Fault feature extraction method and device for variable-speed bearing

Abstract

The invention relates to a variable-speed bearing fault feature extraction method and device, relates to the technical field of bearing fault diagnosis, and aims to solve the problem that the variable-speed bearing fault feature extraction precision is not high. The method comprises the steps that a particle swarm optimization algorithm is introduced into a multi-scale adaptive morphological filtering operator, key parameters of the adaptive morphological filtering operator are obtained, filtering signals of all scales are obtained through calculation according to the multi-scale adaptive morphological filtering operator and the key parameters, and the filtering signals of all the scales are added to obtain a primary filtering signal; introducing a particle swarm optimization algorithm into an adaptive stochastic resonance algorithm to obtain a system cooperative adjustment parameter of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximum, and inputting the primary filtering signal into the adaptive stochastic resonance algorithm to obtain a secondary filtering signal; and calculating the secondary filtering signal by using an angular domain resampling algorithm to obtain an order spectrogram containing bearing fault features. The method has the effect of improving the fault feature extraction precision of the variable-speed bearing.

Description

Technical Field

[0001] This invention relates to the technical field of bearing fault diagnosis, and in particular to a method and apparatus for extracting fault features of variable speed bearings. Background Technology

[0002] Rolling bearings are widely used in supporting rotating machinery due to their superior stability. However, due to the complex and variable working environment and long-term operation, bearings remain a critical component in rotating machinery systems prone to damage. Their failures mainly manifest as localized defects (such as pitting) in the inner ring, outer ring, or rolling elements. Figure 1 As shown. Therefore, rolling bearing condition identification and health diagnosis technology has been a research hotspot in the online equipment monitoring industry in recent years.

[0003] Currently, time-frequency feature analysis based on vibration signals is the most widely used method in the field of bearing fault detection. Due to factors such as load fluctuations and voltage instability, most rotating machinery operates under fluctuating speeds. Under variable speed conditions, the vibration signals of rolling bearings contain various noise components, and strong noise and high-frequency vibrations transmitted from other equipment in the production system can drown out the fault signals in strong background noise. Therefore, proposing a reasonable and accurate bearing fault feature extraction method is an urgent problem to be solved in the field of bearing fault diagnosis.

[0004] Currently, mathematical morphology is widely used in bearing fault detection due to its powerful filtering performance. Existing techniques, such as morphological filtering operators and sensitive multi-scale symbolic dynamic entropy, heavily rely on past experience for bearing detection, resulting in extracted bearing fault features that do not closely approximate the actual situation, thus affecting the accuracy of fault diagnosis. Summary of the Invention

[0005] The purpose of this invention is to provide a method for extracting fault features of variable speed bearings, which can improve the accuracy of fault feature extraction for variable speed bearings.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for extracting fault features of variable speed bearings, characterized in that it includes: The vibration signal of the bearing is acquired, and the vibration signal includes several sets of discrete signals; The morphological operator is used to analyze the one-dimensional discrete signal of the discrete signal to obtain the erosion operator and the dilation operator of the one-dimensional discrete signal; By introducing cascaded operators into the erosion and dilation operators, a multi-scale adaptive morphological filtering operator is obtained. A particle swarm optimization algorithm is introduced into the multi-scale adaptive morphological filtering operator to obtain the key parameters of the adaptive morphological filtering operator. The filtered signals at each scale are calculated based on the multi-scale adaptive morphological filtering operator and the key parameters. The filtered signals at each scale are added together to obtain the first filtered signal of the vibration signal. An adaptive stochastic resonance algorithm is provided, in which a particle swarm optimization algorithm is introduced to obtain the system coordinated adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximized. The primary filtered signal is used as the input signal of the adaptive stochastic resonance algorithm to obtain the secondary filtered signal of the vibration signal. An angular domain resampling algorithm is provided, which is used to calculate the secondary filtered signal to obtain an order spectrum containing bearing fault characteristics.

[0007] A further technical solution is that the corrosion operator is represented by the following formula: ; The expansion operator is expressed by the following equation: ; In the formula, f(n) It is a one-dimensional discrete signal, and its domain is F(0,1,2... N-1) and n∈0,1,2...N-1 ,in g(m) It is a structure operator of morphological operators, and its domain is G(0,1,2... M-1) ,and N>=M , m∈0,1,2... M- 1 .

[0008] A further technical solution is that the adaptive filtering operator is represented by the following formula: ; In the formula, λ It is the control scale for multi-scale filtering.

[0009] A further technical solution involves introducing a particle swarm optimization algorithm into the multi-scale adaptive morphological filter operator to obtain the key parameters of the adaptive morphological filter operator, specifically including: The size O of the optimized population is defined as 20; the learning factor c1=c2=0.5; the inertia factor is 0.7; the weight factor of the PSO optimization filter is [0,1], the particle motion range is [0,50]; the velocity limit range is [-10,10]; the length of the structuring element of the multi-scale adaptive morphological filter operator is an integer in the range of [3,21]; the maximum number of iterations is 30. The vibration signal is input into a multi-scale adaptive morphological filter operator, and the particle swarm optimization algorithm initializes and measures the position and velocity of the parameters to be optimized. The measured particle data is input into the fitness function of the multi-scale adaptive morphological filter operator, and the extreme values ​​of individual particle velocity and population position are obtained through iterative calculation. If any of the set iteration stopping conditions are met, the iteration stops and the iteration curves of the key parameters and the fitness function of the multi-scale adaptive morphological filter operator are output; otherwise, the loop returns to the previous step and continues until the iteration stopping condition is met.

[0010] A further technical solution is that the iteration stopping condition includes: Reaching the maximum number of iterations; The fitness function value of the multi-scale adaptive morphological filter operator is less than the preset threshold; The positional changes of all particles are less than the set minimum value.

[0011] A further technical solution is that the adaptive stochastic resonance algorithm is expressed by the following formula: ; In the formula, a , b These represent the system's coordinated adjustment parameters; A The signal amplitude; ω The characteristic frequency of the signal; φ This represents the initial phase of the signal; τ Indicates a time delay; D This represents the fractal dimension.

[0012] A further technical solution involves introducing a particle swarm optimization algorithm into the adaptive stochastic resonance algorithm to obtain the system coordinated adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximized. Specifically, this includes: Define the population size O as 30; learning factors c1=c2=0.2; inertia factor as 0.6; velocity limit range as [-10, 10]; maximum number of iterations as 50; iteration stops when the maximum evolutionary generation is calculated. Otherwise, return to this step and continue the loop until the maximum evolutionary generation is calculated.

[0013] Compared with existing technologies, the variable speed bearing fault feature extraction method provided by this invention has the following advantages: By introducing particle swarm optimization (PSO) into the multi-scale adaptive morphological filtering operator, the original vibration signal can achieve convergence and filter out a certain number of harmonic interference components. By introducing PSO into the adaptive stochastic resonance algorithm, the ability to extract fault features can be improved. Combining the multi-scale adaptive morphological filtering operator with the adaptive stochastic resonance algorithm effectively filters out the clutter components of the original signal. At the same time, the order of fault features is significantly enhanced in the final order spectrum, which has the ability to improve the accuracy of fault feature extraction for variable speed bearings.

[0014] The present invention also provides a variable speed bearing fault feature extraction device, which employs the aforementioned variable speed bearing fault feature extraction method. The device includes: The acquisition module is used to acquire the vibration signal of the bearing, which includes several sets of discrete signals; An adaptive morphological filtering module is used to analyze the one-dimensional discrete signal of the discrete signal using morphological operators to obtain the erosion operator and the dilation operator of the one-dimensional discrete signal; cascaded operators are introduced into the erosion operator and the dilation operator to obtain a multi-scale adaptive morphological filtering operator; An adaptive morphological filtering optimization module is used to introduce a particle swarm optimization algorithm into the multi-scale adaptive morphological filtering operator to obtain the key parameters of the adaptive morphological filtering operator, calculate the filtered signals at each scale based on the multi-scale adaptive morphological filtering operator and the key parameters, and add the filtered signals at each scale to obtain the first filtered signal of the vibration signal. An adaptive stochastic resonance optimization module is used to introduce a particle swarm optimization algorithm into the adaptive stochastic resonance algorithm to obtain the system coordinated adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximized, and to use the primary filtered signal as the input signal of the adaptive stochastic resonance algorithm to obtain the secondary filtered signal of the vibration signal. The angular domain resampling module is used to calculate the secondary filtered signal using the angular domain resampling algorithm to obtain an order spectrum containing bearing fault characteristics.

[0015] Compared with the prior art, the beneficial effects of the variable speed bearing fault feature extraction device provided by the present invention are the same as those of the method described in the above technical solution, and will not be repeated here. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method provided by the present invention; Figure 2 This is a structural diagram of a basic stochastic resonance (SR) model provided by the present invention; Figure 3 This is a diagram of the test bench layout in the experiment of this invention; Figure 4 This is a comparison diagram of vibration signals before and after filtering in various states of an outer ring fault bearing provided by this invention; Figure 5 This is a comparison diagram of vibration signals before and after filtering in various states of an inner ring fault bearing provided by the present invention; Figure 6 This is a comparison diagram of vibration signals before and after filtering in various states of a rolling element fault bearing provided by the present invention; Figure 7 This is a comparison chart of quantitative indicators of filtering performance provided by the present invention; Figure 8 This is the original signal spectrum diagram of the bearing with outer ring failure provided by the present invention; Figure 9 This is the original signal spectrum diagram of the bearing with inner ring failure provided by the present invention; Figure 10 This is the original signal spectrum diagram of a bearing with rolling element failure provided by the present invention; Figure 11 This invention provides the order spectrum of outer ring fault bearings processed by the MSAMF-ASR method under different speed variation conditions; Figure 12 This is the order spectrum of inner ring fault bearings processed by the MSAMF-ASR method under different speed variation conditions provided by the present invention; Figure 13 This is a rolling element fault bearing order spectrum processed by the MSAMF-ASR method under different speed variation conditions provided by the present invention. Detailed Implementation

[0017] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

[0018] The embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0019] This invention provides a method for extracting fault features of variable speed bearings, such as... Figure 1 As shown, the main process of the method is described below.

[0020] Step S1: Obtain the vibration signal of the bearing, which includes several sets of discrete signals.

[0021] Step S2: Analyze the one-dimensional discrete signal of the discrete signal using morphological operators to obtain the erosion operator and dilation operator of the one-dimensional discrete signal; introduce cascaded operators into the erosion operator and dilation operator to obtain the multi-scale adaptive morphological filter operator MSAMF.

[0022] The most important aspect of mathematical morphology is the design of operators. A morphological operator is defined as follows: Assume there exists a one-dimensional discrete signal. f(n) Let its domain be F(0,1,2... N-1) and n∈0,1,2...N-1 ,in g(m) It is a structure operator, and its domain is defined as follows. G(0,1,2... M-1) ,and N>=M and m∈0,1,2...M-1 Then a one-dimensional signal f(n) Compared to g (m) The mathematical expressions for the erosion operator and the dilation operator are represented by the following equations: Corrosion operator: ; Dilation operator: .

[0023] When filtering signals, the erosion and dilation operators alone are ineffective for feature extraction from noisy signals due to noise and other factors. To improve feature extraction performance, this invention proposes the concept of cascaded operators. Cascaded operators use erosion and dilation operators as core computational units, deriving composite operators such as closing and opening operations through cascaded operations of "erosion-dilation" and "dilation-erosion." This addresses the problem of insufficient feature extraction from noisy signals by simple basic operators, enhancing noise filtering while preserving the impact characteristics of bearing faults. It represents an engineering extension of basic operators, reflecting the technological evolution from "basic operations" to "composite optimization."

[0024] Common cascading operators are shown in the following equations: Closing operator: ; Open the arithmetic unit: ; Here, • and º represent closing and opening operations, respectively. In signal feature extraction, closing operations are scalable, while opening operations are contractive.

[0025] The multi-scale adaptive filtering operator is derived from the opening and closing operators. It progresses from basic morphological operators to composite morphological operators to cascaded composite operators. Through further cascading operations on the opening and closing operators, it enhances the preservation of fault characteristics and noise suppression of noisy signals from variable-speed bearings. This represents a deep optimization of the morphological operator under complex operating conditions. The multi-scale adaptive filtering operator is expressed by the following equation: ; In the formula, λ It is the control scale for multi-scale filtering.

[0026] Step S3: Introduce the particle swarm optimization algorithm into the multi-scale adaptive morphological filtering operator to obtain the key parameters of the adaptive morphological filtering operator. Calculate the filtered signals at each scale based on the multi-scale adaptive morphological filtering operator and the key parameters. Add the filtered signals at each scale to obtain the first filtered signal of the vibration signal.

[0027] Key parameters include weighting factors, filtering scale, and structuring element length. In the MSAMF operator, the selection of key parameters such as weighting factors and filtering scale has a significant impact on filtering. For the filtering scale, the influence of different scales needs to be weighed. Currently, the method to address this problem in multi-scale filtering algorithms is to introduce weighting factors at different scales. The selection of weighting factors at different scales is one of the difficulties in multi-scale morphology. To address the difficulty in determining weighting factors, this invention proposes a method for selecting the structuring element scale through single-scale optimization and introduces the Particle Swarm Optimization (PSO) algorithm into the adaptive selection of weighting factors and filtering scale. PSO can optimize the values ​​of weighting factors and filtering scales to balance the influence of different scales, thus solving the problem of the difficulty in determining these two parameters in multi-scale morphology.

[0028] Specifically, a particle swarm optimization algorithm is introduced into the multi-scale adaptive morphological filtering operator to obtain the key parameters of the adaptive morphological filtering operator, which includes the following steps: Step S31: Define the size O of the optimization population as 20; learning factor c1=c2=0.5; inertia factor as 0.7; weight factor of PSO optimization filter as [0,1], particle motion range as [0,50]; velocity limit range as [-10,10]; length range of MSAMF operator structuring elements as integers in the range [3,21]; maximum number of iterations as 30. Step S32: Input the vibration signal into the multi-scale adaptive morphological filter operator. The particle swarm optimization algorithm initializes and measures the position and velocity of the parameters to be optimized. Input the measured particle data into the fitness function of the multi-scale adaptive morphological filter operator. The extreme values ​​of individual particle velocity and population position are obtained through iterative calculation. Step S33: If any set iteration stopping condition is met, the iteration stops and the iteration curves of the key parameters and the fitness function of the multi-scale adaptive morphological filter operator are output; otherwise, return to the previous step and continue the loop until the iteration stopping condition is met.

[0029] The iteration stopping conditions include: reaching the maximum number of iterations; the fitness function value of the MSAMF operator being less than a preset threshold; and the position changes of all particles being less than a set minimum value.

[0030] After determining the key parameters of the MSAMF operator at different scales, the filtered signals at each scale are summed to obtain the primary filtered signal of the multi-scale adaptive morphological filtering operator.

[0031] Step S4: Provide an adaptive stochastic resonance algorithm (ASR), introduce a particle swarm optimization algorithm into the adaptive stochastic resonance algorithm, obtain the system coordinated adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximized, use the primary filtered signal as the input signal of the adaptive stochastic resonance algorithm, and obtain the secondary filtered signal of the vibration signal.

[0032] Currently, the Stochastic Resonance (SR) algorithm is widely used for feature extraction in rolling bearing faults. Its basic principle is to use a special filter to convert noise energy into weak signal energy. Therefore, by processing the acquired rolling bearing noise and vibration signals through stochastic resonance, the fault feature energy of the weak signal can be increased, making the corresponding feature information clearer, thereby achieving accurate fault feature extraction. The basic theoretical model is as follows: Figure 2 As shown.

[0033] To address the issues of infinite steady-state structure and antiparticle output saturation in the periodic potential function of the stochastic resonance algorithm, this invention proposes an adaptive stochastic resonance algorithm based on the periodic potential function, the expression of which is: ; ; in, ω ( t ) It is Gaussian white noise with a mean of 0 and a variance of 1, where D represents the noise intensity. a , b These represent the system's coordinated adjustment parameters; Г ( t () is a noise signal that includes both system noise and external environmental noise.

[0034] Among them, parameters a It belongs to the system damping coefficient, and its main function is to regulate the energy dissipation rate of the system, affecting the stability and decay process of the system response, and playing a key role in the dynamic damping characteristics of stochastic resonance systems; parameters b It belongs to the category of nonlinear stiffness coefficients and is used to control the strength of the nonlinear terms in the system. It determines the amplification degree of the periodic signal and the range of the resonant bandwidth, and is one of the core parameters for realizing the adjustment of nonlinear characteristics in adaptive random resonance.

[0035] Input signal s ( t )for: ; Therefore, the expression for the adaptive stochastic resonance algorithm is: ; In the formula, a , b These represent the system's coordinated adjustment parameters; A The signal amplitude; ω The characteristic frequency of the signal; φ This represents the initial phase of the signal; τ It represents time delay and is used in phase space reconstruction or signal analysis to describe the delay of a signal in the time dimension. It is used to construct embedded phase space or analyze the time correlation of signals and is a key parameter for characterizing the temporal relationship of signals in fault feature extraction. D The fractal dimension (box dimension) is used to quantify the fractal characteristics of bearing vibration signals, reflecting the complexity and irregularity of the signal. It is an important characteristic indicator for identifying bearing fault conditions. The fractal dimension of faulty bearings usually differs significantly from that of normal bearings.

[0036] The periodic potential function is the nonlinear core of the stochastic resonant system, used to construct a nonlinear potential field to amplify noise and amplify weak bearing fault characteristics. Its variable-scaling expression adapts to the frequency characteristics of non-stationary signals from variable-speed bearings, ensuring the effectiveness of fault feature extraction under varying operating conditions. The quality of its output is evaluated using the signal-to-noise ratio (SNR), thereby verifying its effectiveness in extracting fault features from variable-speed bearings. The SNR can be expressed as: .

[0037] Specifically, a particle swarm optimization algorithm is introduced into the adaptive stochastic resonance algorithm to obtain the system cooperative adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximized, including: Define the population size O as 30; learning factors c1=c2=0.2; inertia factor as 0.6; velocity limit range as [-10, 10]; maximum number of iterations as 50; iteration stops when the maximum evolutionary generation is calculated. Otherwise, return to this step and continue the loop until the maximum evolutionary generation is calculated.

[0038] Among them, maximum evolution generation refers to the iteration stage when the optimization algorithm reaches the preset maximum number of iterations.

[0039] Step S5: Provide an angular domain resampling algorithm, and use the angular domain resampling algorithm to calculate the secondary filtered signal to obtain an order spectrum containing bearing fault characteristics.

[0040] Among them, the angular domain resampling method is widely used to extract the vibration characteristics of bearings under variable speed conditions. Angular domain resampling acquires equipment speed information through a speed autoencoder, thereby obtaining the time scale of the constant angular increment corresponding to the resampling. Using the time scale, the non-stationary vibration signal is interpolated and fitted at the constant angular moment to obtain the stationary signal in the angular domain.

[0041] Based on the assumption that the instantaneous angular acceleration of the rotating principal axis is constant, the cumulative rotation angle during the rotation process can be expressed by the following formula: ; In the formula, a 0 , a 1 and a 2 These are the coefficients of the cumulative rotation angle equation.

[0042] Assume the pulse time of the three consecutive sampling points is t 1 , t 2 , t 3 The rotation angles corresponding to consecutive time points are respectively 0 , ∆q ,and 2∆q The incremental relationship between a point in time and its corresponding rotation angle can be expressed as: ; For any rotation angle, we have: ; Let the maximum sampling order be O max Angular sampling frequency O r It can be represented as: ; The angle interval between each rotation corresponding to the number of sampling points ∆q The highest-order analysis under total cumulative rotation angle O max Angular domain sampling frequency corresponding to cumulative rotation angle O r They are respectively: ; ; ; In the formula, N θ The total rotation angle; F co The corresponding characteristic frequency.

[0043] The original vibration signal is processed by a multi-scale adaptive morphological filter operator and an adaptive stochastic resonance algorithm (MSAMF-ASR). The resulting secondary filtered signal is then input into an angular domain resampling algorithm for calculation, resulting in an order spectrum containing bearing fault characteristics.

[0044] Experimental verification: Bearing fault data sampling platform such as Figure 3 As shown, vibration signals were collected for bearings with inner ring failure, outer ring failure, and rolling element failure, respectively. To verify the effectiveness of this method, Gaussian white noise with a signal-to-noise ratio of 0.01 dB was added to the vibration signals of different failure categories. The input speed was manually adjusted using a speed controller, with a range of 20 r / s to 30 r / s; vibration signals were collected from vibration sensors using a data acquisition unit. Deep groove ball bearings with 9 rolling elements, a pitch circle diameter of 38.521 mm, and a rolling element diameter of 7.942 mm were used in the experiment. Specific parameters are listed in Table 1. The test results are as follows: Figures 4-13 As shown.

[0045] Table 1: Main Failure Sequence of Target Bearing

[0046] Figures 4-6 The original signals of bearing outer ring faults, inner ring faults, and rolling element faults are displayed. After noise processing under four operating conditions, the signals are: A: speed increases, B: speed decreases, C: speed first increases then decreases, and D: speed first decreases then increases. The signals processed by the proposed method are compared. Due to the fluctuation of speed, the time interval of the time-domain signal is not fixed, and the trend of amplitude change is related to the trend of speed change, which is reflected both before and after using the proposed method. Furthermore, regardless of the bearing fault type, the vibration signal processed by this invention exhibits a significant convergence effect, and the signal amplitude is reduced after effectively removing background noise.

[0047] To quantify the filtering performance of this invention, the root mean square error (RMSE) was used as the evaluation metric to assess the filtering effect. The results are as follows: Figure 7 As shown: O represents an outer ring fault, I represents an inner ring fault, and B represents a rolling element fault. By comparing the RMSE of different test signals, this invention shows the best filtering effect on the vibration signal of an inner ring fault bearing under conditions of deceleration followed by acceleration, with the lowest corresponding RMSE. Meanwhile, compared with other tests, the filtering effect on the vibration signal of an inner ring fault bearing under deceleration conditions is generally average. (Comparison) Figure 5 (b) The signal processed by this method is smaller than that of the original signal compared to other test groups, which verifies the feasibility of using the RMSE index to evaluate the filtering performance.

[0048] Figures 8-10 The Fourier transform results of the original signal are presented. Due to the continuous variation of rotational speed, the corresponding pulse impacts also exhibit a non-stationary trend, with no fixed interval between impacts. The frequency components in the spectrum show severe drift and scattering, and the time-varying signal exhibits a strong Doppler smearing effect, making it difficult to distinguish the corresponding characteristic frequencies. Therefore, the spectral analysis method based on Fourier transform is not suitable for variable speed operating conditions. These characteristics are reflected in the spectra of bearings with outer ring failure, inner ring failure, and rolling element failure under different rotational speed conditions.

[0049] Figures 11-13 The order spectra of various bearing fault signals processed by the MSAMF-ASR algorithm are presented. The vibration signals processed by the MSAMF algorithm have filtered out some background noise, while the ASR algorithm utilizes residual noise energy to enhance the fault characteristic order, resulting in a clearer overall order spectrum. Specifically, the vibration spectra of different faulty bearings all display the fault characteristic order and its second harmonic (2×Order), while other interference sidebands are negligible relative to the fault orders; the corresponding amplitudes are clearly displayed, but the amplitude energy attenuation between fault orders is relatively severe, and noise and frequency conversion interference are significantly reduced. The MSAMF-ASR combination method demonstrates excellent filtering and fault enhancement effects.

[0050] This invention also discloses a variable speed bearing fault feature extraction device, comprising: The acquisition module is used to acquire the vibration signal of the bearing, which includes several sets of discrete signals; An adaptive morphological filtering module is used to analyze the one-dimensional discrete signal of the discrete signal using morphological operators to obtain the erosion operator and the dilation operator of the one-dimensional discrete signal; cascaded operators are introduced into the erosion operator and the dilation operator to obtain a multi-scale adaptive morphological filtering operator; An adaptive morphological filtering optimization module is used to introduce a particle swarm optimization algorithm into the multi-scale adaptive morphological filtering operator to obtain the key parameters of the adaptive morphological filtering operator, calculate the filtered signals at each scale based on the multi-scale adaptive morphological filtering operator and the key parameters, and add the filtered signals at each scale to obtain the first filtered signal of the vibration signal. An adaptive stochastic resonance optimization module is used to introduce a particle swarm optimization algorithm into the adaptive stochastic resonance algorithm to obtain the system coordinated adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximized, and to use the primary filtered signal as the input signal of the adaptive stochastic resonance algorithm to obtain the secondary filtered signal of the vibration signal. The angular domain resampling module is used to calculate the secondary filtered signal using the angular domain resampling algorithm to obtain an order spectrum containing bearing fault characteristics.

[0051] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0052] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system.

[0053] The above description of the embodiments is only used to provide a detailed introduction to the technical solutions of this application. However, the description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention, and should not be construed as a limitation of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention should be covered within the protection scope of this invention.

Claims

1. A variable speed bearing fault feature extraction method, characterized in that, The application relates to a bearing fault diagnosis method based on a multi-scale adaptive morphological filter and an adaptive stochastic resonance algorithm. The application comprises the following steps: acquiring a bearing vibration signal, wherein the vibration signal comprises several groups of discrete signals; analyzing one-dimensional discrete signals of the discrete signals by using a morphological operator to obtain an erosion operator and a dilation operator of the one-dimensional discrete signals; introducing a cascade operator into the erosion operator and the dilation operator to obtain a multi-scale adaptive morphological filter operator; introducing a particle swarm optimization algorithm into the multi-scale adaptive morphological filter operator to obtain key parameters of the adaptive morphological filter operator, calculating filter signals of various scales according to the multi-scale adaptive morphological filter operator and the key parameters, and adding the filter signals of various scales to obtain a first filter signal of the vibration signal; providing an adaptive stochastic resonance algorithm, introducing a particle swarm optimization algorithm into the adaptive stochastic resonance algorithm to obtain system cooperative regulation parameters of the adaptive stochastic resonance algorithm when a signal-to-noise ratio is maximum, taking the first filter signal as an input signal of the adaptive stochastic resonance algorithm to obtain a second filter signal of the vibration signal; 2. The method of claim 1, wherein, providing an angular domain resampling algorithm, and calculating the second filter signal by using the angular domain resampling algorithm to obtain an order spectrum diagram containing a bearing fault feature. ； The erosion operator is represented by the following formula: ； wherein The dilation operator is represented by the following formula: is a one-dimensional discrete signal, whose domain is f(n) and F(0,1,2... N-1) wherein g (m) is a structure operator of the shape operator, whose domain is n[0,1,2... N-1] and G(0,1,2... M-1) , N>=M .

3. The method of claim 1, wherein, m[0,1,2... M-1] ； In the formula, The adaptive filter operator is represented by the following formula: is the control scale of the multi-scale filtering.

4. The method of claim 1, wherein, The adaptive filter operator is represented by the following formula: The particle swarm optimization algorithm is introduced into the multi-scale adaptive morphological filter operator to obtain the key parameters of the adaptive morphological filter operator, and the particle swarm optimization algorithm comprises the following steps: defining the size O of an optimization population as 20, learning factors c1=c2=0.5, an inertia factor as 0.7, a weight factor of a PSO optimization filter as [0,1], a particle motion range as [0,50], a speed limit range as [-10,10], a length range of a multi-scale adaptive morphological filter operator structure element as an integer in [3,21], and a maximum iteration number as 30 times; inputting the vibration signal into the multi-scale adaptive morphological filter operator, initializing and measuring positions and speeds of parameters needing optimization by using the particle swarm optimization algorithm, inputting measured particle data into a fitness function of the multi-scale adaptive morphological filter operator, and obtaining particle individual speed extreme values and population position extreme values through iterative calculation; 5. The variable speed bearing fault feature extraction method of claim 4, wherein, if any set iteration stop condition is reached, the iteration is stopped, and the key parameters and an iteration curve of the fitness function of the multi-scale adaptive morphological filter operator are output; otherwise, the last step is repeated until the iteration stop condition is met. The iteration stop condition comprises the following steps: reaching the maximum iteration number; the fitness function value of the multi-scale adaptive morphological filter operator being less than a preset threshold value; 6. The method of claim 1, wherein, the position change amount of all particles being less than a set minimum value. ； wherein a , b respectively represent system coordination parameters; A is the signal amplitude; The adaptive stochastic resonance algorithm is represented by the following formula: is the characteristic frequency of the signal; The particle swarm optimization algorithm is introduced into the adaptive stochastic resonance algorithm to obtain the system cooperative regulation parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximum, and the particle swarm optimization algorithm comprises the following steps: is the initial phase of the signal; ​ represents the time delay; D represents the fractal dimension.

7. The method of claim 1, wherein, ​ The size O of the optimized population is defined as 30; the learning factor c1=c2=0.2; the inertia factor is 0.6; the speed limit range is [-10, 10]; the maximum iteration number is 50 times; when the maximum evolution generation is calculated, the iteration stops. Otherwise, return to this step and continue to loop until the maximum evolution generation is calculated.

8. A device for extracting a fault feature of a variable speed bearing, which employs a method for extracting a fault feature of a variable speed bearing according to any one of claims 1 to 7, characterized by Comprise: The acquisition module is used for acquiring the vibration signal of the bearing, and the vibration signal includes a plurality of groups of discrete signals. The adaptive morphological filtering module is used for analyzing one-dimensional discrete signals of the discrete signals by using a morphological operator to obtain an erosion operator and a dilation operator of the one-dimensional discrete signals; a cascade operator is introduced into the erosion operator and the dilation operator to obtain a multi-scale adaptive morphological filtering operator. The adaptive morphological filtering optimization module is used for introducing a particle swarm optimization algorithm into the multi-scale adaptive morphological filtering operator to obtain key parameters of the adaptive morphological filtering operator, calculating a filtering signal of each scale according to the multi-scale adaptive morphological filtering operator and the key parameters, and adding the filtering signals of each scale to obtain a first filtering signal of the vibration signal. The adaptive stochastic resonance optimization module is used for introducing a particle swarm optimization algorithm into the adaptive stochastic resonance algorithm to obtain system coordination adjustment parameters of the adaptive stochastic resonance algorithm when the signal-to-noise ratio is maximum, taking the first filtering signal as an input signal of the adaptive stochastic resonance algorithm, and obtaining a second filtering signal of the vibration signal. The angular domain resampling module is used for calculating the second filtering signal by using the angular domain resampling algorithm to obtain an order spectrum diagram containing bearing fault characteristics.

Images