Wind turbine planetary gear train crack fault diagnosis method based on dynamic simulation and dual GRA

Abstract

The present invention discloses a crack fault diagnosis method for a planetary gear system of a wind turbine based on a combination of dynamic simulation and dual GRA. First, a three-dimensional model of the sun gear under different crack depth fault states is constructed, and the flexible body file of the gear is generated using finite element software. The rigid-flexible coupling model is constructed in Adams, and dynamic analysis is performed. Second, the time domain simulation signal is extracted based on the dynamic simulation, and the corresponding frequency domain signal is obtained using FFT, and multiple characteristic parameters related to the change of gear damage fault in the time and frequency domain are extracted. Then, parameters related to the calculation sequence are introduced to construct a correlation function related to the sequence elements, thereby simplifying the calculation sequence and improving the algorithm accuracy. The method proposed by the present invention can effectively calculate the threshold value of the crack depth, shorten the calculation sequence length, and at the same time, the improved calculation results have greater discrimination, the relationship between the fault sample and the standard data sample is more prominent, and the difference between the maximum value of the gray correlation degree and the overall average value is increased by 1.65 times.

Description

Technical Field

[0001] The present invention relates to a method for diagnosing crack depth of a wind turbine gear by combining a gear rigid-flexible coupling model with dual grayscale association, and relates to the field of gear crack fault diagnosis. Background Art

[0002] The study of gear crack faults has always been an important research direction that many scholars have paid attention to. The effective diagnosis of cracks can avoid many mechanical failures. Reference [Sun Canfei, Wang Youren, Xia Yubin. Helicopter planetary gear crack fault diagnosis based on SCAE-ACGAN [J]. Vibration. Testing and Diagnosis, 2021, 41(03): 495-502+620-621.] The diagnosis of helicopter planetary gear crack faults was studied. Reference [Ning Shaohui, Han Zhennan, Wu Xuefeng, Wang Zhijian. Gear crack fault diagnosis based on embedded sensors [J]. Vibration and Shock, 2018, 37(11): 42-47.] A new method for gear vibration signal detection was proposed, which shortened the path length of the vibration signal during transmission and effectively reduced the amplitude attenuation of the vibration signal during transmission. References [Gui Yong, Han Qinkai, Li Zheng, Chu Fulei. Gear crack fault diagnosis of wind turbine planetary gear system [J]. Vibration. Testing and Diagnosis, 2016, 36(01): 169-175+205.] disclose the establishment of a wind turbine planetary gear model based on the errors that occur during production and processing in various situations. References [Long Haiyang, Liu Chang, Li Yaogang, Yang Jue, Zhang Shuo, Hui Xuewen. Dynamic simulation of gearbox planetary gear system [J]. Science, Technology and Engineering, 2020, 20(26): 10934-10941.] and [Xiang Ling, Chen Tao. Research on contact force simulation of faulty planetary gear system [J]. Combined Machine Tool and Automated Machining Technology, 2015(02): 97-99+103.] disclose the research on planetary gear contact force based on Adams. Reference [Chen Xiangmin, Duan Meng, Li Qi, et al. Variable speed gear fault diagnosis based on ATF and ASAD [J]. Mechanical Transmission, 2021, 45(10): 144-150.] proposed a gear fault diagnosis method based on adaptive time-varying filtering and angular domain synchronous averaging noise reduction. Reference [Wang L, Shao Y, Cao Z. Optimal demodulation sub band selection for sun gear crack fault diagnosis in planetary gearbox [J]. Measurement, 2018: 554-563.] proposed a new optimal demodulation sub-band selection method for planetary gearbox fault diagnosis, which overcomes the problem of finding the sub-band containing the most fault-related modulation components. The degree of gear crack damage is also one of the key points of gear fault detection.Literature [Sun Qi, Liu Xinchang, Zhang Bing, et al. Detection method of tooth root crack damage degree in spur gear system [J]. Vibration. Testing and Diagnosis, 2019, 39(2):9.], [Liu Jie, Li Huanyu, Zhao Weiqiang. Identification of tooth root crack damage degree based on PCA and grey correlation [J]. Mechanical Transmission, 2020, 44(09):133-139+152.], [Liu XC, Sun Q, Chen C J. Damage Degree Detection of Cracks in a Locomotive Gear Transmission System [J]. Shock and Vibration, 2018, 2018(PT.11):1-14.] The threshold of spur gear crack damage degree was detected by combining principal component analysis (PCA) with grey correlation analysis (GRA). Reference [Zhang W, Tan Y, Pu YA New Gear Fault Identification MethodBased on EEMD Permutation Entropy and Grey Relation Degree[C]2020 13thInternational Congress on Image and Signal Processing,BioMedical Engineeringand Informatics(CISP-BMEI).2020.] proposed a new fault identification method based on EEMD permutation entropy and grey relation degree. Reference [Wu Shoujun, Feng Fuzhou, Wu Chunzhi, Jia Ziyong. A new feature for planetary gear crack depth assessment and its application[J]. Journal of Beijing Institute of Technology, 2022, 42(01):28-35.] proposed the ratio of the cumulative amount of fault frequency and multiple frequency amplitude to the meshing frequency amplitude (RCMFAM) as a new crack depth assessment feature. Reference [Liu Haohua, Li Fangyi, Li Guoyan, et al. Planetary gearbox gear crack damage location based on EEMD [J]. Vibration. Testing and Diagnosis, 2017, 37(3): 6.] proposed a gear local damage frequency demodulation analysis method based on ensemble average empirical mode decomposition (EEMD), which solved the problem of difficulty in extracting characteristic frequencies and locating gear crack damage in planetary gearboxes.

[0003] These methods have highlighted certain advantages in gear crack fault diagnosis, but in actual diagnosis, gear crack fault diagnosis requires a large amount of data support, and the implementation of algorithms such as neural networks often requires a complex network and complicated calculations. The present invention combines gear dynamics, three-dimensional modeling, finite element analysis and other methods to construct a rigid-flexible coupling model, simplifying the gear crack model, and using dynamic simulation analysis to obtain the dynamic time-frequency domain signal of the gear. The degree of gear fault is characterized by the change of the characteristic parameters of the signal, simplifying the complex calculation of crack faults based on deep learning and other methods. The present invention uses a rigid-flexible coupling model and a double grayscale association set to calculate the crack depth threshold of the gear. At the same time, by utilizing the relationship between signal parameters and crack faults, grayscale association is used to optimize the parameter sequence, thereby improving the concentration and accuracy of the grayscale association calculation results. Summary of the Invention

[0004] Gear crack fault diagnosis has always been one of the important research directions in the field of gear fault research. In view of the complexity of the wind turbine gear system, the gear system is studied independently, and the planetary gear system is independently modeled and studied. When based on simulation models or numerical simulations, it is often considered to simplify the gear cracks and ignore the influence of crack irregularities. In reality, gears are subject to factors such as impact and torsion, and are not completely rigid, so it is necessary to consider the occurrence of torsion and extrusion deformation. In response to this problem, the faulty gear is made flexible, and a planetary gear system model is constructed using a rigid-flexible coupling method. Through simulation, a large amount of gear-related data is obtained. In order to simplify the complexity of the data, based on the dynamic simulation signal, the signal state of different faults is characterized by extracting the time-frequency domain signal characteristic parameters, and the grayscale correlation is used to screen out the signal characteristic values ​​that have a greater impact on the gear fault. The present invention realizes the diagnosis of the gear crack depth threshold based on the combination of the rigid-flexible coupling model and the double grayscale correlation.

[0005] The technical solution adopted by the present invention is to construct a gear rigid-flexible coupling model, analyze and extract characteristic parameters of dynamic simulation results, and implement a gear crack depth threshold diagnosis method in combination with grayscale correlation analysis. First, a planetary gear system dynamic simulation model is constructed using a rigid-flexible coupling method to extract simulation signals in the time-frequency domain. Then, based on the simulation signal, multiple signal characteristic parameters are extracted as information factors expressing different crack depth values. A parameter matrix is ​​constructed and combined with grayscale correlation to implement the diagnosis of the gear crack depth threshold. Crack length parameters associated with characteristic sample parameters are introduced, and a correlation weight matrix is ​​constructed to screen characteristic parameters that have a greater impact on the results, thereby simplifying the calculation sequence and improving the algorithm accuracy.

[0006] Furthermore, the construction of the planetary gear train dynamics simulation model includes:

[0007] Use SolidWorks to build a gear crack failure model and complete the assembly of the planetary gear train. At the same time, test and verify whether the fit between the gears is reasonable and ensure that there is no interference.

[0008] Convert the rigid body into a flexible body in ANSYS, set the gear shaft hole as a rigid area, and generate the MNF file of the faulty gear;

[0009] Build a simulation model in Adams, import the MNF file into Adams, replace the original rigid sun gear component, and complete the addition and setting of corresponding constraints, meshing forces and related parameters.

[0010] Furthermore, the constraints described in the above scheme include: ① using the ground as a reference base, adding a revolving pair at the center of mass of the sun gear and the planetary carrier, and adding a fixed pair at the center of mass of the inner ring gear; ② setting the connection relationship between the planetary gear and the planetary carrier as a revolving pair; ③ adding contact collision between gears.

[0011] Furthermore, considering the extrusion deformation of gears under load, a faulty planetary gear system was constructed based on a rigid-flexible coupling approach, and dynamic simulation was used to identify the faulty gear states at different crack depths. Based on dynamic simulation, characteristic parameters of the time-frequency domain signals were extracted, and a parameter matrix was constructed as a standard data matrix. Simulation signals of faulty gears with the crack depth to be measured were then collected to form the test samples. Data screening and matrix dimensionality reduction were achieved by combining primary grayscale correlation, and gear crack depth threshold diagnosis was achieved by combining secondary grayscale correlation.

[0012] Based on the characteristic parameters of the time-frequency domain signal in the dynamic simulation of the faulty gear to express the characteristics of the gear crack depth, the intrinsic relationship between the characteristic parameters and different crack depths is combined with the grayscale correlation analysis method to perform depth solution, which simplifies the tedious and complicated calculations brought by a large amount of data.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned wind turbine planetary gear system crack fault diagnosis method based on the combination of dynamic simulation and dual grayscale correlation analysis (GRA).

[0014] Compared with the prior art, the present invention has the following advantages:

[0015] (1) A gear fault model with cracks was constructed based on rigid-flexible coupling for dynamic simulation, taking into account the effects of extrusion and torsion caused by load on the gear;

[0016] (2) Based on the time-frequency domain simulation signal of gear dynamics, the signal characteristic value is used to characterize the gear crack depth fault under different crack depths. A total of 11 different signal parameter characterization values ​​are extracted, and the parameter matrix is ​​constructed. The crack depth is solved using grayscale correlation;

[0017] (3) Based on the relationship between the signal characteristic value itself and the crack depth, correlation analysis is performed to screen and sequence optimize the extracted signal characteristic parameters, thereby reducing the length of the characteristic parameter sequence and optimizing the accuracy and precision of the grayscale correlation algorithm calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a rigid-flexible coupling model of a planetary gear train with cracks;

[0019] Figure 2 This is the time-frequency domain diagram of the dynamic simulation of the planetary gear train with cracks;

[0020] Figure 3 Modeling and algorithm execution flow chart;

[0021] Figure 4 Correlation between time-frequency domain characteristic parameters and crack depth;

[0022] Figure 5 Calculate the crack depth threshold map based on grayscale correlation analysis. DETAILED DESCRIPTION

[0023] (1) Planetary gear dynamic structure model

[0024] This study utilizes a combination of SolidWorks, ANSYS, and Adams software to establish a gear failure model for a 1.5MW wind turbine transmission system planetary gear train. SolidWorks was used to construct a gear crack failure model and complete the planetary gear train assembly. The gears were then tested to verify proper fit and interference-free operation. The file was exported in Parasolid format to ANSYS / Workbench. The planetary gear train structural parameters are shown in Table 1. In ANSYS / Workbench, the rigid body was converted to a flexible body, and rigid regions were set. Considering the transmission shaft as rigid, the gear shaft hole was designated as a rigid region. The APDL command, a parametric design language within ANSYS, was used to generate an MNF file for the faulty gear.

[0025] Table 1 Design parameters of planetary gear train

[0026]

[0027] In Adams, the constraint relationships between the gears in the model, such as rotating pairs and fixed pairs, are constructed. The MNF file of the faulty gear is imported into Adams, the original rigid sun gear component is replaced, and the corresponding constraints, meshing forces, and related parameters are added and set.

[0028] In order to simplify the model calculation and reduce the simulation time, the 3D model is simplified, such as ignoring the gasket, the outer friction plate of the gear ring, simplifying the planetary carrier structure and the planetary gear support frame, and ignoring the influence of the bearing connection. That is, the state of the bearing is considered to be ideal, and the bearing connection between the planetary carrier and the planetary gear is replaced by a revolute pair. Add relevant constraints between the gears, such as Figure 1 As shown:

[0029] ① Using the ground as a reference, add a revolute pair at the center of mass of the sun gear and planet carrier, and a fixed pair at the center of mass of the inner ring gear. ② Set the connection between the planet gear and the planet carrier to a revolute pair. ③ Add contact collisions between the gears.

[0030] The meshing and squeezing of the gears generates contact force. Considering that there is no elastic fluctuation when the two gears mesh, and no gap exists between the kinematic pairs, the impact function based on Hertz collision theory is used for the present study.

[0031] The relationship between the normal contact force on the tooth surface and the tooth surface deformation δ is obtained from the Hertz collision theory:

[0032]

[0033] Where F is the normal contact force, R is the comprehensive curvature radius, and E * is the comprehensive elastic modulus.

[0034] The solution for R is as follows:

[0035]

[0036] Where r1 and r2 are the equivalent curvature radii of the two meshing gears, respectively, and “±” represents the meshing form between the gears, which is internal and external meshing.

[0037] The equivalent curvature radius r of the gear is:

[0038] r=r b *sin(α) (3)

[0039] Where r b is the gear pitch circle radius, α is the gear pressure angle;

[0040] E * The solution is as follows:

[0041]

[0042] Where E1, E2, ν1, ν2 are the elastic modulus and Poisson's ratio of the two meshing gears; the contact stiffness K is:

[0043]

[0044] Set the gear material in Adams to steel, and its material properties are: ① Poisson's ratio ν = 0.29, ② Elastic modulus 2.07E+05N / mm 2 The stiffness coefficients between the sun gear and the planet gear and between the planet gear and the inner gear ring are 1.1447E+06N / mm respectively. 2 、2.3828e+06N / mm.

[0045] (2) Damage detection index extraction

[0046] In order to make the data sample reflect the signal characteristics more completely, the present invention selected a total of 11 different time-frequency domain eigenvalues. Through simulation, the angular velocity signal of the sun gear of the planetary gear system was obtained and the signal characteristics were extracted as follows:

[0047] Variance V ar for:

[0048]

[0049] Where x(i) is the time domain signal sequence, is the mean, N s is the number of time domain signal samples, and the following equations (9)-(14) are the same;

[0050] Standard deviation S td for:

[0051]

[0052] Effective value R ms for:

[0053]

[0054] Kurtosis K s for:

[0055]

[0056] Peak Factor P ef for:

[0057]

[0058] Pulse factor P uf for:

[0059]

[0060] Margin factor M rf for:

[0061]

[0062] Use FFT transform to process the simulated time domain signal, obtain its corresponding spectrum diagram, and extract signal features as follows:

[0063] The center frequency FC is:

[0064]

[0065] Where f(n) is the frequency domain signal sequence, represents the mean value of the frequency domain signal, N is the number of frequency domain signal samples, and the following equations (16)-(18) are the same;

[0066] Mean frequency MF:

[0067]

[0068] RMS frequency F RMSF for:

[0069]

[0070] Center of gravity frequency F RMSF for:

[0071]

[0072] (3) Improved grayscale correlation algorithm and crack depth solution

[0073] Traditional grayscale correlation analysis simply relies on the degree of correlation between sequences to solve the problem, but does not consider the influence of elements in the sequence. This results in insignificant differences in the calculated results and inaccurate numerical ranges. To address the problem of insufficient optimization of sequence feature parameters and inaccurate results of a single grayscale correlation analysis, this paper proposes to introduce a parameter to optimize the original sequence parameters to achieve an improved algorithm. The specific process is as follows:

[0074] Let the comparison sequence be X i ={x i (k)|k=1,2,…,n},i=1,2,…,m, the reference sequence is X0={x0(k)|k=1,2,…,n}, at this time, a parameter related to the standard sequence, i.e., the subsequence, is introduced. The parameter introduced in the present invention is the crack length l i , construct the standard sequence element x i (k) is the correlation function with the length parameter. That is:

[0075] w ij =Gray(L ij |X ij (k)) (19)

[0076] Where, L ij is the length parameter matrix, X ij (k) is the sequence element matrix, w ij is the correlation function value between each element in the sequence and the crack length.

[0077] The above function can be used to solve the correlation between the influence of each element on the length. ij The size of the elements that have a greater impact on the crack length is obtained, namely:

[0078]

[0079] At this time, the comparison sequence is X i ={x i (t), t∈[1,p];p<n}, i∈[1,m], the reference sequence is X0={x0(t), t∈[1,p];p<n}, t is the number of the element in the reduced sequence, p is the length of the reduced sequence, n is the length of the original sequence, m is the number of matrix rows, x i (t) is the number of elements in the i-th comparison sequence after reduction, x0(t) is the number of elements in the reference sequence after reduction, and at the same time, let L(x) = X i T ,i∈[0,m], then the correlation coefficient is solved as:

[0080]

[0081] Where, L su (x) is the absolute difference of each point in the transposed sequence, ρ∈[0,1] is the resolution coefficient, which is 0.5 in the present invention. 0k (x) is the absolute difference between the element of the standard sequence and the element of the standard sequence in the k dimension.

[0082] Grey relational degree r i for:

[0083]

[0084] According to the change of crack depth, the damaged sun gear is modeled m times in three dimensions, and dynamic simulation analysis is performed to obtain the dynamic time domain characteristics of the sun gear, such as angular acceleration and angular velocity. At the same time, the corresponding frequency domain characteristic signal is solved by using FFT changes. The simulation results are as follows: Figure 2, extract a total of n time-frequency domain feature parameters, take the crack depth as the mother sequence, and form a matrix of m rows and n+1 columns with different states and their corresponding fault features, as shown below:

[0085]

[0086] Where l1, l2, …, l m Represent the crack depths, a1, a2,…, a n They represent the time-frequency domain characteristic parameters of the corresponding signals at different depths.

[0087] Grayscale correlation analysis is used to filter the feature data and obtain i characteristic parameters with a high correlation with gear crack changes, i.e., data dimensionality reduction. The characteristic parameters of the filtered signal to be detected are added to the standard sequence feature values ​​to form a matrix with i rows and m + 1 columns, as shown in Equation (22).

[0088]

[0089] Where a 01 ,a 02 ,…a 0i are the time-frequency domain characteristic parameters of the sequence to be detected.

[0090] The rows and columns of the original matrix subsequence are transformed, and the length of the standard signal characteristic parameter sequence is changed from n to i (i<n). The characteristic parameters of the signal to be detected are used as the parent sequence, and the grey correlation algorithm is used again to calculate the correlation between the characteristic parameter set of the optimized standard signal and the characteristic parameters of the signal to be detected. The crack degree with the maximum value in the correlation is the approximate value of the damage degree of the faulty gear to be detected. The detection process is as follows: Figure 3 The results of the correlation between each characteristic parameter and crack depth are as follows: Figure 4 .

[0091] like Figure 5 , it can be seen that when directly calculating the gray correlation between the eigenvalue and the crack depth, the differentiation between the various crack degrees is not large, and most are above 0.9. After performing a gray correlation optimization to reduce the data dimension, it can be seen that the overall growth trend is more obvious. The correlation is the largest at 12mm, with a gray correlation value of 0.9438. It also reduces the correlation values ​​of the remaining crack depths and the target to be inspected, all below 0.85. After using GRA to reduce the data dimension, the difference between the maximum gray correlation value and the overall gray correlation value increased by approximately 1.65 times.

Claims

1. A wind turbine planetary gear train crack fault diagnosis method based on the combination of dynamic simulation and dual GRA is characterized by: The following steps are involved: Firstly, a planetary gear train dynamics simulation model is constructed using a rigid-flexible coupling approach to extract simulation signals in the time-frequency domain. Then, based on the simulation signal, multiple signal characteristic parameters are extracted as information factors expressing different crack depth values. A parameter matrix is ​​constructed and combined with grayscale correlation to realize the diagnosis of gear crack depth threshold. The crack length parameter associated with the characteristic sample parameter is introduced to construct a correlation weight matrix. The specific process of introducing crack length parameters is as follows: Let the comparison sequence be X i ={x i (k)|k=1,2,…,n},i=1,2,…,m, the reference sequence is X0={x0(k)|k=1,2,…,n}, at this time, the introduced parameter is the crack length l i , construct the standard sequence element x i (k) is related to the length parameter, namely: w ij =Gray(L ij |X ij (k)) Where, L ij | is the length parameter matrix, X ij (k) is the sequence element matrix, w ij Indicates the correlation function value between each element in the sequence and the crack length; The above function can be used to solve the correlation between the influence of each element on the length. ij The size of the elements that have a greater impact on the crack length is obtained, namely: At this time, the comparison sequence is X i ={x i (t), t∈[1,p];p<n}, i∈[1,m], the reference sequence is X0={x0(t), t∈[1,p];p<n}, t is the number of the element in the reduced sequence, p is the length of the reduced sequence, n is the length of the original sequence, m is the number of matrix rows, x i (t) is the number of elements in the i-th comparison sequence after reduction, x0(t) is the reference sequence element after reduction, and at the same time, let L(x) = X i Τ ,i∈[0,m], then the correlation coefficient is solved as: Where, L su (x) is the absolute difference of each point in the transposed sequence, L 0k (x) is the absolute difference between the standard sequence element in the k dimension, ρ∈[0,1] is the resolution coefficient; Grey relational degree r i for: According to the change of crack depth, m three-dimensional modeling is performed and dynamic simulation analysis is carried out to obtain the dynamic time domain characteristics of the sun gear. At the same time, the corresponding frequency domain characteristic signal is solved by using FFT changes, and a total of n time-frequency domain characteristic parameters are extracted. The crack depth is used as the parent sequence, and the different states and their corresponding fault characteristics are combined into a matrix with m rows and n + 1 columns, as shown below: Where l1, l2, …, l m Represent the crack depths, a1, a2,…, a n Represent the time-frequency domain characteristic parameters of the corresponding signals at different depths; Grayscale correlation analysis is used to screen the characteristic data and obtain i characteristic parameters with a high correlation with the gear crack change, that is, data dimensionality reduction. The characteristic parameters of the signal to be detected after screening are added to form a matrix with i rows and m + 1 columns with the characteristic quantities of the standard sequence, as shown in the following formula: Where a 01 ,a 02 ,…a 0i is the time-frequency domain characteristic parameter of the sequence to be detected; The rows and columns of the original matrix subsequence are transformed, and the length of the standard signal characteristic parameter sequence is changed from n to i. The characteristic parameters of the signal to be detected are used as the parent sequence, and the grey correlation algorithm is used again to calculate the correlation between the characteristic parameter set of the optimized standard signal and the characteristic parameters of the signal to be detected. The crack degree with the maximum value in the correlation is the approximate value of the damage degree of the faulty gear to be detected.

2. The wind turbine planetary gear train crack fault diagnosis method based on the combination of dynamic simulation and dual GRA according to claim 1 is characterized in that: The construction of the planetary gear train dynamics simulation model includes: Use SolidWorks to build a gear crack failure model and complete the assembly of the planetary gear train. At the same time, test and verify whether the fit between the gears is reasonable and ensure that there is no interference. Convert the rigid body into a flexible body in ANSYS, set the gear shaft hole as a rigid area, and generate the MNF file of the faulty gear; Build a simulation model in Adams, import the MNF file into Adams, replace the original rigid sun gear component, and complete the addition and setting of corresponding constraints, meshing forces and related parameters.

3. The wind turbine planetary gear train crack fault diagnosis method based on the combination of dynamic simulation and dual GRA according to claim 2 is characterized in that: The constraints include: ① using the ground as a reference base, adding a revolving pair at the mass center of the sun gear and the planet carrier, and adding a fixed pair at the mass center of the inner gear ring; ② Set the connection between the planet gear and the planet carrier as a revolute pair; ③Add contact collision between gears.

4. The wind turbine planetary gear train crack fault diagnosis method based on the combination of dynamic simulation and dual GRA according to any one of claims 1 to 3, characterized in that: Based on dynamic simulation, the characteristic parameters of time-frequency domain signals are extracted and a parameter matrix is ​​constructed as the standard data matrix. Then, the simulation signal of the gear with crack depth fault to be tested is collected to form the test sample. The data screening and matrix dimensionality reduction are realized by combining the primary grayscale correlation, and the gear crack depth threshold diagnosis is realized by combining the secondary grayscale correlation.

5. The wind turbine planetary gear train crack fault diagnosis method based on the combination of dynamic simulation and dual GRA according to claim 4 is characterized in that: The extraction of the time-frequency domain simulation signal includes obtaining the angular velocity signal of the sun gear of the planetary gear train through simulation, and extracting the signal characteristics including: variance V ar , standard deviation S td , effective value R ms 、 Kurtosis K s , peak factor P ef , pulse factor P uf and margin factor M rf ; The FFT transform is used to process the simulated time domain signal to obtain its corresponding spectrum diagram, and the signal features extracted include: center frequency FC, average frequency MF, root mean square frequency F RMSF and the center of gravity frequency F RMSF .

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind turbine planetary gear train crack fault diagnosis method based on the combination of dynamic simulation and dual GRA according to any one of claims 1 to 5 are implemented.

Images