Planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM

Abstract

The invention provides a planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM, and the method comprises the steps: selecting a sensor and an installation position thereof, and setting parameters; acquiring synchronous vibration data; decomposing the synchronous vibration data to obtain a synchronous component and a non-synchronous component; respectively carrying out symbolic noise reduction processing on the synchronous component and the asynchronous component to obtain corresponding synchronous and asynchronous symbol sequences; reconstructing a phase space of the synchronous and asynchronous symbol sequences; decomposing the synchronous symbol sequence phase space and the asynchronous symbol sequence phase space again to obtain four phase spaces; fuzzy similarities of the four phase spaces are calculated respectively, and the fuzzy similarities are used as fault features; and inputting the four fault features into a pre-trained SVM (Support Vector Machine) model for fault diagnosis.

Description

Technical Field

[0001] The present invention belongs to the technical field of rotating machinery fault diagnosis, and in particular relates to a planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM (Support Vector Machine). Background Art

[0002] In modern industrial systems, gears are core components of mechanical transmission systems, and their performance is directly related to the overall system's efficiency and safety. Due to harsh operating environments and increasing operating speeds, gears are susceptible to various forms of damage, including wear, cracks, missing teeth, and breakage. These failures not only impair mechanical system performance but can also lead to serious safety incidents, threatening personnel and property. With increasing automation and advancements in precision manufacturing technology, monitoring gear health and diagnosing faults are becoming increasingly important. Effective fault diagnosis not only improves production efficiency but also mitigates economic losses.

[0003] Fault diagnosis typically involves the following aspects: (1) fault detection, i.e., identifying abnormal conditions; (2) fault classification, i.e., determining the specific type of fault through data analysis; (3) fault assessment, i.e., using specific indicators to measure the severity of the fault; and (4) fault tracing, i.e., aiming to accurately identify the root cause of the fault. Among these tasks, fault feature extraction is a key link and the focus of fault diagnosis research.

[0004] To address the problem of extracting fault signatures from complex vibration signals, researchers have introduced signal processing techniques. When a gear fails, the signal captured by the sensor will contain fault components, causing the signal complexity to change. Consequently, some researchers have proposed using complexity analysis methods for gear fault diagnosis. For example, these methods, such as calculating entropy, correlation dimensions, or applying symbolic dynamics, can quantify the signal complexity before and after a fault.

[0005] Entropy-based fault diagnosis methods are widely adopted due to their ability to efficiently capture fault information. Common entropy methods in this category include sample entropy, fuzzy entropy, and permutation entropy. However, existing entropy methods still have limitations in feature extraction. For example, they focus primarily on the overall complexity of the signal, while ignoring the periodic variations of fault components based on the gear rotation period and gear meshing period, which are crucial for identifying gear fault types.

[0006] In addition, existing entropy methods often lack effective noise reduction measures, making it difficult to separate weak fault signals from strong background noise, especially in the early stages of a fault, which poses a challenge to capturing dynamic changes. In order to diagnose gear faults more accurately, new fault diagnosis methods need to be developed. The method of the present invention not only takes into account synchronous and asynchronous fault information related to the gear rotation frequency, but also pays attention to synchronous and asynchronous fault information related to the gear meshing frequency. In addition, the method of the present invention adopts symbolic noise reduction processing technology to suppress the influence of disturbance noise on fault characteristics, thereby obtaining a higher fault diagnosis reliability. Summary of the Invention

[0007] Purpose of the invention: The present invention provides a planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM, which can generate four fault features, thereby detecting whether a planetary gearbox fault occurs and diagnosing the fault category.

[0008] Technical solution: The present invention proposes a planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM, which specifically includes the following steps:

[0009] (1) Select the sensor and its installation location, and set the parameters;

[0010] (2) Obtaining synchronous vibration data X;

[0011] (3) Decompose the synchronous vibration data X to obtain the synchronous component X T and the asynchronous component X F ;

[0012] (4) For the synchronization component X T and the asynchronous component X F Perform symbolic noise reduction processing to obtain the corresponding symbol sequence S T and S F ;

[0013] (5) Reconstruct the symbol sequence S respectively T and S F The phase space of P is obtained. T and P F ;

[0014] (6) Decomposition of phase space P T , and obtain its synchronous component phase space P TT and the non-synchronous component phase space P FT , decompose the phase space P F , and obtain its synchronous component phase space P FT and the non-synchronous component phase space P FF ;

[0015] (7) Calculate the fuzzy similarity of the above four phase spaces respectively and use it as the fault feature of the planetary gearbox;

[0016] (8) The fault feature data obtained in step (7) is input into the pre-trained SVM model for planetary gearbox fault diagnosis.

[0017] Furthermore, the sensors selected in step (1) are a speed sensor and a vibration sensor; the vibration sensor is installed on the planetary gearbox, and the speed sensor is installed near the main shaft of the sun gear; the parameters that need to be set include: the number of equal-angle sampling points M per rotation of the sun gear, the sampling length, the number of sun gear teeth c, the embedding dimension m, the time delay τ, and the number of symbols ε; the number of equal-angle sampling points M per rotation of the sun gear must be an integer multiple of the number of sun gear teeth c; the sampling length is the data length kM collected within k rotations of the sun gear; the time delay τ is set according to the period of operation of the sun gear per rotation and the number of sun gear teeth; the embedding dimension m and the number of symbols ε are determined according to the gear fault characteristics.

[0018] Furthermore, the implementation process of step (2) is as follows:

[0019] A speed sensor is used to measure the sun gear speed signal, and the speed signal is used as a reference signal. The equi-angle sampling vibration data during the operation of the planetary gearbox is obtained, that is, the synchronous vibration data X. The obtained synchronous vibration data X can be expressed as:

[0020] X={x(1),x(2),…,x(i),…,x(kM)},

[0021] Furthermore, the implementation process of step (3) is as follows: taking the rotation period of the sun gear as a reference, the synchronous vibration data X is processed using the angular domain synchronous averaging algorithm to obtain the synchronous component X of the synchronous vibration data X. T , the angular domain synchronous averaging process is as follows:

[0022]

[0023] Among them, x(i) is the vibration signal collected at equal angles; through angular domain synchronous averaging processing, the synchronous component X of the equal angle vibration data is obtained. T ={x T (1),x T (2),…,x T (i),...,x T (kM)}; Subtract the synchronous component X from the synchronous vibration data X T , get the asynchronous component X of the synchronous vibration data X F , the process is as follows:

[0024]

[0025] Furthermore, the implementation process of step (4) is as follows:

[0026] For the synchronization component X T and the asynchronous component X F Sort in ascending order, and then group the sorted sequence evenly according to the number of symbols ε to form ε intervals, each of which has data, and the symbol value of each interval is the mean value of all data in the interval; after symbolic coding processing, the symbol sequence S of the synchronization component is obtained respectively. T ={s T (1),s T (2),…,s T (i),...,s T (kM)} and the symbol sequence S of the asynchronous component F ={s F (1),s F (2),…,s F (i),...,s F (kM)}.

[0027] Furthermore, the implementation process of step (5) is as follows:

[0028] According to the preset time delay τ and embedding dimension m, the symbol sequence S T and S F Reconstruct the phase space and obtain the phase space P T and P F :

[0029]

[0030] Among them, the time delay That is, the number of sampling points within the sun gear meshing period.

[0031] Furthermore, the implementation process of step (6) is as follows:

[0032] According to the meshing period of the sun gear, the angular domain synchronous averaging method is used to process the phase space P T For each m-dimensional vector in, we get the symbol sequence S T The synchronous component phase space P TT , and in the phase space P T Subtract the phase space P from TT , get the symbol sequence S T The phase space P of the non-synchronous component TF :

[0033]

[0034]

[0035] According to the meshing period of the sun gear, the phase space P is processed by angular domain synchronous averaging. F For each m-dimensional vector in, we get the symbol sequence S F The synchronous component phase space P FT , in the phase space P F Subtract the phase space P from FT , get the symbol sequence S F The phase space P of the non-synchronous component FF :

[0036]

[0037] Furthermore, the implementation process of step (7) is as follows:

[0038] For the phase space P TT For all m-dimensional vectors in , calculate the m-dimensional vector P i TT and The Chebyshev distance between:

[0039]

[0040] Among them, P i TT and Represent the phase space P TT The i-th and j-th m-dimensional vectors in ;

[0041] Calculate the phase space P TT The standard deviation of all m-dimensional vector data in As shown below:

[0042]

[0043] Among them, A TT Represents the phase space P TT The mean of all data in :

[0044]

[0045] According to the calculated phase space P TT The Chebyshev distance between each m-dimensional vector in the phase space P is calculated TT The similarity between each m-dimensional vector in:

[0046]

[0047] According to the calculated phase space P TT The similarity between each m-dimensional vector Compute the first fuzzy similarity for vibration data:

[0048]

[0049] For the phase space P TF For all m-dimensional vectors in , calculate the m-dimensional vector P i TF and The Chebyshev distance between:

[0050]

[0051] Among them, P i TF and Represent the phase space P TF The i-th and j-th m-dimensional vectors in ;

[0052] Calculate the phase space P TF The standard deviation of all m-dimensional vector data in As shown below:

[0053]

[0054] Among them, A TF Represents the phase space P TF The mean of all data in :

[0055]

[0056] According to the calculated phase space P TF The Chebyshev distance between each m-dimensional vector in the phase space P is calculated TF The similarity between each m-dimensional vector in:

[0057]

[0058] According to the calculated phase space P TF The similarity between each m-dimensional vector Calculate the second fuzzy similarity for the vibration data:

[0059]

[0060] For the phase space P FT For all m-dimensional vectors in , calculate the m-dimensional vector P i FT and The Chebyshev distance between:

[0061]

[0062] Among them, P iFT and Represent the phase space P FT The i-th and j-th m-dimensional vectors in ;

[0063] Calculate the phase space P FT The standard deviation of all m-dimensional vector data in As shown below:

[0064]

[0065] Among them, A FT Represents the phase space P FT The mean of all data in :

[0066]

[0067] According to the calculated phase space P FT The Chebyshev distance between each m-dimensional vector in the phase space P is calculated FT The similarity between each m-dimensional vector in:

[0068]

[0069] According to the calculated phase space P FT The similarity between each m-dimensional vector Calculate the third fuzzy similarity of the vibration data:

[0070]

[0071] For the phase space P FF For all m-dimensional vectors in , calculate the m-dimensional vector P i FF and The Chebyshev distance between:

[0072]

[0073] Among them, P i FF and Represent the phase space P FF The i-th and j-th m-dimensional vectors in ;

[0074] Calculate the phase space P FF The standard deviation of all m-dimensional vector data in As shown below:

[0075]

[0076] Among them, A FF Represents the phase space P FF The mean of all data in :

[0077]

[0078] According to the calculated phase space P FF The Chebyshev distance between each m-dimensional vector in the phase space P is calculated FF The similarity between each m-dimensional vector in:

[0079]

[0080] According to the calculated phase space P FF The similarity between each m-dimensional vector Calculate the fourth fuzzy similarity of vibration data:

[0081]

[0082] The four fuzzy similarities TTΦ(m), TFΦ(m), FTΦ(m) and FFΦ(m) are used as the four fault features of the gearbox.

[0083] Furthermore, the implementation process of step (8) is as follows:

[0084] The four calculated fault feature data are input into a pre-trained SVM fault diagnosis model to perform planetary gearbox fault diagnosis; the pre-trained SVM fault diagnosis model is obtained by training the SVM model based on the health status of the planetary gearbox and its corresponding four fuzzy similarities.

[0085] Beneficial effects

[0086] Compared with the prior art, the present invention mainly has the following beneficial effects:

[0087] (1) The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM disclosed in the present invention suppresses the influence of disturbance noise on different fault information through two signal decomposition processes and one symbolic filtering process, highlights the fault information component in the vibration signal, and thus better characterizes the gear fault characteristics;

[0088] (2) The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM disclosed in the patent of this invention takes into account the relationship between the fault characteristics and the gear rotation frequency and gear meshing frequency, thereby obtaining characteristic values ​​of four dimensions. Compared with the traditional fuzzy entropy method, the method of this invention has higher fault diagnosis accuracy and more stable performance;

[0089] (3) The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM disclosed in the patent of this invention takes into account the characteristics of gear faults when establishing the phase space and adopts the gear meshing period as the time delay, thereby overcoming the disadvantage of traditional entropy that there is no unified determination of time delay.

[0090] (4) The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM disclosed in the patent of this invention is not affected by changes in the gearbox speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 This is a flow chart of the planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM disclosed in the present invention;

[0092] Figure 2 This is a photo of the WT-planetary gearbox test bench in an embodiment of the present invention;

[0093] Figure 3 Gears for five health states;

[0094] Figure 4 are the four fuzzy similarities TT, TF, FT and FF of the method of the present invention;

[0095] Figure 5 is the confusion matrix of the SVM fault diagnosis model training set of the method of the present invention;

[0096] Figure 6 is the confusion matrix of the SVM fault diagnosis model test set of the method of the present invention;

[0097] Figure 7 The confusion matrix of the SVM fault diagnosis model test set when the four fuzzy similarities of the method of the present invention are used as fault features separately; DETAILED DESCRIPTION

[0098] The present invention will be described in further detail below with reference to the accompanying drawings. Figure 1 The process of the planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM disclosed in the present invention is shown, which specifically includes the following steps:

[0099] Step (1): Select the sensor and its installation location, and set the parameters.

[0100] The selected sensors are a vibration sensor and a speed sensor; the vibration sensor is installed on the planetary gearbox, and the speed sensor is installed near the sun gear main shaft; the parameters that need to be set include: the number of equal-angle sampling points M per sun gear rotation, the sampling length, the number of sun gear teeth c, the embedding dimension m, the time delay τ, and the number of symbols ε; the number of equal-angle sampling points M per sun gear rotation must be an integer multiple of the number of sun gear teeth c; the sampling length is the data length kM collected within k rotations of the sun gear; the time delay τ is set according to the period of operation per sun gear rotation and the number of sun gear teeth; the embedding dimension m and the number of symbols ε are determined according to the gear fault characteristics.

[0101] The gear fault dataset used in this embodiment is the WT-planetary gearbox dataset published in the literature by Liu Dongdong et al. from Beijing University of Technology.

[0102] Figure 2 This is a WT planetary gearbox test bench. It 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.

[0103] The test bench simulates the working process of the sun gear under five different working conditions. The working conditions of the sun gear include gear health, gear damage, tooth root fracture, tooth surface wear and gear tooth missing. Figure 3 shown.

[0104] During the experiment, the sun gear rotates at 20Hz, and a total of 100 sets of sample data are used, with each set of sample data being five revolutions long. The number of sun gear teeth c = 28, and the data is resampled to obtain data M = 2800 per gear revolution through resampling technology to ensure that the number of sampling points in the gear meshing cycle is an integer, and the time delay is The embedding dimension m is set to 2 and the number of symbols ε is set to 20.

[0105] Step (2): Obtain synchronous vibration data X.

[0106] A speed sensor is used to measure the sun gear speed signal, and this speed signal is used as a reference signal. A vibration sensor is used to obtain vibration data during the operation of the planetary gearbox. According to the set M, the vibration data is resampled according to the speed signal to obtain the synchronous vibration data X. The synchronous vibration data can be expressed as follows:

[0107] X={x(1),x(2),…,x(i),…,x(kM)}.

[0108] Since the sun gear rotates at 20 Hz, the data length of each time window is the data length within five rotations of the sun gear, that is, the number of rotations k = 5. Therefore, the synchronous vibration data X can be expressed as follows:

[0109] X={x(1),x(2),…,x(i),…,x(14000)}.

[0110] Step (3): Decompose the synchronous vibration data X to obtain the synchronous component X T and the asynchronous component X F .

[0111] Taking the sun gear rotation period as a reference, the synchronous vibration data X is processed using the angular domain synchronous averaging algorithm to obtain the synchronous component X of the synchronous vibration data X. T , the angular domain synchronous averaging process is as follows:

[0112]

[0113] Among them, x(i) is the vibration signal collected at equal angles; through angular domain synchronous averaging processing, the synchronous component X of the equal angle vibration data is obtained. T ={x T (1),x T (2),…,x T (i),…,x T (kM)}; Subtract the synchronous component X from the synchronous vibration data X T , get the asynchronous component X of the synchronous vibration data X F , the process is as follows:

[0114]

[0115] Step (4): For the synchronization component X T and the asynchronous component X F Perform symbolic noise reduction processing to obtain the corresponding symbol sequence S T and S F .

[0116] For the synchronization component X T and the asynchronous component X F Sort in ascending order, and then group the sorted sequence evenly according to the number of symbols ε to form ε intervals, each of which has data, and the symbol value of each interval is the mean value of all data in the interval; after symbolic coding processing, the symbol sequence S of the synchronization component is obtained respectively. T ={s T (1),s T (2),…,s T (i),…,s T (kM)} and the symbol sequence S of the asynchronous component F ={s F (1),s F (2),…,s F (i),…,s F (kM)}.

[0117] Step (5): Reconstruct the symbol sequence S respectively T and S F The phase space of P is obtained. T and P F .

[0118] According to the preset time delay τ and embedding dimension m, the symbol sequence S T and S F Reconstruct the phase space and obtain the phase space P T and PF :

[0119]

[0120] Among them, the time delay That is, the number of sampling points within the sun gear meshing period.

[0121] Step (6): Decompose the phase space P T , and obtain its synchronous component phase space P TT and the non-synchronous component phase space P FT , decompose the phase space P F , and obtain its synchronous component phase space P FT and the non-synchronous component phase space P FF .

[0122] According to the meshing period of the sun gear, the angular domain synchronous averaging method is used to process the phase space P T For each m-dimensional vector in, we get the symbol sequence S T The synchronous component phase space P TT , and in the phase space P T Subtract the phase space P from TT , get the symbol sequence S T The phase space P of the non-synchronous component TF :

[0123]

[0124]

[0125] According to the meshing period of the sun gear, the phase space P is processed by angular domain synchronous averaging. F For each m-dimensional vector in, we get the symbol sequence S F The synchronous component phase space P FT , in the phase space P F Subtract the phase space P from FT , get the symbol sequence S F The phase space P of the non-synchronous component FF :

[0126]

[0127] Step (7): Calculate the fuzzy similarities of the above four phase spaces respectively and use them as the fault features of the planetary gearbox.

[0128] For the phase space P TT For all m-dimensional vectors in , calculate the m-dimensional vector P i TT and The Chebyshev distance between:

[0129]

[0130] Among them, P i TT and Represent the phase space P TT The i-th and j-th m-dimensional vectors in ;

[0131] Calculate the phase space P TT The standard deviation of all m-dimensional vector data in As shown below:

[0132]

[0133] Among them, A TT Represents the phase space P TT The mean of all data in :

[0134]

[0135] According to the calculated phase space P TT The Chebyshev distance between each m-dimensional vector in the phase space P is calculated TT The similarity between each m-dimensional vector in:

[0136]

[0137] According to the calculated phase space P TT The similarity between each m-dimensional vector Compute the first fuzzy similarity for vibration data:

[0138]

[0139] For the phase space P TF For all m-dimensional vectors in , calculate the m-dimensional vector P i TF and The Chebyshev distance between:

[0140]

[0141] Among them, P i TF and Represent the phase space P TF The i-th and j-th m-dimensional vectors in ;

[0142] Calculate the phase space P TF The standard deviation of all m-dimensional vector data in As shown below:

[0143]

[0144] Among them, A TF Represents the phase space P TF The mean of all data in :

[0145]

[0146] According to the calculated phase space P TF The Chebyshev distance between each m-dimensional vector in the phase space P is calculated TF The similarity between each m-dimensional vector in:

[0147]

[0148] According to the calculated phase space P TF The similarity between each m-dimensional vector Calculate the second fuzzy similarity for the vibration data:

[0149]

[0150] For the phase space P FT For all m-dimensional vectors in , calculate the m-dimensional vector P i FT and The Chebyshev distance between:

[0151]

[0152] Among them, P i FT and Represent the phase space P FT The i-th and j-th m-dimensional vectors in ;

[0153] Calculate the phase space P FT The standard deviation of all m-dimensional vector data in As shown in the following formula:

[0154]

[0155] Among them, A FT Represents the phase space P FT The mean of all data in :

[0156]

[0157] According to the calculated phase space P FT The Chebyshev distance between each m-dimensional vector in the phase space P is calculated FT The similarity between each m-dimensional vector in:

[0158]

[0159] According to the calculated phase space P FT The similarity between each m-dimensional vector Calculate the third fuzzy similarity of the vibration data:

[0160]

[0161] For the phase space P FF For all m-dimensional vectors in , calculate the m-dimensional vector P i FF and The Chebyshev distance between:

[0162]

[0163] Among them, P i FF and Represent the phase space P FF The i-th and j-th m-dimensional vectors in ;

[0164] Calculate the phase space P FF The standard deviation of all m-dimensional vector data in As shown in the following formula:

[0165]

[0166] Among them, A FF Represents the phase space P FF The mean of all data in :

[0167]

[0168] According to the calculated phase space P FF The Chebyshev distance between each m-dimensional vector in the phase space P is calculated FF The similarity between each m-dimensional vector in:

[0169]

[0170] According to the calculated phase space P FF The similarity between each m-dimensional vector Calculate the fourth fuzzy similarity of vibration data:

[0171]

[0172] The four fuzzy similarities TTΦ(m), TFΦ(m), FTΦ(m) and FFΦ(m) are used as the four fault features of the gearbox.

[0173] Step (8): Input the fault feature data obtained in step (7) into the pre-trained SVM model to perform planetary gearbox fault diagnosis.

[0174] The four fuzzy similarities under the five health states of the sun wheel are as follows Figure 4 As shown in the figure, it is difficult to accurately distinguish the five different health states using any one dimension of fuzzy similarity. However, combining the four fuzzy similarities can achieve accurate fault diagnosis. Therefore, by using the four fuzzy similarities and the SVM fault diagnosis model, we can achieve higher diagnostic accuracy.

[0175] When training the SVM fault diagnosis model, we selected 50 groups of sample results as the training set, trained the SVM diagnosis model, and then used the remaining 50 groups of sample results as the test set to verify the performance of the method of the present invention. The confusion matrices of the method of the present invention on the training set and the test set are shown as follows: Figure 5 and Figure 6 As shown. Figure 5 and Figure 6 We can see that, benefiting from the eigenvalues ​​of the four dimensions, the patented method has a higher fault diagnosis accuracy. Figure 7 The confusion matrix for the test set is shown when the four fuzzy similarities are used individually as fault signatures. As can be seen from the figure, no single fuzzy similarity can accurately distinguish the fault type. This patented method can not only determine whether a gear fault has occurred, but also distinguish the fault type, thus efficiently completing gear fault diagnosis.

Claims

1. A planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM, characterized by: The following steps are involved: (1) Select the sensor and its installation location, and set the parameters; (2) Obtaining synchronous vibration data X; (3) Decompose the synchronous vibration data X to obtain the synchronous component X T and the asynchronous component X F ; (4) For the synchronization component X T and the asynchronous component X F Perform symbolic noise reduction processing to obtain the corresponding symbol sequence S T and S F ; (5) Reconstruct the symbol sequence S respectively T and S F The phase space of P is obtained. T and P F ; (6) Decomposition of phase space P T , and obtain its synchronous component phase space P TT and the non-synchronous component phase space P FT , decompose the phase space P F , and obtain its synchronous component phase space P FT and the non-synchronous component phase space P FF ; (7) Calculate the fuzzy similarity of the above four phase spaces respectively and use it as the fault feature of the planetary gearbox; (8) The fault feature data obtained in step (7) is input into the pre-trained SVM model for planetary gearbox fault diagnosis.

2. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1 is characterized in that: The sensors described in step (1) are a vibration sensor and a speed sensor; the vibration sensor is installed on the planetary gearbox, and the speed sensor is installed near the sun gear main shaft; the parameters that need to be set include: the number of equal-angle sampling points M per sun gear rotation, the sampling length, the number of sun gear teeth c, the embedding dimension m, the time delay τ, and the number of symbols ε; the number of equal-angle sampling points M per sun gear rotation must be an integer multiple of the number of sun gear teeth c; the sampling length is the data length kM collected within k rotations of the sun gear; the time delay τ is set according to the period of operation of the sun gear per rotation and the number of sun gear teeth; the embedding dimension m and the number of symbols ε are determined according to the gear fault characteristics.

3. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1 is characterized in that: The implementation process of step (2) is as follows: using a speed sensor to measure the sun gear speed signal, and using the speed signal as a reference signal; obtaining equal-angle sampling vibration data during the operation of the planetary gearbox, that is, synchronous vibration data X; the obtained synchronous vibration data X can be expressed as: X={x(1),x(2),…,x(i),…,x(kM)}, Where k is the total number of revolutions of the sun gear, and M is the number of synchronous sampling points per revolution of the sun gear.

4. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1 is characterized in that: The implementation process of step (3) is as follows: using the sun gear rotation period as a reference, the synchronous vibration data X is processed using the angular domain synchronous averaging algorithm to obtain the synchronous component X of the synchronous vibration data X. T , the angular domain synchronous averaging process is as follows: Among them, x(i) is the vibration signal collected at equal angles; through angular domain synchronous averaging processing, the synchronous component X of the equal angle vibration data is obtained. T ={x T (1),x T (2),…,x T (i),…,x T (kM)}; Subtract the synchronous component X from the synchronous vibration data X T , get the asynchronous component X of the synchronous vibration data X F , the process is as follows: X F ={x(1)-x T (1),x(2)-x T (2),…,x(i)-x T (i),…,x(kM)-x T (kM)} ={x F (1),x F (2),…,x F (i),…,x F (kM)}。 5. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1 is characterized in that: The implementation process of step (4) is as follows: T and the asynchronous component X F Sort in ascending order, and then group the sorted sequence evenly according to the number of symbols ε to form ε intervals, each of which has data, and the symbol value of each interval is the mean value of all data in the interval; after symbolic coding processing, the symbol sequence S of the synchronization component is obtained respectively. T ={s T (1),s T (2),…,s T (i),…,s T (kM)} and the symbol sequence S of the asynchronous component F ={s F (1),s F (2),…,s F (i),…,s F (kM)}.

6. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1, characterized in that: The implementation process of step (5) is as follows: according to the preset time delay τ and embedding dimension m, the symbol sequence S T and S F Reconstruct the phase space and obtain the phase space P T and P F : Among them, the time delay That is, the number of sampling points within the sun gear meshing period.

7. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1, characterized in that: The implementation process of step (6) is as follows: According to the meshing period of the sun gear, the phase space P is processed by the angular domain synchronous averaging method. T For each m-dimensional vector in, we get the symbol sequence S T The synchronous component phase space P TT , and in the phase space P T Subtract the phase space P from TT , get the symbol sequence S T The phase space P of the non-synchronous component TF : According to the meshing period of the sun gear, the phase space P is processed by angular domain synchronous averaging. F For each m-dimensional vector in, we get the symbol sequence S F The synchronous component phase space P FT , in the phase space P F Subtract the phase space P from FT , get the symbol sequence S F The phase space P of the non-synchronous component FF :

8. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1, characterized in that: The implementation process of step (7) is as follows: For the phase space P TT For all m-dimensional vectors in , calculate the m-dimensional vector P i TT and The Chebyshev distance between: Among them, P i TT and Represent the phase space P TT The i-th and j-th m-dimensional vectors in ; Calculate the phase space P TT The standard deviation of all m-dimensional vector data in As shown below: Among them, A TT Represents the phase space P TT The mean of all data in : According to the calculated phase space P TT The Chebyshev distance between each m-dimensional vector in the phase space P is calculated TT The similarity between each m-dimensional vector in: According to the calculated phase space P TT The similarity between each m-dimensional vector in Compute the first fuzzy similarity for vibration data: For the phase space P TF For all m-dimensional vectors in , calculate the m-dimensional vector P i TF and The Chebyshev distance between: Among them, P i TF and Represent the phase space P TF The i-th and j-th m-dimensional vectors in ; Calculate the phase space P TF The standard deviation of all m-dimensional vector data in As shown below: Among them, A TF Represents the phase space P TF The mean of all data in : According to the calculated phase space P TF The Chebyshev distance between each m-dimensional vector in the phase space P is calculated TF The similarity between each m-dimensional vector in: According to the calculated phase space P TF The similarity between each m-dimensional vector Calculate the second fuzzy similarity for the vibration data: For the phase space P FT For all m-dimensional vectors in , calculate the m-dimensional vector P i FT and The Chebyshev distance between: Among them, P i FT and Represent the phase space P FT The i-th and j-th m-dimensional vectors in ; Calculate the phase space P FT The standard deviation of all m-dimensional vector data in As shown below: Among them, A FT Represents the phase space P FT The mean of all data in : According to the calculated phase space P FT The Chebyshev distance between each m-dimensional vector in the phase space P is calculated FT The similarity between each m-dimensional vector in: According to the calculated phase space P FT The similarity between each m-dimensional vector in Calculate the third fuzzy similarity of the vibration data: For the phase space P FF For all m-dimensional vectors in , calculate the m-dimensional vector P i FF and The Chebyshev distance between: Among them, P i FF and Represent the phase space P TF The i-th and j-th m-dimensional vectors in ; Calculate the phase space P FF The standard deviation of all m-dimensional vector data in As shown below: Among them, A FF Represents the phase space P FF The mean of all data in : According to the calculated phase space P FF The Chebyshev distance between each m-dimensional vector in the phase space P is calculated FF The similarity between each m-dimensional vector in: According to the calculated phase space P FF The similarity between each m-dimensional vector Calculate the fourth fuzzy similarity of vibration data: The four fuzzy similarities TTΦ(m), TFΦ(m), FTΦ(m) and FFΦ(m) are used as the four fault features of the gearbox.

9. The planetary gearbox fault diagnosis method based on multi-fuzzy similarity and SVM according to claim 1, characterized in that: The implementation process of step (8) is as follows: The four calculated fault feature data are input into a pre-trained SVM fault diagnosis model to perform planetary gearbox fault diagnosis; the pre-trained SVM fault diagnosis model is obtained by training the SVM model based on the health status of the planetary gearbox and its corresponding four fuzzy similarities.

Images