Bearing fault signal enhancement method based on speed pause particle swarm parameter optimization

Abstract

The invention discloses a bearing fault signal enhancement method based on speed pause particle swarm parameter optimization, and belongs to the technical field of mechanical fault intelligent diagnosis, and the method comprises the steps: setting a high-dimensional orthogonal optimization space as a search space of an SA algorithm, so as to carry out cooling circulation; a VPPSO algorithm is adopted to optimize the cooling rate in the SA algorithm; then, the SA algorithm iteratively uses a multi-dimensional norm as a judgment standard to search an optimal coordinate, and after an optimal filter coefficient is obtained through projection, blind deconvolution processing and envelope spectrum analysis are carried out on a vibration signal. The method has the beneficial effects that the capability of extracting the composite fault characteristic frequency of the bearing in a strong noise environment is remarkably improved, the fault frequency amplitude is remarkably improved, the monitoring time is short, and the method is suitable for a state monitoring system of rotating machinery such as a wind driven generator and an aero-engine.

Description

Technical Field

[0001] This invention relates to the field of intelligent mechanical fault diagnosis technology, and in particular to a bearing fault signal enhancement method based on speed-pause particle swarm optimization. Background Technology

[0002] As a crucial component of mechanical equipment, rolling bearings play a vital role in ensuring the long-term stable operation of machinery. Therefore, accurately extracting the fault characteristic components of rolling bearings is of great significance in fault diagnosis. However, in actual industrial production, the vibrations reflecting fault characteristics are often affected by interference signals such as environmental noise.

[0003] To effectively enhance weak fault impact signals submerged in noise, researchers have developed various signal enhancement methods based on blind deconvolution. However, traditional mainstream methods have significant limitations in practical applications: the MED method (Minimum Entropy Deconvolution) is sensitive to random pulses and struggles to extract periodic pulses; the MCKD method (Maximum Correlation Kurtosis Deconvolution) requires precise pre-setting of fault period parameters; the MOMEDA method improves pulse localization capabilities, but its convergence speed is slow and it relies on prior knowledge; and FDS-MOMEDA's use of spherical coordinate transformation leads to low efficiency in high-dimensional space search. Therefore, a technique is needed that can adaptively optimize the relevant parameters of the optimization algorithm to effectively enhance fault signals and improve search efficiency. Blind deconvolution enhancement of bearing composite fault features under strong noise environments is proposed for effective and timely extraction of weak faults. Summary of the Invention

[0004] The purpose of this invention is to provide a bearing fault signal enhancement method based on velocity-pause particle swarm optimization, which enables blind deconvolution enhancement of bearing composite fault features under strong noise environment, effectively and timely extracts weak bearing faults, and is applicable to condition monitoring systems of rotating machinery such as wind turbines and aero engines.

[0005] The objective of this invention is achieved through the following technical solution: A bearing fault signal enhancement method based on velocity-pause particle swarm optimization parameters is proposed. The enhancement method sets a high-dimensional orthogonal optimization space as the search space of the SA algorithm for cooling cycle; the VPPSO algorithm is used to optimize the cooling rate in the SA algorithm; then the SA algorithm iteratively uses the multidimensional norm as the judgment criterion to search for the optimal coordinates, and after obtaining the optimal filter coefficients by projection, the vibration signal is subjected to blind deconvolution processing and envelope spectrum analysis.

[0006] Furthermore, the enhancement method specifically includes the following steps: S1. Input the original vibration signal; S2. Construct a finite high-dimensional orthogonal optimization space as the search domain, with the spatial dimension boundary being [-∞, ∞]. S3. The cooling rate parameter α of the SA algorithm is dynamically optimized using the VPPSO algorithm; S4. Randomly initialize the filter coefficient vector in a finite high-dimensional orthogonal optimization space, set the initial temperature, and use the maximization of MDN as the objective function. S5. Perform annealing iterations via SA; S6, when the temperature drops to T min When the maximum number of iterations is reached, output the coordinates of each dimension as the optimal filter coefficients; S7. Use the optimal filter coefficients to perform blind deconvolution processing on the vibration signal to enhance the fault pulse signal; S8. Perform envelope spectrum analysis and trigger an alarm when the characteristic frequency amplitude exceeds the preset value.

[0007] Furthermore, the VPPSO algorithm employs a dual-swarm strategy. Swarm 1 updates velocity and position based on the improved velocity equation and velocity pause mechanism; swarm 2 updates position based on the globally optimal particle. The particle velocity update formula in swarm 1 is: The particle velocity update formula in Particle Swarm 1 is: The particle position update formula in Particle Swarm 2 is: in, The historical velocity of particles in particle swarm one. It represents the updated velocity of particles in particle swarm optimization, where a(t) is the dynamic coefficient. It is the best individual particle in history. It is the historical position of a particle in particle swarm optimization. It is the historically globally optimal particle, z1 is the cognitive acceleration coefficient, and z2 is the social acceleration coefficient. It is the updated position of the particle in particle swarm optimization. This refers to the updated positions of particles in Particle Swarm 2. It is a random factor, a random variable uniformly distributed in the range [0,1]. , , Corresponding to the randomness of dynamic coefficients, Corresponding to the randomness of the cognitive acceleration coefficient, Corresponding to the randomness of the social acceleration coefficient, To correspond to the randomness of the speed selection, rand is a random number generation function that generates random numbers uniformly distributed on [0,1], and β is the speed pause parameter, which is a preset hyperparameter.

[0008] Furthermore, the formula for the dynamic coefficient is: Where b is a constant, T is the maximum number of iterations, and t is the current number of iterations.

[0009] Furthermore, step S3 includes the following steps: X1. Input vibration signal to obtain corresponding health indicators; X2. Initialize the VPPSO algorithm; X3. Set dynamic coefficients; X4. Particle swarm optimization updates the velocity and position of the particles within it; X5, Particle Swarm 2 updates the positions of the particles; X6. Update the individual best particle and the global best particle based on the health indicators of each particle in Particle Swarm 1. X7. Update the global optimal particle based on the health indicators of each particle in Particle Swarm 2. X8. Repeat steps X4 to X7 until the maximum number of iterations is reached, then output the globally optimal particle, i.e., the cooling rate parameter α.

[0010] Furthermore, in step S4, the objective function formula is: in, It is the largest multidimensional norm. It is the target vector that determines the signal. The weight and position of the mid-pulse component, It is a filtered signal. It is the transpose of the target vector.

[0011] Furthermore, step S5 includes the following steps: Y1. Generate a new solution by applying random perturbations to the filter coefficients at the current temperature; Y2. Based on the Metropolis criterion and rand, decide whether to accept the new solution. The formula for calculating the Metropolis criterion is as follows: in, It is the probability of accepting the new solution. It is the difference between the objective function values ​​of the new solution and the current solution. This represents the temperature value at the k-th iteration; Y3. The cooling rate parameter α, optimized using the VPPSO algorithm, is calculated according to the cooling function using the following formula: in, It is the temperature value at the (k+1)th iteration.

[0012] Furthermore, in step X1, the function formula for the health indicator is: Where n is the dimension of the health indicator function, It is an input variable.

[0013] Furthermore, in step X6, the formula for updating the optimal individual particle is: The formula for updating the globally optimal particle in Particle Swarm 1 is: In step X7, the formula for updating the globally optimal particle in Particle Swarm 2 is: Where Vb(t+1) is the updated individual optimal particle. Here, f(Vb(t)) represents the health index corresponding to the particle position in the updated particle swarm 1, f(Vb(t)) represents the health index corresponding to the historical best individual particle, and f(Vb(t+1)) represents the health index corresponding to the updated best individual particle. is the health index corresponding to the particle position in the updated particle swarm 2, and f(Sb(t)) is the health index corresponding to the historical global best particle.

[0014] Furthermore, in step S7, the formula for calculating blind deconvolution is: in, It is the signal after blind deconvolution. These are the optimal filter coefficients. It is the input vibration signal.

[0015] The present invention has the following advantages: 1. The VPPSO algorithm dynamically optimizes the simulated annealing cooling rate parameter α, solving the problem that the parameters in the SA algorithm still depend on experience. It also enhances the fault characteristics under strong noise conditions and solves the problem of low search efficiency of FDS-MOMEDA in high-dimensional space.

[0016] 2. The method of this application can effectively and promptly detect minor bearing faults, shortening monitoring time. It is applicable to condition monitoring systems for rotating machinery such as wind turbines and aero engines. (See attached figures for details.) Figure 1 This is a flowchart of the present invention.

[0017] Figure 2 The envelope graph after processing the dataset using the MOMEDA method.

[0018] Figure 3 Envelope graph of the dataset after processing by the FDS-MOMEDA method.

[0019] Figure 4 The envelope graph after processing the dataset using this method.

[0020] Figure 5 This is the envelope diagram after processing the dataset with 4dB Gaussian white noise added using this method. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in this invention can be combined with each other.

[0024] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0025] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0026] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0027] refer to Figure 1 As shown, one embodiment of the present invention is as follows: A bearing fault signal enhancement method based on velocity-pause particle swarm optimization parameters is proposed. The enhancement method sets a high-dimensional orthogonal optimization space as the search space of the SA algorithm for cooling cycle; the VPPSO algorithm is used to optimize the cooling rate in the SA algorithm; then the SA algorithm iteratively uses the multidimensional norm as the judgment criterion to search for the optimal coordinates, and after obtaining the optimal filter coefficients by projection, the vibration signal is subjected to blind deconvolution processing and envelope spectrum analysis.

[0028] The enhancement method specifically includes the following steps: S1. Input the original vibration signal.

[0029] This invention uses the HUSTbearing dataset to perform fault enhancement testing on rolling bearings. The HUSTbearing dataset has a sampling frequency of 25.6 kHz, a sampling time of 10.2 seconds for each fault at each speed, and a sampling number of 262,144.

[0030] The HUSTbearing dataset introduces faults into the test bearings via electrical discharge machining (EDM), with fault diameters including 0.007 inches, 0.014 inches, 0.021 inches, 0.028 inches, and 0.040 inches. The dataset includes bearing data for nine health states: normal, moderate inner ring fault, severe inner ring fault, moderate outer ring fault, severe outer ring fault, moderate rolling element fault, severe rolling element fault, moderate combined outer and inner ring fault, and severe combined outer and inner ring fault.

[0031] The HUSTbearing dataset tests bearings of type ER-16K, and the specific parameters are shown in Table 1.

[0032] parameter numerical values Inner ring diameter 30.59mm Outer ring diameter 46.47mm Contact angle 0° pitch diameter 39.65mm Rolling element diameter 7.94mm Number of rolling elements 9 Table 1: Parameters of the tested bearing for composite fault signal processing Data files with a combined moderate fault in the outer and inner rings were selected for processing, under operating conditions of 65Hz (3900rpm).

[0033] Characteristic frequencies of outer ring rolling bearing failure Characteristic frequencies of inner ring rolling bearing failure The calculation formula is as follows:

[0034]

[0035] in, The bearing rotation speed is 3900 rpm in the selected dataset. Substituting this into the formula for calculating the characteristic frequency of rolling bearing failure, the characteristic frequency of the outer ring failure is 233.415 Hz, and the characteristic frequency of the inner ring failure is 351.52 Hz. S2. Construct a finite high-dimensional orthogonal optimization space as the search domain, with the spatial dimension boundary being [-∞, ∞], and the specific dimension in the implementation is 50.

[0036] S3. The cooling rate parameter α of the SA algorithm is dynamically optimized using the VPPSO algorithm; The VPPSO algorithm in this embodiment introduces dynamic coefficients, a velocity pause mechanism, and employs a dual-swarm strategy. Particle swarm one updates its velocity and position based on the improved velocity equation and velocity pause mechanism; particle swarm two updates its position based on the globally optimal particle. One cycle mainly consists of three stages: particle swarm one updates its state, particle swarm two updates its state, and the health indicators of all particles are evaluated to determine the globally optimal particle, i.e., the cooling rate parameter α.

[0037] The specific steps are as follows: X1. Input vibration signal to obtain corresponding health indicators.

[0038] Using function F6 from the CEC2005 standard test set as the health indicator function, the formula for the health indicator function is as follows: Where n is the dimension of the health indicator function, It is an input variable.

[0039] X2. Initialize the VPPSO algorithm.

[0040] In this embodiment, the inertial weight component in PSO is removed before the loop begins, and a dynamic coefficient is introduced.

[0041] X3. Set dynamic coefficients.

[0042] The formula for the dynamic coefficient is: Where b is a constant, T is the maximum number of iterations, and t is the current number of iterations. Dynamic coefficients are used to replace the inertial weight components to avoid rapid changes in particle velocity, thereby better controlling the particle's search behavior, achieving stable control, and effectively preventing excessively fast convergence.

[0043] X4, Particle Swarm 1 updates the velocity and position of the particles.

[0044] Pausing provides particles with more motion options and avoids premature convergence. The particle pauses its velocity relative to the parameter β and rand, maintaining the velocity from the previous iteration with a probability. In this embodiment, β is set to 0.3, which effectively avoids local optima during the search process and significantly increases the probability of finding the global optimum.

[0045] The particle velocity update formula in Particle Swarm 1 is: When rand < β, the positions of particles in particle swarm 1 remain unchanged; when rand > β, particle swarm 1 updates its positions according to the velocity pause mechanism. The particle position update formula in Particle Swarm 1 is: in, The historical velocity of particles in particle swarm one. It represents the updated velocity of particles in particle swarm optimization, where a(t) is the dynamic coefficient. It is the best individual particle in history. It is the historical position of a particle in particle swarm optimization. It is the historically globally optimal particle, z1 is the cognitive acceleration coefficient, and z2 is the social acceleration coefficient. It is the updated position of the particle in particle swarm optimization. is a random factor, a random variable uniformly distributed in the range [0,1], rand is a random number generation function, a random number uniformly distributed in [0,1], and β is a speed pause parameter, a preset hyperparameter.

[0046] X5, Particle Swarm 2 updates the positions of the particles.

[0047] By introducing a dynamic coefficient, Particle Swarm 2 updates its position only based on the globally optimal particle. By increasing population diversity, Particle Swarm 2 avoids getting trapped in local optima, thus optimizing global performance and search capability. The particle position update formula in Particle Swarm 2 is: in, This refers to the updated positions of particles in Particle Swarm 2. , , Corresponding to the randomness of dynamic coefficients, Corresponding to the randomness of the cognitive acceleration coefficient, Corresponding to the randomness of the social acceleration coefficient, The randomness of the corresponding speed selection.

[0048] X6. Update the individual best particle and the global best particle based on the health indicators of each particle in Particle Swarm 1.

[0049] The formula for updating the best individual particle is: The formula for updating the globally optimal particle in Particle Swarm 1 is:

[0050] in, It is the updated individual optimal particle. It is the health indicator corresponding to the particle position in the updated particle swarm. It is the health indicator corresponding to the best individual in history. It is the updated health indicator corresponding to the individual's optimal particle.

[0051] X7. Update the global optimal particle based on the health indicators of each particle in Particle Swarm 2.

[0052] The formula for updating the global best particle in Particle Swarm 2 is: in, It is the health indicator corresponding to the particle position in the updated particle swarm 2. It is the health indicator corresponding to the historical global best particle.

[0053] X8. Repeat steps X4 to X7 until the maximum number of iterations is reached, then output the globally optimal particle, i.e., the cooling rate parameter α. In this implementation, the output cooling rate parameter is 0.998003.

[0054] S4. Randomly initialize the filter coefficient vector in a finite high-dimensional orthogonal optimization space, set the initial temperature, and use the maximization of MDN as the objective function.

[0055] Initialization is performed in the optimization space, with the multi-D norm used as the objective function: in, It is the largest multidimensional norm. It is the target vector that determines the signal. The weight and position of the mid-pulse component, It is a filtered signal. It is the target vector The transpose of .

[0056] S5. Annealing iteration via SA, specifically including the following steps: The following steps: Y1. Apply random perturbations to the filter coefficients at the current temperature to generate a new solution.

[0057] Y2. Based on the Metropolis criterion and rand, decide whether to accept the new solution. The formula for calculating the Metropolis criterion is as follows: Where P is the probability of accepting the new solution, and ΔK is the difference between the objective function values ​​of the new solution and the current solution. This represents the temperature value at the k-th iteration.

[0058] During implementation, the multi-D norm of the corresponding point under the current perturbation is calculated and compared with the health index of the current optimal point. If the multi-D norm of the current perturbation is greater than the health index of the current optimal point, the multi-D norm of the current perturbation is selected as the new health index; if the multi-D norm of the current perturbation is less than the health index of the current optimal point, a comparison between P and rand is used to determine whether to accept the new solution. When the probability P of accepting the new solution is greater than the random number generation function rand, the new solution is accepted.

[0059] Y3. The cooling rate parameter α, optimized using the VPPSO algorithm, is calculated according to the cooling function using the following formula: in, It is the temperature value at the (k+1)th iteration.

[0060] S6. When the temperature drops to Tmin or the maximum number of iterations is reached, output the coordinates of each dimension as the optimal filter coefficients. After the temperature reaches Tmin or the iteration is completed, project the point onto each coordinate to obtain the optimal filter coefficients. .

[0061] S7. Use the optimal filter coefficients to perform blind deconvolution processing on the vibration signal to enhance the fault pulse signal.

[0062] The input dataset is divided equally, and data of length 25600 is randomly selected for blind deconvolution. 4dB Gaussian white noise is added to the original data to simulate random interference in real-world signals.

[0063] The formula for calculating blind deconvolution is:

[0064] Where h(t) is the signal after blind deconvolution, y(t) is the optimal filter coefficient, and y(t) is the input vibration signal.

[0065] S8. Perform envelope spectrum analysis with a frequency range of 0-1500Hz. Trigger an alarm when the amplitude of the characteristic frequency exceeds the preset value.

[0066] exist Figure 2-5 In the envelope diagram shown, f i f0 represents the inner loop fault, and f0 represents the outer loop fault. The inner loop fault frequency and its harmonics are observed in the envelope diagram within the range of 0 to 1500 Hz. This method is compared with existing blind deconvolution methods MOMEDA and FDS-MOMEDA.

[0067] like Figure 2 As shown, in the envelope spectrum of the signal processed by the MOMEDA method, the peak values ​​corresponding to the 1st and 4th times inner ring fault characteristic frequencies and the 2nd and 6th times outer ring fault characteristic frequencies are completely masked by irrelevant signals and cannot be extracted normally. Figure 3 As shown, although the fault features in the envelope spectrum of the signal processed by the FDS-MOMEDA method are enhanced compared to the MOMEDA method, the peak values ​​corresponding to the 1st and 4th times inner ring fault feature frequencies and the 2nd and 6th times outer ring fault feature frequencies are still severely affected by irrelevant signals and cannot be extracted normally. Figure 4As shown, in the envelope spectrum of the signal processed by this method, the peak values ​​corresponding to the inner circle fault characteristic frequencies of 1 to 4 times and the outer circle fault characteristic frequencies of 1 to 6 times are significant, and the periodicity of the fault characteristic frequencies is obvious. This indicates that the envelope spectrum successfully extracted the fault characteristic frequencies of the composite fault signal, demonstrating the advantage of this method for composite fault diagnosis.

[0068] The dataset with 4dB Gaussian white noise was processed using this method, and the results are as follows: Figure 5 As shown, the inner loop fault characteristic frequency and its harmonics still exhibit significant peak values, indicating that this method is resistant to the influence of noise and can effectively extract the composite fault characteristics of bearings even in high-noise environments.

[0069] As can be seen from the above, by introducing the VPPSO algorithm and the SA algorithm in synergy, the disturbance in the cooling space becomes more obvious, which can effectively avoid the occurrence of local optima during the search process, thereby significantly increasing the probability of finding the global optimal solution, increasing the probability of finding the optimal inverse filter, and thus improving the ability to extract weak signal fault features. Finally, by enhancing the periodic pulse component in the vibration signal through the inverse filter, the problem of the weak fault features of bearings being indistinct under strong noise environment is solved.

[0070] The method proposed in this application maps filter coefficients to spatial coordinates within a constructed finite high-dimensional orthogonal space. By dynamically optimizing the cooling rate parameter α using the VPPSO algorithm, it solves the problem of traditional methods relying on prior experience. This results in faster temperature changes during the number of cycles and a quicker optimization process for the cooling cycle, effectively addressing the low search efficiency of the FDS-MOMEDA algorithm in high-dimensional spaces, a problem that becomes more pronounced with large datasets. Since the MOMEDA algorithm does not involve iterative algorithms, no comparison is made. Table 2 shows a comparison of the average computation time for completing iterations between the FDS-MOMEDA algorithm and the method proposed in this application.

[0071] method Average computation time FDS-MOMEDA 18.1458 seconds This application 16.9675 seconds Table 2: Average computation time for different methods to complete iterations As can be seen from the above, the method of this application has higher search efficiency and more obvious fault characteristics compared with the FDS-MOMEDA algorithm. It can effectively and timely extract weak faults and is suitable for condition monitoring systems of rotating machinery such as wind turbines and aero engines.

[0072] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A bearing fault signal enhancement method based on velocity-pause particle swarm optimization, characterized in that: The enhancement method sets the high-dimensional orthogonal optimization space as the search space of the SA algorithm for cooling cycle; it uses the VPPSO algorithm to optimize the cooling rate in the SA algorithm; then the SA algorithm iteratively uses the multidimensional norm as the judgment criterion to search for the optimal coordinates, and after obtaining the optimal filter coefficients by projection, it performs blind deconvolution processing and envelope spectrum analysis on the vibration signal.

2. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 1, characterized in that: The enhancement method specifically includes the following steps: S1. Input the original vibration signal; S2. Construct a finite high-dimensional orthogonal optimization space as the search domain, with the spatial dimension boundary being [-∞, ∞]. S3. The cooling rate parameter α of the SA algorithm is dynamically optimized using the VPPSO algorithm; S4. Randomly initialize the filter coefficient vector in a finite high-dimensional orthogonal optimization space, set the initial temperature, and use the maximization of MDN as the objective function. S5. Perform annealing iterations via SA; S6, when the temperature drops to T min When the maximum number of iterations is reached, output the coordinates of each dimension as the optimal filter coefficients; S7. Use the optimal filter coefficients to perform blind deconvolution processing on the vibration signal to enhance the fault pulse signal; S8. Perform envelope spectrum analysis and trigger an alarm when the characteristic frequency amplitude exceeds the preset value.

3. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 2, characterized in that: The VPPSO algorithm employs a dual-swarm strategy: swarm one updates velocity and position based on the improved velocity equation and velocity pause mechanism; swarm two updates position based on the globally optimal particle. The particle velocity update formula in Particle Swarm 1 is: When rand < β, the positions of particles in particle swarm 1 remain unchanged; when rand > β, particle swarm 1 updates its positions according to the velocity pause mechanism. The particle position update formula in Particle Swarm 1 is: The particle position update formula in Particle Swarm 2 is: in, The historical velocity of particles in particle swarm one. It represents the updated velocity of particles in particle swarm optimization, where a(t) is the dynamic coefficient. It is the best individual particle in history. It is the historical position of a particle in particle swarm optimization. It is the historically globally optimal particle, z1 is the cognitive acceleration coefficient, and z2 is the social acceleration coefficient. It is the updated position of the particle in particle swarm optimization. This refers to the updated positions of particles in Particle Swarm 2. is a random factor, a random variable uniformly distributed in the range [0,1], rand is a random number generation function, a random number uniformly distributed in [0,1], and β is a speed pause parameter, a preset hyperparameter.

4. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 3, characterized in that: The formula for the dynamic coefficient is: Where b is a constant, T is the maximum number of iterations, and t is the current number of iterations.

5. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 4, characterized in that: Step S3 includes the following steps: X1. Input vibration signal to obtain corresponding health indicators; X2. Initialize the VPPSO algorithm; X3. Set dynamic coefficients; X4. Particle swarm optimization updates the velocity and position of the particles within it; X5, Particle Swarm 2 updates the positions of the particles; X6. Update the individual best particle and the global best particle based on the health indicators of each particle in Particle Swarm 1. X7. Update the global optimal particle based on the health indicators of each particle in Particle Swarm 2. X8. Repeat steps X4 to X7 until the maximum number of iterations is reached, then output the globally optimal particle, i.e., the cooling rate parameter α.

6. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 2, characterized in that: In step S4, the objective function formula is: in, It is the largest multidimensional norm. It is the target vector that determines the signal. The weight and position of the mid-pulse component, It is a filtered signal. It is the target vector The transpose of .

7. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 2, characterized in that: Step S5 includes the following steps: Y1. Generate a new solution by applying random perturbations to the filter coefficients at the current temperature; Y2. Based on the Metropolis criterion and rand, decide whether to accept the new solution. The formula for calculating the Metropolis criterion is as follows: in, It is the probability of accepting the new solution. It is the difference between the objective function values ​​of the new solution and the current solution. This represents the temperature value at the k-th iteration; Y3. The cooling rate parameter α, optimized using the VPPSO algorithm, is calculated according to the cooling function using the following formula: in, It is the temperature value at the (k+1)th iteration.

8. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 4, characterized in that: In step X1, the function formula for the health indicator is: Where n is the dimension of the health indicator function, It is an input variable.

9. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 8, characterized in that: In step X6, the formula for updating the optimal personal particle is: The formula for updating the globally optimal particle in Particle Swarm 1 is: In step X7, the formula for updating the globally optimal particle in particle swarm 2 is: in, It is the updated individual optimal particle. It is the health indicator corresponding to the particle position in the updated particle swarm. It is the health indicator corresponding to the best individual in history. It is the updated health metric corresponding to the individual's optimal particle. It is the health indicator corresponding to the particle position in the updated particle swarm 2. It is the health indicator corresponding to the historical global best particle.

10. The bearing fault signal enhancement method based on velocity-pause particle swarm optimization according to claim 2, characterized in that: In step S7, the formula for calculating blind deconvolution is: in, It is the signal after blind deconvolution. These are the optimal filter coefficients. It is the input vibration signal.

Images