Early collision detection method and device for power equipment, electronic equipment and storage medium

Abstract

The invention discloses an early collision detection method and device for power equipment, electronic equipment and a storage medium, and the detection method comprises the steps: carrying out the full-period sampling processing of an initial signal collected in advance, and obtaining a monitoring matrix D; decomposing D into a left sub-matrix U, a right sub-matrix V and a diagonal sub-matrix Lambda; non-zero diagonal values in lambda are divided into three diagonal intervals: a large value area, a transition area and a small value area, and the large value area, the transition area and the small value area reflect a main vibration signal, a periodic collision signal and a noise signal in the initial signal respectively; determining a collision submatrix D2 according to D; periodic signals in the D2 are extracted, and an early collision signal graph is obtained. According to the method, the monitoring matrix is divided into a plurality of intervals through the non-zero diagonal values, and the early collision signals in the monitoring matrix are extracted based on the intervals, so that early weak collision signals between the power equipment can be extracted through the method, and early collision fault detection is guided to be carried out.

Description

Technical Field

[0001] This application relates to the technical field of equipment fault detection, and particularly relates to an early collision detection method, device, electronic device, and storage medium for power equipment. Background Art

[0002] Collision is the main cause of the shaft bending of large power equipment such as steam turbines, generators, and synchronous condensers. During the collision process, a temperature gradient will be generated on the shaft, causing thermal deformation. After the unbalance amount generated by the thermal deformation and the original unbalance amount on the rotor are combined, the vibration amplitude will become unstable, which is likely to lead to equipment failure.

[0003] Currently, vibration analysis methods are widely used in collision fault diagnosis. This method diagnoses collision faults based on the fluctuation, climb, and divergence characteristics of the vibration amplitude. However, this method can only detect the large shaft bending in the middle and late stages of the collision, and cannot detect weak collisions in the early stage. Summary of the Invention

[0004] Based on this, the purpose of this application is to provide an early collision detection method, device, electronic device, and storage medium for power equipment to detect periodic collision signals of power equipment in the early stage.

[0005] In a first aspect, an embodiment of this application provides an early collision detection method for power equipment. The detection method includes: performing full-cycle sampling processing on a pre-collected initial signal to obtain a monitoring matrix D, where the initial signal includes a main vibration signal, a periodic collision signal, and a noise signal; based on the matrix singular value decomposition method, decomposing the monitoring matrix D into a left sub-matrix U, a right sub-matrix V, and a diagonal sub-matrix Λ; where the diagonal sub-matrix Λ includes multiple non-zero diagonal values, and the main vibration signal includes multiple harmonics; any harmonic in the main vibration signal corresponds to 2 non-zero diagonal values in the diagonal sub-matrix Λ; dividing the non-zero diagonal values in the diagonal sub-matrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area, where the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal in the initial signal; based on the diagonal intervals and the decomposed left sub-matrix U, right sub-matrix V, and diagonal sub-matrix Λ, determining a collision sub-matrix D2 according to the monitoring matrix D; extracting the periodic signal in the collision sub-matrix D2 to obtain an early collision signal diagram.

[0006] Further, the step of performing full-cycle sampling processing on the pre-collected initial signal to obtain a monitoring matrix D includes: within a preset time range, full-cycle collecting the signal waveform corresponding to the initial signal; continuously intercepting k segments of signals with a sample length of n at equal lengths to obtain a monitoring matrix D, where each sample includes N sampling points.

[0007] Further, for the step of decomposing the above monitoring matrix D into a left sub - matrix U, a right sub - matrix V, and a diagonal sub - matrix Λ based on the matrix singular value decomposition method: Based on the matrix singular value decomposition method, the expression of the above monitoring matrix D is: D = U×V×Λ T , where T is the transpose symbol.

[0008] Further, for the step of dividing the non - zero diagonal values in the above diagonal sub - matrix Λ into three diagonal intervals: a large - value region, a transition region, and a small - value region: Based on a preset partitioning principle, the non - zero diagonal values in the above diagonal sub - matrix Λ are divided into three diagonal intervals: a large - value region, a transition region, and a small - value region; where the above partitioning principle is obtained based on the fact that the large - value region, the transition region, and the small - value region respectively reflect the main vibration signal, the periodic collision signal, and the noise signal of the initial signal.

[0009] Further, for the step of determining the collision sub - matrix D2 according to the above monitoring matrix D based on the above diagonal intervals and the decomposed left sub - matrix U, right sub - matrix V, and diagonal sub - matrix Λ: The expression of the above monitoring matrix D is: D = D1 + D2 + D3; The expression of the main vibration sub - matrix D1 is: The expression of the collision sub - matrix D2 is: The expression of the noise sub - matrix D3 is: where, σ i , σ j , σ k are respectively the non - zero diagonal values in the diagonal sub - matrix Λ of the large - value region, the transition region, and the small - value region; u i , u j , u k are respectively the matrix values in the left sub - matrix U of the large - value region, the transition region, and the small - value region; v i , v j , v k are respectively the matrix values in the right sub - matrix V of the large - value region, the transition region, and the small - value region; m, J, K are respectively the number of harmonics in the main vibration sub - matrix D1, the collision sub - matrix D2, and the noise sub - matrix D3; T is the transpose symbol.

[0010] Further, for the step of extracting the periodic signal in the above collision sub - matrix D2 to obtain an early - stage collision signal diagram: Retain the non - zero diagonal values in the above transition region, set the non - zero diagonal values in the above large - value region and the above small - value region to zero, and extract the periodic signal to obtain an early - stage collision signal diagram.

[0011] Further, the numerical value of the above non - zero diagonal value only depends on the amplitude of the harmonics in the above initial signal.

[0012] Further, for each harmonic in the main vibration signal of the above initial signal, the non-zero diagonal values in the corresponding diagonal sub-matrix Λ and their sorting only depend on the amplitude of the corresponding harmonic.

[0013] In a second aspect, an embodiment of the present application provides an early collision detection device for a power device. The detection device includes: a first detection module for performing full-cycle sampling processing on a pre-acquired initial signal to obtain a monitoring matrix D, where the initial signal includes a main vibration signal, a periodic collision signal, and a noise signal; a second detection module for decomposing the monitoring matrix D into a left sub-matrix U, a right sub-matrix V, and a diagonal sub-matrix Λ based on the matrix singular value decomposition method; where the diagonal sub-matrix Λ includes multiple non-zero diagonal values, and the main vibration signal includes multiple harmonics; any harmonic in the main vibration signal corresponds to 2 non-zero diagonal values in the diagonal sub-matrix Λ; a third detection module for dividing the non-zero diagonal values in the diagonal sub-matrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area, where the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal in the initial signal; a fourth detection module for determining a collision sub-matrix D2 based on the monitoring matrix D according to the diagonal interval and the decomposed left sub-matrix U, right sub-matrix V, and diagonal sub-matrix Λ; a fifth detection module for extracting the periodic signal in the collision sub-matrix D2 to obtain an early collision signal diagram.

[0014] Further, the first detection module is further configured to: within a preset time range, collect the signal waveform corresponding to the initial signal in a full cycle; continuously intercept k segments of signals with a sample length of n at equal lengths to obtain a monitoring matrix D, where each sample includes N sampling points.

[0015] Further, the second detection module is further configured to: based on the matrix singular value decomposition method, the expression of the monitoring matrix D is: D = U × V × Λ T , where T is the transpose symbol.

[0016] Further, the second detection module is further configured to: based on a preset partitioning principle, divide the non-zero diagonal values in the diagonal sub-matrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area; where the partitioning principle is based on the fact that the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal of the initial signal.

[0017] Further, the fourth detection module is further configured to: the expression of the monitoring matrix D is: D = D1 + D2 + D3; the expression of the main vibration sub-matrix D1 is: The expression of the collision sub-matrix D2 is: The expression of the noise sub-matrix D3 is: Among them, σ i 、σ j 、σ k are the non-zero diagonal values in the diagonal sub-matrix Λ of the large value region, the transition region, and the small value region respectively; u i 、u j 、u k are the matrix values in the left sub-matrix U of the large value region, the transition region, and the small value region respectively; v i 、v j 、v k are the matrix values in the right sub-matrix V of the large value region, the transition region, and the small value region respectively; m, J, and K are the number of harmonics in the main vibration sub-matrix D1, the collision sub-matrix D2, and the noise sub-matrix D3 respectively; T is the transpose symbol.

[0018] Further, the fifth detection module is further configured to: retain the non-zero diagonal values of the above transition region, set the non-zero diagonal values of the above large value region and the above small value region to zero, extract the periodic signal, and obtain the early collision signal diagram.

[0019] Further, the value of the above non-zero diagonal value only depends on the amplitude of the harmonics in the above initial signal.

[0020] Further, the non-zero diagonal value and its sorting in the diagonal sub-matrix Λ corresponding to each harmonic in the main vibration signal in the above initial signal only depend on the amplitude of the corresponding harmonic.

[0021] In a third aspect, an embodiment of the present application provides an electronic device, including: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to execute the above-mentioned early collision detection method for power equipment.

[0022] In a fourth aspect, an embodiment of the present application provides a storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the above-mentioned early collision detection method for power equipment.

[0023] The beneficial effects of the embodiments of the present application are as follows:

[0024] The present application discloses an early collision detection method, device, electronic device, and storage medium for power equipment. The detection method includes: performing full-cycle sampling processing on a pre-collected initial signal to obtain a monitoring matrix D, where the initial signal includes a main vibration signal, a periodic collision signal, and a noise signal; based on the matrix singular value decomposition method, decomposing the monitoring matrix D into a left sub-matrix U, a right sub-matrix V, and a diagonal sub-matrix Λ; where the diagonal sub-matrix Λ includes multiple non-zero diagonal values, and the main vibration signal includes multiple harmonics; any harmonic in the main vibration signal corresponds to 2 non-zero diagonal values in the diagonal sub-matrix Λ; dividing the non-zero diagonal values in the diagonal sub-matrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area, where the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal in the initial signal; based on the diagonal intervals and the decomposed left sub-matrix U, right sub-matrix V, and diagonal sub-matrix Λ, determining a collision sub-matrix D2 according to the monitoring matrix D; extracting the periodic signal in the collision sub-matrix D2 to obtain an early collision signal diagram. The present application divides the monitoring matrix into multiple intervals through non-zero diagonal values and extracts the early collision signal in the monitoring matrix based on this. Through this method, weak early collision signals between power equipment can be extracted, thereby guiding the development of early collision fault detection. Description of the Drawings

[0025] Figure 1 It is a flowchart of an early collision detection method for power equipment;

[0026] Figure 2 It is a schematic diagram of a static-dynamic collision;

[0027] Figure 3 It is a schematic diagram of the measured collision force signal of a certain test bench in the state of static-dynamic component collision;

[0028] Figure 4 It is a vibration waveform diagram collected with collision;

[0029] Figure 5 It is a frequency spectrum diagram of the vibration signal with collision;

[0030] Figure 6 It is a schematic diagram of the signal waveform, frequency spectrum, and diagonal value distribution of a signal with 3 harmonic components;

[0031] Figure 7 It is a schematic diagram of the waveform, frequency spectrum, and diagonal value distribution of a periodic collision signal;

[0032] Figure 8 It is a schematic diagram of the waveform, frequency spectrum, and diagonal value distribution of a noise signal;

[0033] Figure 9Schematic diagram of the diagonal value partition of the vibration signal monitoring matrix containing collisions;

[0034] Figure 10 Schematic diagram of the reconstructed periodic collision signal;

[0035] Figure 11 Schematic diagram of the reconstructed harmonic signal;

[0036] Figure 12 Schematic diagram of the reconstructed noise signal;

[0037] Figure 13 Schematic diagram of an electronic device. Specific implementation manners

[0038] The present application will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0039] Collision is a common fault of large-scale power equipment such as steam turbines, generators, and synchronous condensers. Such large-scale power equipment all belongs to rotating machinery and is composed of rotating components and stationary components. In order to save energy and reduce leakage, the gap between the moving and stationary components is designed to be small. During rotation, collisions between the moving and stationary components may occur due to reasons such as uneven moving and stationary clearances, and deviation of the cylinder and stator during operation. In addition to damaging the moving and stationary components, collisions will also generate a temperature gradient on the cross-section of the rotating shaft, causing thermal deformation of the rotor. When the deformation amount increases to a certain extent, the elastic deformation turns into plastic deformation, resulting in bending of the large shaft. Collision is a common fault of large-scale power equipment such as steam turbines, generators, and synchronous condensers. Such large-scale power equipment all belongs to rotating machinery and is composed of rotating components and stationary components. In order to save energy and reduce leakage, the gap between the moving and stationary components is designed to be small. During rotation, collisions between the moving and stationary components may occur due to reasons such as uneven moving and stationary clearances, and deviation of the cylinder and stator during operation. In addition to damaging the moving and stationary components, collisions will also generate a temperature gradient on the cross-section of the rotating shaft, causing thermal deformation of the rotor. When the deformation amount increases to a certain extent, the elastic deformation turns into plastic deformation, resulting in bending of the large shaft.

[0040] Statistical data shows that collision is the main cause of the bending of the rotating shaft of large-scale power equipment. According to the vibration climbing or divergence trend of power equipment, collision faults can be diagnosed. However, in the existing technology, the collision faults have entered the middle and late stages, and there is a possibility of triggering malignant accidents such as large shaft bending and cracking of the bearing bush babbit.

[0041] The existing collision detection methods include the following:

[0042] (1) Acoustic emission detection method. When metal components collide during high-speed rotation, abnormal sounds will be generated. According to the acoustic emission characteristics monitored by sensors, collision faults can be detected. The acoustic emission characteristics include: acoustic emission intensity, acoustic emission frequency, etc. Currently, the main research content focuses on how to eliminate the influence of noise in the detection signals of acoustic emission sensors.

[0043] (2) Vibration analysis method. The vibration analysis method is widely used in collision fault diagnosis. In the collision state, a temperature gradient will be generated on the rotating shaft, causing thermal deformation. After the unbalance amount generated by the thermal deformation and the original unbalance amount on the rotor are combined, the vibration amplitude will become unstable. According to the fluctuation, climb, and divergence characteristics of the vibration amplitude, the collision fault can be diagnosed. However, when the above vibration phenomena occur, the rotating shaft has already undergone large thermal deformation, and the collision fault has entered the middle and late stages, with a high risk of large shaft bending fault.

[0044] (3) Collision force detection method. In the collision state, the stator components will be subjected to the collision action of the collision force, including: radial force and tangential force, and this collision force has the characteristics of periodicity and discontinuity. Install force sensors on the stator components such as the back of the bearing or the cylinder to measure the exciting force received on the rotating shaft and the stator components in the collision state. According to the discontinuity and collision characteristics of the collision force, the collision fault can be analyzed.

[0045] Disadvantages of the prior art:

[0046] (1) Most rotating machinery works in a strong noise environment, and the abnormal sound generated by the collision is relatively weak and masked by the environmental noise. When using the acoustic emission method to diagnose the collision fault, the misdiagnosis rate and missed diagnosis rate are relatively high.

[0047] (2) Vibration analysis method. According to the fluctuation, climb, and divergence characteristics of the vibration amplitude, the collision fault can be diagnosed. However, when the above vibration phenomena occur, the rotating shaft has already undergone large thermal deformation, and the collision fault has entered the middle and late stages, with a high risk of large shaft bending fault. This method is not very suitable for fault detection in the early collision state.

[0048] (3) Collision force detection method. The collision force is relatively weak. After being transmitted outward through components such as the machine shell, the collision force is further attenuated. Monitoring the internal collision force at external shell parts and other locations has limited sensitivity. Due to various factors such as the structure, it is also very difficult to arrange force sensors at the collision location.

[0049] This application is applied to the scenario of collision detection for mechanical equipment.

[0050] In this embodiment, as Figure 1 shown, it is a flowchart of an early collision detection method for power equipment.

[0051] Before starting the operation, first explain the basic starting point of this application:

[0052] When the vibration amplitude of the rotating shaft of the power equipment exceeds the static-dynamic clearance, as Figure 2 (schematic diagram of static-dynamic collision) shown, a collision will occur between the static and dynamic components. After a collision fault occurs during the rotation state, the collision cross-section is subjected to a positive pressure f nand the tangential collision force f t of the collision effect.

[0053] For large power equipment, the static-dynamic clearance is relatively small near the high vibration point, and the collision is the most serious. Therefore, the collision fault is mainly manifested as local collision. The collision force does not exist throughout the entire rotation cycle but has strong periodic collision characteristics. Figure 3 Figure 7 shows the schematic diagram of the measured collision force signal of a certain test bench under the collision state of the static and dynamic components.

[0054] Under the periodic collision action of the collision force, the vibration signal will also show periodic collision characteristics, which can be expressed as:

[0055]

[0056] where Δy(t) is the periodic collision signal, x p (t) is the single and real-time collision signal, A 5 is the preset collision amplitude, σ is the attenuation coefficient, ω 2 is the frequency excited by the collision, T i is the i-th collision moment.

[0057] The vibration signals of power equipment such as steam turbine generator sets and synchronous condensers mainly include components such as half of the rotational frequency, 1 times the rotational frequency, 2 times the rotational frequency, collision, and noise. Without loss of generality, the vibration signal y(t) can be expressed as follows:

[0058]

[0059] where A 1 -A 4 are the amplitudes of the half-frequency component, 1 times frequency component, 2 times frequency component, and noise component respectively; ω is the rotational frequency, R(t) is a random function, A 4 R(t) is the vibration signal caused by noise; Δy(t) is the vibration signal caused by collision.

[0060] Theoretically, if the characteristics of the early periodic collision signal can be detected from the vibration signal in Equation (2), the early collision fault can be diagnosed. The purpose of this embodiment is to illustrate how to extract the characteristics of the early periodic collision signal from the collected vibration data.

[0061] In the prior art, by performing waveform analysis and spectrum analysis calculation processing on the originally collected vibration data including the collision signal (such as the data given in Equation (2)), the Figure 4 shown waveform and Figure 5 shown spectrum diagram can be obtained, and the characteristics of the early periodic collision signal cannot be reflected from Figure 4 and Figure 5 ​

[0062] In the early stage of the collision, although the rotating shaft is subjected to the impact of the collision force, there is no thermal deformation on the rotating shaft at this time, and there are no fluctuations and divergences in the vibration yet, but the rotating shaft has already been subjected to the action of the collision force. In summary, the Figure 4 and Figure 5 both cannot extract the characteristics of the early periodic collision signal. In the early stage of the fault, the amplitude and duration of the periodic collision detected by this method can be used to monitor and diagnose the early collision fault, extract the early periodic collision signal, better detect the power equipment, and improve the diagnostic accuracy of the early collision fault. The present application provides the following method.

[0063] The detection method of the embodiment of the present application includes:

[0064] S101: Perform full-cycle sampling processing on the pre-collected initial signal to obtain a monitoring matrix D, where the initial signal includes a main vibration signal, a periodic collision signal, and a noise signal.

[0065] S101 includes: within a preset time range, collect the signal waveform corresponding to the initial signal in full cycle; continuously intercept k segments of signals with a sample length of n at equal length to obtain a monitoring matrix D, where each sample includes N sampling points.

[0066] Specifically, here the collected signal sequence is continuously intercepted in several segments at equal length to obtain a monitoring matrix D.

[0067] Before collecting the initial signal, vibration sensors should be arranged at the parts of the power equipment where collisions may occur first. The vibration sensors include eddy current displacement type, magnetoelectric velocity type, and piezoelectric acceleration type. In order to efficiently monitor the collision fault, the piezoelectric acceleration type sensor is preferably selected to test the vibration acceleration signal of the collision part. Connect the output signal of the vibration sensor to the dynamic signal analyzer and continuously collect the vibration signal for a period of time.

[0068] The time series of the vibration signal continuously collected by the dynamic signal analyzer for a period of time is:

[0069] x = [x 1 x 2 …x N Formula (3);

[0070] In the formula, N is the total number of sample points, and x is the set of vibration data collected at equal time intervals, that is, Figure 4 the vibration waveform diagram in.

[0071] During the vibration test, the full-cycle sampling method is adopted, and the number of sampling points N is taken as a power of 2. For example, the full-cycle sampling number and the number of sampling points can be set to 8 cycles and 1024 points respectively.

[0072] As Figure 4 shown, take a positive integer n, and continuously intercept k segments of the time series x i at equal length n to obtain the monitoring matrix D:

[0073]

[0074] Specifically, k samples are collected, each sample has a length of n, and each sample has N sampling points.

[0075] To maximize the dimension q = min(k, n) of the monitoring matrix, k = N / 2 + 1 and n = N / 2 can be taken.

[0076] S102: Based on the matrix singular value decomposition method, decompose the monitoring matrix D into a left sub-matrix U, a right sub-matrix V, and a diagonal sub-matrix Λ; among them, the diagonal sub-matrix Λ includes multiple non-zero diagonal values, and the main vibration signal includes multiple harmonics; any harmonic in the main vibration signal of the initial signal corresponds to 2 non-zero diagonal values in the diagonal sub-matrix Λ.

[0077] Specifically, for example, if the number of harmonics collected in the main vibration signal of the initial signal is A, then the diagonal sub-matrix Λ will include 2A non-zero diagonal values.

[0078] Specifically, the monitoring matrix D can be decomposed into the product of the left sub-matrix U, the right sub-matrix V, and the transpose of the diagonal sub-matrix Λ.

[0079] It is found through theoretical proof in this application that when the number of sampling points N of the initial signal is fixed, the non-zero diagonal values and their distributions depend on the amplitudes of the harmonic components of the main vibration signal in the initial signal. Any harmonic in the initial signal corresponds to 2 adjacent non-zero diagonal values in the diagonal sub-matrix Λ. For the specific demonstration process, see Formulas (6)-(10).

[0080] Specifically, the monitoring matrix D is a real matrix of k×n, and D can be decomposed as follows:

[0081]

[0082] where R is a real matrix, I is an identity matrix, and diag represents a diagonal matrix; the diagonal elements satisfy σ 1 ≥σ 2 ≥…≥σ p >0. The monitoring matrix D includes multiple harmonics. The diagonal sub-matrix Λ of the monitoring matrix D includes a series of non-zero diagonal values. Every 2 adjacent non-zero diagonal values form a group. Specifically, each harmonic component in the vibration signal will generate and only generate two non-zero diagonal values in the diagonal matrix Λ; through derivation, when the number of sampling points N is fixed, these two non-zero diagonal values σ l ,σ l+1Its size and sorting are completely determined by the harmonic component amplitude aj, where aj includes A1 - A5 in formulas (1) - (2).

[0083] More specifically, the derivation process is as follows:

[0084] 2 non - zero diagonal values σ l , σ l+1 Satisfy the following relationship:

[0085]

[0086] In the formula, ω j Is the harmonic component (including the rotational frequency ω), f s Is the sampling frequency, Is the phase of the harmonic component ω j , k and n are the coefficients before the harmonic fundamental frequency, and q and p are defined by k and n.

[0087] If the sampling frequency f s Is taken as M times the rotational frequency ω, at this time,

[0088]

[0089] When N is an even multiple of M,

[0090]

[0091] Furthermore, the difference between two adjacent non - zero diagonal values is equal to 0.5 times the harmonic component amplitude aj, that is:

[0092] σ l - σ l+1 = a j / 2 formula (9);

[0093] From formulas (8) and (9), it can be further obtained that:

[0094]

[0095] Through the above analysis, it can be seen that when the number of sampling points N is fixed, the diagonal element values in the diagonal matrix Λ corresponding to the harmonic component ω j Are only related to the harmonic component amplitude aj of this harmonic component, and the sorting of non - zero diagonal values is completely determined by the harmonic component amplitude aj in the signal. The above derivation process is one of the innovation points of this application.

[0096] Specifically, Figure 6 Is a schematic diagram of the signal waveform, spectrum, and diagonal value distribution with 3 harmonic components (i.e., the main vibration signal). The spectrum of the signal is a discrete spectrum, and the non - zero diagonal values are relatively large, existing in a very narrow interval in the first half.

[0097] Specifically,Figure 7 It is a schematic diagram of the waveform, spectrum, and diagonal value distribution of the periodic collision signal. The spectrum of the periodic collision signal is a discrete spectrum, and the interval of the discrete spectrum is the rotation frequency. The diagonal values widely exist in the first half interval, gradually decreasing, but there is no cliff-like decrease.

[0098] Specifically, Figure 8 It is a schematic diagram of the waveform, spectrum, and diagonal value distribution of the noise signal. The spectrum of the noise signal is a continuous spectrum, the diagonal values are small, exist in a very wide interval, and the amplitude attenuation is small.

[0099] Figures 6 - 8 It shows that the singular value distribution characteristics obtained by singular value decomposition of vibration waveforms of different types or different characteristics are different, which provides a theoretical basis and support for reconstructing the collision signal using the singular values in the middle region later.

[0100] S103: Divide the non-zero diagonal values in the diagonal submatrix Λ into three diagonal intervals: the large value area, the transition area, and the small value area.

[0101] S103 includes: Based on a preset partitioning principle, divide the non-zero diagonal values in the diagonal submatrix Λ into three diagonal intervals: the large value area, the transition area, and the small value area; among them, the partitioning principle is obtained based on the fact that the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal of the initial signal.

[0102] Specifically, perform a fast Fourier transform on the Figure 4 waveform diagram to extract the corresponding spectrogram, as shown in Figure 5 . Figure 5 It is mainly manifested as the first three harmonic frequencies.

[0103] According to the vibration spectrogram ( Figure 5 ), select the number of non-zero diagonal values corresponding to the harmonic components. In Figure 5 , the number of harmonic components can be selected as 3, and each harmonic component corresponds to 2 non-zero diagonal values. Therefore, the number of diagonals of the corresponding harmonic components can be taken as 6. According to the characteristics of the periodic collision signal, select the non-zero diagonal values distributed in the 7th and a subsequent interval, which is set as [7, N / 8] in this example; according to the characteristics of the noise signal, select the non-zero diagonal values distributed in the second half interval, which is set as [N / 8 + 1, N / 2] in this example. The above selected 6, 7, 8 are the so-called "partitioning principle", and the determination of the partitioning principle is also related to the signal type.

[0104] Taking the "partitioning principle" as the benchmark to divide the interval, the non-zero diagonal values in the diagonal submatrix Λ can be divided into three diagonal intervals: the large value area, the transition area, and the small value area, as shown in Figure 9 . Figure 9 The singular values in

[0105] S104: Based on the left sub-matrix U, right sub-matrix V, and diagonal sub-matrix Λ obtained from the decomposition of the diagonal interval, determine the collision sub-matrix D2 according to the monitoring matrix D.

[0106] S104 includes: According to the non-zero diagonal value distribution characteristics, the monitoring matrix D can be expressed as the superposition of three sub-matrices, and the expression is: D = D1 + D2 + D3;

[0107] The expression of the main vibration sub-matrix D1 is:

[0108] The expression of the collision sub-matrix D2 is:

[0109] The expression of the noise sub-matrix D3 is:

[0110] Among them, σ i , σ j , σ k are the non-zero diagonal values in the diagonal sub-matrix Λ of the large value area, transition area, and small value area respectively; u i , u j , u k are the matrix values in the left sub-matrix U of the large value area, transition area, and small value area respectively; v i , v j , v k are the matrix values in the right sub-matrix V of the large value area, transition area, and small value area respectively; m, J, and K are the number of harmonics in the main vibration sub-matrix D1, collision sub-matrix D2, and noise sub-matrix D3 respectively; T is the transpose symbol.

[0111] S105: Extract the periodic signal in the collision sub-matrix D2 to obtain the early collision signal diagram.

[0112] Specifically, extract the element values of the first row and the first column in the collision sub-matrix D2, plot the periodic collision signal, and obtain the early collision signal diagram, that is Figure 10 .

[0113] S105 is also a process of signal reconstruction. S105 includes: Retain the non-zero diagonal values in the transition area, set the non-zero diagonal values in the large value area and the small value area to zero, extract the periodic signal, and obtain the early collision signal diagram.

[0114] Specifically, retain the large non-zero diagonal values in the front section, set the non-zero diagonal values in the transition section and the small interval section to zero, that is, D = D1, and the harmonic signal can be extracted, that is, obtain Figure 11 .

[0115] Retain the non - zero diagonal values in the transition section, set the non - zero diagonal values in the large and small interval sections to zero, that is, D = D2. Extract the element values of the first row and the first column in the collision sub - matrix D2 based on Formula 12, and draw the periodic collision signal, then the periodic fault collision signal can be extracted, that is Figure 10 。

[0116] Similarly, retain the small non - zero diagonal values in the latter section, set the non - zero diagonal values in the transition section and the former section to zero, that is, D = D3, and the noise signal can be extracted, that is Figure 12 。

[0117] The signals reconstructed from the non - zero diagonal values in the three different sections respectively reflect the main harmonic components (i.e., the main vibration components), periodic collisions and noise signals in the vibration signal. The main purpose of this application is to obtain Figure 10 the periodic fault collision signal diagram of

[0118] The beneficial effects of this embodiment are as follows:

[0119] In the early stage of the collision, the collision force acts on the rotating shaft, but at this time, there is no thermal deformation on the rotating shaft, and the vibration has not fluctuated and diverged, but the rotating shaft has already been affected by the collision force. From the reconstructed collision diagram obtained after the processing of this application (such as Figure 10 ), the characteristics of early periodic collisions can be clearly seen. According to the amplitude and duration of the periodic collisions detected by this method, the early collision faults can be monitored and diagnosed, improving the diagnostic accuracy of early collision faults. This method can extract the weak periodic collision signals (i.e., Figure 10 ) hidden in the harmonic and noise signals from the measured vibration signals, and carry out early collision fault detection based on this, improving the detection ability of collision faults in the early state. This method is based on the signals obtained by vibration sensors and does not require additional sensors such as sound and force sensors.

[0120] In this application, the fault signal is regarded as being composed of several main harmonic components, fault collision signals, and noise signals. According to the measured vibration signal time series, a monitoring matrix is obtained, and the monitoring matrix is decomposed into the product of three sub-matrices: a left matrix, a right matrix, and a diagonal matrix. Under the conditions of full-cycle sampling and the number of sampling points being an even multiple of the highest sampling frequency, each harmonic component can be represented by 2 non-zero diagonal values. The fault characteristics reflected by the non-zero diagonal values in different intervals are different. The amplitude of the harmonic component is usually large, corresponding to large non-zero diagonal values, and can be represented by several main non-zero diagonal values in the first half of the non-zero diagonal value distribution space; the non-zero diagonal values of the collision signal gradually decay and can be represented by several non-zero diagonal values in the transition section of the non-zero diagonal value distribution space; the non-zero diagonal values of the noise signal are small and widely exist in the non-zero diagonal value space, and can be represented by several small non-zero diagonal values in the second half of the non-zero diagonal value distribution space. Based on formulas (11)-(13), several main non-zero diagonal values in the first half (i.e., the large value area) and the non-zero diagonal values in the second half (i.e., the small value area) are set to zero. In the reconstructed fault signal (i.e., Figure 10 ), early collision signals can be detected more obviously, thereby guiding the early collision fault detection.

[0121] This embodiment provides an early collision detection device for power equipment. The detection device includes:

[0122] A first detection module for performing full-cycle sampling processing on a pre-collected initial signal to obtain a monitoring matrix D, where the initial signal includes a main vibration signal, a periodic collision signal, and a noise signal; a second detection module for decomposing the monitoring matrix D into a left sub-matrix U, a right sub-matrix V, and a diagonal sub-matrix Λ based on the matrix singular value decomposition method; where the diagonal sub-matrix Λ includes multiple non-zero diagonal values, and the main vibration signal includes multiple harmonics; any harmonic in the main vibration signal corresponds to 2 non-zero diagonal values in the diagonal sub-matrix Λ; a third detection module for dividing the non-zero diagonal values in the diagonal sub-matrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area, where the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal in the initial signal; a fourth detection module for determining a collision sub-matrix D2 based on the diagonal interval and the decomposed left sub-matrix U, right sub-matrix V, and diagonal sub-matrix Λ according to the monitoring matrix D; a fifth detection module for extracting the periodic signal in the collision sub-matrix D2 to obtain an early collision signal diagram.

[0123] Further, the first detection module is further configured to: within a preset time range, collect the signal waveform corresponding to the initial signal in a full cycle; continuously intercept k segments of signals with a sample length of n at equal lengths to obtain a monitoring matrix D, where each sample includes N sampling points.

[0124] Further, the second detection module is further configured to: based on the matrix singular value decomposition method, the expression of the above monitoring matrix D is: D = U×V×Λ T , where T is the transpose symbol.

[0125] Further, the second detection module is further configured to: based on a preset partitioning principle, divide the non-zero diagonal values in the above diagonal sub-matrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area; wherein, the above partitioning principle is based on the fact that the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal of the initial signal.

[0126] Further, the fourth detection module is further configured to: the expression of the above monitoring matrix D is: D = D1 + D2 + D3; the expression of the main vibration sub-matrix D1 is: The expression of the collision sub-matrix D2 is: The expression of the noise sub-matrix D3 is: where σ i , σ j , σ k are respectively the non-zero diagonal values in the diagonal sub-matrix Λ of the large value area, the transition area, and the small value area; u i , u j , u k are respectively the matrix values in the left sub-matrix U of the large value area, the transition area, and the small value area; v i , v j , v k are respectively the matrix values in the right sub-matrix V of the large value area, the transition area, and the small value area; m, J, K are respectively the number of harmonics in the main vibration sub-matrix D1, the collision sub-matrix D2, and the noise sub-matrix D3; T is the transpose symbol.

[0127] Further, the fifth detection module is further configured to: retain the non-zero diagonal values in the above transition area, set the non-zero diagonal values in the above large value area and the above small value area to zero, extract the periodic signal, and obtain the early collision signal diagram.

[0128] Further, the numerical values of the above non-zero diagonal values only depend on the amplitudes of the harmonics in the above initial signal.

[0129] Further, for each harmonic in the main vibration signal in the above initial signal, the non-zero diagonal value in the corresponding diagonal sub-matrix Λ and its sorting only depend on the amplitude of the corresponding harmonic.

[0130] The beneficial effects of the early collision detection device for power equipment provided by this application are the same as those of the foregoing embodiments of the early collision detection method for power equipment, and will not be elaborated in this embodiment.

[0131] An embodiment of the present application also provides an electronic device 130, such as Figure 13 As shown, it is a schematic structural diagram of the electronic device 130 provided by an embodiment of the present application, including: a processor 131, a memory 132, and a bus 133. The memory 123 stores machine-readable instructions executable by the processor 131. When the electronic device 130 runs, the processor 131 communicates with the memory 123 through the bus 133, and the machine-readable instructions are executed by the processor 131 to perform the method described in the above method embodiment.

[0132] An embodiment of the present application also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is run by a processor, it executes the above-mentioned method for early collision detection of power equipment. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems and devices can refer to the corresponding processes in the method embodiments, which will not be elaborated in this application. In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms. In addition, in each embodiment of the present application, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0133] The above embodiments are merely examples for clear illustration and not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or alterations can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. And the obvious changes or alterations derived therefrom still fall within the protection scope of this application's creation.

Claims

1. An early collision detection method for electric power equipment, characterized in that: The detection method comprises: The initial signal collected in advance is sampled in full cycle to obtain a monitoring matrix D, wherein the initial signal includes a main vibration signal, a periodic collision signal and a noise signal; Based on the matrix singular value decomposition method, the monitoring matrix D is decomposed into a left submatrix U, a right submatrix V and a diagonal submatrix Λ; wherein the diagonal submatrix Λ includes a plurality of non-zero diagonal values, and the main vibration signal includes a plurality of harmonics; any harmonic in the main vibration signal corresponds to two non-zero diagonal values ​​in the diagonal submatrix Λ; Dividing the non-zero diagonal values ​​in the diagonal submatrix Λ into three diagonal intervals: a large value area, a transition area, and a small value area, wherein the large value area, the transition area, and the small value area respectively reflect the main vibration signal, the periodic collision signal, and the noise signal in the initial signal; Based on the diagonal interval and the decomposed left submatrix U, right submatrix V, and diagonal submatrix Λ, the collision submatrix D2 is determined according to the monitoring matrix D; The periodic signal in the collision sub-matrix D2 is extracted to obtain an early collision signal graph.

2. The early collision detection method for electric equipment according to claim 1, characterized in that: The step of performing full-cycle sampling processing on the pre-collected initial signal to obtain the monitoring matrix D comprises: Within a preset time range, collecting the signal waveform corresponding to the initial signal in a whole period; K segments of signals with a sample length of n are continuously intercepted at equal length to obtain a monitoring matrix D, in which each sample includes N sampling points.

3. The early collision detection method for electric power equipment according to claim 2, characterized in that: The step of decomposing the monitoring matrix D into a left submatrix U, a right submatrix V and a diagonal submatrix Λ based on the matrix singular value decomposition method is as follows: Based on the matrix singular value decomposition method, the expression of the monitoring matrix D is: D=U×V×Λ T , where T is the transposition symbol.

4. The early collision detection method for electric equipment according to claim 3, characterized in that: The step of dividing the non-zero diagonal values ​​in the diagonal submatrix Λ into three diagonal intervals: a large value area, a transition area and a small value area: Based on a preset partitioning principle, the non-zero diagonal values ​​in the diagonal submatrix Λ are divided into three diagonal intervals: a large value area, a transition area, and a small value area; The partitioning principle is based on the fact that the large value area, the transition area and the small value area respectively reflect the main vibration signal, the periodic collision signal and the noise signal of the initial signal.

5. The early collision detection method for electric equipment according to claim 3, characterized in that: The step of determining the collision submatrix D2 according to the monitoring matrix D based on the left submatrix U, the right submatrix V, and the diagonal submatrix Λ obtained by the decomposition of the diagonal interval: the expression of the monitoring matrix D is: D=D1+D2+D3; The expression of the main oscillator matrix D1 is: The expression of the collision submatrix D2 is: The expression of the noise submatrix D3 is: Among them, σ i , σ j , σ k are the non-zero diagonal values ​​in the diagonal submatrix Λ in the large value region, transition region, and small value region respectively; u i 、u j 、u k are the matrix values ​​in the left submatrix U of the large value area, transition area, and small value area respectively; v i 、v j 、v k are the matrix values ​​in the right submatrix V in the large value area, transition area, and small value area respectively; m, J, and K are the number of harmonics in the main oscillation submatrix D1, collision submatrix D2, and noise submatrix D3 respectively; T is the transpose symbol.

6. The early collision detection method for electric power equipment according to claim 5, characterized in that: The step of extracting the periodic signal in the collision sub-matrix D2 to obtain an early collision signal graph: The non-zero diagonal values ​​of the transition region are retained, the non-zero diagonal values ​​of the large value region and the small value region are set to zero, and the periodic signal is extracted to obtain an early collision signal graph.

7. The early collision detection method for electric equipment according to claim 1, characterized in that: The magnitude of the non-zero diagonal values ​​depends only on the amplitudes of the harmonics in the initial signal.

8. The early collision detection method for electric equipment according to claim 7, characterized in that: The non-zero diagonal values ​​and their order in the diagonal submatrix Λ corresponding to each harmonic in the main vibration signal in the initial signal depend only on the amplitude of the corresponding harmonic.

9. An early collision detection device for electric power equipment, characterized in that: The detection device comprises: The first detection module is used to perform full-cycle sampling processing on the pre-collected initial signal to obtain a monitoring matrix D, wherein the initial signal includes a main vibration signal, a periodic collision signal and a noise signal; The second detection module is used to decompose the monitoring matrix D into a left submatrix U, a right submatrix V and a diagonal submatrix Λ based on a matrix singular value decomposition method; wherein the diagonal submatrix Λ includes a plurality of non-zero diagonal values, the main vibration signal includes a plurality of harmonics; any harmonic in the main vibration signal corresponds to two non-zero diagonal values ​​in the diagonal submatrix Λ; A third detection module is used to divide the non-zero diagonal values ​​in the diagonal submatrix Λ into three diagonal intervals: a large value area, a transition area and a small value area, wherein the large value area, the transition area and the small value area respectively reflect the main vibration signal, the periodic collision signal and the noise signal in the initial signal; A fourth detection module, configured to determine a collision submatrix D2 according to the monitoring matrix D based on the diagonal interval and the decomposed left submatrix U, right submatrix V, and diagonal submatrix Λ; The fifth detection module is used to extract the periodic signal in the collision sub-matrix D2 to obtain an early collision signal graph.

10. The early collision detection device for electric power equipment according to claim 9, characterized in that: The first detection module is further used for: Within a preset time range, collecting the signal waveform corresponding to the initial signal in a whole period; K segments of signals with a sample length of n are continuously intercepted at equal length to obtain a monitoring matrix D, in which each sample includes N sampling points.

11. The early collision detection device for electric equipment according to claim 10, characterized in that: The second detection module is further used for: Based on the matrix singular value decomposition method, the expression of the monitoring matrix D is: D=U×V×Λ T , where T is the transposition symbol.

12. The early collision detection device for electric power equipment according to claim 11, characterized in that: The second detection module is further used for: Based on a preset partitioning principle, the non-zero diagonal values ​​in the diagonal submatrix Λ are divided into three diagonal intervals: a large value area, a transition area, and a small value area; The partitioning principle is based on the fact that the large value area, the transition area and the small value area respectively reflect the main vibration signal, the periodic collision signal and the noise signal of the initial signal.

13. The early collision detection device for electric power equipment according to claim 11, characterized in that: The fourth detection module is further used for: the expression of the monitoring matrix D is: D=D1+D2+D3; The expression of the main oscillator matrix D1 is: The expression of the collision submatrix D2 is: The expression of the noise submatrix D3 is: Among them, σ i , σ j , σ k are the non-zero diagonal values ​​in the diagonal submatrix Λ in the large value region, transition region, and small value region respectively; u i 、u j 、u k are the matrix values ​​in the left submatrix U of the large value area, transition area, and small value area respectively; v i 、v j 、v k are the matrix values ​​in the right submatrix V in the large value area, transition area, and small value area respectively; m, J, and K are the number of harmonics in the main oscillation submatrix D1, collision submatrix D2, and noise submatrix D3 respectively; T is the transpose symbol.

14. The early collision detection device for electric equipment according to claim 13, characterized in that: The fifth detection module is further used for: The non-zero diagonal values ​​of the transition region are retained, the non-zero diagonal values ​​of the large value region and the small value region are set to zero, and the periodic signal is extracted to obtain an early collision signal graph.

15. The early collision detection device for electric equipment according to claim 9, characterized in that: The magnitude of the non-zero diagonal values ​​depends only on the amplitudes of the harmonics in the initial signal.

16. The early collision detection device for electric equipment according to claim 15, characterized in that: The non-zero diagonal values ​​and their order in the diagonal submatrix Λ corresponding to each harmonic in the main vibration signal in the initial signal depend only on the amplitude of the corresponding harmonic.

17. An electronic device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to execute the early collision detection method for electric equipment as described in any one of claims 1 to 8.

18. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the early collision detection method for electric equipment according to any one of claims 1 to 8 is executed.

Images