Gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity

Abstract

The invention discloses an asynchronous and synchronous fuzzy similarity-based gear fault feature extraction method. The method comprises the following steps of: determining a sampling frequency, a sampling length, a synchronous sampling point number in each rotation of a gear, an embedded dimension m, a time delay tau and a similarity tolerance threshold r of synchronous data acquisition; acquiring synchronous vibration data X by using a vibration sensor and a rotating speed sensor by taking the gear rotating speed as a reference; performing phase-space reconstruction on the synchronous vibration data according to a preset time delay tau and an embedded dimension m; performing centralization processing on each m-dimensional vector in the phase space P by using a centralization processing method to obtain an asynchronous component phase space P1 of the vibration data; calculating the distance between m-dimensional vectors in the asynchronous component phase space P1; calculating an asynchronous fuzzy similarity; performing synchronous average processing on each m-dimensional vector in the phase space P by using a synchronous average processing method to obtain a synchronous component phase space P2 of the vibration data; calculating the distance between m-dimensional vectors in the synchronous component phase space P2; calculating synchronous fuzzy similarity; and taking the asynchronous fuzzy similarity and the synchronous fuzzy similarity as gear fault features. The method can be used for gear fault feature extraction, and the extracted fault features can be used for diagnosing gear fault types.

Description

Technical Field

[0001] The invention belongs to the technical field of rotating machinery fault diagnosis, and in particular relates to a gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity. Background Art

[0002] Rotating machinery and equipment are the core components of the modern industrial production system. Their efficient and stable operation is crucial to ensuring production efficiency and product quality. As a key link in the transmission of rotating machinery, the gear transmission system has been widely used in many fields such as automobiles, aviation, energy and manufacturing, and undertakes important functions such as power transmission and motion form conversion. In the long-term operation of gears, various faults such as pitting, broken teeth, cracks, and wear are prone to occur due to factors such as material fatigue, wear, overload, poor lubrication, and assembly errors. These faults will not only reduce the working efficiency of mechanical equipment, but may also cause major safety accidents, which will have a serious impact on production safety and economic benefits.

[0003] Traditional gear fault diagnosis methods rely on manual experience and regular shutdown inspections. This method is not only time-consuming and labor-intensive, but also often fails to diagnose weak faults, making it difficult to achieve early fault diagnosis. With the development of science and technology, especially the rapid development of sensor technology, signal processing technology, and artificial intelligence algorithms, modern gear fault diagnosis technology is gradually shifting towards online fault diagnosis, aiming to effectively reduce unplanned downtime, thereby improving equipment reliability and maintenance efficiency.

[0004] Fault feature extraction is a key step in gear fault diagnosis. At present, a variety of signal analysis methods have been applied to gear fault feature extraction. These signal analysis methods can be divided into the following three categories: time domain analysis method, frequency domain analysis method and time-frequency domain analysis method.

[0005] The fault feature extraction method based on complexity and similarity measurement is a time domain analysis method that can effectively extract gear fault features. During the gear operation process, gear faults will cause fault components to appear in the gear vibration signal, which in turn causes changes in the complexity and similarity of the vibration signal. Therefore, engineers often use complexity or similarity measurement methods to extract fault features.

[0006] In recent years, complexity measurement methods such as fuzzy entropy and permutation entropy have been widely used in gear fault feature extraction due to their superiority in processing non-stationary signals. However, the gear speed and gear meshing frequency in the signal and their harmonics seriously affect the performance of the above methods in fault extraction. To address this problem, the present invention patent discloses a gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity. This method can divide the vibration signal into asynchronous and synchronous components, and then calculate the fuzzy similarity of each component respectively, so as to extract the gear fault features more comprehensively. Summary of the invention

[0007] Purpose of the invention: The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed in the patent of this invention can extract the asynchronous and synchronous fault features of the gear, thereby distinguishing different gear fault types.

[0008] Technical solution: A gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity, specifically including the following steps:

[0009] (1) Determine the sampling frequency, sampling length, number of synchronous sampling points per gear revolution, embedding dimension m, time delay τ, and similarity tolerance threshold r for synchronous data acquisition;

[0010] (2) Using the gear speed as a reference, the vibration sensor and the speed sensor are used to collect synchronous vibration data X;

[0011] (3) Reconstructing the phase space P of the synchronous vibration data X according to the preset time delay τ and embedding dimension m;

[0012] (4) Centralize each m-dimensional vector in the phase space P to obtain the phase space P1 of the asynchronous component;

[0013] (5) Calculate the distance between each m-dimensional vector in the phase space P1 of the asynchronous component;

[0014] (6) Calculate asynchronous fuzzy similarity;

[0015] (7) synchronously average each m-dimensional vector in the phase space P to obtain the synchronous component phase space P2;

[0016] (8) Calculate the distance between each m-dimensional vector in the synchronization component phase space P2;

[0017] (9) Calculate synchronous fuzzy similarity;

[0018] (10) Asynchronous fuzzy similarity and synchronous fuzzy similarity are used as gear fault features.

[0019] Furthermore, the embedding dimension m in step (1) is determined according to the gear fault characteristics; and the time delay τ is set according to the gear rotation period.

[0020] Furthermore, the implementation process of step (2) is as follows:

[0021] The gear speed is used as a reference signal to synchronously obtain the vibration data of the gear during operation; the speed sensor is used to measure the gear speed signal, and the vibration data obtained during the gear operation is displacement data, velocity data or acceleration data; the obtained synchronous vibration data is expressed as:

[0022] X={x(1),x(2),…,x(i),…,x(kM)}

[0023] Where M is the number of data points collected per revolution, and k is the total number of gear revolutions for which the signal is collected.

[0024] Furthermore, the implementation process of step (3) is as follows:

[0025] The phase space of the synchronous vibration data is reconstructed according to the time delay τ and embedding dimension m pre-set in step (1). The reconstructed phase space P is Among them, the m-dimensional embedding vector It is expressed as:

[0026]

[0027] Furthermore, the implementation process of step (4) is as follows:

[0028] For each m-dimensional vector in the phase space P, the centralization method is used to centralize it to ensure that the mean of each m-dimensional vector is 0; therefore, the embedding vector Centralized into:

[0029]

[0030] in, is the embedding vector The mean is as follows:

[0031]

[0032] The phase space P1 of the non-synchronous component of the vibration data obtained by the central processing method is:

[0033] Furthermore, the implementation process of step (5) is as follows:

[0034] For the m-dimensional vector in the phase space P1 after centralization, calculate the m-dimensional vector and The mutual distance between:

[0035]

[0036] in, and They represent the i-th and j-th m-dimensional vectors after centralization in the phase space P1 respectively.

[0037] Furthermore, the implementation process of step (6) is as follows:

[0038] Calculate the standard deviation of all m-dimensional vector data in the phase space P1 after centralization As shown below:

[0039]

[0040] Wherein, N represents the total length of the vibration data X, that is, N = kM; A represents the mean of all data in the phase space P1 after centralization, as shown in the following formula:

[0041]

[0042] in, Represents the i-th m-dimensional vector after centralization The (λ+1)th data in .

[0043] According to the calculated mutual distance between each m-dimensional vector in the phase space Calculate the similarity between each m-dimensional vector in the phase space:

[0044]

[0045] Among them, r is the similarity tolerance threshold.

[0046] According to the calculated similarity between each m-dimensional vector in the phase space P1 Compute asynchronous fuzzy similarity for vibration data:

[0047]

[0048] Furthermore, the implementation process of step (7) is as follows:

[0049] For each m-dimensional vector in the phase space P, the synchronous averaging method is used to process it and eliminate the asynchronous component; therefore, the embedding vector is processed as:

[0050]

[0051] in, is the embedding vector The synchronous component phase space P2 of the synchronous vibration signal obtained by synchronous averaging is

[0052] Furthermore, the implementation process of step (8) is as follows:

[0053] For the m-dimensional vector in the phase space P2 after synchronous averaging, calculate the m-dimensional vector and The distance between them:

[0054]

[0055] in, and They represent the m-dimensional vectors after synchronous averaging of the i-th and j-th vectors in the phase space respectively.

[0056] Furthermore, the implementation process of step (9) is as follows:

[0057] Calculate the standard deviation of all m-dimensional vector data in the phase space P2 after synchronous averaging As shown below:

[0058]

[0059] Where N represents the total length of the vibration data X, that is, N = kM; B represents the mean of all data in the phase space after synchronous averaging, as shown in the following formula:

[0060]

[0061] Among them, z i (i+λM) represents the i-th m-dimensional vector after synchronous averaging The (λ+1)th data in .

[0062] According to the calculated mutual distance between each m-dimensional vector in the phase space P2 Calculate the similarity between each m-dimensional vector in the phase space:

[0063]

[0064] Among them, r is the similarity tolerance threshold.

[0065] According to the calculated similarity between each m-dimensional vector in the phase space P2 Compute the synchronized fuzzy similarity of vibration data:

[0066]

[0067] Furthermore, the implementation process of step (10) is as follows:

[0068] The asynchronous fuzzy similarity and the synchronous fuzzy similarity of the vibration data are combined, and the calculated asynchronous fuzzy similarity and the synchronous fuzzy similarity are used as the gear fault features extracted by the present invention.

[0069] Beneficial effects: Compared with the prior art, the present invention mainly includes the following beneficial effects: (1) The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed in the patent of the present invention can effectively extract gear fault features. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed in the patent of the present invention considers the fuzzy similarity of the asynchronous component and the synchronous component when extracting the gear fault feature, so as to better characterize the gear fault feature; (2) Compared with the traditional fuzzy entropy method, the gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed in the patent of the present invention requires less computational complexity; (3) The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed in the patent of the present invention is not affected by the gear speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flow chart of the gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed by the present invention;

[0071] Figure 2 It is a WT-planetary gearbox test bench in an embodiment of the present invention;

[0072] Figure 3 Gears for four health states;

[0073] Figure 4 is the gear fault characteristic value extracted in the embodiment of the present invention; DETAILED DESCRIPTION

[0074] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0075] The calculation process of the gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity disclosed in the present invention is as follows: Figure 1 As shown, the specific steps include:

[0076] Step (1): Determine the sampling frequency, sampling length, number of synchronous sampling points per gear revolution, embedding dimension m, time delay, and similarity tolerance threshold r for synchronous data acquisition.

[0077] The number of sampling points in each rotation period of the gear is determined by the sampling frequency. The sampling frequency is set to Mn to ensure that sufficient data M can be obtained in each rotation period of the gear, where n represents the gear speed (Hz). In this embodiment, the amount of data collected per gear rotation is M, so the time delay τ of phase space reconstruction is set to the time used to collect M data points, that is, the rotation period of the detected gear. The embedding dimension is set to 4, and the similarity tolerance threshold r is set to 0.15.

[0078] Step (2): Using the gear speed as a reference, a vibration sensor and a speed sensor are used to collect synchronous vibration data X.

[0079] The gear fault data set used in this embodiment is the WT-planetary gearbox data set published in the literature by Liu Dongdong et al. of Beijing University of Technology.

[0080] Figure 2 This is a WT-planetary gearbox test bench. The test bench consists of a motor, a planetary gearbox, a fixed-axis gearbox and a loading device. In a planetary gearbox, four planetary gears rotate around a sun gear.

[0081] The test bench simulates the working process of the sun gear under different health conditions. The health status of the sun gear includes gear health, gear damage, tooth surface wear and gear missing teeth. Figure 3 shown.

[0082] During the experiment, the sampling frequency was set to 48KHz, the sun gear speed was set to 20Hz, and the data length was set to twenty turns.

[0083] Through the resampling technology, the data is resampled to obtain data M = 2400 per gear revolution, so the time delay of phase space reconstruction is set to the time used for 2400 sampling intervals.

[0084] The phase space reconstruction of the synchronous vibration data is performed based on the phase space time delay τ=2400 and the embedding dimension m=4.

[0085] Step (3): Reconstruct the phase space P of the synchronous vibration data according to the preset time delay τ and embedding dimension m.

[0086] The phase space of the synchronous vibration data X is reconstructed according to the embedding dimension m and time delay τ set in steps (1) and (2):

[0087]

[0088] Among them, the embedding vector is an m-dimensional vector represented by:

[0089]

[0090] Step (4): Centralize each m-dimensional vector in the phase space P to obtain the phase space P1 of the asynchronous component.

[0091] For each m-dimensional vector in the phase space P, a centralization method is used to centralize it to ensure that the mean value of each m-dimensional vector is 0;

[0092] Therefore, the embedding vector Centralized into:

[0093]

[0094] in, is the embedding vector The mean is as follows:

[0095]

[0096] The phase space P1 of the non-synchronous component is as follows:

[0097]

[0098] Step (5): Calculate the distance between each m-dimensional vector in the phase space P1 of the asynchronous component.

[0099] Calculate the standard deviation of the asynchronous component phase space P1 data As shown below:

[0100]

[0101] Wherein, N represents the total length of the vibration data X, that is, N = kM; A represents the mean value of all data in the phase space P1, as shown in the following formula:

[0102]

[0103] in, Represents the i-th m-dimensional vector in phase space P1 The (λ+1)th data in .

[0104] For the m-dimensional vector in phase space P1, calculate the m-dimensional vector and The mutual distance between:

[0105]

[0106] in, and Represent the i-th and j-th m-dimensional vectors in the phase space P1 respectively.

[0107] Step (6): Calculate the asynchronous fuzzy similarity.

[0108] According to the mutual distance between the m-dimensional vectors obtained in step (5) Calculate the similarity between each m-dimensional vector in the phase space P1, as shown in the following formula:

[0109]

[0110] According to the similarity between the m-dimensional vectors in the calculated phase space Compute asynchronous fuzzy similarity for vibration data:

[0111]

[0112] Step (7): Synchronously average each m-dimensional vector in the phase space P to obtain the phase space P2 of the synchronous component.

[0113] For each m-dimensional vector in the phase space P, a synchronous averaging method is used to process it and eliminate the asynchronous component;

[0114] Therefore, the m-dimensional vector in the phase space P is processed as:

[0115]

[0116] in, is the embedding vector The mean of .

[0117] The phase space P2 of the synchronization component is as follows:

[0118]

[0119] Step (8): Calculate the distance between each m-dimensional vector in the synchronization component phase space P2.

[0120] Calculate the standard deviation of all m-dimensional vector data in phase space P2 As shown below:

[0121]

[0122] Wherein, N represents the total length of the vibration data X, that is, N = kM; B represents the mean of all data in the phase space P2 after synchronous averaging processing, as shown in the following formula:

[0123]

[0124] Among them, z i (i+λM) represents the i-th m-dimensional vector in the synchronization component phase space P2 The (λ+1)th data in .

[0125] For the m-dimensional vector in phase space P2, calculate the m-dimensional vector and The distance between them:

[0126]

[0127] in, and Represent the i-th and j-th m-dimensional vectors in the phase space P2 respectively.

[0128] Step (9): Calculate the synchronous fuzzy similarity.

[0129] According to the mutual distance between the m-dimensional embedding vectors obtained in step (8) Calculate the similarity between each embedding vector in the m-dimensional phase space as shown in the following formula:

[0130]

[0131] According to the similarity between the m-dimensional vectors in the calculated phase space Compute the synchronized fuzzy similarity of vibration data:

[0132]

[0133] Step (10): Use the asynchronous fuzzy similarity and the synchronous fuzzy similarity as gear fault features.

[0134] The asynchronous fuzzy similarity and synchronous fuzzy similarity values ​​of the four health status gears are shown as follows: Figure 4 As shown, the plus sign represents tooth surface wear, the square represents gear damage, the star represents gear missing teeth, and the circle represents healthy gear. Figure 4 It can be seen that the fault features of gear health, tooth surface wear and gear damage are well extracted; the fuzzy similarity values ​​of the gear in the state of missing teeth are relatively scattered. This is because when the gear is in the state of missing teeth, the probability of random situations occurring during meshing is greater, and the collected vibration data usually shows significant disorder.

[0135] From the results, the method of the present invention can well extract the fault characteristics of gears in four healthy states: healthy gear, tooth surface wear, gear damage and gear missing teeth. The method of the present invention has a small computational complexity. In this embodiment, the average computational cost of extracting asynchronous fuzzy similarity and synchronous fuzzy similarity by the method of the present invention is 0.249 seconds, while the average computational cost of the traditional fuzzy entropy method is 1.449 seconds. It can be seen that compared with the traditional fuzzy entropy method, the method of the present invention has a lower computational complexity. The method of the present invention can extract gear fault characteristics in real time.

[0136] In summary, compared with the gear fault feature extraction method based on traditional fuzzy entropy, the method of the present invention can effectively extract gear fault features. At the same time, the present invention has lower computational complexity, significantly saves computational costs, and can extract gear fault features in real time.

Claims

1. A gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity, characterized in that: The following steps are involved: (1) Determine the sampling frequency, sampling length, number of synchronous sampling points per gear revolution, embedding dimension m, time delay τ, and similarity tolerance threshold r for synchronous data acquisition; (2) Using the gear speed as a reference, the vibration sensor and the speed sensor are used to collect synchronous vibration data X; (3) Reconstruct the phase space P of the synchronous vibration data according to the preset time delay τ and embedding dimension m; (4) Centralize each m-dimensional vector in the phase space P to obtain the phase space P1 of the asynchronous component; (5) Calculate the distance between each m-dimensional vector in the phase space P1 of the asynchronous component; (6) Calculate asynchronous fuzzy similarity; (7) synchronously average each m-dimensional vector in the phase space P to obtain the phase space P2 of the synchronous component; (8) Calculate the distance between each m-dimensional vector in the synchronization component phase space P2; (9) Calculate synchronous fuzzy similarity; (10) Asynchronous fuzzy similarity and synchronous fuzzy similarity are used as gear fault features.

2. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The embedding dimension m in step (1) is determined according to the gear fault characteristics; the time delay τ is set according to the gear rotation period.

3. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (2) is as follows: The gear speed is used as a reference signal to synchronously obtain the vibration data of the gear during operation; the speed sensor is used to measure the gear speed signal, and the vibration data obtained during the gear operation is displacement data, velocity data or acceleration data; the obtained synchronous vibration data is expressed as: X={x(1),x(2),…,x(i),…,x(kM)} Where M is the number of data points collected per revolution, and k is the total number of gear revolutions for which the signal is collected.

4. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (3) is as follows: The phase space of the synchronous vibration data is reconstructed according to the time delay τ and embedding dimension m pre-set in step (1). The reconstructed phase space P is Among them, the m-dimensional embedding vector x i m It is expressed as:

5. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (4) is as follows: For each m-dimensional vector in the phase space, a centralization method is used to centralize it to ensure that the mean value of each m-dimensional vector is 0; Therefore, the embedding vector in, is the embedding vector The mean is as follows: The phase space P1 of the non-synchronous component of the vibration data obtained by the central processing method is:

6. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (5) is as follows: For the m-dimensional vector in the phase space P1 after centralization, calculate the m-dimensional vector and The mutual distance between: in, and They represent the i-th and j-th m-dimensional vectors after centralization in the phase space P1 respectively.

7. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (6) is as follows: Calculate the standard deviation of all m-dimensional vector data in the phase space P1 after centralization As shown below: Wherein, N represents the total length of the vibration data X, that is, N = kM; A represents the mean of all data in the phase space P1 after centralization, as shown in the following formula: in, Represents the i-th m-dimensional vector after centralization The (λ+1)th data in . According to the calculated mutual distance between each m-dimensional vector in the phase space P1 Calculate the similarity between each m-dimensional vector in the phase space P1: Among them, r is the similarity tolerance threshold. According to the calculated similarity between each m-dimensional vector in the phase space P1 Compute asynchronous fuzzy similarity for vibration data:

8. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (7) is as follows: For each m-dimensional vector in the phase space P, the synchronous averaging method is used to process it and eliminate the asynchronous component; therefore, the embedding vector is processed as: in, is the embedding vector The synchronous component phase space P2 of the synchronous vibration signal obtained by synchronous averaging is 9. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (8) is as follows: For the m-dimensional vector in the phase space P2 after synchronous averaging, calculate the m-dimensional vector and The mutual distance between: in, and They represent the m-dimensional vectors after synchronous averaging of the i-th and j-th vectors in the phase space P2 respectively.

10. The gear fault feature extraction method based on asynchronous and synchronous fuzzy similarity according to claim 1 is characterized in that: The implementation process of step (9) is as follows: Calculate the standard deviation of all m-dimensional vector data in the phase space P2 after synchronous averaging As shown below: Where N represents the total length of the vibration data X, that is, N = kM; B represents the mean of all data in the phase space after synchronous averaging, as shown in the following formula: Among them, z i (i+λM) represents the i-th m-dimensional vector after synchronous averaging The (λ+1)th data in . According to the calculated mutual distance between each m-dimensional vector in the phase space P2 Calculate the similarity between each m-dimensional vector in the phase space: Among them, r is the similarity tolerance threshold. According to the calculated similarity between each m-dimensional vector in the phase space P2 Compute the synchronized fuzzy similarity of vibration data:

Images