Rolling Bearing Fault Diagnosis Method Based on Tunable Quality Factor Wavelet Threshold Denoising

Abstract

The present invention relates to a rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising, which includes: 1) Collecting the state signals during the operation of the rolling bearing and calculating the corresponding fault characteristic frequencies when damage occurs to each component of the test bearing; 2) Constructing a continuous threshold function including a transition band; 3) Adaptively constructing adjustable quality factor wavelet parameters matching the fault characteristic signals based on the vibration test signals; 4) Obtaining the signal after optimal adjustable quality factor wavelet threshold denoising; 5) Diagnosing the fault type of the rolling bearing based on the envelope spectrum of the denoised signal. The present invention combines the advantages of the adjustable quality factor wavelet with flexible oscillation attributes of the basis function and the powerful denoising ability of threshold denoising, and can analyze and diagnose the vibration signals of rolling bearings in a strong noise environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of rolling bearing fault diagnosis, and particularly relates to a method for analyzing and diagnosing rolling bearing vibration signals and faults based on adjustable quality factor wavelet threshold denoising. Background Art

[0002] Rolling bearings are known as the "joints of industry" and are key supporting and transmission components of major electromechanical equipment such as aero-engines and wind turbines. Monitoring the vibration state of rolling bearings using acceleration sensors has the advantages of convenient installation and low cost, and has become a common method for monitoring the operating state of rolling bearings.

[0003] Local damage to the components of rolling bearings will generate periodic impact fault components in the vibration signal. However, these fault components are often submerged in the vibration signals of other components and a large amount of background noise. Extracting fault components from the bearing vibration monitoring signal through signal processing methods has become the key to rolling bearing fault diagnosis. Wavelet threshold denoising is a widely used fault feature extraction technology and has also been verified in actual cases. Wavelet threshold denoising is essentially developed based on the inner product matching principle, and its fault feature extraction effect mainly depends on two aspects: one is the similarity between the wavelet basis function and the impact fault feature signal. The more similar the wavelet basis function is to the fault impact signal, the decomposition coefficient of the fault feature signal based on the inner product transformation will be significantly greater than the decomposition coefficients of the noise and other vibration signals; the other is the selection of the threshold function. The most common threshold functions are the hard threshold and soft threshold functions. By setting the wavelet decomposition coefficients below the threshold to zero, the purpose of denoising is achieved. However, both the hard threshold and soft threshold functions have certain limitations, including the discontinuity points of the hard threshold will cause oscillations in the reconstructed signal, and the overall reduction of the wavelet coefficients by the soft threshold makes the reconstructed fault feature signal lower than the actual value. The existing research work on wavelet threshold denoising mainly focuses on the transformation of the threshold function, and the wavelet basis function is mostly selected as the dyadic wavelet basis function. The dyadic wavelet basis function has fixed basis function oscillation attributes and time-frequency division characteristics, and it is difficult to adaptively match fault feature signals with different oscillation attributes, thus weakening the ability of wavelet threshold denoising to extract rolling bearing fault features in a strong noise environment. Therefore, simultaneously solving the problems of adaptive matching construction of the wavelet basis function and the selection of the threshold function is very important for extracting the fault features of rolling bearings using the wavelet threshold denoising framework. Summary of the Invention

[0004] Technical Problems to be Solved

[0005] Aiming at the problems existing in the existing wavelet threshold denoising technology that the fixed wavelet basis function is difficult to adaptively match the fault feature signals with different oscillation attributes and the ability to extract weak fault features of rolling bearings in a strong noise environment is weak, the present invention provides a rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising.

[0006] Technical solution

[0007] A rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising, characterized by the following steps:

[0008] Step 1: Collect the vibration signal x(n) during the operation of the rolling bearing through the acceleration sensor installed on the bearing housing, where n = 1, 2,..., N, and N is the length of the discrete signal; based on the pitch diameter, contact angle, number of rollers, and rotational speed of the shaft of the test bearing, calculate the corresponding fault characteristic frequencies f i of different components of the rolling bearing in the case of local damage, where i = 1, 2, 3, 4 correspond to the outer ring, inner ring, rolling element, and cage components of the rolling bearing respectively;

[0009] Step 2: Construct a threshold function that is continuous at the threshold and has the same large amplitude coefficient as the original value, expressed as follows:

[0010]

[0011] In the formula: λ is the threshold adaptively determined by the Stein unbiased likelihood estimation principle for the adjustable quality factor wavelet decomposition coefficients; α is the transition band adjustment coefficient and satisfies 0 < α < 1;

[0012] Step 3: Based on the vibration signal of the rolling bearing, adaptively construct a matching adjustable quality factor wavelet basis function First, construct the parameter space Γ{Q, r} of the adjustable quality factor wavelet transform, where Q is the quality factor and r is the redundancy; for each group of parameter combinations (Q, r) in the parameter space, construct the corresponding adjustable quality factor wavelet basis function ψ Q,r , and then decompose the vibration signal x(n) to obtain the wavelet decomposition coefficients ω j of each layer, where j = 1, 2,...J is the wavelet decomposition layer and J is the maximum decomposition layer; use the threshold function in Step 2 to reset the wavelet decomposition coefficients ω j of each layer and obtain the denoised signal through the inverse adjustable quality factor wavelet transform Calculate the envelope signal harmonic-to-noise ratio of, and draw the envelope signal harmonic-to-noise ratio spectrum diagram under different parameter combinations (Q, r); select the parameter combination (Q*, r*) corresponding to the maximum envelope signal harmonic-to-noise ratio and as the optimally matched adjustable quality factor wavelet basis function;

[0013] Step 4: Adopt the most matched wavelet basis function with adjustable quality factor Select the corresponding denoised signal in Step 3 as the wavelet threshold denoised signal with adjustable quality factor

[0014] Step 5: Calculate the Hilbert envelope spectrum of the wavelet threshold denoised signal with adjustable quality factor and determine the faulty components of the rolling bearing according to the main frequency components in the envelope spectrum

[0015] A further technical solution of the present invention: The specific process of the said Step 3 is as follows

[0016] 3.1) The control parameters of the wavelet with adjustable quality factor include the quality factor Q and the redundancy r, and construct the corresponding parameter space Γ

[0017] Γ = {(Q, r)|Q = 2, 3,... 15; r = 2, 3,... 10}

[0018] 3.2) For each parameter combination (Q, r) in the parameter space Γ, construct the corresponding wavelet basis function ψ Q,r of the wavelet with adjustable quality factor, and then perform J-layer decomposition on the vibration monitoring signal x(n) to obtain the wavelet coefficients ω j of each layer

[0019] ω j = TQWT[x(n)], j = 1, 2,..., J

[0020] where: TQWT[·] is the wavelet transform with adjustable quality factor, j is the wavelet decomposition layer, J is the maximum decomposition layer, and each layer of wavelet basis function can be expressed as ψ Q,r (j);

[0021] 3.3) Based on the modulus |ψ Q,r (j)| of each layer of wavelet basis function, normalize the wavelet coefficients of each layer to obtain the normalized wavelet coefficients ω' j of each layer

[0022] ω' j = ω j / |ψ Q,r (j)|, j = 1, 2,..., J

[0023] For the set {ω' j} of the normalized wavelet coefficients of each layer, adaptively determine the normalized threshold λ' through the Stein unbiased likelihood estimation principle, and the threshold of each layer of coefficients of the adjustable quality wavelet is calculated as

[0024] λ j= λ′|ψ Q,r (j)|, j = 1, 2, ..., J

[0025] 3.4) Select the transition band adjustment coefficient as α = 0.5 - 0.6, and reset each layer of coefficients ω j through the threshold function constructed in step 2. The reset coefficients of each layer are denoted as ω j :

[0026]

[0027] After resetting each layer of coefficients through the inverse wavelet transform with an adjustable quality factor, the denoised signal is reconstructed as:

[0028]

[0029] where: iTQWT[·] is the inverse wavelet transform with an adjustable quality factor;

[0030] 3.5) Calculate the envelope signal X(n) of, and then calculate the harmonic signal-to-noise ratio of the envelope signal

[0031]

[0032] In the formula: corr<·> is the autocorrelation function, and max(·) is the operation of taking the maximum value; characterizes the energy ratio of the periodic component to the noise component in the envelope signal;

[0033] 3.6) Calculate the harmonic signal-to-noise ratio of the envelope signal for each parameter combination (Q, r) in the parameter space Γ in sequence Respectively, with the quality factor Q and redundancy r as the coordinate axes, plot the corresponding harmonic signal-to-noise ratio spectrogram of the envelope signal, and select the parameter combination corresponding to the maximum harmonic signal-to-noise ratio in the figure as the adjustable quality factor wavelet parameters (Q*, r*) matching the fault characteristic signal:

[0034]

[0035] A further technical solution of the present invention: The judgment criterion for the faulty component of the rolling bearing in step 5 is: Based on the Hilbert envelope spectrum of the signal after adjustable quality factor wavelet threshold denoising, if its main frequency component is a certain fault characteristic frequency f i of the rolling bearing calculated in step 1 and its multiple frequency components, it indicates that the component corresponding to this fault characteristic frequency in the rolling bearing has local damage.

[0036] A further technical solution of the present invention: α = 0.5 - 0.6 described in step 2.

[0037] A computer system, characterized in that it includes: one or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the above method.

[0038] A computer-readable storage medium, characterized in that it stores computer-executable instructions which, when executed, are used to implement the above method.

[0039] Beneficial effects

[0040] A rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising provided by the present invention, based on the advantage that the adjustable quality factor wavelet has a flexible oscillation property of the wavelet basis function, can adaptively construct an adjustable quality factor wavelet basis function matching the fault feature signal based on the vibration test signal, and at the same time constructs a continuous threshold function, which can realize the extraction of the fault features of the rolling bearing under a strong noise environment and has a wide application prospect.

[0041] Compared with the prior art, the beneficial effects of the present method are as follows:

[0042] a. The constructed optimal adjustable quality factor wavelet basis function is an adjustable quality factor wavelet basis function that is driven by the rolling bearing vibration monitoring signal, adaptively constructed and matching the oscillation property of the fault feature signal; based on the inner product matching principle, its decomposition coefficients for the fault feature signal and the noise signal have a more significant discrimination degree compared with the traditional method.

[0043] b. The constructed threshold function ensures the continuity of the threshold function through the setting of the transition band, avoiding the discontinuous points of the traditional hard threshold function; at the same time, taking the original value for the large amplitude coefficients also avoids the coefficient reduction in the traditional soft threshold function; it can make the reconstructed signal better restore the fault feature signal. The present invention combines the advantage of the adjustable quality factor wavelet with a flexible oscillation property of the basis function and the powerful noise reduction ability of threshold denoising, and can analyze and diagnose the vibration signal of the rolling bearing under a strong noise environment. Description of the drawings

[0044] The drawings are only for the purpose of showing specific embodiments and are not considered to be a limitation of the present invention. Throughout the drawings, the same reference signs denote the same components.

[0045] Figure 1 is a flowchart of the method of the present invention.

[0046] Figure 2 is the vibration test signal and its spectrogram of a certain test cylindrical roller bearing. Among them, (a) is the time-domain waveform of the vibration acceleration signal; (b) is the spectrum of the acceleration signal.

[0047] Figure 3 is Figure 2 (a) The envelope signal harmonic-to-noise ratio spectrogram after threshold denoising of the signal under different adjustable quality factor wavelet control parameters.

[0048] Figure 4 is the preferred adjustable quality factor wavelet basis function and its normalized frequency band division characteristics.

[0049] Figure 5 is Figure 2 (a) The reconstructed signal of the signal based on adjustable quality factor wavelet threshold denoising.

[0050] Figure 6 is Figure 5 The Hilbert envelope spectrum of the denoised signal, and the fault characteristics of the outer ring of the rolling bearing can be detected in the figure. Detailed implementation manners

[0051] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0052] Combined with Figures 1 to 6 , the content of the present invention will be further described as follows:

[0053] A rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising, as Figure 1 shown, the main steps are as follows:

[0054] Step 1: Acquisition of rolling bearing vibration signals and calculation of fault characteristic frequencies, including:

[0055] 1.1) Install an acceleration sensor on the bearing housing where the rolling bearing is located to collect the vibration signal during the operation of the rolling bearing, denoted as x(n), n = 1, 2,..., N, where N is the length of the discrete signal.

[0056] 1.2) Based on the pitch diameter, contact angle, number of rollers, roller diameter and rotational speed of the rolling bearing, calculate the fault characteristic frequencies f corresponding to different components of the rolling bearing in the case of local damage, where i (Guo Liang, Chen Zhiqiang, Gao Hongli, Feng Liwen, Li Changgen, Lei Yuncong. A method and system for monitoring the faults of train bogie bearings based on dictionary learning, CN 111307455A), where i = 1, 2, 3, 4 correspond to the outer ring, inner ring, rolling element, and cage components of the rolling bearing respectively.

[0057] Step 2: Construct a threshold function that is continuous at the threshold and has the same large amplitude coefficients as the original values, which is expressed as follows:

[0058]

[0059] In the formula: λ is the threshold adaptively determined by the Stein unbiased risk estimator (SURE) principle for the wavelet decomposition coefficients of the tunable quality factor; α is the transition band adjustment coefficient and satisfies 0 < α < 1, and generally takes the value α = 0.5 - 0.6.

[0060] The constructed threshold function divides the threshold into three intervals: The wavelet decomposition coefficients of the tunable quality factor in the interval (-αλ,αλ) are set to zero; the intervals (-λ / α,-αλ) ∪ (αλ,λ / α) are the transition bands, and in this interval, the absolute values of the wavelet decomposition coefficients of the tunable quality factor gradually increase. The setting of the transition band ensures the continuity of the threshold function and avoids the discontinuity points of the traditional hard threshold function; the wavelet decomposition coefficients of the tunable quality factor in the intervals (-∞,-λ / α) ∪ (λ / α,∞) take the original values, avoiding the coefficient reduction in the traditional soft threshold function.

[0061] Step 3: Adaptive construction of a tunable quality factor wavelet basis function that matches the vibration signal of the rolling bearing Including:

[0062] 3.1) The control parameters of the tunable quality factor wavelet include the quality factor Q and the redundancy r, and construct the corresponding parameter space Γ:

[0063] Γ = {(Q,r)|Q = 2,3,...15; r = 2,3,...10}

[0064] The quality factor Q controls the oscillation property of the wavelet basis function, and the wavelet basis function corresponding to a higher quality factor Q has a higher oscillation property; the redundancy r determines the width of the frequency domain transition band of the wavelet basis function; there are 126 parameter combinations (Q,r) in the constructed parameter space.

[0065] 3.2) For each parameter combination (Q,r) in the parameter space Γ, construct the corresponding tunable quality factor wavelet basis function ψ Q,r (Ivan W. Selesnick. Wavelet transform with tunable Q - factor. IEEE Transactions on Signal Processing, 2011), and then perform J - layer decomposition on the vibration monitoring signal x(n) to obtain the wavelet coefficients ω of each layer j :

[0066] ω j= TQWT[x(n)], j = 1, 2, ..., J

[0067] where: TQWT[·] is the tunable quality factor wavelet transform, j = 1, 2, ..., J is the wavelet decomposition layer, J is the maximum decomposition layer, and the wavelet basis function of each layer can be expressed as ψ Q,r (j).

[0068] 3.3) Based on the modulus of the wavelet basis function of each layer |ψ Q,r (j)|, normalize the wavelet coefficients of each layer to obtain the normalized wavelet coefficients ω′ j :

[0069] ω′ j = ω j / |ψ Q,r (j)|, j = 1, 2, ..., J

[0070] For the set of normalized wavelet coefficients of each layer {ω′ j}, adaptively determine the normalized threshold λ′ through the Stein's unbiased risk estimator (SURE) principle. On this basis, the coefficient threshold of each layer of the tunable quality wavelet can be calculated as:

[0071] λ j = λ′|ψ Q,r (j)|, j = 1, 2, ..., J

[0072] 3.4) Select the transition band adjustment coefficient as α = 0.5 - 0.6, and reset the coefficients ω of each layer through the threshold function constructed in step 2. The coefficients of each layer after reset are expressed as j After resetting, the coefficients of each layer are[[ID=3�]]

[0073]

[0074] Reconstruct the denoised signal through the inverse tunable quality factor wavelet transform of the coefficients of each layer after reset

[0075]

[0076] where: iTQWT[·] is the inverse tunable quality factor wavelet transform. [[ID=5६]]

[0077] 3.5) Calculate the envelope signal X(n) of the denoised signal , and then calculate the harmonic signal-to-noise ratio of the envelope signal

[0078]

[0079] Where: corr<·> is the autocorrelation function, and max(·) is the operation of taking the maximum value. It characterizes the energy ratio between the periodic component and the noise component in the envelope signal.

[0080] 3.6) Calculate the harmonic signal-to-noise ratio of the envelope signal for each parameter combination (Q, r) in the parameter space Γ in sequence Taking the quality factor Q and the redundancy r as the coordinate axes respectively, plot the corresponding harmonic signal-to-noise ratio spectrum of the envelope signal, and select the parameter combination corresponding to the maximum harmonic signal-to-noise ratio in the figure as the adjustable quality factor wavelet parameter (Q*, r*) matching the fault feature signal:

[0081]

[0082] Step 4: Adopt the adjustable quality factor wavelet basis function with the best match Select the corresponding denoised signal in Step 3 as the adjustable quality factor wavelet threshold denoised signal

[0083] Step 5: Calculate the Hilbert envelope spectrum of the adjustable quality factor wavelet threshold denoised signal and determine the faulty components of the rolling bearing according to the main frequency components in the envelope spectrum, including:

[0084] 5.1) Based on the Hilbert envelope spectrum of the signal after adjustable quality factor wavelet threshold denoising, if its main frequency components are a certain fault feature frequency f i calculated in Step 1 of the rolling bearing and its multiple frequency components, it indicates that the components (outer ring, inner ring, rolling element, cage) corresponding to this fault feature frequency in the rolling bearing have local damage.

[0085] 5.2) Based on the Hilbert envelope spectrum of the signal after adjustable quality factor wavelet threshold denoising, if there are no obvious fault feature frequencies and their multiple frequency components, it indicates that the rolling bearing is in good operating condition.

[0086] Embodiment

[0087] The present invention will be further described below in conjunction with specific embodiments.

[0088] This embodiment is a double-row cylindrical roller bearing with the outer ring fixed to the bearing housing and the inner ring connected to the rotating shaft. Each row contains 16 cylindrical rollers. Before the experiment started, a sound rolling bearing was installed. The rotating shaft and the bearing were driven by a motor to conduct a full-life experiment. After the experiment ended, it was confirmed by disassembling the parts that spalling damage occurred on the outer ring of the bearing. During the experiment, the rotating speed of the rotating shaft was kept constant at 2000 revolutions per minute. The vibration signal during the operation of the bearing was monitored by an acceleration sensor installed on the outer side of the bearing housing shell, and the sampling frequency was 20 kHz. According to the geometric dimensions of the rolling bearing used in the experiment and the rotating speed of the rotating shaft, the corresponding fault characteristic frequencies of each component of the rolling bearing were calculated as shown in Table 1.

[0089] Table 1 Fault characteristic frequencies of each component of the cylindrical roller rolling bearing used in the experiment

[0090]

[0091] Figure 2 (a) and 2(b) are respectively a segment of the vibration acceleration signal of the rolling bearing collected during the experiment and its spectrogram. The test signal lasts for 0.15 s. It can be seen that the time-domain waveform and spectral components of the vibration signal are relatively complex, and significant periodic fault impacts and other characteristics cannot be observed. Therefore, the fault characteristic signal of the outer ring of the bearing at this acquisition moment may be submerged in other vibrations and environmental noises of the test bench and is difficult to directly detect from the time domain or frequency domain. Next, the proposed adjustable quality factor wavelet threshold denoising based on the present invention is used to analyze the test vibration signal.

[0092] Construct an adjustable quality factor wavelet parameter space Γ = {(Q, r)|Q = 2, 3,... 15; r = 2, 3,... 10}. For each group of parameter combinations (Q, r) in the parameter space, establish the corresponding adjustable quality factor wavelet basis function ψ Q,r , and then Figure 2 The vibration signal in (a) is decomposed into 15 layers to obtain the wavelet coefficients of each layer. The wavelet coefficients of each layer are normalized according to the modulus of the wavelet basis function of each layer. Thus, the threshold of the normalized wavelet coefficients is estimated by unbiased likelihood, the transition band adjustment coefficient α = 0.5 is selected, and then the wavelet coefficients of each layer are reset through the threshold function. The signal after threshold denoising can be obtained from the reset wavelet coefficients through the adjustable quality factor wavelet inverse transform. Calculate the envelope signal harmonic-to-noise ratio index of this signal, and plot the envelope signal harmonic-to-noise ratio calculated under different parameter combinations (Q, r) on a graph with the redundancy r as the abscissa and the quality factor Q as the ordinate, as Figure 3 shown.

[0093] Figure 3The maximum envelope signal-to-noise ratio is obtained under the parameters Q* = 2 and r* = 3. Therefore, these parameters are determined as the adjustable quality factor wavelet parameters that best match the fault impact characteristics. The adjustable quality factor wavelet basis function constructed from these parameters and its frequency band division characteristics are as Figure 4 shown. Using the adjustable quality factor wavelet basis function ψ(Q* = 2, r* = 3) optimized in Figure 4 , the adjustable quality factor wavelet threshold denoising is performed on the vibration signal in Figure 2 (a), and the result is as Figure 5 shown.

[0094] It can be seen that: Figure 5 Periodic impact signals can be clearly observed in the signal obtained after adjustable quality factor wavelet threshold denoising in Figure 6 . is the Hilbert envelope spectrum of the denoised signal, and its main frequency components are the fault characteristic frequency of the outer ring of the rolling bearing and its multiple frequency components, indicating that local damage has occurred on the outer ring of the rolling bearing at the vibration signal test time.

[0095] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising, characterized in that: The steps are as follows: Step 1: Collect the vibration signal x(n) during the operation of the rolling bearing through the acceleration sensor installed on the bearing housing, where n = 1, 2, …, N and N is the length of the discrete signal; based on the pitch diameter, contact angle, number of rollers, and rotational speed of the shaft of the test bearing, calculate the corresponding fault characteristic frequencies f i , where i = 1, 2, 3, 4 correspond to the outer ring, inner ring, rolling element, and cage components of the rolling bearing, respectively; Step 2: Construct a threshold function that is continuous at the threshold and has the same large amplitude coefficient as the original value, which is expressed as follows: In the formula: λ is the threshold adaptively determined by the Stein unbiased likelihood estimation principle for the wavelet decomposition coefficients of the adjustable quality factor; α is the transition band adjustment coefficient and satisfies 0 < α < 1; Step 3: Adaptively construct an adjustable quality factor wavelet basis function that matches the rolling bearing vibration signal First, construct the parameter space Γ{Q, r} of the adjustable quality factor wavelet transform, where Q is the quality factor and r is the redundancy; for each set of parameter combinations (Q, r) in the parameter space, construct the corresponding adjustable quality factor wavelet basis function ψ Q,r , and then decompose the vibration signal x(n) to obtain the wavelet decomposition coefficients ω of each layer j , where j = 1, 2,... J is the wavelet decomposition layer and J is the maximum decomposition layer; use the threshold function in Step 2 to reset the wavelet decomposition coefficients ω of each layer j and obtain the denoised signal through the inverse transform of the adjustable quality factor wavelet Calculate the envelope signal harmonic-to-noise ratio of, and plot the envelope signal harmonic-to-noise ratio spectrum under different parameter combinations (Q, r); select the parameter combination (Q*, r*) and ψ corresponding to the maximum envelope signal harmonic-to-noise ratio as the optimally matched adjustable quality factor wavelet basis function; Q*,r* ​ Step 4: Adopt the most matched wavelet basis function ψ with adjustable quality factor Q*,r* , and select the corresponding denoised signal in Step 3 as the wavelet threshold denoised signal with adjustable quality factor Step 5: Calculate the Hilbert envelope spectrum of the adjustable quality factor wavelet threshold denoised signal, and determine the faulty component of the rolling bearing according to the main frequency components in the envelope spectrum. ​ 2. A rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising according to claim 1, characterized in that: The specific process of the said Step 3 is as follows: 3.1) The control parameters of the adjustable quality factor wavelet include the quality factor Q and the redundancy r, and construct the corresponding parameter space Γ: Γ = {(Q, r)|Q = 2, 3,... 15; r = 2, 3,... 10} 3.2) For each parameter combination (Q, r) in the parameter space Γ, construct the corresponding adjustable quality factor wavelet basis function ψ Q,r , and then perform J-level decomposition on the vibration monitoring signal x(n) to obtain wavelet coefficients ω at each level j : ω j = TQWT[x(n)], j = 1, 2, ..., J Among them: TQWT[·] is the tunable quality factor wavelet transform, j is the wavelet decomposition layer, J is the maximum decomposition layer, and the wavelet basis function of each layer is expressed as ψ Q,r (j); 3.3) Based on the modulus |ψ Q,r (j)| of the wavelet basis functions of each layer, the wavelet coefficients of each layer are normalized to obtain the normalized wavelet coefficients ω j ′: ω′ j = ω j / |ψ Q,r (j)|, j = 1, 2,..., J For each set of normalized wavelet coefficient {ω′ j}, the normalized threshold λ′ is adaptively determined by the principle of Stein's unbiased risk estimator, and the coefficient thresholds of each layer of the adjustable quality wavelet are calculated as follows: λ j = λ′|ψ Q,r (j)|, j = 1, 2, ..., J 3.4) Select the transition zone adjustment coefficient as α = 0.5 - 0.6, and reset each layer coefficient ω through the threshold function constructed in step 2. j After resetting, each layer coefficient is expressed as The coefficients of each layer after reset Through the inverse wavelet transform with an adjustable quality factor, the denoised signal is reconstructed as follows: Where: iTQWT[·] is the inverse transform of the adjustable quality factor wavelet; 3.5) Calculate the envelope signal X(n), and then calculate the harmonic signal-to-noise ratio of the envelope signal wherein: corr<·> is the autocorrelation function, and max(·) is the operation of taking the maximum value; characterizes the energy ratio between the periodic component and the noise component in the envelope signal; 3.6) Calculate the harmonic signal-to-noise ratio of the envelope signal for each parameter combination (Q, r) in the parameter space Γ in sequence Taking the quality factor Q and the redundancy r as the coordinate axes respectively, plot the corresponding harmonic signal-to-noise ratio spectrum of the envelope signal, and select the parameter combination corresponding to the maximum harmonic signal-to-noise ratio in the figure as the adjustable quality factor wavelet parameter (Q*, r*) matching the fault feature signal:

3. A rolling bearing fault diagnosis method based on adjustable quality factor wavelet threshold denoising according to claim 1, characterized in that: The judgment criterion for the faulty components of the rolling bearing in Step 5 is as follows: In the Hilbert envelope spectrum of the signal after wavelet threshold denoising with an adjustable quality factor, if its main frequency components are a certain fault characteristic frequency f of the rolling bearing calculated in Step 1 i and its multiple frequency components, it indicates that local damage has occurred to the components corresponding to the fault characteristic frequency in the rolling bearing.

4. A computer system, characterized in that Including: One or more processors, a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in claim 1.

5. A computer-readable storage medium, characterized in that Stored with computer-executable instructions, the instructions are used to implement the method described in claim 1 when executed.

Images