Bearing fault weak signal detection method and system based on stochastic resonance model

By using adaptive mutant particle swarm algorithm to optimize the coupling system parameters in bearing fault detection, a new random resonance model is formed, which solves the problems of low frequency extraction efficiency and noise utilization rate of bearing faults in the existing technology, and realizes complex mechanical vibration Effective matching and identification of signals.

CN118153415BActive Publication Date: 2024-11-29BEIFANG UNIV OF NATITIES

Patent Information

Application Number
CN202311654317.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-05
Publication Date
2024-11-29
Estimated Expiration
2043-12-05

AI Technical Summary

Technical Problem

The prior art is difficult to effectively match complex and diverse mechanical vibration signals, resulting in low frequency extraction efficiency and noise utilization rate of bearing failure.

Method used

The bearing fault detection method based on the stochastic resonance model is adopted, and the coupling system parameters are optimized through the adaptive mutant particle swarm algorithm to improve the signal-to-noise ratio, and the signal-to-noise ratio is introduced as a metric index to form a new coupled multi-steady-state stochastic resonance model to match different periods System parameters under the synergistic effect of force and noise force.

Benefits of technology

It improves the extraction efficiency and noise utilization of fault characteristic frequency, enhances the signal-to-noise ratio, and can more effectively identify bearing fault signals, which is suitable for the detection of complex mechanical vibration signals.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118153415B_ABST
    Figure CN118153415B_ABST
Patent Text Reader

Abstract

The invention discloses a bearing fault weak signal detection method based on a random resonance model, relates to the technical field of bearing fault detection, and comprises the following steps: S1: sampling the bearing to obtain fault data; S2: preprocessing the fault data by means of a secondary sampling technique; S3: sending the preprocessed fault data into a coupling system with a signal-to-noise ratio as an optimization index to obtain the optimal values ​​of a, b, v, r and c by means of an adaptive mutation particle swarm algorithm; S4: substituting the obtained optimal values ​​of a, b, v, r and c into the coupling system to generate random resonance, and obtaining an output signal and a spectrum; The invention also discloses a bearing fault weak signal detection system based on a random resonance model, comprising: a sensor for collecting the vibration signal of the bearing; a processor for executing the above steps S2 to S4; and a display for displaying the output signal and the spectrum, which can match complex and diverse mechanical vibration signals, and improve the efficiency of extracting fault characteristic frequencies and the utilization rate of noise.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of bearing fault detection, and in particular to a bearing fault weak signal detection method and system based on a stochastic resonance model. Background Art

[0002] The concept of stochastic resonance was first proposed by Benzi in 1981 when he studied the interaction between the ancient meteorological glacial period and the warm climate period. Subsequently, Fauve et al. first verified the existence of stochastic resonance in actual physical systems through experiments on Schmidt trigger circuits. Since then, stochastic resonance theory has been widely used to solve key technical problems in different disciplines, such as fault diagnosis, signal enhancement, image enhancement and treatment of hearing loss. Traditional signal detection mainly improves the signal-to-noise ratio by removing or suppressing noise, thereby achieving the purpose of signal detection. However, fault signals are often drowned out by strong background noise from other coupled machine components and the working environment, making some initial faults impossible to identify. Through stochastic resonance, the output signal of a nonlinear dynamic system can enhance the signal-to-noise ratio by adding noise to the system, thereby enabling weak signals to be identified.

[0003] At present, with the rapid development of industry, weak signal detection is becoming more and more important, and nonlinear weak signal detection technology based on stochastic resonance has been widely used and promoted; in practical engineering applications, input signals and noise are not easy to control, so the nonlinear system of stochastic resonance has become a research hotspot; due to the difference in potential functions, stochastic resonance has many nonlinear models, the most classic of which is the one-dimensional Langevin equation model, that is, the classic bistable stochastic resonance system model. In order to solve the problem of output saturation of the classic bistable system, Shi Boqiang et al. proposed a segmented bistable stochastic resonance model, which was applied to the weak fault feature extraction of bearings and achieved good results. However, the classic bistable system generally has only a single type of potential well structure, which is not enough to match complex and diverse mechanical vibration signals, and its efficiency in fault feature extraction and noise utilization are both low. Summary of the invention

[0004] The purpose of the present invention is to provide a bearing fault weak signal detection method based on a random resonance model to solve the problems existing in the above-mentioned prior art, which can match complex and diverse mechanical vibration signals and improve the efficiency of fault characteristic frequency extraction and noise utilization.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The present invention provides a bearing fault weak signal detection method based on a stochastic resonance model, comprising the following steps:

[0007] S1: Sample the vibration signal of the bearing to obtain fault data;

[0008] S2: preprocess the fault data using secondary sampling technology;

[0009] S3: The pre-processed fault data is sent into the coupling system with the signal-to-noise ratio as the optimization index, and the optimal values ​​of a, b, v, r and c are found by the adaptive mutation particle swarm algorithm, where the coupling system is:

[0010]

[0011] Where a and b are non-negative system parameters, v is the potential well depth, r is the potential well width, and c is the steepness of the potential well;

[0012] S4: Substitute the optimal values ​​of a, b, v, r and c into the coupled system to generate random resonance and obtain the output signal and spectrum.

[0013] Preferably, in step S2, the sampling compression ratio R=2400 in the secondary sampling technology.

[0014] Preferably, in step S4, the motion trajectory of the Brownian particle in the potential field in the stochastic resonance is expressed as:

[0015]

[0016] Where a, b, v, r and c are the parameters in the coupling system, S(t) = Asin(ωt), ω is the driving frequency, A is the signal amplitude; ζ(t) is Gaussian white noise, D is the noise intensity amplification factor; sgn(x) is the sign function.

[0017] Preferably, the Runge-Kutta equation is used to find the solution of the expression of the motion trajectory of the Brownian particle in the potential field; wherein the Runge-Kutta equation is:

[0018]

[0019] In the formula, f(x i ,y i ) represents the right-hand side of the expression of the motion trajectory of the Brownian particle in the potential field, and h represents the step length.

[0020] Preferably, in step S3, during the operation of the adaptive mutation particle swarm algorithm, the search range of parameters a, b, v, r and c in the coupling system is [0.1, 10].

[0021] Preferably, in step S3, during the operation of the adaptive mutation particle swarm algorithm, the initialization population size is 40 and the number of iterations is 200.

[0022] The present invention also provides a bearing fault weak signal detection system based on a stochastic resonance model, comprising:

[0023] A sensor for collecting vibration signals of the bearing;

[0024] A processor, used to execute steps S2 to S4 in the bearing fault weak signal detection method based on the stochastic resonance model as described above;

[0025] and a display for displaying the output signal and spectrum.

[0026] Compared with the prior art, the present invention has achieved the following technical effects:

[0027] The present invention provides a bearing fault weak signal detection method and system based on a stochastic resonance model, which adopts a novel coupling system, takes advantage of the adaptive variation particle swarm algorithm, and introduces the signal-to-noise ratio as a measurement index to obtain system parameters of a novel coupled multi-stable stochastic resonance model under the synergistic action of different periodic forces and noise forces, retains the continuity and constraint advantages of the Woods-Saxon potential well model, and has a higher noise utilization rate; compared with the output signal-to-noise ratio of a classical bistable model system, the output signal-to-noise ratio of the coupling system model in the present invention is improved; the coupling system model in the present invention is more efficient in extracting signal characteristic frequencies than the classical bistable system, is conducive to the application of stochastic resonance in engineering practice, and can match complex and diverse mechanical vibration signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 The time domain diagram and frequency domain analysis diagram of the single-frequency input noise signal in Example 3;

[0030] Figure 2 The time domain diagram and frequency domain analysis diagram of the output of the classic bistable system in Example 3;

[0031] Figure 3 It is an iterative curve diagram of the output of the classic bistable system in Example 3;

[0032] Figure 4 The time domain diagram and frequency domain analysis diagram of the coupling system output in Example 3;

[0033] Figure 5 It is an iterative curve diagram of the coupled bistable system output in Example 3;

[0034] Figure 6 The time domain diagram and frequency domain analysis diagram of the multi-frequency input noise signal in Example 4;

[0035] Figure 7 The time domain diagram and frequency domain analysis diagram of the output of the classic bistable system in Example 4;

[0036] Figure 8 It is an iterative curve diagram of the output of the classic bistable system in Example 4;

[0037] Fig. 9 The time domain diagram and frequency domain analysis diagram of the coupling system output in Example 4;

[0038] Fig.10 It is an iterative curve diagram of the coupled bistable system output in the fourth embodiment;

[0039] Fig.11 The time domain diagram and frequency domain analysis diagram of the impact input noise signal in Example 5;

[0040] Fig.12 The time domain diagram and frequency domain analysis diagram of the output of the classic bistable system in Example 5;

[0041] Fig.13 It is an iterative curve diagram of the output of the classic bistable system in Example 5;

[0042] Fig.14 The time domain diagram and frequency domain analysis diagram of the coupling system output in Example 5;

[0043] Fig.15 It is an iterative curve diagram of the coupled bistable system output in Example 5;

[0044] Fig.16 The time domain diagram and frequency domain analysis diagram of the bearing outer ring fault signal in experimental verification 1;

[0045] Fig.17 The time domain diagram and frequency domain analysis diagram of the output of the classic bistable system in experimental verification 1;

[0046] Fig.18 It is the iterative curve diagram of the output of the classic bistable system in experimental verification 1;

[0047] Fig.19 The time domain diagram and frequency domain analysis diagram of the coupled system output in experimental verification 1;

[0048] Fig. 20 It is the iterative curve diagram of the output of the coupled bistable system in experimental verification 1;

[0049] Fig.21 The time domain diagram and frequency domain analysis diagram of the bearing inner ring fault signal in experimental verification 2;

[0050] Fig. 22 The time domain diagram and frequency domain analysis diagram of the output of the classic bistable system in experimental verification 2;

[0051] Fig.23 It is the iterative curve diagram of the output of the classic bistable system in experimental verification 2;

[0052] Fig.24 The time domain diagram and frequency domain analysis diagram of the coupled system output in experimental verification 2;

[0053] Fig.25 Iterative curve diagram of the output of the coupled bistable system in experimental verification 2. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The purpose of the present invention is to provide a bearing fault weak signal detection method based on a random resonance model to solve the problems existing in the above-mentioned prior art, which can match complex and diverse mechanical vibration signals and improve the efficiency of fault characteristic frequency extraction and noise utilization.

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Embodiment 1

[0058] The present invention provides a bearing fault weak signal detection method based on a stochastic resonance model. The bearing fault weak signal detection method based on a stochastic resonance model specifically comprises the following steps:

[0059] S1: Sample the vibration signal of the bearing through the vibration sensor to obtain fault data;

[0060] S2: preprocess the fault data by secondary sampling technology, where the sampling compression ratio R=2400;

[0061] Specifically, the secondary sampling technology is an existing technology, which has been introduced in detail in the paper "Preliminary Study on the Research and Application of Secondary Sampling Stochastic Resonance Spectrum" published by Leng Yonggang, Wang Taiyong, Qin Xuda, etc.

[0062] S3: The pre-processed fault data is sent into the coupling system with the signal-to-noise ratio as the optimization index, and the optimal values ​​of a, b, v, r and c are found by the adaptive mutation particle swarm algorithm, where the coupling system is:

[0063]

[0064] Where a and b are non-negative system parameters, v is the potential well depth, r is the potential well width, and c is the steepness of the potential well;

[0065] Specifically, during the operation of the adaptive mutation particle swarm algorithm, the search range of the parameters a, b, v, r and c in the coupling system is [0.1, 10], the initialization population size is 40, and the number of iterations is 200.

[0066] S4: Substitute the optimal values ​​of a, b, v, r and c into the coupled system to generate stochastic resonance and obtain the output signal and spectrum;

[0067] Among them, in stochastic resonance, the motion trajectory of Brownian particles in the potential field is described by the Langevin equation, which describes the relationship between the velocity, acceleration and random force of Brownian particles. The Langevin equation is:

[0068]

[0069] Where U(x) is the coupling system; S(t) = Asin(ωt), ω is the driving frequency, A is the signal amplitude; ζ(t) is Gaussian white noise, and D is the noise intensity amplification factor;

[0070] Substituting the coupled system into the above equation, the expression of the motion trajectory of the Brownian particle in the potential field in stochastic resonance is obtained as follows:

[0071]

[0072] Where a, b, v, r and c are the parameters in the coupling system, S(t) = Asin(ωt), ω is the driving frequency, A is the signal amplitude; ζ(t) is Gaussian white noise, D is the noise intensity amplification factor; sgn(x) is the sign function.

[0073] Use the Runge-Kutta equation to find the solution to the expression of the motion trajectory of a Brownian particle in a potential field; the Runge-Kutta equation is:

[0074]

[0075] In the formula, f(x i ,y i ) represents the right-hand side of the expression of the motion trajectory of the Brownian particle in the potential field, and h represents the step length.

[0076] The above-mentioned bearing fault weak signal detection method based on the stochastic resonance model adopts a new coupling system, takes advantage of the adaptive mutation particle swarm algorithm, and introduces the signal-to-noise ratio as a measurement indicator to obtain the system parameters of the new coupled multi-stable stochastic resonance model under the synergistic action of different periodic forces and noise forces, retains the continuity and constraint advantages of the Woods-Saxon potential well model, and has a higher noise utilization rate; compared with the output signal-to-noise ratio of the classical bistable model system, the output signal-to-noise ratio of the coupled system model in the present invention is improved; the coupled system model in the present invention is more efficient in extracting signal characteristic frequencies than the classical bistable system, which is helpful for the application of stochastic resonance in engineering practice and can match complex and diverse mechanical vibration signals.

[0077] Embodiment 2

[0078] This embodiment provides a bearing fault weak signal detection system based on a stochastic resonance model, including:

[0079] A sensor is used to collect the vibration signal of the bearing. Specifically, the sensor is a vibration sensor that can be purchased on the market;

[0080] A processor, used to execute steps S2 to S4 in the bearing fault weak signal detection method based on the stochastic resonance model in the first embodiment; wherein the processor is a computer, a single-chip microcomputer or a chip, etc.;

[0081] And a display, for displaying the output signal and spectrum; the display is specifically a computer screen, an LED display screen, etc.

[0082] Embodiment 3

[0083] This embodiment provides an experimental simulation of single-frequency signal feature extraction using the bearing fault weak signal detection method based on the stochastic resonance model in the first embodiment. The preset conditions are: the input signal is a single-frequency signal: S(t) = Asin(2πf0t), where A = 0.5, f0 = 0.02HZ, and the sampling frequency is f s =5HZ, the sampling point is 4096, and a Gaussian white noise with an intensity of 3dB is added. The signal-to-noise ratio is 17.14dB at this time. The initialization population is 40 and the number of iterations is 200. The search range of the parameters of the classical bistable system is [0.1,10], and the search range of the parameters in the coupled system is also [0.1,10].

[0084] The specific process is: the noisy single-frequency signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Figure 4 The coupled system output noise signal time domain, frequency domain diagram and Figure 5 Iteration graph of the coupled system shown.

[0085] The comparative example is: the coupling system in the first embodiment is replaced by a classical bistable system, and the noisy single-frequency signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Figure 2 The coupled system output noise signal time domain, frequency domain diagram and Figure 3 Iteration curve diagram of the classic bistable system shown.

[0086] Figure 1 It is the time domain diagram and frequency domain analysis diagram of a single-frequency input noise signal. In the time domain diagram, the signal is submerged by noise and the characteristics of the signal cannot be distinguished. From the frequency domain diagram, it can be seen that there is a peak with a spectrum amplitude of 0.311 at 0.02HZ, but there are some interference frequencies around it, and the characteristic frequency is not very obvious. The noise signal is input into the classic bistable system, and the parameters are optimized through the adaptive mutation particle swarm optimization algorithm. Figure 2 The figure shows the time domain and frequency domain diagrams of the output noise signal of the classic bistable system. From the time domain diagram, the waveform is significantly improved. In the frequency domain diagram, the spectrum amplitude at 0.02HZ is increased to 1.12, which is a significant improvement compared with the previous one. The interference frequency of the surrounding noise is greatly reduced. Figure 3 This is the iteration curve of the classic bistable system. The algorithm converges at T=46, and the fitness value is 18.72dB, corresponding to parameters a=3.01, b=1.67. Figure 4 The figure shows the time domain and frequency domain diagrams of the coupled system output plus noise signal. From the time domain diagram, the waveform is further improved compared with the classical bistable system. In the frequency domain diagram, the spectrum amplitude at 0.02HZ is increased to 2.25, which is 1.13 higher than the output of the classical bistable system, and the characteristic frequency can be better identified. Figure 5 It is the iterative curve diagram of the coupled system. The algorithm converges at T=88, and the fitness value is 21.41dB, which is 2.69dB higher than the classical bistable system. The corresponding parameters are a=0.11, b=0.10, v=6.70, c=0.12, and r=6.47. Therefore, the coupled system of the present invention is significantly better than the classical bistable system in extracting the characteristic frequency of the single-frequency signal.

[0087] Embodiment 4

[0088] This embodiment provides an experimental simulation of multi-frequency signal feature extraction using the bearing fault weak signal detection method based on the random resonance model in Embodiment 1. The preset conditions are: the input signal is a multi-frequency signal: S(t)=Asin(2πf0t)+Bsin(2πf1t)+Csin(2πf2t), where A=0.5, B=0.5, C=0.5, f0=0.01HZ, f1=0.05HZ, f2=0.08HZ, and the sampling frequency is f s=5HZ, the sampling point is 4096, and a Gaussian white noise with an intensity of 3dB is added. The signal-to-noise ratio is 12.21dB, the initialization population is 40, and the number of iterations is 200. The search range of a and b in the classical bistable system is [0.1, 10], and the search range of a, b, v, c, r in the coupled system is also [0.1, 10].

[0089] The specific process is: the noisy multi-frequency signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Fig. 9 The coupled system output noise signal time domain, frequency domain diagram and Fig.10 Iteration graph of the coupled system shown.

[0090] The comparative example is: the coupling system in the first embodiment is replaced by a classical bistable system, and the noisy single-frequency signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Figure 7 The coupled system output noise signal time domain, frequency domain diagram and Figure 8 Iteration curve diagram of the classic bistable system shown.

[0091] Figure 6 It is the time domain diagram and frequency domain analysis diagram of the multi-frequency input noise signal. In the time domain diagram, the signal is submerged by the noise and the characteristics of the signal cannot be distinguished; from the frequency domain diagram, it can be seen that there are peaks with spectrum amplitudes of 0.36, 0.53, and 0.41 at 0.01Hz, 0.05Hz, and 0.08Hz, respectively, but there are still some interference frequencies around them, especially the noise intensity after 0.1HZ is too large, and the characteristic frequency is not very obvious. The noise signal is input into the classic bistable system, and the particle swarm optimization algorithm is used to optimize the parameters. Figure 7 The figure shows the time domain and frequency domain diagrams of the output noise-added signal of the classic bistable system. From the time domain diagram, the waveform is significantly improved. In the frequency domain diagram, the spectrum amplitude at 0.1Hz is increased to 1.15, which is significantly improved compared with the previous one, and the interference frequency of the surrounding noise is greatly reduced, but there is no obvious improvement at 0.05Hz and 0.08Hz. Figure 8 It is the iteration curve of the classic bistable system. The algorithm converges at T=138, the fitness value is 21.35Hz, and the corresponding parameters are a=3.86 and b=1.54. Fig. 9 The figure shows the time domain and frequency domain diagrams of the coupled system output noise signal. From the time domain diagram, the waveform is further improved compared with the classical bistable system. In the frequency domain diagram, the spectrum amplitudes at 0.01Hz, 0.05Hz, and 0.08Hz are increased to 1.89, 1.51, and 0.74, respectively, which are 0.74, 1.01, and 0.31 higher than the characteristic frequencies extracted from the classical bistable system, respectively, and the characteristic frequencies can be better identified. Fig.10It is the iterative curve diagram of the coupled system. The algorithm converges at T=92, and the fitness value is 23.19dB, which is 1.84dB higher than the classical bistable system. The corresponding parameters are a=0.10, b=0.10, v=3.76, c=5.34, and r=7.92. Therefore, the coupled system in the present invention is significantly better than the classical bistable system in extracting the characteristic frequency of the multi-frequency signal.

[0092] Embodiment 5

[0093] This embodiment provides an experimental simulation of extracting the characteristics of an impact signal using the bearing fault weak signal detection method based on the stochastic resonance model in the first embodiment. The preset conditions are: the input signal is an impact signal: Where A is the displacement constant, ω n is the natural frequency, ξ is the damping coefficient, A=0.5, ω n =0.01,ξ=0.05, the sampling frequency is f s =5HZ, the sampling point is 4096, and a Gaussian white noise with an intensity of 2dB is added. The signal-to-noise ratio is -21.97dB at this time. The initialization population is 40 and the number of iterations is 200. The search range of a and b in the classical bistable system is [0.1, 10], and the search range of a, b, v, c, r in the coupled system is also [0.1, 10].

[0094] The specific process is: the noisy impulse signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Fig.14 The coupled system output noise signal time domain, frequency domain diagram and Fig.15 Iteration graph of the coupled system shown.

[0095] The comparative example is: the coupling system in the first embodiment is replaced by a classical bistable system, and the noisy impulse signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Fig.12 The coupled system output noise signal time domain, frequency domain diagram and Fig.13 Iteration curve diagram of the classic bistable system shown.

[0096] Fig.11 The time domain diagram and frequency domain analysis diagram of the input noise impact signal. In the time domain diagram, the signal is submerged by Gaussian white noise, and the characteristics of the signal cannot be distinguished; the peak of the characteristic frequency of 0.1Hz is almost invisible in the frequency domain diagram, and there are a lot of interference frequencies around it. The noise impact signal is input into the classic bistable system, and the particle swarm optimization algorithm is used to optimize the parameters. Fig.12 The figure shows the time domain and frequency domain diagrams of the output noise-added signal of the classic bistable system. From the time domain diagram, the waveform is improved. In the frequency domain diagram, the spectrum amplitude at 0.1Hz is increased to 0.44, which is an increase of 0.22. Moreover, the interference frequency of the noise is greatly reduced after 0.05Hz. Fig.13It is the iteration curve of the classic bistable system. The algorithm converges at T=161, the fitness value is 19.88dB, and the corresponding parameters are a=3.21 and b=2.18. Fig.14 The figure shows the time domain and frequency domain diagrams of the coupled system output noise signal. From the time domain diagram, the waveform is further improved compared with the classical bistable system. In the frequency domain diagram, the spectrum amplitude at 0.1Hz is increased to 1.57, which is 1.13 higher than the characteristic frequency extracted by the classical bistable system, and the characteristic frequency can be better identified. Fig.15 It is the iteration curve of the coupled system. The algorithm converges at T=145, and the fitness value is 21.04dB, which is 1.16dB higher than the classical bistable system. The corresponding parameters are a=0.10, b=0.10, v=2.25, c=1.66, and r=9.00. Therefore, the coupled system in the present invention is significantly better than the classical bistable system in extracting the characteristic frequency of the multi-frequency signal.

[0097] Embodiment 6

[0098] This embodiment provides an experimental verification of a deep groove bearing vibration fault using the bearing fault weak signal detection method based on the stochastic resonance model in the first embodiment.

[0099] The outer and inner ring data of deep groove ball bearings of model 6205-2RS JEM SKF were selected for analysis experiments to verify the practicality and efficiency of the coupling system in the present invention in bearing fault detection, and the detection effect of bearing fault signals was compared with that of the classic bistable system; the bearing fault data came from the electrical engineering laboratory of Case Western Reserve University (CWRU), and the main parameters of the deep groove ball bearings are shown in Table 1; the sampling frequency f s =12kHZ, number of sampling points N=10000, bearing speed r=1797r / min. The fault characteristic frequencies that may appear in rolling bearing fault detection are mostly in the range of tens to hundreds of Hz, which does not meet the adiabatic approximation theory. Therefore, secondary sampling technology must be used to pre-process the high-frequency signal, and its sampling compression ratio R=2400. In physics, when a bearing fails, periodic pulses of vibration signals will be generated. Therefore, the fault type can be determined by detecting and analyzing the pulse period; the fault characteristic frequencies of the bearings are shown in Table 2, where the natural frequency f S =29.93Hz.

[0100] Table 1 Main calculation parameters of rolling bearings

[0101]

[0102] Table 2 Characteristic frequency of rolling bearings

[0103] Bearing elements Inner Circle Outer ring Cage Rolling element Fault frequency / HZ <![CDATA[5.4152f r ]]> <![CDATA[3.5848f r ]]> <![CDATA[0.3928f r ]]> <![CDATA[1.7135f r ]]>

[0104] This article only uses the inner and outer ring fault signals of rolling bearings for analysis. The fault signal frequency calculation formula is as follows:

[0105]

[0106] In the formula, f r is the rotation frequency of the bearing, α is the contact angle of the rolling element, and the value is generally α=0. In addition, n and D correspond to the bearing parameters in Table 1. The fault frequency f of the outer ring of the bearing can be calculated from the above formula. BPFO =107.3HZ, the fault frequency of the inner ring is f BPFI =162.2HZ.

[0107] Experimental verification 1 ~ Bearing outer ring failure example

[0108] The specific process is: the bearing outer ring fault signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Fig.19 The coupled system output noise signal time domain, frequency domain diagram and Fig. 20 Iteration graph of the coupled system shown.

[0109] The comparative example is: the coupling system in the first embodiment is replaced by a classic bistable system, and the bearing outer ring fault signal is processed according to steps S2 to S4 in the first embodiment, and the following is obtained: Fig.17 The coupled system output noise signal time domain, frequency domain diagram and Fig.18 Iteration curve diagram of the classic bistable system shown.

[0110] Fig.16 The time domain waveform and spectrum of the bearing outer ring fault signal are shown in Figure 2. The fault type cannot be identified, and there are a lot of spike interferences in the high frequency band of 2500-3500 Hz. The high frequency outer ring envelope signal is subsampled (subsampling frequency fsr = 5 Hz) and sent to the classic bistable system to generate random resonance with the signal-to-noise ratio as the optimization index. The optimal parameters are a = 1, b = 1, and the signal-to-noise ratio is 0.34 dB. The output signal and spectrum are shown in Figure 2. Fig.17 As shown in the figure, it can be seen that the spectrum has a peak at f = 108Hz (within the allowable error range). After the fault signal is sent to the coupling system in the same way, the optimal system parameters are found to be a = 0.10, b = 0.10, v = 7.10, c = 5.93, r = 8.45, and the signal-to-noise ratio SNR = 22.37dB; the output signal and spectrum are shown in Fig.19 As shown, compared with the classic bistable system, the interference frequencies around the fault frequency are eliminated a lot, and the burrs in the high-frequency band are almost gone, proving that there is a fault in the outer ring. This verifies the effectiveness of the coupling system in weak signal detection, and it is better than the classic bistable system.

[0111] Experimental verification 2 ~ Bearing inner ring failure example

[0112] The specific process is: the bearing inner ring fault signal is processed according to steps S2 to S4 in the first embodiment to obtain the following: Fig.24 The coupled system output noise signal time domain, frequency domain diagram and Fig.25 Iteration graph of the coupled system shown.

[0113] The comparative example is: the coupling system in the first embodiment is replaced by a classic bistable system, and the bearing inner race fault signal is processed according to steps S2 to S4 in the first embodiment, and the following is obtained: Fig. 22 The coupled system output noise signal time domain, frequency domain diagram and Fig.23 Iteration curve diagram of the classic bistable system shown.

[0114] Fig.21 The time domain waveform and spectrum diagram of the bearing inner ring fault signal. From the time domain diagram and spectrum diagram, the fault frequency f is almost unrecognizable. BPFI =162Hz. Similarly, the high-frequency inner envelope signal is sent to the classic bistable system after secondary sampling, and the self-mutating particle swarm algorithm is used to find the optimal system parameters a=1, b=1, signal-to-noise ratio SNR=22.75dB. The output signal and spectrum are as follows Fig. 22 As shown in the figure, it can be seen that the spectrum has a peak at f = 162Hz (within the allowable error range). After the fault signal is sent to the coupling system in the same way, the optimal system parameters are found to be a = 0.10, b = 3.54, v = 9.91, c = 1.59, r = 4.95, SNR = 24.29dB, and the output signal and spectrum are shown in Fig.24 As shown, compared with the classic bistable system, the interference frequencies around the fault frequency are eliminated a lot, and the burrs in the high-frequency band are almost gone, which proves that there is a fault in the inner ring. This once again verifies the effectiveness of the proposed coupling system in weak signal detection, and it is better than the classic bistable system.

[0115] In summary, the present invention provides a new coupling system model, which uses an adaptive mutation particle swarm algorithm to find the optimal parameters of the system in single-frequency, multi-frequency and impact signal environments, and is finally verified in the bearing fault signal; and has the following advantages: (1) The model retains the continuity and constraint advantages of the Woods-Saxon potential well model, and has a higher noise utilization rate; (2) Compared with the output signal-to-noise ratio of the classical bistable model system, the output signal-to-noise ratio of the new coupling system model is improved; (3) The new coupling model system is more efficient in extracting signal characteristic frequencies than the classical bistable system, which is conducive to the application of stochastic resonance in engineering practice.

[0116] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A bearing fault weak signal detection method based on a stochastic resonance model, characterized in that: The steps include: S1: Sample the vibration signal of the bearing to obtain fault data; S2: preprocess the fault data using secondary sampling technology; S3: The pre-processed fault data is sent into the coupling system with the signal-to-noise ratio as the optimization index, and the optimal values ​​of a, b, v, r and c are found by the adaptive mutation particle swarm algorithm, where the coupling system is: Where a and b are non-negative system parameters, v is the potential well depth, r is the potential well width, and c is the steepness of the potential well; S4: Substitute the optimal values ​​of a, b, v, r and c into the coupled system to generate stochastic resonance and obtain the output signal and spectrum; The expression of the motion trajectory of Brownian particles in the potential field in stochastic resonance is: Where a, b, v, r and c are the parameters in the coupling system, S(t) = Asin(ωt), ω is the driving frequency, A is the signal amplitude; is Gaussian white noise, D is the noise intensity amplification factor; sgn(x) is the sign function; Use the Runge-Kutta equation to find the solution to the expression of the motion trajectory of a Brownian particle in a potential field; the Runge-Kutta equation is: In the formula, f(x i ,y i ) represents the right-hand side of the expression of the motion trajectory of the Brownian particle in the potential field, and h represents the step length.

2. The method for detecting weak signals of bearing faults based on a stochastic resonance model according to claim 1 is characterized in that: In step S2, the sampling compression ratio R=2400 in the secondary sampling technique.

3. The method for detecting weak signals of bearing faults based on a stochastic resonance model according to claim 1 is characterized in that: In step S3, during the operation of the adaptive mutation particle swarm algorithm, the search range of the parameters a, b, v, r and c in the coupling system is [0.1, 10].

4. The method for detecting weak signals of bearing faults based on a stochastic resonance model according to claim 1 is characterized in that: In step S3, during the operation of the adaptive mutation particle swarm algorithm, the initial population size is 40 and the number of iterations is 200.

5. A bearing fault weak signal detection system based on a stochastic resonance model, characterized in that: include: A sensor for collecting vibration signals of the bearing; A processor, configured to execute steps S2 to S4 of the method for detecting weak signals of bearing faults based on a stochastic resonance model as claimed in any one of claims 1 to 4; and a display for displaying the output signal and spectrum.

Citation Information

Patent Citations

  • Fault signal detection method of plasticity controllable slope stochastic resonance

    CN115935143A

  • Crystal growing furnace bearing fault detection method based on adaptive multistable stochastic resonance

    CN116990024A

Cited By

  • Weak signal detection method based on improved under-damping unsaturated bistable stochastic resonance model

    CN121327464A