the spring image sequence. The details are as follows.

[0069] Step S1: Obtain the spring image sequence during the normal operation of the circuit breaker in advance.

[0070] Hardware device part:

[0071] As Figure 2As shown in the figure, the hardware device includes a computer, a camera, an LED lamp, and a controller. The computer is connected to the controller and communicates bidirectionally. The controller is electrically connected to the closing and opening coils of the high-voltage circuit breaker. The computer is electrically connected to the camera through a USB video data cable.

[0072] The camera is a high-speed camera, and the LED lamp is used for supplementary lighting during photography.

[0073] The computer synchronously triggers the controller and the camera. When the controller outputs the current of the closing and opening coils to start the action of the circuit breaker, it triggers the high-speed camera to start capturing the high-speed image sequence of the deformation of the energy storage spring. Thousands of frames of images can be obtained during one operation process and stored in the relational database of the computer for spring performance evaluation.

[0074] The operating performance of the high-voltage circuit breaker is closely related to the transient deformation of its spring. Defects such as spring relaxation, creep, and fracture cause changes in characteristic parameters such as the stretching time, stroke, pitch, and speed during the closing and opening processes. How to accurately measure the spring deformation is the key to scientifically evaluating its state and ensuring the reliable operation of the circuit breaker.

[0075] One end of the circuit breaker spring is fixed to one end of the operating mechanism, and the other end of the circuit breaker spring is fixed to the other end of the operating mechanism. When the circuit breaker operates, one end of the circuit breaker spring is the fixed end, and the other end of the spring is the free end.

[0076] Step S2: Process the spring image sequence during the normal operation of the circuit breaker through the correlation filtering algorithm DSST to obtain the displacement-time curve X(t) of the first turn of the spring. Extract the displacement curve y(t) of the first turn of the spring in the spring stretching direction from the displacement-time curve X(t) of the first turn of the spring. Decompose the displacement curve y(t) of the first turn of the spring in the spring stretching direction through the empirical mode decomposition algorithm EMD to obtain the first intrinsic mode function IMF1(t), the second intrinsic mode function IMF2(t), the third intrinsic mode function IMF3(t), and the residual component r3(t). Use the discrete Fourier transform to extract the frequency spectrum diagram of the third intrinsic mode function IMF3(t) from the third intrinsic mode function IMF3(t).

[0077] Intrinsic mode function, the horizontal axis is time, the vertical axis is amplitude, unit pixel. Frequency spectrum diagram, the horizontal axis is frequency, the vertical axis is power ratio.

[0078] The division of step S2 includes the following steps:

[0079] Step S201: Process the spring image sequence during the normal operation of the circuit breaker through the correlation filtering algorithm DSST to obtain the displacement-time curve X(t) of the first turn of the spring.

[0080] One turn of the spring on the side of the free end of the spring is the first turn of the spring. The joint where the first turn of the spring is fixedly connected to the operating mechanism is used as the marking point of the first turn of the spring. The displacement-time curve X(t) of the first turn of the spring is the curve of the displacement of the marking point of the first turn of the spring changing with time during the spring deformation process.

[0081] The correlation filtering algorithm DSST for the energy storage spring image target is a tracking algorithm, which is described in detail as follows.

[0082] During the energy release process of the circuit breaker energy storage spring, in different stages, the spring swings violently back and forth with the movement of the load-bearing moving parts. The background and illumination of the captured image target are different, and the angle deviation causes changes in the target scale. Therefore, the high-speed image sequence inter-frame target tracking algorithm can adapt to various complex situations. The accuracy and speed of pixel coordinate detection determine the accuracy of spring mechanical property evaluation.

[0083] As a correlation filter tracking algorithm with scale information, the correlation filtering algorithm DSST adapts to target scale changes by training a classifier separately, extracts the Histogram of Oriented Gradients (HOG) of the target direction to describe the contour of the target, and solves problems such as illumination changes, background movement, and geometric deformation.

[0084] The correlation filtering algorithm DSST designs two independent correlation filters, namely the position filter and the scale filter, to achieve target tracking and scale estimation respectively. Through Fourier transform, the dot product calculation in the frequency domain is used to replace the complex time-domain convolution operation, which greatly reduces the calculation amount and speeds up the tracking speed of the spring moving end as a deformation target.

[0085] The position filter is described in detail as follows.

[0086] To achieve the purpose that the output response value is the largest at the center position of the target and the output response value is smaller the farther away from the center position of the target, a filter h with a Gaussian output response to the response of the sample feature f is designed. By minimizing the cost function, the optimal filter is obtained:

[0087]

[0088] In Equation (1), d represents the d-dimensional feature vector, l represents a dimension of the d-dimensional feature, h is the filter with a Gaussian output response, * is the convolution, f is the sample feature, g is the Gaussian function, and λ is the regularization term coefficient.

[0089] The convolution operation in the time domain in Equation (1) is transformed into a dot product calculation in the frequency domain through discrete Fourier transform for solution, and we get:

[0090]

[0091] In Equation (2), is the conjugate transpose of f in the frequency domain, is the conjugate transpose of g in the frequency domain.

[0092] In the new frame of image, the position filter trained with the target information of the previous frame is used to determine the target position of the current frame, and the output response value of the filter is:

[0093]

[0094] In formula (3), y is the output response value of the filter, is the feature channel of the l-th dimension at time t, and the center position of the tracking target is determined according to the maximum value of the output response.

[0095] The scale filter is described in detail as follows.

[0096] The scale estimation process is basically the same as the target tracking process. The difference is that the scale estimation uses a one-dimensional filter, and the position tracking process uses a two-dimensional filter. The new position of the target in the next frame of image is determined by the position filter, but the target scale has not been updated at this time. Therefore, when performing scale estimation through the scale filter, with the current target position as the center and the non-linear exponential relationship of the target box in the previous frame as the candidate scale, the target sample scale selection principle is:

[0097]

[0098] In formula (4), P and R respectively represent the width and height of the target box in the previous frame of image; a is the scale factor; s is the number of scales. Finally, the target center position and the optimal target scale are combined to predict the target in the new frame.

[0099] The correlation filtering algorithm DSST itself is an existing technology and will not be elaborated here.

[0100] The calculation process of the spring deformation quantization is described in detail as follows.

[0101] Such as Figure 3As shown in the figure, it is a flowchart for calculating the displacement-time curve of the first turn of the spring based on the correlation filtering algorithm DSST. During the operation of the circuit breaker, the control coil core moves to the position to trigger the release of the detent. The deformation of the spring essentially detects the movement of the target position in the image. To avoid the time-consuming of redundant high-speed image sequence analysis before the action and after reaching the static state, the frame difference method is adopted in this application to quickly search for and eliminate redundant images. The gray value of the z-th frame image in the image sequence is subtracted from the gray value of the corresponding pixel point in the z+1-th frame image, and then the absolute value is taken to obtain the difference image, which is then thresholded and binarized. If the white pixel value after binarization exceeds the set threshold, it is determined that this frame is a redundant frame and is eliminated. Then, the image is preprocessed to reduce the interference of on-site light changes, electromagnetic interference, and equipment noise on the image sequence and improve the quality of the image. The position-time curve of the first turn of the spring is obtained by the correlation filtering algorithm DSST.

[0102] Step S202: Extract the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction from the displacement-time curve X(t) of the first turn of the spring.

[0103] During actual measurement, the spring is vertically distributed and expands or contracts in the Y-axis direction. Therefore, during actual analysis, the displacement fluctuation in the X-axis direction does not need to be considered, which simplifies the calculation amount.

[0104] Step S203: Decompose the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction by the empirical mode decomposition (EMD) algorithm to obtain the first intrinsic mode function IMF1(t), the second intrinsic mode function IMF2(t), the third intrinsic mode function IMF3(t), and the residual component r3(t).

[0105] The modal analysis of the spring displacement is described in detail as follows.

[0106] Analysis of the spring movement process:

[0107] The spring is responsible for providing the kinetic energy for closing and opening operations to drive the operating mechanism to perform actions. At this time, the spring load is all the moving parts of the mechanism, and the energy transfer process is affected by the mass of the parts themselves and the lubrication degree between the parts. When the closing action is nearly completed, the contacts collide and exert a huge impact force on the spring, and the spring limit device is activated. At this time, the spring only has stable loads such as buffers, and thus enters the free damping stage, and then the movement speed gradually decreases until it reaches a relatively static state. In addition, the spring can be simplified as a multi-degree-of-freedom spring oscillator subjected to an impact force, and it expands and contracts under the influence of non-linear damping. In the energy transfer, the impact force will excite longitudinal waves and transverse waves in the spring oscillator.

[0108] For such a spring system, its dynamic equation can be simplified and expressed in the following form:

[0109] M*X”(t)+C*X'(t)+K*X(t)=F(t)(5)

[0110] In Equation (5), M, C, and K are the mass matrix, damping matrix, and stiffness matrix respectively, X(t), X'(t), and X”(t) represent the displacement, velocity, and acceleration vectors respectively, and F(t) is the external impact force applied to the spring. After the limit device is triggered, the interaction between the spring device and the transmission mechanism can be regarded as a negligible micro quantity. In this case, the spring is only affected by the damping of the buffer spring, so the load is relatively simple and stable. This application precisely evaluates the health state of the spring based on this free damping stage.

[0111] Spring deformation - modal analysis:

[0112] Since the spring exhibits non - linear characteristics during the operation of the circuit breaker, the system response shows significant complexity, just like the transverse and longitudinal waves generated by adding an impact force in the sine oscillation function analysis. Traditional linear analysis methods have limited effects in revealing the complex characteristics of the system, while modern deep - learning - based analysis methods require a large amount of sample data and have poor model interpretability, making it difficult to achieve general analysis of different types of spring oscillators. To address these challenges, this application performs Empirical Mode Decomposition (EMD, for short) on the displacement curve y(t) of the first - turn spring in the Y - axis direction. First, the EMD process splits the displacement signal into a finite number of Intrinsic Mode Functions (IMFs, for short). These functions exhibit excellent local characteristics and can capture the changes in the instantaneous frequency of the signal within a specific range. Specifically, for the displacement curve y(t) of the first - turn spring in the Y - axis direction, the EMD method can be expressed as:

[0113] y(t) = IMF1(t) + IMF2(t) + IMF3(t) + r3(t) (6)

[0114] In Equation (6), IMF1(t) is the time - domain signal of the first Intrinsic Mode Function IMF, IMF2(t) is the time - domain signal of the second Intrinsic Mode Function IMF, IMF3(t) is the time - domain signal of the third Intrinsic Mode Function IMF, and r3(t) is the time - domain signal of the residual component.

[0115] Step S204: Use the discrete Fourier transform to extract the spectrogram of the third Intrinsic Mode Function IMF3(t) from the third Intrinsic Mode Function IMF3(t).

[0116] Spring displacement modal spectrum analysis is described in detail as follows.

[0117] EMD decomposition can capture the instantaneous changes of the spring-damping model, describe the vibration characteristics of the spring under different working conditions, and detect the impact points in the time series. By applying the discrete Fourier transform, the spectrum of each IMF, including information such as frequency and power ratio, can be further extracted, thus showing the dynamic characteristics of the spring system in the time-frequency domain.

[0118] IMF1(t) is the position where the first strong impact appears, revealing the instantaneous impact of the spring contacting the limit device. The impact phenomenon is closely related to the force exerted on the spring. After fatigue of the old spring, the impact force significantly decreases.

[0119] IMF2(t) mainly reflects the longitudinal wave excited by the impact force in the spring, revealing the distribution and transmission characteristics of the impact force in the spring structure. From a physical perspective, it represents the vibration mode excited inside the spring, and further can be understood as the mechanical response and propagation process of the spring.

[0120] IMF3(t) mainly describes the spring damping displacement after removing the influence of most of the impact force and longitudinal wave, and can interpret the damping characteristics of the spring after being subjected to the impact force, as well as the gradual dissipation over time.

[0121] To deeply explore the fault characteristics of the spring, the discrete Fourier transform is performed on IMF3(t) here. Since only one IMF is discretely Fourier-transformed, the algorithm is simplified, the calculation amount is reduced, and the efficiency is improved. Since IMF3(t) correspondingly reflects the spring damping displacement after removing the influence of most of the impact force and longitudinal wave and can interpret the damping characteristics of the spring after being subjected to the impact force, it is more scientific, more reasonable, and has better technical effects to detect and evaluate the health state of the high-voltage circuit breaker spring based on the third intrinsic mode function IMF3(t).

[0122] The discrete Fourier transform is specifically expressed as:

[0123] C i (f) = ∫c i (t)e -j2πft dt (7)

[0124] In Equation (7), C i (f) is the spectrum of the i-th IMF, and c i (t) is the time-domain signal of the i-th IMF.

[0125] By analyzing the power ratio and frequency characteristics after decomposition, the vibration behavior of the spring system under non-linear damping conditions can be described. For old springs that have been used for a long time, their physical properties change due to factors such as long-term use, fatigue damage, and material aging, thus affecting the output performance and health state. By decomposing the deformation displacement mode into specific responses and then Fourier-transforming to the frequency domain to characterize these changes of the spring.

[0126] Specifically, at the peak frequency, the power ratio and frequency of the old spring may be lower than those of the new spring. This is because the stiffness of the old spring decreases, resulting in a weakened response and a lower natural frequency at the peak frequency. That is, the damping characteristics of the old spring also change, which will affect the distribution of the power ratio. By analyzing the spectrograms under different working conditions, the fault modes and performance changes of the spring system are quantitatively evaluated, thus providing strong support for the design, diagnosis, and maintenance of the spring.

[0127] The detection and evaluation steps include three steps: the first step is to obtain the spring image sequence when the circuit breaker to be tested operates, the second step is to extract the spectrogram of the third intrinsic mode function IMF3(t) from the spring image sequence, and the third step is to judge the health status of the spring of the circuit breaker to be tested, which are detailed as follows.

[0128] Step S1’: Obtain the spring image sequence when the circuit breaker to be tested operates.

[0129] Step S1’ of the detection and evaluation steps is the same as Step S1 of the preparation steps. The difference is that Step S1 obtains the spring image sequence when the normal circuit breaker operates, and Step S1’ obtains the spring image sequence when the circuit breaker to be tested operates. The same points will not be elaborated.

[0130] Step S2’: Process the spring image sequence when the circuit breaker to be tested operates through the correlation filtering algorithm DSST to obtain the displacement-time curve X(t) of the first turn of the spring. Extract the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction from the displacement-time curve X(t) of the first turn of the spring. Decompose the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction through the empirical mode decomposition algorithm EMD to obtain the first intrinsic mode function IMF1(t), the second intrinsic mode function IMF2(t), the third intrinsic mode function IMF3(t), and the residual component r3(t). Use the discrete Fourier transform to extract the spectrogram of the third intrinsic mode function IMF3(t) from the third intrinsic mode function IMF3(t).

[0131] Step S2’ of the detection and evaluation steps is the same as Step S2 of the preparation steps. The difference is that Step S2 inputs the spring image sequence when the normal circuit breaker operates into Step S2 to obtain the spectrogram of the third intrinsic mode function IMF3(t) of the normal circuit breaker as the output, and Step S2’ inputs the spring image sequence when the circuit breaker to be tested operates into Step S2 to obtain the spectrogram of the third intrinsic mode function IMF3(t) of the circuit breaker to be tested as the output. The same points will not be elaborated.

[0132] Step S3: Compare the spectrogram of the third intrinsic mode function IMF3(t) of the spring of the circuit breaker to be measured with the spectrogram of the third intrinsic mode function IMF3(t) of the spring of a normal circuit breaker. If the deviation between the frequencies corresponding to the peak power ratios in the spectrogram of the spring of the circuit breaker to be measured and the spectrogram of the spring of a normal circuit breaker exceeds the set comparison threshold, it is determined as a fault.

[0133] The comparison threshold is set at 5%.

[0134] Embodiment 2:

[0135] The present invention discloses a method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker, including a preparation step and a detection and evaluation step. The same parts as those in Embodiment 1 will not be described in detail.

[0136] The comparison threshold is set at 3%.

[0137] Embodiment 3:

[0138] The present invention discloses a method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker, including a preparation step and a detection and evaluation step. The same parts as those in Embodiment 1 will not be described in detail.

[0139] The comparison threshold is set at 7%.

[0140] Relative to the above embodiments, the comparison threshold is 3% - 7%, which is set as needed.

[0141] Evaluation experiment and result analysis:

[0142] 1. Experimental process:

[0143] Taking the spring of the CT4 operating mechanism of the LW8 type 35kV outdoor SF6 circuit breaker as the test object, using the intelligent control device to control the opening and closing operations of the circuit breaker, and carrying out the experiment of evaluating the spring health state according to the above evaluation scheme. The high-speed camera for detecting the spring displacement is set at a shooting frame rate of 3500FPS. Adjust the position angle of the camera and the intensity of the auxiliary light source of the LED lamp. After photometric focusing, the image of the closing spring of the circuit breaker in the field of view is clear, and the camera is set to the hardware trigger mode. The control device outputs a current signal to start the closing of the circuit breaker, and at the same time triggers the high-speed camera to shoot the movement process of the energy storage spring.

[0144] In the experiment, first obtain the high-speed motion sequence images of the normal spring during closing, and then replace the normal spring with an old spring with stress relaxation failure for the closing test. The normal spring and the old spring are of the same model.

[0145] The normal and old springs are each tested multiple times. The high-speed camera records the image sequences of the tests in different health states, and obtains multiple groups of spring deformation samples for comparative experiments.

[0146] 2. Spring displacement tracking:

[0147] The lower end of the closing spring is fixed at the bottom of the mechanism. When calculating the deformation, the first turn of the spring's moving end is used as the tracking target. The DSST (Discriminative Scale Space Tracking) algorithm is used to identify and track the movement process of a specific target of the spring, and the pixel coordinates of the target centroid are obtained to draw the deformation-time curve.

[0148] As Figure 4a shown, it is the response map of the position filter. The X-axis represents the X direction, the Y-axis represents the Y direction, and the Z-axis represents the output response value.

[0149] As Figure 4b shown, it is the response map of the scale filter. The left marked box part is the first turn of the spring, and the data bar on the right represents the output response value of the scale filter.

[0150] As Figure 5a shown, it is the displacement-time curve of the first turn of the closing spring. The horizontal axis of the displacement curve is time, with the unit of ms, and the vertical axis is displacement, with the unit of pixel. The target center coordinates are determined according to the maximum response values output by the two filters for each frame of the image. By calculating the relative displacement between the moving target and the fixed background of the image, the deformation displacement-time curve of the first turn of the moving end of the spring can be drawn, and the derivative can be used to obtain the velocity-time curve. The vertical coordinate in the figure represents the vertical distance between the moving target in the image and the fixed reference point of the image. In the displacement-time curve, the target starts to move from point A, and at point C, the toggle arm drives the moving contact to close to the predetermined limit position. The movement of the spring is blocked and impacted. After reaching the maximum displacement point D, it undergoes several expansions and contractions during the free damping stage and finally stops at point E.

[0151] As Figure 5b shown, it is the velocity-time curve of the closing spring. The horizontal axis of the velocity curve is time, with the unit of ms, and the vertical axis is velocity, with the unit of pixel / ms. In the velocity-time curve, the target starts to move from point A, reaches the maximum velocity at point B, and at point C, the toggle arm drives the moving contact to close to the predetermined limit position. The movement of the spring is blocked and impacted, and the velocity drops instantaneously. After reaching the maximum displacement point D, the velocity drops instantaneously.

[0152] In this experiment, the closing of the circuit breaker is recorded as the 0 moment, and the energy storage spring instantaneously releases energy. After 25.5 ms, the spring reaches the maximum speed of 16.9 pixels / ms. At 40.4 ms, the moving and static contacts collide. There is a rigid connection between the toggle arm and the moving contact, and the deformation of the spring is blocked, resulting in an instantaneous decrease in the spring movement speed, and the closing process ends. After that, the spring starts to perform forced vibration affected by non-linear damping. Due to inertia, the spring is compressed to the limit position at 49.3 ms, and the maximum telescopic length can be calculated from the starting position of energy release. During the oscillation time of 567.2 ms, it goes through 12 telescopic cycles, and finally the spring reaches the equilibrium position at 384.4 pixels, and the deformation of the spring ends with the completion of the closing energy release process. Each experimental sample can analyze the deformation process of the spring, but relying solely on the criteria at each moment is not scientific because the process time of the on-site circuit breaker is also related to the switching current load and the mechanism lubrication condition.

[0153] 3. Deformation-modal characteristic evaluation of spring health status:

[0154] To avoid the influence of mechanism jamming and different loads on the spring health evaluation results, this application conducts deformation-modal analysis on the free damping process after the spring completes its output. That is, the characteristics are explored from the free relaxation stage after the spring is stretched to the maximum deformation to determine its state.

[0155] Take the displacement data from the maximum displacement point D to the stationary point E of the position-time curve, which essentially removes the variable load stage of spring energy release.

[0156] As Figure 6 shown, the comparison diagram of the displacement-time curves of normal springs and faulty springs. Comparing the displacement-time curves of normal and old springs, aligning the time scales at the maximum displacement point, it can be seen that the trajectories of normal springs and old springs no longer coincide.

[0157] As Figure 7a shown, it is the curve diagram of the first intrinsic mode function IMF1(t) of the spring of a normal high-voltage circuit breaker. The horizontal axis is time, in milliseconds, and the vertical axis is amplitude, in pixels.

[0158] As Figure 7b shown, it is the curve diagram of the second intrinsic mode function IMF2(t) of the spring of a normal high-voltage circuit breaker. The horizontal axis is time, in milliseconds, and the vertical axis is amplitude, in pixels.

[0159] As Figure 7c shown, it is the curve diagram of the third intrinsic mode function IMF3(t) of the spring of a normal high-voltage circuit breaker. The horizontal axis is time, in milliseconds, and the vertical axis is amplitude, in pixels.

[0160] As Figure 8aAs shown, it is a graph of the first intrinsic mode function IMF1(t) of the spring of an old and faulty high-voltage circuit breaker. The horizontal axis is time, with the unit of milliseconds, and the vertical axis is amplitude, with the unit of pixels. The first intrinsic mode function IMF1(t) is the position where the first strong impact appears, revealing the instantaneous impact of the spring contact limit device. The impact phenomenon is closely related to the force exerted on the spring. After the old spring is fatigued, the impact force significantly decreases.

[0161] As Figure 8b shown, it is a graph of the second intrinsic mode function IMF2(t) of the spring of an old and faulty high-voltage circuit breaker. The horizontal axis is time, with the unit of milliseconds, and the vertical axis is amplitude, with the unit of pixels. The second intrinsic mode function IMF2(t) mainly reflects the longitudinal wave excited by the impact force in the spring, revealing the distribution and transmission characteristics of the impact force in the spring structure. Analyzed from a physical perspective, it represents the vibration mode excited inside the spring, and further can be understood as the mechanical response and propagation process of the spring.

[0162] As Figure 8c shown, it is a graph of the third intrinsic mode function IMF3(t) of the spring of an old and faulty high-voltage circuit breaker. The horizontal axis is time, with the unit of milliseconds, and the vertical axis is amplitude, with the unit of pixels. The third intrinsic mode function IMF3(t) mainly describes the damping displacement of the spring after removing the influence of most of the impact force and longitudinal wave, and can interpret the damping characteristics of the spring after being subjected to the impact force, as well as gradually dissipating over time.

[0163] Scientifically evaluate the two situations of the curve, and use EMD to decompose the displacement curves of the normal spring and the old and faulty spring to obtain the main three IMF components.

[0164] Deeply explore the fault characteristics of the spring, and perform discrete Fourier transform on the third intrinsic mode function IMF3(t) here.

[0165] As Figure 9a shown, it is the spectrogram of the normal spring, that is, the transformation result. The horizontal axis is frequency, with the unit of kHz, and the vertical axis is the power ratio. The normal reaches the peak value of the power ratio of 6503 at a frequency of 0.21132 kHz.

[0166] As Figure 9b shown, it is the spectrogram of the old and faulty spring, that is, the transformation result. The horizontal axis is frequency, with the unit of kHz, and the vertical axis is the power ratio. The old and faulty spring reaches the peak value of the power ratio of 5285 at 0.20823 kHz.

[0167] By comparing and observing the frequencies and power ratios corresponding to the third intrinsic mode function IMF3(t), there are significant differences in the spectral distributions of normal springs and old and faulty springs throughout the entire frequency spectrum. The peak power ratio and peak frequency of the old and faulty springs are lower than those of the normal springs. The reasons for the decrease in the peak power ratio and peak frequency of the old and faulty springs are the relaxation and aging of the springs, which result in a decrease in their stiffness.

[0168] The modal interpretation of the spring deformation explains its own health condition. Using high-speed image recognition to detect the deformation to evaluate the state of the circuit breaker spring, the defect or fault criterion has a significant effect, solving the problem of defect detection of the on-site high-voltage circuit breaker spring under live and non-disassembly conditions.

[0169] Experimental induction:

[0170] Use a high-speed camera to capture the image sequence of the energy storage spring's telescopic deformation during the operation of the circuit breaker. Based on the correlation filtering algorithm DSST of HOG features, track the spring displacement to obtain the displacement-time curve that characterizes the dynamic information of the spring standard. Perform modal decomposition based on the frequency spectrum interval on the displacement-time curve to extract the damping section characteristics of the spring's free release, and then the health status evaluation of the circuit breaker's energy storage spring can be realized. The correlation filtering algorithm DSST improves the speed of image deformation detection and is not affected by the variable scale of the spring during the instant of energy release. Eliminate the influence of the complex variable-speed movement of the operating mechanism in the spring output section. Use the frequency-domain characteristics of the free damping section of the energy storage spring to form a feature vector. The discrimination of the spring performance takes into account the operation timing and component attributes, and also avoids the influence of the load and jamming problems of the mechanism components themselves. The experimental results show that the correlation filtering algorithm DSST is used for target tracking, which can quickly and quantitatively evaluate the spring's telescopic deformation. The modal characteristic spectrum of the displacement-time deformation accurately characterizes the health status of the spring, meeting the requirements of on-site disassembly-free inspection of the mechanical properties of key components of the circuit breaker.

[0171] Conclusion:

[0172] Evaluating the health status of the circuit breaker spring under non-disassembly conditions is one of the difficult problems in on-site operation and maintenance. Aiming at the blank of the lack of spring detection technology on site, this application deeply studies the image target tracking and modal feature extraction optimization algorithms to realize the non-contact detection of circuit breaker spring defects, and draws the following conclusions:

[0173] 1. A new method for evaluating the health status of high-voltage circuit breaker springs under non-disassembly conditions is proposed. Use a high-speed camera to capture the spring deformation image sequence during the opening and closing processes of the circuit breaker. Based on the correlation filtering algorithm DSST, track the image target to obtain the deformation curve, and then comprehensively evaluate the spring defect state based on the deformation-modal decomposition features.

[0174] 2. The DSST tracking and recognition algorithm based on HOG features is used to track and identify the position of the moving end of the spring, and a displacement-time curve depicting the dynamic information of the spring scale is obtained. This target tracking algorithm is fast and not affected by the size, shape, and grayscale changes of the energy storage spring's deformation action, meeting the requirements for on-site spring state detection.

[0175] 3. A method for constructing a spring state discrimination feature vector based on the frequency domain features of the free damping section is proposed. It takes into account the operation timing, component attributes, and load changes, avoiding the influence of the complex variable-speed motion of components and mechanism jamming in the spring output section, and improving the accuracy of spring health state evaluation.

[0176] Application prospect:

[0177] Based on the scientific evaluation of the health state through spring deformation-modal feature analysis, the technical problem of being unable to judge the existence of fatigue, creep, and defects in the spring without disassembly is solved, and it has broad application prospects in the field of on-site detection of circuit breaker states.

Claims

1. A method for non-dismantling detection of spring deformation-modal evaluation of the health state of high-voltage circuit breakers, characterized in that: It includes the following steps. Step 1: Obtain the spring image sequence when the circuit breaker operates. Step 2: Process the spring image sequence when the circuit breaker operates through the correlation filtering algorithm DSST to obtain the displacement-time curve of the first turn of the spring. Extract the displacement curve of the first turn of the spring in the spring expansion and contraction direction from the displacement-time curve of the first turn of the spring. Decompose it through the empirical mode decomposition algorithm EMD to obtain the first intrinsic mode function, the second intrinsic mode function, the third intrinsic mode function, and the residual component. Use the discrete Fourier transform to extract the frequency spectrum diagram of the third intrinsic mode function from the third intrinsic mode function. Step 3: Based on Step 1 and Step 2, obtain the frequency spectrum diagrams of the third intrinsic mode functions of the spring of the circuit breaker to be tested and the spring of the normal circuit breaker. When the deviation between the frequency corresponding to the peak power ratio in the frequency spectrum diagram of the spring of the circuit breaker to be tested and the frequency corresponding to the peak power ratio in the frequency spectrum diagram of the spring of the normal circuit breaker exceeds the set comparison threshold, it is determined as a fault.

2. The method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, characterized in that: In Step 1, a computer, a camera, an LED lamp, and a controller are used to obtain the spring image sequence when the circuit breaker operates. The computer is connected to the controller and communicates bidirectionally. The controller is electrically connected to the closing and opening coils of the high-voltage circuit breaker. The computer is electrically connected to the camera through a USB video data cable.

3. The method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, characterized in that: The specific division of Step 2 includes the following steps. Step 201: Process the spring image sequence when the circuit breaker operates through the correlation filtering algorithm DSST to obtain the displacement-time curve X(t) of the first turn of the spring. Step 202: Extract the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction from the displacement-time curve X(t) of the first turn of the spring. Step 203: Decompose the displacement curve y(t) of the first turn of the spring in the spring expansion and contraction direction through the empirical mode decomposition algorithm EMD to obtain the first intrinsic mode function IMF1(t), the second intrinsic mode function IMF2(t), the third intrinsic mode function IMF3(t), and the residual component r3(t). Step S204: Use the discrete Fourier transform to extract the frequency spectrum diagram of the third intrinsic mode function IMF3(t) from the third intrinsic mode function IMF3(t).

4. The method for non-dismantling detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, wherein: In Step 2, the position filter of the correlation filtering algorithm DSST is a two-dimensional filter and is used for position tracking. The scale filter of the correlation filtering algorithm DSST is a one-dimensional filter and is used for scale estimation. The position filter is represented by Equation (1). In Equation (1), d represents a d-dimensional feature vector, l represents a dimension of the d-dimensional feature, h is the filter of the Gaussian output response, * is convolution, f is the sample feature, g is the Gaussian function, and λ is the regularization term coefficient. Through the discrete Fourier transform, the convolution operation in the time domain in Equation (1) is converted into a point multiplication calculation in the frequency domain for solution, and the following is obtained: In formula (2), is the conjugate transpose of f in the frequency domain, is the conjugate transpose of g in the frequency domain; In the new frame of image, use the position filter trained with the target information of the previous frame to determine the current frame target position. The output response value of the position filter is: In Equation (3), y is the output response value of the filter, and Z t l is the feature channel of the l-th dimension at time t, and the center position of the tracking target is determined according to the maximum value of the output response.

5. The method for non-dismantling detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 4, wherein: In Step 2, when performing scale estimation through the scale filter, with the current target position as the center and the previous frame target box having a non-linear exponential relationship as the candidate scale, the target sample scale selection principle is: In formula (4), P and R respectively represent the width and height of the target box in the previous frame image; a is the scale factor; s is the number of scales; n is a constant between ; Combine the obtained target center position and the target optimal scale to predict the target of the new frame.

6. The method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, wherein: In step 2, before inputting the spring image sequence during the breaker operation into the correlation filtering algorithm DSST, the frame difference method is used to search in the spring image sequence to obtain and remove redundant images.

7. The method for non-dismantling detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 6, characterized in that: In step 2, the gray value of the z-th frame image in the spring image sequence is subtracted from the gray value of the corresponding pixel point of the (z + 1)-th frame image and then the absolute value is taken to obtain a difference image, which is then subjected to threshold binaryzation. If the white pixel value after binaryzation exceeds the set threshold, then this frame image is determined to be a redundant image.

8. The method for non-dismantling detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, characterized in that: In step 2, the modal decomposition algorithm EMD is expressed by Equation (6). y(t) = IMF1(t) + IMF2(t) + IMF3(t) + r3(t) (6) In Equation (6), y(t) is the displacement curve of the first turn of the spring in the spring expansion and contraction direction, IMF1(t) is the time-domain signal of the first intrinsic mode function IMF, IMF2(t) is the time-domain signal of the second intrinsic mode function IMF, IMF3(t) is the time-domain signal of the third intrinsic mode function IMF, and r3(t) is the time-domain signal of the residual component.

9. The method for non-dismantling detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, characterized in that: In step 2, the discrete Fourier transform is expressed by Equation (7). In formula (7), C i (f) is the spectrum of the i-th IMF, and c i (t) is the time-domain signal of the i-th IMF.

10. The method for non-disassembly detection of spring deformation-modal evaluation of the health state of a high-voltage circuit breaker according to claim 1, characterized in that: In step 3, the set comparison threshold is 3% - 7%.

Images