A bearing fault intelligent analysis method and device

Through local mean decomposition and similarity matching technology, the problems of subjectivity and noise interference in bearing failure analysis are solved, the analysis accuracy is improved, and it is suitable for intelligent diagnosis of bearing failures.

CN118484730BActive Publication Date: 2024-11-12WUXI KEY SENSOR PHOTONICS TECH CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410942021.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-12
Estimated Expiration
2044-07-15

AI Technical Summary

Technical Problem

The accuracy of intelligent analysis of bearing faults is poor, mainly because the interpretation of vibration data depends on professional knowledge, subjectivity and noise interference. The existing analysis methods have limitations in non-stationary signal processing, resulting in aliasing phenomenon and spectrum leakage.

Method used

The local mean decomposition method is used to denoising and reconstruct the bearing vibration signal, match it with the normal and abnormal signal databases through similarity calculation, determine the fault type, and optimize the signal processing with sliding average and Fourier transform technology.

Benefits of technology

It improves the accuracy of bearing failure analysis, reduces labor costs, reduces misjudgment rate, and enhances the value of industrial applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118484730B_ABST
    Figure CN118484730B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of mechanical fault detection, and specifically discloses a bearing fault intelligent analysis method and device. The vibration signal to be measured of the target bearing is denoised and reconstructed by a local mean decomposition method, and the reconstructed signal to be measured is input into a normal operating signal library for similarity matching to determine whether the target bearing is abnormal. When the reconstructed signal to be measured is abnormal, the reconstructed signal to be measured is input into an abnormal operating signal library for further similarity matching to determine the abnormality type. The present invention effectively solves the problem that it is difficult for ordinary operators to effectively understand the use of bearing vibration data for bearing fault analysis, and the judgment of bearing faults is highly subjective and inaccurate, thereby improving the accuracy of bearing fault analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault detection, and in particular to a bearing fault intelligent analysis method and a bearing fault intelligent analysis device. Background Art

[0002] Bearings are important parts of mechanical equipment. Their main function is to support mechanical rotating bodies and reduce the mechanical load friction coefficient of mechanical equipment during transmission. Therefore, the quality of bearings greatly affects the performance of mechanical equipment. During the operation of mechanical equipment, bearings need to be tested to monitor the bearing status and avoid the operation of mechanical equipment affected by bearing failure. However, in actual operation, there are many factors that lead to poor accuracy of intelligent analysis of bearing failures, mainly as follows:

[0003] (1) Since bearing vibration data often contains a lot of information, including frequency, amplitude, phase, etc., accurate interpretation of these data requires rich professional knowledge and experience. It is difficult for ordinary operators to effectively understand and use the above bearing vibration data for bearing fault analysis, and the judgment of bearing faults is highly subjective and inaccurate;

[0004] (2) Since there are a lot of noise and vibration sources in the operating environment of mechanical equipment, these external factors may interfere with the measurement of bearing vibration, resulting in misjudgment or omission;

[0005] (3) Currently, the commonly used analysis methods for bearing vibration data include windowed Fourier transform, wavelet analysis, and Wigner-Ville distribution (WVD). Since these analysis methods are based on Fourier transform as the ultimate theoretical basis, Fourier transform has limitations when analyzing non-stationary signals. It may cause aliasing and spectrum leakage, thus affecting the extraction and judgment of bearing vibration data.

[0006] In summary, how to provide a bearing fault intelligent analysis method with higher accuracy has become a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0007] The present invention provides a bearing fault intelligent analysis method and a bearing fault intelligent analysis device, which solve the problem of great subjectivity and inaccuracy in the judgment of bearing faults existing in the related art.

[0008] As a first aspect of the present invention, a bearing fault intelligent analysis method is provided, comprising:

[0009] Obtain the vibration signal to be measured during the daily operation of the target bearing;

[0010] The vibration signal to be measured is subjected to denoising and reconstruction processing by using a local mean decomposition method to obtain a reconstructed signal to be measured;

[0011] According to the type of the target bearing corresponding to the reconstructed signal to be tested, a number of target normal reconstructed signals corresponding to the target sample bearing in the normal operating signal library are obtained, and the target sample bearing is of the same type as the target bearing; wherein the normal operating signal library includes several types of sample bearings, and each sample bearing corresponds to several normal reconstructed signals that have been subjected to denoising and reconstruction processing;

[0012] Calculate the similarity between the reconstructed signal to be tested and the target normal reconstructed signal respectively. If the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is less than the normal similarity threshold, determine that the reconstructed signal to be tested is an abnormal signal, and input the reconstructed signal to be tested into the abnormal operation signal library;

[0013] According to the type of the target bearing corresponding to the reconstructed signal to be tested, a number of target abnormal reconstructed signals corresponding to the target sample bearing in the abnormal operation signal library are obtained, and the target sample bearing is of the same type as the target bearing; wherein the abnormal operation signal library includes several types of sample bearings, each sample bearing corresponds to several abnormal types, and each abnormal type corresponds to several abnormal reconstructed signals that have been subjected to denoising and reconstruction processing;

[0014] The similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is calculated respectively. If the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal is greater than or equal to the abnormal similarity threshold, the abnormality type of the reconstructed signal to be tested is determined according to the abnormality type of the corresponding target abnormal reconstructed signal.

[0015] Furthermore, the vibration signal to be measured is subjected to denoising and reconstruction processing by a local mean decomposition method to obtain a reconstructed signal to be measured, including:

[0016] The vibration signal to be measured is sampled according to a preset sampling frequency to obtain a discrete signal to be measured;

[0017] Obtain all local extreme points in the discrete signal to be measured;

[0018] Calculate the local mean and local amplitude of each local extreme point;

[0019] Perform sliding average processing on the local mean and local amplitude of each local extreme point respectively to obtain the local mean function and local amplitude function of the discrete signal to be measured;

[0020] According to the local mean function and the local amplitude function, a zero mean function and a pure frequency modulation function of the discrete signal to be measured are obtained;

[0021] Extracting the product function and the corresponding residual signal in the discrete signal to be measured according to the zero mean function and the pure frequency modulation function of the discrete signal to be measured;

[0022] The discrete signal to be measured is reconstructed according to the product function and the residual signal to obtain the reconstructed signal to be measured.

[0023] Furthermore, obtaining the local extreme point of the discrete signal to be measured includes:

[0024] The discrete signal to be measured is amplified and enhanced by using a loss function to obtain the amplified and enhanced discrete signal to be measured, and the signal-to-noise ratio of the amplified and enhanced discrete signal to be measured is equal to the signal-to-noise ratio of the discrete signal to be measured;

[0025] Obtain all local extreme points in the discrete signal to be measured after amplification and enhancement.

[0026] Furthermore, the local mean and local amplitude of each local extreme point are respectively subjected to sliding average processing to obtain the local mean function and local amplitude function of the discrete signal to be measured, including:

[0027] Calculate the initial step size between adjacent local extreme points;

[0028] Calculate the step size mean of the sliding average processing according to the initial step size between adjacent local extreme value points;

[0029] Calculate the step length standard deviation of the sliding average processing based on the step length mean;

[0030] The step length of the sliding average processing is obtained according to the step length mean and the step length standard deviation;

[0031] According to the step size of the sliding average processing, the local mean value and the local amplitude of each local extreme point are respectively subjected to sliding average processing to obtain the local mean function and the local amplitude function of the discrete signal to be measured.

[0032] Furthermore, the present invention provides a bearing fault intelligent analysis method, which further includes:

[0033] Merge and optimize several normal reconstruction signals corresponding to each type of sample bearing in the normal operation signal library to obtain normal comparison signals corresponding to each type of sample bearing;

[0034] According to the normal comparison signal corresponding to each type of sample bearing, an optimized normal operation signal library is formed, wherein the optimized normal operation signal library includes several types of sample bearings, and each type of sample bearing corresponds to a normal comparison signal;

[0035] Merge and optimize a plurality of abnormal reconstruction signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library to obtain an abnormal comparison signal corresponding to each abnormal type of each type of sample bearing;

[0036] An optimized abnormal operation signal library is formed according to the abnormal comparison signal corresponding to each abnormal type of each type of sample bearing, wherein the optimized abnormal operation signal library includes several types of sample bearings, each sample bearing corresponds to several abnormal types, and each abnormal type corresponds to an abnormal comparison signal.

[0037] Furthermore, a plurality of normal reconstructed signals corresponding to each type of sample bearing in the normal operation signal library are merged and optimized to obtain a normal comparison signal corresponding to each type of sample bearing, including:

[0038] transforming each normal reconstructed signal in the normal operating signal library into a normal frequency domain signal by fast Fourier transform processing;

[0039] Extracting the maximum intensity normal frequency domain signal and the minimum intensity normal frequency domain signal from a number of normal frequency domain signals corresponding to each type of sample bearing in the normal operation signal library;

[0040] Calculate the average value of the maximum intensity normal frequency domain signal and the minimum intensity normal frequency domain signal of each type of sample bearing to obtain the normal comparison signal of each type of sample bearing;

[0041] The abnormal reconstruction signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library are merged and optimized to obtain the abnormal comparison signals corresponding to each abnormal type of each type of sample bearing, including:

[0042] Each abnormal reconstructed signal in the abnormal operation signal library is transformed into an abnormal frequency domain signal by fast Fourier transform processing;

[0043] Extracting the maximum intensity abnormal frequency domain signal and the minimum intensity abnormal frequency domain signal from a number of abnormal frequency domain signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library;

[0044] The average values ​​of the maximum intensity abnormal frequency domain signal and the minimum intensity abnormal frequency domain signal of each abnormal type of each type of sample bearing are calculated to obtain the abnormal comparison signal of each abnormal type of each type of sample bearing.

[0045] Furthermore, after performing denoising and reconstruction processing on the vibration signal to be measured and obtaining the reconstructed signal to be measured, the method further includes:

[0046] The reconstructed signal to be measured is converted into a frequency domain signal to be measured by fast Fourier transform processing;

[0047] According to the type of the target bearing corresponding to the frequency domain signal to be measured, a target normal comparison signal corresponding to a target sample bearing of the same type as the target bearing in the optimized normal operating signal library is obtained;

[0048] Calculate the similarity between the frequency domain signal to be tested and the target normal comparison signal. If the similarity between the frequency domain signal to be tested and the target normal comparison signal is less than the normal similarity threshold, determine that the frequency domain signal to be tested is an abnormal signal, and input the frequency domain signal to be tested into the optimized abnormal operation signal library;

[0049] According to the type of the target bearing corresponding to the frequency domain signal to be measured, a number of target abnormal comparison signals corresponding to the target sample bearings of the same type as the target bearing in the optimized abnormal operation signal library are obtained;

[0050] The similarity between the frequency domain signal to be tested and each target abnormal comparison signal is calculated respectively. If the similarity between the frequency domain signal to be tested and the target abnormal comparison signal is greater than or equal to the abnormal similarity threshold, the abnormal type of the frequency domain signal to be tested is determined according to the abnormal type of the corresponding target abnormal comparison signal.

[0051] Furthermore, the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is calculated respectively, including:

[0052] The similarity between the reconstructed signal to be tested and each target normal reconstructed signal is calculated according to the Spearman coefficient;

[0053] The similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is calculated respectively, including:

[0054] The similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is calculated according to the Spearman coefficient.

[0055] Furthermore, the present invention provides an intelligent analysis method for bearing faults, which also includes: if the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is less than the normal similarity threshold, and the similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is less than the abnormal similarity threshold, then the reconstructed signal to be tested is determined as the unknown abnormal type of the corresponding target sample bearing and added to the abnormal operation signal library.

[0056] As another aspect of the present invention, there is provided a bearing fault intelligent analysis device, comprising:

[0057] A signal acquisition module is used to obtain the vibration signal to be measured during the daily operation of the target bearing;

[0058] A denoising and reconstruction module is used to perform denoising and reconstruction processing on the vibration signal to be measured to obtain a reconstructed signal to be measured;

[0059] A normal operating signal library search module is used to obtain a number of target normal reconstruction signals corresponding to target sample bearings in the normal operating signal library according to the type of the target bearing corresponding to the reconstructed signal to be tested, and the target sample bearings are of the same type as the target bearings; wherein the normal operating signal library includes a number of types of sample bearings, and each sample bearing corresponds to a number of normal reconstruction signals that have been subjected to denoising and reconstruction processing;

[0060] A normal operation similarity calculation module is used to calculate the similarity between the reconstructed signal to be tested and each target normal reconstructed signal respectively. If the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is less than the normal similarity threshold, the reconstructed signal to be tested is determined to be an abnormal signal, and the reconstructed signal to be tested is input into the abnormal operation signal library;

[0061] The abnormal operation signal library search module is used to obtain a number of target abnormal reconstruction signals corresponding to the target sample bearings in the abnormal operation signal library according to the type of the target bearing corresponding to the reconstructed signal to be tested, and the target sample bearings are of the same type as the target bearings; wherein the abnormal operation signal library includes a number of types of sample bearings, each sample bearing corresponds to a number of abnormal types, and each abnormal type corresponds to a number of abnormal reconstruction signals that have been subjected to denoising and reconstruction processing;

[0062] The abnormal operation similarity calculation module is used to calculate the similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal respectively. If the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal is greater than or equal to the abnormal similarity threshold, the abnormal type of the reconstructed signal to be tested is determined according to the abnormal type of the corresponding target abnormal reconstructed signal.

[0063] The present invention denoises and reconstructs the vibration signal to be measured of the target bearing through a local mean decomposition method, and inputs the denoised and reconstructed signal to be measured into a normal operating signal library, matches it through similarity calculation, and determines whether the target bearing is abnormal. When the reconstructed signal to be measured is abnormal, the reconstructed signal to be measured is input into the abnormal operating signal library for further matching through similarity calculation to determine the abnormal type of the target bearing. The bearing fault intelligent analysis method provided by the embodiment of the present invention effectively solves the problem that it is difficult for ordinary operators to effectively understand and use bearing vibration data for bearing fault analysis, and there is a large subjectivity and inaccuracy in the judgment of bearing faults, thereby improving the accuracy of bearing fault analysis and greatly saving labor costs, and having great industrial application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.

[0065] Figure 1The present invention provides a flow chart of a bearing fault intelligent analysis method.

[0066] Figure 2 A specific flow chart of a bearing fault intelligent analysis method provided by the present invention.

[0067] Figure 3 A flow chart of obtaining a reconstructed signal to be tested provided by the present invention.

[0068] Figure 4 The present invention provides a flow chart for obtaining local extreme value points of a discrete signal to be measured.

[0069] Figure 5 The present invention provides , and Schematic diagram of the relationship.

[0070] Figure 6 The present invention provides a flow chart for obtaining a local mean value function and a local amplitude function of a discrete signal to be measured.

[0071] Figure 7 It is a histogram provided by the present invention for auxiliary analysis of sliding average processing.

[0072] Figure 8 It is a schematic diagram of a bearing fault intelligent analysis device provided by the present invention. DETAILED DESCRIPTION

[0073] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0074] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment 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 making creative work should fall within the scope of protection of the present invention.

[0075] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0076] The embodiment of the present invention provides a bearing fault intelligent analysis method, such as Figure 1 and Figure 2 As shown, including:

[0077] S100, obtaining a vibration signal to be measured during daily operation of a target bearing;

[0078] In the embodiment of the present invention, the vibration signal to be measured during the daily operation of the target bearing can be obtained by a vibration signal acquisition card, a vibration sensor, etc. arranged on the target bearing. The vibration signals to be measured obtained in this embodiment are all time domain signals.

[0079] S200, performing denoising and reconstruction processing on the vibration signal to be measured by using a local mean decomposition method to obtain a reconstructed signal to be measured;

[0080] As an adaptive time-frequency analysis method, the local mean decomposition method can gradually decompose a complex multi-component signal into the sum of several product functions and a residual component through multiple loop iterations according to the complexity and changing rules of the signal itself. Compared with traditional Fourier and wavelet analysis, the embodiment of the present invention removes Gaussian white noise from the vibration signal to be measured by the local mean decomposition method and reconstructs it to obtain the reconstructed signal to be measured. It has very good adaptability to the analysis of non-stationary signals. The following intelligent analysis of bearing faults is performed based on the reconstructed signal to be measured, which can improve the accuracy of fault analysis.

[0081] In addition, it should be understood that the reconstructed signal to be measured obtained after denoising and reconstructing the vibration signal to be measured by the local extreme value decomposition method is still a time domain signal.

[0082] S300, according to the type of the target bearing corresponding to the reconstructed signal to be tested, obtain a number of target normal reconstructed signals corresponding to the target sample bearing in the normal operation signal library, the target sample bearing and the target bearing are of the same type; wherein the normal operation signal library includes a number of types of sample bearings, each of which corresponds to a number of normal reconstructed signals that have been subjected to denoising and reconstruction processing;

[0083] The normal operating signal library in the embodiment of the present invention includes several types of sample bearings, each type of sample bearing corresponds to several types of normal reconstructed signals that have been denoised and reconstructed, wherein the sample bearing of the same type as the target bearing is the target sample bearing, and the normal reconstructed signal corresponding to the target sample bearing is the target normal reconstructed signal.

[0084] The type of the target bearing should be one of the sample bearings in the normal operation signal library.

[0085] It should be understood that the signal of a certain sample bearing operating normally under no-load conditions and the signal of a certain sample bearing operating normally under heavy load conditions, after denoising and reconstruction processing, both belong to the normal reconstructed signal of the corresponding sample bearing, and the normal reconstructed signals of other situations are not listed here one by one.

[0086] S400, respectively calculating the similarity between the reconstructed signal to be tested and the target normal reconstructed signal, if the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is less than the normal similarity threshold, determining that the reconstructed signal to be tested is an abnormal signal, and inputting the reconstructed signal to be tested into the abnormal operation signal library;

[0087] Specifically, the embodiment of the present invention sequentially calculates the similarity between the reconstructed signal to be tested and the target normal reconstructed signal, and compares the similarity with a normal similarity threshold value, where the normal similarity threshold value can be set according to actual needs.

[0088] When the similarity between the reconstructed signal to be tested and a target normal reconstructed signal is greater than or equal to the normal similarity threshold, the reconstructed signal to be tested is determined to be a normal signal, subsequent calculations are stopped, and a detection result is formed and output.

[0089] If the similarity between the reconstructed signal to be tested and each target normal reconstructed signal after traversal calculation is less than the normal similarity threshold, it means that the reconstructed signal to be tested fails to match in the normal operation signal library, and the reconstructed signal to be tested is determined to be an abnormal signal. The reconstructed signal to be tested is input into the abnormal operation signal library for further comparison to determine the abnormality type.

[0090] In addition, the similarity calculation method may be a cosine similarity calculation method, a Pearson correlation coefficient, etc., which may be selected according to actual needs.

[0091] S500, according to the type of the target bearing corresponding to the reconstructed signal to be tested, obtain a number of target abnormal reconstructed signals corresponding to the target sample bearing in the abnormal operation signal library, the target sample bearing and the target bearing are of the same type; wherein the abnormal operation signal library includes a number of types of sample bearings, each sample bearing corresponds to a number of abnormal types, and each abnormal type corresponds to a number of abnormal reconstructed signals that have been subjected to denoising and reconstruction processing;

[0092] The abnormal operation signal library in the embodiment of the present invention includes several types of sample bearings that are the same as the types of sample bearings in the normal operation signal library. Each type of sample bearing in the abnormal operation signal library corresponds to several abnormal types, such as bearing raceway surface fracture, insufficient bearing lubricant, bearing ball fracture, etc. Each abnormal type corresponds to multiple situations. For example, the abnormal type of bearing raceway surface fracture of a certain bearing may include abnormal reconstruction signals of bearing raceway surface fracture under no-load conditions, and abnormal reconstruction signals of bearing raceway surface fracture under heavy load conditions, etc.

[0093] When the reconstructed signal to be tested that is determined to be abnormal is input into the abnormal operation signal library, according to the type of the target bearing corresponding to the reconstructed signal to be tested, several target abnormal reconstruction signals corresponding to the target sample bearings of the same type as the target bearing in the abnormal operation signal library are obtained. These target abnormal reconstruction signals belong to different abnormal types. The abnormal type of the target bearing is determined by calculating the similarity between the reconstructed signal to be tested and the target abnormal reconstruction signal.

[0094] In addition, the abnormal operation signal library can be further divided into the inner ring abnormal signal library, the outer ring abnormal signal library, etc. Here, the inner ring abnormal signal library and the outer ring abnormal signal library are taken as examples. The inner ring abnormal signal library includes several types of sample bearings, each sample bearing corresponds to several inner ring abnormal types, and each inner ring abnormal type corresponds to several inner ring abnormal reconstructed signals after denoising and reconstruction; the outer ring abnormal signal library includes several types of sample bearings, each sample bearing corresponds to several outer ring abnormal types, and each outer ring abnormal type corresponds to several outer ring abnormal reconstructed signals after denoising and reconstruction. By further refining the signal library, the efficiency of bearing fault analysis can be effectively improved.

[0095] S600, respectively calculating the similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal; if the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal is greater than or equal to an abnormal similarity threshold, determining the abnormality type of the reconstructed signal to be tested according to the abnormality type of the corresponding target abnormal reconstructed signal.

[0096] Specifically, the embodiment of the present invention sequentially calculates the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal, and compares the similarity with the abnormal similarity threshold. The abnormal similarity threshold here can be set according to actual needs.

[0097] When the similarity between the reconstructed signal to be tested and a target abnormal reconstructed signal is greater than or equal to the abnormal similarity threshold, the abnormal type of the reconstructed signal to be tested is determined according to the abnormal type of the corresponding target abnormal reconstructed signal.

[0098] Different from the matching calculation logic of the normal operation signal library, the embodiment of the present invention traverses and calculates the similarity between the reconstructed signal to be tested and all target abnormal reconstructed signals.

[0099] If after traversal calculation, the similarity between the reconstructed signal to be tested and only one target abnormal reconstructed signal is greater than or equal to the abnormal similarity threshold, a fault analysis result is formed and output according to the abnormal type corresponding to the target abnormal reconstructed signal.

[0100] If after traversal calculation, the similarities between the reconstructed signal to be tested and multiple target abnormal reconstructed signals are greater than or equal to the abnormal similarity threshold, the abnormal types corresponding to the above target abnormal reconstructed signals are sorted according to the similarity and the fault analysis results are formed and output. The user can further troubleshoot and analyze according to the probability order of the abnormal types given by the fault analysis results, which can greatly reduce the time cost of equipment fault troubleshooting.

[0101] In summary, the embodiment of the present invention denoises and reconstructs the vibration signal to be measured of the target bearing through the local mean decomposition method, and inputs the denoised and reconstructed signal to be measured into the normal operation signal library, matches it through similarity calculation, and determines whether the target bearing is abnormal. When the reconstructed signal to be measured is abnormal, the reconstructed signal to be measured is input into the abnormal operation signal library for further matching through similarity calculation to determine the abnormal type of the target bearing. The bearing fault intelligent analysis method provided by the embodiment of the present invention effectively solves the problem that ordinary operators find it difficult to effectively understand and use bearing vibration data for bearing fault analysis, and there is a large subjectivity and inaccuracy in the judgment of bearing faults. It improves the accuracy of bearing fault analysis, and greatly saves labor costs, and has great industrial application value.

[0102] Furthermore, if Figure 3 As shown, the vibration signal to be measured is denoised and reconstructed by the local mean decomposition method to obtain the reconstructed signal to be measured, including:

[0103] S210, sampling the vibration signal to be measured according to a preset sampling frequency to obtain a discrete signal to be measured;

[0104] Specifically, the vibration signal to be measured during the daily operation of the target bearing obtained by the embodiment of the present invention is: , according to the preset sampling frequency Vibration signal to be measured Sampling is performed to obtain the discrete signal to be measured ,in, , is the total number of sampling points.

[0105] S220, obtaining all local extreme value points in the discrete signal to be measured;

[0106] Find the discrete signal to be measured All local extreme points in , , , is the total number of local extreme points, .

[0107] S230, calculating the local mean and local amplitude of each local extreme point;

[0108] Specifically, the local mean calculation formula of the local extreme point is shown in the following formula (1):

[0109] (1);

[0110] in, .

[0111] Specifically, the local amplitude of the local extreme point is as shown in the following formula (2):

[0112] (2);

[0113] in, .

[0114] S240, performing sliding average processing on the local mean and local amplitude of each local extreme point respectively, to obtain the local mean function and local amplitude function of the discrete signal to be measured;

[0115] Specifically, by taking all local means At the corresponding local extreme point and The local mean function is obtained by extending the straight line between the two values ​​and smoothing the extended straight line using the sliding average method. ;

[0116] Similarly, all local amplitudes At the corresponding local extreme point and The local amplitude function is obtained by extending the straight line between the two values ​​and smoothing the extended straight line using the sliding average method. .

[0117] S250, obtaining a zero mean function and a pure frequency modulation function of the discrete signal to be measured according to the local mean function and the local amplitude function;

[0118] Specifically, according to the local mean function , local amplitude function Get the discrete signal to be measured The zero mean function and pure frequency modulation function , zero mean function The expression of is shown in the following formula (3):

[0119] (3);

[0120] Pure frequency modulation function The expression of is shown in the following formula (4):

[0121] (4).

[0122] S260, extracting a product function and a corresponding residual signal in the discrete signal to be measured according to the zero mean function and the pure frequency modulation function of the discrete signal to be measured;

[0123] Discrete signal to be measured There may be multiple product functions in the matrix, which need to be extracted one by one. Specifically, first, the discrete signal to be measured is Analyze and identify and extract the first product function , extract the first product function From the discrete signal to be measured Subtract it from the residual signal Then repeat the above steps, that is, after each iteration, a new product function is extracted from the current residual signal, and the newly obtained product function is subtracted from the residual signal to obtain the updated residual signal, and the above steps are repeated until the stop condition is met.

[0124] The stop condition set in this embodiment is to repeat the loop Stop at this time. is a preset number of iterations, It can be set according to specific needs. Generally speaking, The settings should often satisfy the iteration The remaining signal after becomes a constant or no longer contains obvious oscillation components, The specific value of is not limited here.

[0125] Specifically, the discrete signal to be measured The specific expression of the product function in is shown in the following formula (5):

[0126] (5);

[0127] Iteration The remaining signal after this time is , The specific expression of is shown in the following formula (6):

[0128] (6);

[0129] S270 , reconstruct the discrete signal to be measured according to the product function and the residual signal to obtain a reconstructed signal to be measured.

[0130] Specifically, the expression of the reconstructed signal to be measured is shown in the following formula (7):

[0131] (7).

[0132] Furthermore, if Figure 4 As shown, the local extreme points of the discrete signal to be measured are obtained, including:

[0133] S221, performing amplification and enhancement processing on the discrete signal to be measured by using a loss function to obtain the amplified and enhanced discrete signal to be measured, wherein the signal-to-noise ratio of the amplified and enhanced discrete signal to be measured is equal to the signal-to-noise ratio of the discrete signal to be measured;

[0134] In the embodiment of the present invention, the discrete signal to be measured is analyzed by using the SI-SNR loss function. Amplification and enhancement processing is performed, and the signal-to-noise ratio is kept unchanged while amplifying the enhanced signal.

[0135] Specifically, as shown in the following formula (8):

[0136] (8);

[0137] in, represents the discrete signal to be measured, , It represents the discrete signal to be measured after amplification and enhancement. represents the proportionality coefficient, , It means that the discrete signal to be measured after amplification is The projection on represents the error signal, .

[0138] , as well as The relationship between the three can be referred to Figure 5 .

[0139] S222, obtaining all local extreme value points in the discrete signal to be measured after amplification and enhancement.

[0140] Through the SI-SNR loss function to measure the discrete signal After amplification and enhancement, the discrete signal to be measured can be better captured Based on the slight changes in the All local extreme points in It can reduce the loss of information and further improve the accuracy of bearing fault analysis.

[0141] Furthermore, if Figure 6 As shown, the local mean and local amplitude of each local extreme point are respectively subjected to sliding average processing to obtain the local mean function and local amplitude function of the discrete signal to be measured, including:

[0142] S241, calculating the initial step length between adjacent local extreme value points;

[0143] Specifically, calculate and All initial step lengths in The value of the step size depends on The value of is different. .

[0144] S242, calculating the step size mean of the sliding average processing according to the initial step size between adjacent local extreme value points;

[0145] Specifically, Figure 7 As shown, this embodiment uses a bar chart-assisted analysis method to calculate the step mean of the sliding average processing, the X-axis represents the initial step value, and the Y-axis represents the probability of the initial step falling in different intervals. For example, the probability of the initial step falling in the interval of 15 to 20 is 0.07.

[0146] Specifically, this embodiment uses the histcounts function (a one-dimensional array histogram bin counting function) integrated in MATLAB to automatically calculate and determine the column width. , that is, the length of the initial step value interval, the column width in the embodiment of the present invention Specifically 5, according to the column width Calculate the probability corresponding to each initial step value interval ,in, Indicates the index of the bar chart.

[0147] The specific expression of the step length mean is shown in the following formula (9):

[0148] (9);

[0149] in, represents the mean step length, .

[0150] S243, calculating the step length standard deviation of the sliding average processing according to the step length mean;

[0151] Specifically, the expression of the step size standard deviation of the moving average processing is shown in the following formula (10):

[0152] (10);

[0153] in, Represents the step standard deviation of the moving average process.

[0154] S244, obtaining the step length of the sliding average processing according to the step length mean and the step length standard deviation;

[0155] Specifically, the step length of the sliding average process is expressed as follows:

[0156] (11);

[0157] in, represents the step size of the sliding average processing. It should be understood that the step size of the sliding average processing determined by the above method is the optimal step length.

[0158] S245 , performing sliding average processing on the local mean and local amplitude of each local extreme point according to the step size of the sliding average processing, to obtain the local mean function and local amplitude function of the discrete signal to be measured.

[0159] Specifically, according to the optimal step size Perform sliding average processing on the local mean and local amplitude of each local extreme point to obtain the local mean function of the discrete signal to be measured and the local amplitude function .

[0160] In summary, since the accuracy of the local extreme point has a decisive influence on the result of the entire local mean decomposition, which in turn affects the accuracy of the bearing fault analysis, the embodiment of the present invention uses the SI-SNR loss function to analyze the discrete signal to be measured. On the basis of extracting local extreme points after amplification and enhancement processing, the step size of the sliding average processing was further optimized, which further improved the accuracy and reliability of the processing results of the local mean decomposition method, and further improved the reliability of the bearing fault analysis results.

[0161] Furthermore, the bearing fault intelligent analysis method provided by the embodiment of the present invention further includes:

[0162] A100, respectively merging and optimizing a plurality of normal reconstruction signals corresponding to each type of sample bearing in the normal operating signal library to obtain a normal comparison signal corresponding to each type of sample bearing;

[0163] A200, forming an optimized normal operation signal library according to the normal comparison signal corresponding to each type of sample bearing, wherein the optimized normal operation signal library includes several types of sample bearings, and each type of sample bearing corresponds to a normal comparison signal;

[0164] Through merging and optimizing processing, several normal reconstructed signals corresponding to each sample bearing in the normal operating signal library are merged into a normal comparison signal, which greatly reduces the data volume of the normal operating signal library, streamlines the normal operating signal library to the core content, and realizes lightweight operation, thereby reducing the retrieval and matching time of the reconstructed signal to be tested in the optimized normal operating signal library, significantly improving the retrieval speed, and effectively avoiding the accumulation of redundant data, ensuring the efficient use of the normal operating signal library, and improving the efficiency and real-time performance of bearing fault analysis.

[0165] B100, respectively merging and optimizing a plurality of abnormal reconstruction signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library to obtain an abnormal comparison signal corresponding to each abnormal type of each type of sample bearing;

[0166] B200. An optimized abnormal operation signal library is formed according to the abnormal comparison signal corresponding to each abnormal type of each type of sample bearing, wherein the optimized abnormal operation signal library includes several types of sample bearings, each sample bearing corresponds to several abnormal types, and each abnormal type corresponds to an abnormal comparison signal.

[0167] Similarly, through merging and optimizing processing, several abnormal reconstruction signals corresponding to each abnormal type of each sample bearing in the abnormal operation signal library are merged into one abnormal comparison signal, which greatly reduces the data volume of the abnormal operation signal library, streamlines the abnormal operation signal library to the core content, and realizes lightweight operation, thereby reducing the retrieval and matching time of the reconstructed signal to be tested in the optimized abnormal operation signal library, significantly improving the retrieval speed, and effectively avoiding the accumulation of redundant data, ensuring the efficient use of the abnormal operation signal library, and improving the efficiency and real-time performance of bearing fault analysis.

[0168] It should be understood that the normal operation signal library and the abnormal operation signal library can be optimized according to a preset period, or after new sample data is added to the normal operation signal library and the abnormal operation signal library, the normal operation signal library or the abnormal operation signal library can be optimized according to the newly added sample data. The embodiments of the present invention are not limited to this.

[0169] Furthermore, a plurality of normal reconstructed signals corresponding to each type of sample bearing in the normal operation signal library are merged and optimized to obtain a normal comparison signal corresponding to each type of sample bearing, including:

[0170] transforming each normal reconstructed signal in the normal operating signal library into a normal frequency domain signal by fast Fourier transform processing;

[0171] Specifically, the normal frequency domain signal obtained by fast Fourier transform is shown in the following formula (12):

[0172] (12);

[0173] in, represents a normal frequency domain signal, represents the complex exponential function, represents the angular frequency, represents the imaginary unit, Indicates time.

[0174] Extracting the maximum intensity normal frequency domain signal and the minimum intensity normal frequency domain signal from a number of normal frequency domain signals corresponding to each type of sample bearing in the normal operation signal library;

[0175] Specifically, the maximum intensity normal frequency domain signal among the normal frequency domain signals corresponding to each type of sample bearing in the normal operation signal library is as shown in the following formula (13):

[0176] (13);

[0177] The minimum intensity normal frequency domain signal among several normal frequency domain signals corresponding to each type of sample bearing in the normal operation signal library is shown in the following formula (14):

[0178] (14);

[0179] Calculate the average value of the maximum intensity normal frequency domain signal and the minimum intensity normal frequency domain signal of each type of sample bearing to obtain the normal comparison signal of each type of sample bearing;

[0180] Specifically, the embodiment of the present invention performs the maximum intensity normal frequency domain signal and minimum intensity normal frequency domain signal Averaging is performed to obtain normal comparison signals for each type of sample bearings.

[0181] The abnormal reconstruction signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library are merged and optimized to obtain the abnormal comparison signals corresponding to each abnormal type of each type of sample bearing, including:

[0182] Each abnormal reconstructed signal in the abnormal operation signal library is transformed into an abnormal frequency domain signal by fast Fourier transform processing;

[0183] Specifically, the abnormal frequency domain signal obtained by fast Fourier transform is shown in the following formula (15):

[0184] (15);

[0185] in, represents an abnormal frequency domain signal, represents the complex exponential function, represents the angular frequency, represents the imaginary unit, Indicates time.

[0186] Extracting the maximum intensity abnormal frequency domain signal and the minimum intensity abnormal frequency domain signal from a number of abnormal frequency domain signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library;

[0187] Specifically, the maximum intensity abnormal frequency domain signal among a number of abnormal frequency domain signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library is expressed as the following formula (16):

[0188] (16);

[0189] The minimum intensity abnormal frequency domain signal among several normal frequency domain signals corresponding to each abnormal type of each type of sample bearing in the normal operation signal library is shown in the following formula (17):

[0190] (17);

[0191] The average values ​​of the maximum intensity abnormal frequency domain signal and the minimum intensity abnormal frequency domain signal of each abnormal type of each type of sample bearing are calculated to obtain the abnormal comparison signal of each abnormal type of each type of sample bearing.

[0192] Specifically, the embodiment of the present invention performs the maximum intensity abnormal frequency domain signal and minimum intensity abnormal frequency domain signal An averaging process is performed to obtain an abnormal comparison signal of each abnormal type for each type of sample bearing.

[0193] Furthermore, after performing denoising and reconstruction processing on the vibration signal to be measured and obtaining the reconstructed signal to be measured, the method further includes:

[0194] C100, converting the reconstructed signal to be measured into the frequency domain signal to be measured by fast Fourier transform processing;

[0195] In one embodiment provided by the present invention, the reconstructed signal to be tested is also converted into a frequency domain signal to be tested and input into an optimized normal operating signal library and an optimized abnormal operating signal library for similarity matching, thereby reducing the difficulty of matching and improving the efficiency of bearing fault detection.

[0196] C200, according to the type of the target bearing corresponding to the frequency domain signal to be measured, obtaining a target normal comparison signal corresponding to a target sample bearing of the same type as the target bearing in the optimized normal operating signal library;

[0197] Specifically, according to the type of the target bearing corresponding to the frequency domain signal to be measured, a search is performed in the optimized normal operation signal library to obtain a target normal comparison signal corresponding to the target sample bearing of the same type as the target bearing. It should be understood that the number of target normal comparison signals is 1.

[0198] C300, calculating the similarity between the frequency domain signal to be measured and the target normal comparison signal. If the similarity between the frequency domain signal to be measured and the target normal comparison signal is less than the normal similarity threshold, the frequency domain signal to be measured is determined to be an abnormal signal, and the frequency domain signal to be measured is input into the optimized abnormal operation signal library;

[0199] Specifically, in this embodiment, it is only necessary to calculate the similarity between the frequency domain signal to be measured and the target normal comparison signal corresponding to the target sample bearing, and compare it with the normal similarity threshold, where the normal similarity threshold can be set according to actual needs.

[0200] When the similarity between the frequency domain signal to be tested and the target normal comparison signal is greater than or equal to the normal similarity threshold, the frequency domain signal to be tested is determined to be a normal signal, subsequent calculations are stopped, and a detection result is formed and output based on the similarity calculation result.

[0201] If the similarity between the frequency domain signal to be tested and the target normal comparison signal is less than the normal similarity threshold, the frequency domain signal to be tested is determined to be an abnormal signal, and the frequency domain signal to be tested is input into the optimized abnormal operation signal library for further comparison to determine the abnormal type.

[0202] C400, according to the type of the target bearing corresponding to the frequency domain signal to be measured, obtain a number of target abnormal comparison signals corresponding to the target sample bearings of the same type as the target bearing in the optimized abnormal operation signal library;

[0203] Specifically, according to the type of the target bearing corresponding to the frequency domain signal to be measured, a search is performed in the optimized abnormal operation signal library to obtain a target abnormal comparison signal corresponding to a target sample bearing of the same type as the target bearing. It should be understood that the number of target abnormal comparison signals depends on the abnormal type corresponding to the target sample bearing.

[0204] C500, respectively calculating the similarity between the frequency domain signal to be measured and each target abnormal comparison signal; if the similarity between the frequency domain signal to be measured and the target abnormal comparison signal is greater than or equal to the abnormal similarity threshold, determining the abnormality type of the frequency domain signal to be measured according to the abnormality type of the corresponding target abnormal comparison signal.

[0205] Specifically, the similarities between the frequency domain signal to be tested and several target abnormal comparison signals corresponding to the target sample bearing are calculated in turn, and the similarity calculation result is compared with the abnormal similarity threshold, where the abnormal similarity threshold can be selected according to actual needs.

[0206] If there is only one target abnormal comparison signal whose similarity with the frequency domain signal to be measured is greater than or equal to the abnormal similarity threshold, the abnormal type corresponding to the target abnormal comparison signal is formed into the fault analysis result of the target bearing and output.

[0207] If there are multiple target abnormal comparison signals whose similarity with the frequency domain signal to be measured is greater than or equal to the abnormal similarity threshold, the abnormal types corresponding to the above target abnormal comparison signals are sorted according to the similarity and output as the fault analysis results. This arrangement can intuitively reflect the possibility of each abnormal type detected by the target bearing. Users can further investigate and analyze according to the abnormal type sorting in the fault analysis results, which can greatly reduce the time cost of equipment fault troubleshooting.

[0208] Furthermore, the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is calculated respectively, including:

[0209] The similarity between the reconstructed signal to be tested and each target normal reconstructed signal is calculated according to the Spearman coefficient;

[0210] The Spearman correlation is a nonparametric measure of rank correlation, which can reflect the strength of the monotonic relationship between two random variables. The Spearman correlation is nonparametric. Compared with similarity calculation methods such as the Pearson coefficient, it has the advantages of being more accurate and reliable in processing ordered data, being less affected by outliers, and being more robust. Therefore, this embodiment selects the Spearman correlation to calculate the similarity between the reconstructed signal to be tested and each target normal reconstructed signal.

[0211] Specifically, the Spearman coefficient of the reconstructed signal to be tested and the target normal reconstructed signal is calculated, as shown in the following formula (18):

[0212] (18);

[0213] in, Represents the Spearman coefficient of the reconstructed signal to be tested and the target normal reconstructed signal, It indicates the level difference of the amplitude value between the reconstructed signal to be tested and the target normal reconstructed signal at the same time.

[0214] For the embodiment of the present invention, the amplitude values ​​of the reconstructed signal to be tested at the same time are taken in sequence and the amplitude value of the target normal reconstructed signal , get the data in sequence , ,……, ,Will , ,……, Rearrange them in ascending order. , ,……, Also rearranged in ascending order. It is the third one, after rearrangement is the first, then the difference in rank between the two The rest is the same and will not be elaborated here.

[0215] After calculating the Spearman coefficients of the reconstructed signal to be tested and the target normal reconstructed signal, the corresponding similarity is calculated according to the Spearman coefficients, as shown in the following formula (19):

[0216] (19);

[0217] like is greater than the normal similarity threshold, then the reconstructed signal to be tested is determined to be a normal signal. If the reconstructed signal to be tested is less than or equal to the normal similarity threshold, the reconstructed signal to be tested is determined to be an abnormal signal, and the reconstructed signal to be tested determined to be abnormal is input into the abnormal operation signal library for further comparison to determine the abnormal type.

[0218] The similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is calculated respectively, including:

[0219] The similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is calculated according to the Spearman coefficient.

[0220] Specifically, the Spearman coefficient of the reconstructed signal to be tested and the target abnormal reconstructed signal is calculated, as shown in the following formula (20):

[0221] (20);

[0222] in, represents the Spearman coefficient of the reconstructed signal to be tested and the target abnormal reconstructed signal, It indicates the level difference of the amplitude values ​​of the reconstructed signal to be tested and the target abnormal reconstructed signal corresponding to the same time.

[0223] For the embodiment of the present invention, the amplitude values ​​of the reconstructed signal to be tested at the same time are taken in sequence and the amplitude value of the target normal reconstructed signal , get the data in sequence , ,……, ,Will , ,……, Rearrange them in ascending order. , ,……, Also rearranged in ascending order. It is the third one, after rearrangement is the first, then the difference in rank between the two The rest is the same and will not be elaborated here.

[0224] After calculating the Spearman coefficients of the reconstructed signal to be tested and the target normal reconstructed signal, the corresponding similarity is calculated according to the Spearman coefficients, as shown in the following formula (19):

[0225] (19);

[0226] If after traversal calculation, the reconstructed signal to be tested has only a similarity with one target abnormal reconstructed signal If the target abnormality is greater than the abnormality similarity threshold, a fault analysis result is formed and output according to the abnormality type corresponding to the target abnormality reconstruction signal.

[0227] If after traversal calculation, the similarity between the reconstructed signal to be tested and multiple target abnormal reconstructed signals If all of them are greater than the abnormal similarity threshold, the abnormal types corresponding to the target abnormal reconstruction signals are sorted according to the similarity and the fault analysis results are output. Users can further troubleshoot and analyze according to the probability order of the abnormal types given by the fault analysis results, which can greatly reduce the time cost of equipment fault troubleshooting.

[0228] It should be understood that, in another embodiment of the present invention, when the reconstructed signal to be tested is transformed into a frequency domain signal to be tested by fast Fourier transform, input into the optimized normal operating signal library, and similarity matching calculation is performed with the target normal comparison signal in the optimized normal operating signal library, Indicates the level difference in signal strength between the frequency domain signal to be measured and the target normal comparison signal at the same frequency.

[0229] Similarly, when the frequency domain signal to be measured, which is determined to be an abnormal signal in the optimized normal operation signal library, is input into the optimized abnormal operation signal library and the similarity matching calculation is performed with the target abnormal comparison signal in the optimized abnormal operation signal library, Indicates the level difference in signal strength between the frequency domain signal to be tested and the target abnormal comparison signal at the same frequency.

[0230] Furthermore, if the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is less than the normal similarity threshold, and the similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal is less than the abnormal similarity threshold, the reconstructed signal to be tested is determined as the unknown abnormal type of the corresponding target sample bearing and added to the abnormal operation signal library.

[0231] Specifically, when the reconstructed signal to be tested that is determined to be abnormal fails to match in the abnormal operation signal library, the reconstructed signal to be tested is determined to be an unknown abnormal type of the corresponding target sample bearing and is added to the abnormal operation signal library.

[0232] The abnormal operation signal library may store multiple unknown anomalies, such as unknown anomaly 1, unknown anomaly 2, etc., and record the number of occurrences of each unknown anomaly. If the number of occurrences of a certain unknown anomaly exceeds the preset threshold, the technicians will analyze and evaluate the unknown anomaly.

[0233] Specifically, if the unknown abnormality belongs to an existing abnormality type of a sample bearing of the same type, it is updated to an abnormality reconstructed signal under the abnormality type corresponding to the sample bearing of the same type, and is deleted from the unknown abnormality.

[0234] If, after analysis and evaluation, the unknown abnormality does not belong to the existing abnormality type of the sample bearing of the same type, a corresponding abnormality type is added for the sample bearing of the same type for the unknown abnormality, and the reconstructed signal to be tested corresponding to the unknown abnormality is updated to the abnormal reconstructed signal under the newly added abnormality type, and then it is deleted from the unknown abnormality.

[0235] Therefore, the embodiments of the present invention can effectively cope with various sudden new faults and have high practical application value.

[0236] The embodiment of the present invention provides a bearing fault intelligent analysis device 010, such as Figure 8 As shown, including:

[0237] The signal acquisition module 011 is used to obtain the vibration signal to be measured during the daily operation of the target bearing;

[0238] De-noising and reconstruction module 012, used for performing denoising and reconstruction processing on the vibration signal to be measured to obtain a reconstructed signal to be measured;

[0239] The normal operation signal library search module 013 is used to obtain a number of target normal reconstruction signals corresponding to the target sample bearings in the normal operation signal library according to the type of the target bearing corresponding to the reconstructed signal to be measured, and the target sample bearings are of the same type as the target bearings; wherein the normal operation signal library includes a number of types of sample bearings, and each sample bearing corresponds to a number of normal reconstruction signals that have been subjected to denoising and reconstruction processing;

[0240] A normal operation similarity calculation module 014 is used to calculate the similarity between the reconstructed signal to be tested and the target normal reconstructed signal respectively. If the similarity between the reconstructed signal to be tested and each target normal reconstructed signal is less than the normal similarity threshold, the reconstructed signal to be tested is determined to be an abnormal signal, and the reconstructed signal to be tested is input into the abnormal operation signal library;

[0241] The abnormal operation signal library search module 015 is used to obtain a number of target abnormal reconstruction signals corresponding to the target sample bearings in the abnormal operation signal library according to the type of the target bearing corresponding to the reconstructed signal to be tested, and the target sample bearings are of the same type as the target bearings; wherein the abnormal operation signal library includes a number of types of sample bearings, each sample bearing corresponds to a number of abnormal types, and each abnormal type corresponds to a number of abnormal reconstruction signals that have been subjected to denoising and reconstruction processing;

[0242] The abnormal operation similarity calculation module 016 is used to respectively calculate the similarity between the reconstructed signal to be tested and each target abnormal reconstructed signal. If the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal is greater than or equal to the abnormal similarity threshold, the abnormal type of the reconstructed signal to be tested is determined according to the abnormal type of the corresponding target abnormal reconstructed signal.

[0243] Specifically, the embodiment of the present invention denoises and reconstructs the vibration signal to be measured during the daily operation of the target bearing acquired by the signal acquisition module 011 through the denoising and reconstruction module 012 to obtain the reconstructed signal to be measured, and retrieves the target normal reconstructed signal corresponding to the same type of target sample bearing in the normal operation signal library through the normal operation signal library search module 013, and then calculates the similarity between the reconstructed signal to be measured and the target normal reconstructed signal through the normal operation similarity calculation module 014 to determine whether the target bearing is abnormal.

[0244] When the reconstructed signal to be tested is abnormal, the reconstructed signal to be tested is input into the abnormal operation signal library, and the abnormal operation signal library search module 015 retrieves the target abnormal reconstructed signal corresponding to the same type of target sample bearing in the abnormal operation signal library, and then the abnormal operation similarity calculation module 016 calculates the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal to determine the abnormal type of the target bearing.

[0245] In summary, the embodiments of the present invention solve the problem that ordinary operators find it difficult to effectively understand and use the above-mentioned bearing vibration data to perform bearing fault analysis, and that the judgment of bearing faults is highly subjective and inaccurate through intelligent analysis, thereby improving the accuracy of bearing fault analysis.

[0246] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A bearing fault intelligent analysis method, characterized in that: include: Obtain the vibration signal to be measured during the daily operation of the target bearing; Performing denoising and reconstruction processing on the vibration signal to be measured by a local mean decomposition method to obtain a reconstructed signal to be measured; According to the type of the target bearing corresponding to the reconstructed signal to be tested, a number of target normal reconstructed signals corresponding to the target sample bearing in the normal operation signal library are obtained, and the target sample bearing is of the same type as the target bearing; wherein the normal operation signal library includes several types of sample bearings, and each sample bearing corresponds to several normal reconstructed signals that have been subjected to denoising and reconstruction processing; Calculating the similarity between the reconstructed signal to be tested and the target normal reconstructed signal respectively, if the similarity between the reconstructed signal to be tested and each of the target normal reconstructed signals is less than a normal similarity threshold, determining that the reconstructed signal to be tested is an abnormal signal, and inputting the reconstructed signal to be tested into an abnormal operation signal library; According to the type of the target bearing corresponding to the reconstructed signal to be tested, a plurality of target abnormal reconstructed signals corresponding to the target sample bearing in the abnormal operation signal library are obtained, and the target sample bearing is of the same type as the target bearing; wherein the abnormal operation signal library includes a plurality of types of sample bearings, each sample bearing corresponds to a plurality of abnormal types, and each abnormal type corresponds to a plurality of abnormal reconstructed signals that have been subjected to denoising and reconstruction processing; Calculating the similarity between the reconstructed signal to be tested and each of the target abnormal reconstructed signals respectively, if the similarity between the reconstructed signal to be tested and the target abnormal reconstructed signal is greater than or equal to an abnormal similarity threshold, determining the abnormal type of the reconstructed signal to be tested according to the abnormal type of the corresponding target abnormal reconstructed signal; The method of performing denoising and reconstruction processing on the vibration signal to be measured by using a local mean decomposition method to obtain a reconstructed signal to be measured includes: Sampling the vibration signal to be measured according to a preset sampling frequency to obtain a discrete signal to be measured; Obtaining all local extreme value points in the discrete signal to be measured; Calculate the local mean and local amplitude of each local extreme point; Performing sliding average processing on the local mean and local amplitude of each of the local extreme points respectively to obtain the local mean function and local amplitude function of the discrete signal to be measured; According to the local mean function and the local amplitude function, a zero mean function and a pure frequency modulation function of the discrete signal to be measured are obtained; Extracting a product function and a corresponding residual signal in the discrete signal to be measured according to the zero mean function and the pure frequency modulation function of the discrete signal to be measured; Reconstructing the discrete signal to be measured according to the product function and the residual signal to obtain a reconstructed signal to be measured; The obtaining of the local extreme point of the discrete signal to be measured comprises: Amplifying and enhancing the discrete signal to be measured by using a loss function to obtain an amplified and enhanced discrete signal to be measured, wherein the signal-to-noise ratio of the amplified and enhanced discrete signal to be measured is equal to the signal-to-noise ratio of the discrete signal to be measured; Obtaining all local extreme value points in the amplified and enhanced discrete signal to be measured; The step of performing sliding average processing on the local mean and the local amplitude of each local extreme point to obtain the local mean function and the local amplitude function of the discrete signal to be measured comprises: Calculating the initial step length between adjacent local extreme value points; Calculate the step size mean of the sliding average processing according to the initial step size between adjacent local extreme value points; Calculate the step size standard deviation of the sliding average processing according to the step size mean; Obtaining a step length for sliding average processing according to the step length mean and the step length standard deviation; Performing sliding average processing on the local mean and local amplitude of each local extreme point according to the step size of the sliding average processing, to obtain the local mean function and local amplitude function of the discrete signal to be measured; The method further comprises: Merging and optimizing the normal reconstruction signals corresponding to each type of sample bearing in the normal operation signal library to obtain the normal comparison signal corresponding to each type of sample bearing; Forming an optimized normal operation signal library according to the normal comparison signal corresponding to each type of sample bearing, wherein the optimized normal operation signal library includes several types of sample bearings, and each type of sample bearing corresponds to a normal comparison signal; Merging and optimizing a plurality of abnormal reconstruction signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library to obtain an abnormal comparison signal corresponding to each abnormal type of each type of sample bearing; An optimized abnormal operation signal library is formed according to the abnormal comparison signal corresponding to each abnormal type of each type of sample bearing, wherein the optimized abnormal operation signal library includes several types of sample bearings, each sample bearing corresponds to several abnormal types, and each abnormal type corresponds to an abnormal comparison signal; The step of merging and optimizing the normal reconstruction signals corresponding to each type of sample bearing in the normal operation signal library to obtain the normal comparison signal corresponding to each type of sample bearing includes: Transforming each of the normal reconstructed signals in the normal operating signal library into a normal frequency domain signal by fast Fourier transform processing; Extracting the maximum intensity normal frequency domain signal and the minimum intensity normal frequency domain signal from a plurality of normal frequency domain signals corresponding to each type of sample bearing in the normal operation signal library; Calculate the average value of the maximum intensity normal frequency domain signal and the minimum intensity normal frequency domain signal of each type of sample bearing to obtain a normal comparison signal of each type of sample bearing; The step of merging and optimizing a plurality of abnormal reconstruction signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library to obtain an abnormal comparison signal corresponding to each abnormal type of each type of sample bearing includes: Transforming each of the abnormal reconstructed signals in the abnormal operation signal library into an abnormal frequency domain signal by fast Fourier transform processing; Extracting the maximum intensity abnormal frequency domain signal and the minimum intensity abnormal frequency domain signal from a plurality of abnormal frequency domain signals corresponding to each abnormal type of each type of sample bearing in the abnormal operation signal library; The average values ​​of the maximum intensity abnormal frequency domain signal and the minimum intensity abnormal frequency domain signal of each abnormal type of each type of sample bearing are calculated to obtain the abnormal comparison signal of each abnormal type of each type of sample bearing.

2. The method according to claim 1, characterized in that After performing denoising and reconstruction processing on the vibration signal to be measured to obtain the reconstructed signal to be measured, the method further includes: Converting the reconstructed signal to be measured into a frequency domain signal to be measured by fast Fourier transform processing; According to the type of the target bearing corresponding to the frequency domain signal to be measured, a target normal comparison signal corresponding to a target sample bearing of the same type as the target bearing in the optimized normal operating signal library is obtained; Calculating the similarity between the frequency domain signal to be measured and the target normal comparison signal, if the similarity between the frequency domain signal to be measured and the target normal comparison signal is less than a normal similarity threshold, determining that the frequency domain signal to be measured is an abnormal signal, and inputting the frequency domain signal to be measured into the optimized abnormal operation signal library; According to the type of the target bearing corresponding to the frequency domain signal to be measured, a plurality of target abnormal comparison signals corresponding to the target sample bearings of the same type as the target bearing in the optimized abnormal operation signal library are obtained; The similarity between the frequency domain signal to be measured and each of the target abnormal comparison signals is calculated respectively. If the similarity between the frequency domain signal to be measured and the target abnormal comparison signal is greater than or equal to the abnormal similarity threshold, the abnormal type of the frequency domain signal to be measured is determined according to the abnormal type of the corresponding target abnormal comparison signal.

3. The method according to claim 1, characterized in that: The respectively calculating the similarity between the reconstructed signal to be tested and the target normal reconstructed signal comprises: Calculating the similarity between the reconstructed signal to be tested and the target normal reconstructed signal according to the Spearman coefficient; The respectively calculating the similarity between the reconstructed signal to be tested and each of the target abnormal reconstructed signals comprises: The similarity between the reconstructed signal to be tested and each of the target abnormal reconstructed signals is calculated according to the Spearman coefficient.

4. The method according to claim 1, characterized in that: Also includes: If the similarity between the reconstructed signal to be tested and each of the target normal reconstructed signals is less than the normal similarity threshold, and the similarity between the reconstructed signal to be tested and each of the target abnormal reconstructed signals is less than the abnormal similarity threshold, then the reconstructed signal to be tested is determined as the unknown abnormal type of the corresponding target sample bearing and added to the abnormal operation signal library.

5. A bearing fault intelligent analysis device, used to implement the bearing fault intelligent analysis method according to any one of claims 1 to 4, characterized in that: include: A signal acquisition module is used to obtain the vibration signal to be measured during the daily operation of the target bearing; A denoising and reconstruction module, used for performing denoising and reconstruction processing on the vibration signal to be measured to obtain a reconstructed signal to be measured; A normal operating signal library search module is used to obtain a number of target normal reconstruction signals corresponding to a target sample bearing in a normal operating signal library according to the type of the target bearing corresponding to the reconstructed signal to be tested, wherein the target sample bearing is of the same type as the target bearing; wherein the normal operating signal library includes a number of types of sample bearings, and each sample bearing corresponds to a number of normal reconstruction signals that have been subjected to denoising and reconstruction processing; A normal operation similarity calculation module, used to respectively calculate the similarity between the reconstructed signal to be tested and the target normal reconstructed signal, and if the similarity between the reconstructed signal to be tested and each of the target normal reconstructed signals is less than a normal similarity threshold, the reconstructed signal to be tested is determined to be an abnormal signal, and the reconstructed signal to be tested is input into an abnormal operation signal library; An abnormal operation signal library search module is used to obtain a number of target abnormal reconstruction signals corresponding to target sample bearings in the abnormal operation signal library according to the type of the target bearing corresponding to the reconstructed signal to be tested, and the target sample bearing is of the same type as the target bearing; wherein the abnormal operation signal library includes several types of sample bearings, each sample bearing corresponds to several abnormal types, and each abnormal type corresponds to several abnormal reconstruction signals that have been subjected to denoising and reconstruction processing; The abnormal operation similarity calculation module is used to respectively calculate the similarity between the reconstructed signal to be tested and each of the target abnormal reconstruction signals. If the similarity between the reconstructed signal to be tested and the target abnormal reconstruction signal is greater than or equal to the abnormal similarity threshold, the abnormal type of the reconstructed signal to be tested is determined according to the abnormal type of the corresponding target abnormal reconstruction signal.

Citation Information

Patent Citations

  • Power grid monitoring-oriented voiceprint model evolution equipment fault fusion judgment method

    CN115343573A

  • Large-scale equipment fault detection method and system based on data analysis

    CN117091754A

  • Bearing fault diagnosis method, system and equipment in non-Gaussian noise environment

    CN118013223A

  • Rolling bearing fault detection method

    CN118329445A