Method for fault diagnosis of a bearing and computer-readable medium

Abstract

The application relates to a fault diagnosis method and a computer readable medium for a bearing. The fault diagnosis method comprises the following steps: collecting a vibration signal of the bearing; performing N-layer wavelet packet decomposition on the vibration signal to obtain two N subband signals; performing envelope demodulation on each subband signal obtained in the wavelet packet decomposition respectively, and obtaining a corresponding envelope spectrum signal through fast Fourier transform; and performing fault diagnosis according to the envelope spectrum signal. The fault diagnosis method and the computer readable medium have high adaptability and high accuracy.

Description

Technical Field

[0001] This invention relates to the field of bearing testing technology. Specifically, this invention relates to a method for diagnosing bearing faults and a computer-readable medium for performing the method. Background Technology

[0002] Bearings are essential industrial components used to provide rotational support for shafts. For example, wheel bearings are installed in the wheel hubs of vehicles to rotatably support the wheel hubs on the axle. Bearings mainly consist of an outer ring, an inner ring, rolling elements, and a cage. Various types of rolling elements can roll between the outer and inner rings, allowing relative rotation between them. Because these components are subjected to significant alternating loads and frictional forces during bearing operation, they are prone to failure and damage, leading to bearing failure. Common bearing failure modes include pitting and fracture.

[0003] To ensure the safe operation of equipment, it is necessary to monitor the operating condition of bearings. Typically, various types of sensors (such as vibration sensors) are installed at the bearings to collect their operating signals, and signal analysis is used to diagnose whether the bearings have malfunctioned and what kind of malfunction has occurred.

[0004] Currently, bearing fault diagnosis is typically based on envelope demodulation or resonance demodulation methods. Such methods are exemplified in patent documents such as CN 106289775 A. However, for vehicles operating at variable speeds, traditional envelope demodulation methods often require acquiring the original speed signal and then synchronously sampling the vibration signal to accurately identify the fault. Therefore, when using this method for fault diagnosis, in addition to a vibration sensor, a speed sensor is needed to acquire the bearing's original speed signal to achieve speed tracking and overcome the challenges posed by the constantly changing speed of the vehicle during operation. However, due to limitations in production costs and installation space, it is sometimes impossible to install a separate speed sensor in products such as wheel bearings, thus making it difficult to obtain the original speed signal. This makes it difficult for traditional envelope demodulation methods to accurately diagnose bearing faults. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to provide a highly adaptive and accurate fault diagnosis method and computer-readable medium for bearings.

[0006] The aforementioned technical problem is solved by a fault diagnosis method for bearings according to the present invention. This fault diagnosis method includes the following steps:

[0007] Collect vibration signals from the bearing;

[0008] The vibration signal is decomposed into N-level wavelet packets to obtain 2 N Each one carries a signal;

[0009] Each sub-band signal obtained from wavelet packet decomposition is envelope demodulated, and the corresponding envelope spectrum signal is obtained by fast Fourier transform.

[0010] Fault diagnosis is performed based on envelope spectrum signals.

[0011] Wavelet packet decomposition can break down collected vibration signals containing various noises into multiple vibration signals with finite bandwidths. If the number of wavelet packet decomposition levels is appropriately chosen, each sub-band signal can approximate a single-frequency vibration signal, thus decoupling the noisy vibration signal from the target fault vibration signal. Therefore, it is not necessary to synchronize the original vibration signal based on the real-time rotational speed of the bearing to perform envelope spectrum analysis on each sub-band signal, and relatively accurate analysis results can be obtained. Furthermore, for vehicle wheel bearings, this fault diagnosis method is more suitable for execution during vehicle operation.

[0012] According to a preferred embodiment of the present invention, the wavelet packet decomposition can be a five-level wavelet packet decomposition, thereby obtaining 32 sub-band signals. In practice, for bearings, the sub-band signals obtained by the five-level wavelet packet decomposition are basically close to the single-frequency vibration signal, which can meet the general requirements for analytical accuracy.

[0013] According to another preferred embodiment of the present invention, in wavelet packet decomposition, a mother wavelet function with a waveform similar to the target fault impact of the bearing can be selected for wavelet packet decomposition. This gives the wavelet packet decomposition a certain noise reduction effect, thereby further ensuring the accuracy of the diagnostic results. Preferably, the mother wavelet function can be a Morlet function or a Symlet function, or it can be other available functions in the MATLAB program.

[0014] According to another preferred embodiment of the present invention, before performing wavelet packet decomposition on the vibration signal, the fault diagnosis method further includes a step of denoising the acquired vibration signal. Preferably, this denoising process can be based on wavelet soft thresholding, which includes the following steps:

[0015] The vibration signal is decomposed into wavelet coefficients.

[0016] A soft thresholding function is applied to denoise the decomposed wavelet coefficients.

[0017] Wavelet reconstruction is performed on the wavelet coefficients after noise reduction to obtain the noise-reduced vibration signal.

[0018] Various environmental noises often exist in the operating environment of bearings. For example, for wheel hub bearings, various broadband noises from the road surface are generated during vehicle operation. Since the signal after wavelet decomposition is not sensitive to these noises, the original vibration signal is first denoised by wavelet decomposition before wavelet packet decomposition. This attenuates the amplitude of the noisy sub-band signal, thereby effectively removing noise components unrelated to the useful signal from the original signal.

[0019] According to another preferred embodiment of the present invention, fault diagnosis based on envelope spectrum signals may include the following steps:

[0020] The first five peak values ​​are extracted from the envelope spectrum of each sub-band signal, accumulated, and averaged. This average value is then used as the characteristic value of the envelope spectrum signal.

[0021] Fault diagnosis of bearings is performed based on characteristic values.

[0022] In practice, the first five peaks in the envelope spectrum of each sub-band signal are the main peaks, and the average of the first five peaks is generally sufficient to reflect the characteristic values ​​of the sub-band signal relatively accurately.

[0023] According to another preferred embodiment of the present invention, fault diagnosis based on envelope spectrum signals may further include the step of extracting feature values ​​corresponding to the target component based on the fault frequencies of different components of the bearing, thereby performing fault diagnosis on the target component. Therefore, typical faults of each component of the bearing (e.g., outer ring, inner ring, rollers, etc.) can be diagnosed separately.

[0024] According to another preferred embodiment of the present invention, when diagnosing bearing faults based on characteristic values, fault diagnosis can be performed by comparing the characteristic values ​​with a predetermined range of characteristic values. Therefore, bearing faults can be diagnosed more accurately and quickly, and the diagnosis can be easily performed automatically by a computer.

[0025] The aforementioned technical problem is also solved by a computer-readable medium according to the present invention. This computer-readable medium stores a computer program, which, when executed by a computer, implements the fault diagnosis method having the aforementioned features. Attached Figure Description

[0026] The invention is further described below with reference to the accompanying drawings. In the drawings, the same reference numerals represent elements with the same function. Wherein:

[0027] Figure 1 A flowchart of a fault diagnosis method according to a first embodiment of the present invention is shown;

[0028] Figure 2This illustrates a wavelet packet tree model of multi-layer wavelet packet decomposition for a fault diagnosis method according to a first embodiment of the present invention;

[0029] Figure 3 A flowchart illustrating the fault diagnosis steps of a fault diagnosis method according to a first embodiment of the present invention is provided.

[0030] Figure 4 A flowchart illustrating a fault diagnosis method according to a second embodiment of the present invention is shown; and

[0031] Figure 5 A flowchart illustrating the noise reduction processing steps of a fault diagnosis method according to a second embodiment of the present invention is shown. Detailed Implementation

[0032] The following describes specific embodiments of the bearing fault diagnosis method and computer-readable medium according to the present invention, with reference to the accompanying drawings. The detailed description and drawings below are provided to illustrate the principles of the invention, and the invention is not limited to the described preferred embodiments; the scope of protection of the invention is defined by the claims.

[0033] According to embodiments of the present invention, a fault diagnosis method for bearings is provided. This fault diagnosis method, by processing and analyzing bearing vibration signals collected by vibration sensors, can directly diagnose faults in various components of the bearing. The following is a detailed description of this fault diagnosis method based on embodiments of the present invention.

[0034] Figure 1 A flowchart of a fault diagnosis method according to a first embodiment of the present invention is shown. Figure 1 As shown, the fault diagnosis method includes: step S1 of acquiring signals, step S2 of wavelet packet decomposition, step S3 of envelope spectrum analysis, and step S4 of fault diagnosis.

[0035] First, in step S1 of signal acquisition, the raw bearing vibration signal is acquired using a vibration sensor mounted on the bearing. The parameter directly measured by this vibration sensor can be the acceleration of the bearing vibration, which can represent the vibration state of the bearing.

[0036] After acquiring the original vibration signal, step S2 performs N-level wavelet packet decomposition on the vibration signal, thereby decomposing the bearing vibration signal into 2 N These sub-band signals establish 2... N Point wavelet packet tree model. The number of wavelet packet decomposition layers can preferably be selected as five layers, so that 32 decomposed sub-band signals can be obtained from the original vibration signal. Figure 2A wavelet packet tree model with five levels of wavelet packet decomposition is schematically illustrated. In practice, the sub-band signals after five levels of wavelet packet decomposition are essentially close to single-frequency vibration signals and all have finite bandwidth, facilitating further analysis. When performing wavelet packet decomposition, it is necessary to appropriately select the type of mother wavelet function. MATLAB provides several available mother wavelet functions. Preferably, the Morlet function, Symlet function, or other suitable functions can be used. Since the selected mother wavelet functions have waveforms that are quite similar to the target fault impact of the bearing, the wavelet packet decomposition process using these mother wavelet functions also plays a certain role in noise reduction of the vibration signal.

[0037] Next, in step S3, envelope spectrum analysis is performed sequentially on each sub-band signal after wavelet packet decomposition. The specific method for envelope spectrum analysis is as follows: first, envelope demodulation is performed on each sub-band signal obtained from wavelet packet decomposition; then, the corresponding envelope spectrum signal for each sub-band signal is obtained through Fast Fourier Transform (FFT). In this step, since each sub-band signal obtained through wavelet packet decomposition is a signal with finite bandwidth, envelope spectrum analysis can be performed directly on these sub-band signals without the need for preprocessing with bandpass or low-pass filters.

[0038] Finally, in step S4, fault diagnosis can be performed based on the envelope spectrum signal obtained in step S3. In this step, based on the envelope spectrum signal image of each sub-band signal, those skilled in the art can directly analyze and judge various bearing faults based on experience. However, to facilitate the automation of the fault diagnosis method, preferably, this step can also have... Figure 3 The detailed steps are shown below. Specifically, in step S4, the first N peak values ​​related to bearing faults can be extracted from the corresponding envelope spectrum signal of each sub-band signal; then, these N peak values ​​of the envelope spectrum signal are accumulated and averaged, and the average value is used as the feature value of the envelope spectrum signal; finally, the bearing fault is diagnosed based on this feature value. In practice, the first five peak values ​​are sufficient to represent the true characteristics of the envelope spectrum signal, so it is preferable to extract the first five peak values. In step S4, feature values ​​corresponding to the target component can be extracted according to the different fault frequencies of different components of the bearing. Here, the target component is, for example, the outer ring and / or inner ring and / or rolling elements and / or cage of the bearing. Therefore, fault diagnosis can be performed on different target components as needed.

[0039] According to a further preferred embodiment, when diagnosing bearing faults based on characteristic values, fault diagnosis can be performed by comparing the characteristic values ​​of the envelope spectrum signal with a predetermined range of characteristic values. This predetermined range of characteristic values ​​can be obtained empirically and / or calculated; if the characteristic value falls within this predetermined range, it indicates that a corresponding fault has occurred. This predetermined range can be stored in a corresponding system, thus facilitating the automated fault diagnosis process via computer.

[0040] Figure 4 A flowchart of a fault diagnosis method according to a second embodiment of the present invention is shown. Figure 4 As shown, the difference between the second embodiment and the first embodiment is that a noise reduction step is added to the original vibration signal before wavelet packet decomposition.

[0041] In step S1', the original vibration signal of the bearing is acquired in the same manner as in the first embodiment, according to the fault diagnosis method of the second embodiment. In step S2', the acquired original vibration signal is denoised according to the fault diagnosis method of the second embodiment. Then, the denoised vibration signal is subjected to N-level wavelet packet decomposition in step S3', followed by envelope spectrum analysis in step S4', and finally, fault diagnosis is performed based on the envelope spectrum signal in step S5'. In the second embodiment, the steps of signal acquisition, wavelet packet decomposition, envelope spectrum analysis, and fault diagnosis are the same as in the first embodiment, and will not be repeated here. Only the denoising process is described in detail below.

[0042] In this embodiment, the noise reduction process is preferably based on wavelet soft threshold shrinkage technology. Figure 5 A flowchart illustrating the noise reduction process is provided. Figure 5 As shown, the original vibration signal is first decomposed using wavelet decomposition to obtain the decomposed wavelet coefficients. Then, a soft thresholding function is applied to denoise the decomposed wavelet coefficients. Finally, wavelet reconstruction is performed on the denoised wavelet coefficients to obtain the denoised vibration signal. The applied soft thresholding function is expressed as follows:

[0043]

[0044] Here, δ(x) represents the soft thresholding function, x represents the acquired vibration signal, and T represents the threshold. The specific principles and usage of the soft thresholding function are known in the art and will not be elaborated here.

[0045] During vehicle operation, various broadband noises from the road surface are generated, which mix with bearing vibration signals and are collected by sensors. Since the wavelet decomposition result is insensitive to the noise it contains, the amplitude of noisy sub-band signals will attenuate. Therefore, applying a soft thresholding function can effectively remove noise components irrelevant to the useful signal. Thus, adding a noise reduction step can effectively improve the accuracy of fault diagnosis results.

[0046] The fault diagnosis methods according to the various embodiments described above can be executed on a processing device such as a digital signal processor (DSP) and can be stored as a program on various computer-readable media. Accordingly, according to another embodiment of the present invention, a computer-readable medium is also provided, wherein a corresponding computer program is stored, which, when executed by a computer, can implement one or more of the fault diagnosis methods according to the embodiments described above.

[0047] The fault diagnosis method of this invention achieves decoupling between the noise vibration signal and the target fault vibration signal through wavelet packet decomposition. Therefore, envelope spectrum analysis of each sub-band signal can be performed without resampling the original vibration signal using the bearing's real-time rotational speed, yielding relatively accurate analysis results. Consequently, this fault diagnosis method eliminates the need for additional sensors on the axle to detect rotational speed; fault diagnosis can be completed using only existing vibration sensors and the vehicle-provided rotational speed (RPM) signal. This makes the method particularly suitable for space-constrained applications such as vehicle wheel hub bearings and facilitates execution during vehicle operation. Furthermore, this method can effectively extract fault feature values ​​from various bearing components and achieve full-band, multi-scale signal analysis, thereby accurately analyzing and diagnosing bearing faults.

[0048] While possible embodiments have been described exemplarily in the foregoing description, it should be understood that numerous variations of embodiments exist through combinations of all known and readily conceived technical features and implementation methods. Furthermore, it should be understood that the exemplary embodiments are merely examples and do not in any way limit the scope, application, or construction of the invention. The foregoing description is more intended to provide those skilled in the art with technical guidance for transforming at least one exemplary embodiment, wherein various changes, particularly regarding the function and structure of the components, can be made without departing from the scope of the claims.

Claims

1. A method for fault diagnosis of bearings, characterized in that, The fault diagnosis method includes the following steps: The vibration signal of the bearing is collected; The vibration signal is subjected to N-level wavelet packet decomposition to obtain 2 N Each one carries a signal; Each sub-band signal obtained in the wavelet packet decomposition is envelope demodulated, and the corresponding envelope spectrum signal is obtained by fast Fourier transform. Fault diagnosis is performed based on the envelope spectrum signal, which includes the following steps: First, extract the top N peak values ​​related to the bearing fault from the envelope spectrum signal of each sub-band signal; then, accumulate and calculate the average value of the top N peak values ​​of the envelope spectrum signal, and use the average value as the feature value of the envelope spectrum signal; finally, perform fault diagnosis on the bearing based on the feature value.

2. The fault diagnosis method according to claim 1, characterized in that, The wavelet packet decomposition is divided into five layers of wavelet packet decomposition, resulting in 32 sub-band signals.

3. The fault diagnosis method according to claim 1, characterized in that, In the wavelet packet decomposition, a mother wavelet function with a similar waveform to the target fault impact of the bearing is selected for the wavelet packet decomposition.

4. The fault diagnosis method according to claim 3, characterized in that, The mother wavelet function is either a Morlet function or a Symlet function.

5. The fault diagnosis method according to claim 1, characterized in that, Before performing wavelet packet decomposition on the vibration signal, the fault diagnosis method further includes the following steps: The collected vibration signals are subjected to noise reduction processing.

6. The fault diagnosis method according to claim 5, characterized in that, The noise reduction process is based on wavelet soft thresholding, and includes the following steps: The vibration signal is decomposed into wavelet coefficients. A soft thresholding function is applied to denoise the decomposed wavelet coefficients; The wavelet coefficients after noise reduction are reconstructed by wavelet to obtain the noise-reduced vibration signal.

7. The fault diagnosis method according to any one of claims 1 to 6, characterized in that, The first five peak values ​​of each sub-band signal are extracted from the envelope spectrum signal, accumulated, and averaged.

8. The fault diagnosis method according to claim 7, characterized in that, Fault diagnosis based on the envelope spectrum signal also includes the following steps: The feature values ​​corresponding to the target component are extracted based on the failure frequency of different components of the bearing, thereby performing fault diagnosis on the target component.

9. The fault diagnosis method according to claim 7, characterized in that, When diagnosing a bearing fault based on the characteristic value, the fault diagnosis is performed by comparing the characteristic value with a predetermined characteristic value range.

10. A computer-readable medium storing a computer program, characterized in that, When the computer program is executed by a computer, it implements the fault diagnosis method according to any one of claims 1 to 9.

Images