Wind turbine gearbox signal sparse decomposition method, device and storage medium

Through the signal sparse decomposition method based on blind source separation, the lack of noise suppression and characteristic characterization in the signal processing of gearbox of wind turbines is solved, and the sparse decomposition of fault characteristics of complex signals is realized, which improves the adaptability and generalization ability of signal processing , enhances the accuracy and efficiency of fault feature extraction.

CN118568598BActive Publication Date: 2024-11-22ANHUI UNIVERSITY OF ARCHITECTURE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411037656.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-22
Estimated Expiration
2044-07-31

AI Technical Summary

Technical Problem

The existing signal processing technology is greatly affected by noise in the monitoring of gearbox status of wind turbines, complex model construction, poor adaptability to actual working conditions, difficult to effectively remove noise and interference, and reduce fault identification capabilities.

Method used

Using the sparse decomposition method of wind turbine gearbox signal based on blind source separation, the separation matrix, complex and fast independent component analysis, complex local average decomposition and virtual channels are used to separate and denoise the sparse characteristic parameters by using signal amplitude characteristic parameters. , construct a sparse decomposition structure to improve the adaptability and generalization ability of signal processing.

Benefits of technology

Effectively remove noise interference, enhance fault identification capabilities, improve the accuracy and efficiency of extracting fault characteristics of wind turbine gearbox signal, and ensure the efficient implementation of status monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118568598B_ABST
    Figure CN118568598B_ABST
Patent Text Reader

Abstract

The present invention discloses a sparse decomposition method, device and storage medium for a wind turbine gearbox signal, comprising constructing a sparse decomposition structure of a wind turbine gearbox signal, constructing a separation matrix when the aliasing of the original signal of the gearbox is unknown, and separating the sparse feature parameters by the signal amplitude feature parameters; optimizing the construction of the separation matrix by maximizing the objective function, and improving the performance of the sparse decomposition; taking the signal amplitude and time coding as input, performing complex local average decomposition, and constructing a virtual channel, solving the problem of underdetermined blind source separation, enhancing the processing capability of the sparse decomposition for complex signals, and obtaining an output in which the original signal corresponds to the sparse feature parameters one by one; and using historical measured data to train the sparse decomposition structure, so that the trained model can input a specific gearbox signal and obtain the corresponding sparse feature parameters. The present invention effectively improves the effect of noise suppression and characteristic characterization of the wind turbine gearbox signal.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of feature extraction, and in particular to a method, device and storage medium for sparse decomposition of a wind turbine gearbox signal based on blind source separation. Background Art

[0002] The gearbox of a wind turbine generator is one of the key components in a wind power generation system. Due to its harsh working environment and frequent load changes, the gearbox is prone to failure. The condition monitoring of the gearbox of a wind turbine generator is beneficial to ensure the efficiency and stability of the wind turbine generator and avoid high maintenance costs and downtime. Signal processing is one of the key technologies for gearbox condition monitoring. Vibration signals usually contain a lot of noise and interference, and the signal characteristics are relatively complex. It is very challenging to directly analyze and extract fault characteristics. Existing signal processing technology mainly processes non-stationary vibration signals through time domain and frequency domain analysis of signals, provides rich time-frequency characteristics, or establishes a physical model or mathematical model of the gearbox, analyzes the difference between the actual signal and the model signal, and provides an in-depth understanding of the fault mechanism. However, there are problems such as being greatly affected by noise, complex model construction, poor adaptability to actual working conditions, and difficulty in parameter acquisition, which limit the signal processing effect. Therefore, a signal processing method is needed to effectively remove the noise and interference of the gearbox signal, enhance the fault identification capability, and ensure the efficient implementation of gearbox condition monitoring. Summary of the invention

[0003] The present invention proposes a method, device and storage medium for sparse decomposition of wind turbine gearbox signals based on blind source separation, which can solve at least one of the technical problems in the background technology.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] A sparse decomposition method for wind turbine gearbox signal based on blind source separation, which performs the following steps through a computer device:

[0006] Constructing a sparse decomposition structure of a wind turbine gearbox signal, wherein the sparse decomposition structure includes a separation matrix, a complex fast independent component analysis, a complex local average decomposition, and a virtual channel;

[0007] The amplitude characteristic parameters of the wind turbine gearbox under different working conditions are measured, and the input and output vector dimensions of the sparse characteristic parameters are determined based on the measured data;

[0008] The amplitude of the wind turbine gearbox signal and the corresponding time code are used as the input of complex local average decomposition to construct a virtual channel. The characteristic parameters of the virtual channel are used as the input of complex fast independent component analysis to construct an iterative separation rule. The separation matrix is ​​further constructed according to the iterative separation rule to obtain an output with a one-to-one correspondence between the original signal and the sparse characteristic parameters.

[0009] The sparse decomposition structure is trained using historical measured data. After the training, the model can input the specified original signal of the wind turbine gearbox and obtain the corresponding sparse feature parameters.

[0010] Furthermore, the sparse decomposition structure of the wind turbine gearbox signal, and the construction of its separation matrix include the following steps: using a blind source separation algorithm, when the aliasing of the original signal of the wind turbine gearbox is not detected, relying on the signal amplitude characteristic parameters, the sparse characteristic parameters are separated from the original signal, and the processing method is as follows:

[0011]

[0012]

[0013] in y =[ y 1( t ), y 2( t ),…, y n ( t )] T It is obtained after separation n dimensional sparse feature parameters, W is the constructed separation matrix, x =[ x 1( t ), x 2( t ),…, x m ( t )] T It is collected m dimensional original signal, A is the mixing matrix, s =[ s 1( t ), s 2( t ),…, s n ( t )] T yes n dimensional independent unknown source signal.

[0014] Furthermore, the separation matrix W The construction also includes a complex fast independent component analysis estimation step. The specific expression of maximizing the objective function during the estimation process is as follows:

[0015]

[0016] inE [.] represents the mean, G (.) represents any non-quadratic function, w =[ w 1( t ), w 2( t ),…, w m ( t )] T yes m dimensional vector, representing the complex separation matrix W In a row, Represents the centralization and whitening x , the method is as follows:

[0017]

[0018]

[0019]

[0020] in After centralization t samples of time steps, L is the length of the observed signal used for analysis, V is the whitening matrix, is the covariance matrix The eigenvalue diagonal matrix of is the matrix consisting of the corresponding eigenvectors.

[0021] Furthermore, it is also assumed that the amplitude of the source signal is 1, that is, the amplitude of the separated signal is also 1, and the Lagrangian relaxation method is used to reconstruct the objective function as follows:

[0022]

[0023] in is the Lagrange multiplier; the Newton iteration method maximizes the reconstruction of the objective function w The expression is as follows:

[0024] in r Indicates r During the iteration w , G ’ (.)and G ’’ (.) are G The first and second derivatives of (.); Normalization should be performed in the next iteration to ensure the uniqueness of each new row component. The normalization formula is as follows:

[0025]

[0026] Furthermore, a virtual channel is constructed based on complex local average decomposition to solve the underdetermined blind source separation problem when the number of input signals is less than the number of unknown source signals. The amplitude of the wind turbine gearbox signal and the corresponding time encoding The components in each projection direction are respectively projected to the directions of 0 and π / 2, and the components in each projection direction are processed respectively to obtain the local mean function and the local envelope function, and the frequency modulation signal and the envelope signal of the real and imaginary parts. The envelope signal and the pure normalized frequency modulation signal are multiplied to obtain the product function component, which specifically includes the following steps:

[0027] S:41: Projected to the directions of 0 and π / 2 respectively:

[0028]

[0029]

[0030] where Re[·] represents the real part of the complex value;

[0031] S42: Separate deduction and Local mean function , and the local envelope function , , construct the wind turbine gearbox signal amplitude and corresponding time encoding The complex local mean function as follows:

[0032]

[0033] in N is the number of product functions, M is the number of iterations in the decomposition;

[0034] S43: Calculate the local envelope demodulation function as follows:

[0035]

[0036] S44: Calculate frequency modulated rotation signal and as follows:

[0037]

[0038]

[0039] Repeat S42 to S44, and and Bring in and , until and Equal to one or and is a pure normalized frequency modulated signal;

[0040] S45: Calculate the product of all local envelope functions generated during the iteration process and obtain and The envelope signal is as follows:

[0041]

[0042]

[0043] in l is the maximum number of iterations;

[0044] S46: Seek No. N Substituting the components of the complex product function, it is calculated as follows:

[0045]

[0046] S47: Calculate the residual signal :

[0047]

[0048] check Does it contain any rotation mode? If so, As new , and repeat steps S41 to S47; otherwise, terminate the process.

[0049] Furthermore, the method further includes dividing the wind turbine gearbox signal amplitude obtained by historical actual measurement and the corresponding time-coded data into a training set and a test set at a ratio of 4:1, using the training set to train the fault identification of the original data and the decomposed complex product function component, and using the test set to obtain a sparse decomposition structure of the wind turbine gearbox signal with excellent final identification effect, wherein the loss function used in the training process is:

[0050]

[0051] In the formula and Respectively represent i The true output category of the training samples and the identification category of the complex product function component, Indicates error.

[0052] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0053] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0054] It can be seen from the above technical scheme that the sparse decomposition method of the wind turbine gearbox signal of the present invention includes constructing a sparse decomposition structure of the wind turbine gearbox signal, and the structure includes a separation matrix, a complex fast independent component analysis, a complex local average decomposition and a virtual channel construction. Based on the actual collected historical wind turbine gearbox signal, when the aliasing of the original signal of the gearbox is unknown, a separation matrix is ​​constructed, and the sparse feature parameters are separated by the signal amplitude feature parameters; by maximizing the objective function, an independent component analysis is performed on the signal after centralization and whitening, the construction of the separation matrix is ​​optimized, and the performance of sparse decomposition is improved; the signal amplitude and time coding are used as input, and complex local average decomposition is performed to construct a virtual channel, which solves the problem of underdetermined blind source separation, enhances the processing ability of sparse decomposition for complex signals, and obtains an output with a one-to-one correspondence between the original signal and the sparse feature parameters. The present invention utilizes the sparsity of the wind turbine gearbox signal, and directly builds a sparse decomposition model based on historical measured vibration data as a training sample, which can effectively extract the sparse feature parameters corresponding to the original signal of a specific gearbox, and realize signal denoising and fault characteristic characterization.

[0055] In general, the wind turbine gearbox signal sparse decomposition method, device and storage medium based on blind source separation of the present invention overcome the shortcomings of existing signal processing methods in noise suppression and characteristic characterization, realize the sparse decomposition of fault characteristics of complex signals, and at the same time improve the adaptability and generalization ability of signal processing.

[0056] The solution proposed in the present invention has strong noise interference suppression, can extract accurate features from fault feature decomposition, effectively improves the performance of fault feature extraction of wind turbine gearbox signal, and ensures the implementation of wind turbine status monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is an overall block diagram of a sparse decomposition method for a wind turbine gearbox signal according to an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the blind source separation principle of an embodiment of the present invention;

[0059] Figure 3 It is a schematic diagram of sparse feature component extraction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0061] like Figure 1 As shown, a method for sparse decomposition of a wind turbine gearbox signal based on blind source separation described in an embodiment of the present invention first includes constructing a sparse decomposition structure of the wind turbine gearbox signal, constructing a separation matrix when the aliasing of the original signal of the gearbox is unknown, and separating the sparse feature parameters through the signal amplitude feature parameters; by maximizing the objective function, performing independent component analysis on the signal after centering and whitening processing, optimizing the construction of the separation matrix, and improving the performance of sparse decomposition; taking the signal amplitude and time coding as input, performing complex local average decomposition, constructing a virtual channel, solving the problem of underdetermined blind source separation, enhancing the processing capability of sparse decomposition for complex signals, and obtaining an output with a one-to-one correspondence between the original signal and the sparse feature parameters; using historical measured data to train the sparse decomposition structure, the model after training can realize input of a specific gearbox signal to obtain the corresponding sparse feature parameters.

[0062] The specific steps are as follows:

[0063] S1: Using the blind source separation algorithm, when the aliasing of the original signal of the wind turbine gearbox is not detected, the sparse feature parameters are separated from the original signal by relying on the signal amplitude feature parameters. The processing method is as follows:

[0064]

[0065]

[0066] in y =[ y 1( t ), y 2( t ),…, y n ( t )] T It is obtained after separation n dimensional sparse feature parameters, W is the constructed separation matrix, x =[ x 1( t ), x 2(t ),…, x m ( t )] T It is collected m dimensional original signal, A is the mixing matrix, s =[ s 1( t ), s 2( t ),…, s n ( t )] T yes n dimensional independent unknown source signal.

[0067] S2: Separation Matrix W The construction also includes a complex fast independent component analysis estimation step. The specific expression of maximizing the objective function during the estimation process is as follows:

[0068]

[0069] in E [.] represents the mean, G (.) represents any non-quadratic function, w =[ w 1( t ), w 2( t ),…, w m ( t )] T yes m dimensional vector, representing the complex separation matrix W In a row, Represents the centralization and whitening x , the method is as follows:

[0070]

[0071]

[0072]

[0073] in After centralization t The samples of time steps, L is the length of the observed signal used for analysis, V is the whitening matrix, is the covariance matrix The eigenvalue diagonal matrix of is the matrix consisting of the corresponding eigenvectors.

[0074] S3: It also includes assuming that the amplitude of the source signal is 1, that is, the amplitude of the separated signal is also 1, and the Lagrangian relaxation method is used to reconstruct the objective function as follows:

[0075]

[0076] in is the Lagrange multiplier; the Newton iteration method maximizes the reconstruction of the objective function w The expression is as follows:

[0077] in r Indicates r During the iteration w , G ’ (.)and G ’’ (.) are G The first and second derivatives of (.); Normalization should be performed in the next iteration to ensure the uniqueness of each new row component. The normalization formula is as follows:

[0078]

[0079] S4: Constructing virtual channels based on complex local average decomposition to solve the underdetermined blind source separation problem where the number of input signals is less than the number of unknown source signals, the amplitude of the wind turbine gearbox signal and the corresponding time encoding The components in each projection direction are respectively projected to the directions of 0 and π / 2, and the components in each projection direction are processed respectively to obtain the local mean function and the local envelope function, and the frequency modulation signal and the envelope signal of the real and imaginary parts. The envelope signal and the pure normalized frequency modulation signal are multiplied to obtain the product function component, which specifically includes the following steps:

[0080] S:41: Projected to the directions of 0 and π / 2 respectively:

[0081]

[0082]

[0083] where Re[·] represents the real part of the complex value;

[0084] S42: Separate deduction and Local mean function , and the local envelope function , , construct the wind turbine gearbox signal amplitude and corresponding time encoding The complex local mean function as follows:

[0085]

[0086] in N is the number of product functions, M is the number of iterations in the decomposition;

[0087] S43: Calculate the local envelope demodulation function as follows:

[0088]

[0089] S44: Calculate frequency modulated rotation signal and as follows:

[0090]

[0091]

[0092] Repeat S42 to S44, and and Bring in and , until and Equal to one or and is a pure normalized frequency modulated signal;

[0093] S45: Calculate the product of all local envelope functions generated during the iteration process and obtain and The envelope signal is as follows:

[0094]

[0095]

[0096] in l is the maximum number of iterations;

[0097] S46: Seek No. N Substituting the components of the complex product function, it is calculated as follows:

[0098]

[0099] S47: Calculate the residual signal :

[0100]

[0101] check Does it contain any rotation mode? If so, As new , and repeat steps S41 to S47; otherwise, terminate the process.

[0102] S5: It also includes dividing the wind turbine gearbox signal amplitude obtained by historical actual measurement and the corresponding time-coded data into a training set and a test set according to a ratio of 4:1, using the training set to train the fault identification of the original data and the decomposed complex product function component, and using the test set to obtain a sparse decomposition structure of the wind turbine gearbox signal with excellent final identification effect, wherein the loss function used in the training process is:

[0103]

[0104] In the formula and Respectively represent i The true output category of the training samples and the identification category of the complex product function component, Indicates error.

[0105] S6: Input the wind turbine gearbox signal at a certain moment into the trained sparse decomposition model, and the output obtained is the corresponding sparse feature parameter.

[0106] The following are examples:

[0107] like Figure 1 As shown, the present embodiment describes a method for sparse decomposition of a wind turbine gearbox signal based on blind source separation, including constructing a sparse decomposition structure of the wind turbine gearbox signal; optimizing the sparse structure feature extraction performance; and constructing a sparse decomposition model of the original signal of the wind turbine gearbox.

[0108] The sparse decomposition structure for constructing the wind turbine gearbox signal includes separation matrix, complex fast independent component analysis, complex local average decomposition and virtual channel.

[0109] The optimized sparse structure feature extraction performance includes maximizing the objective function, performing independent component analysis on the signal after centering and whitening processing; taking the signal amplitude and time coding as input, performing complex local average decomposition, constructing a virtual channel, solving the underdetermined blind source separation problem, enhancing the processing ability of sparse decomposition for complex signals, and obtaining an output with a one-to-one correspondence between the original signal and the sparse feature parameters.

[0110] The construction of the sparse decomposition model of the original signal of the wind turbine gearbox includes measuring the amplitude characteristic parameters of the wind turbine gearbox under different working conditions, determining the input and output vector dimensions of the sparse characteristic parameters according to the measured data; using the amplitude of the wind turbine gearbox signal and the corresponding time code as the input of the complex local average decomposition, constructing a virtual channel, using the virtual channel characteristic parameters as the input of the complex fast independent component analysis, constructing an iterative separation rule, and further constructing a separation matrix according to the iterative separation rule to obtain an output in which the original signal corresponds to the sparse characteristic parameters one by one; using the historical measured data to train the sparse decomposition structure, and the model after the training is completed can realize the input of the specified wind turbine gearbox original signal to obtain the corresponding sparse characteristic parameters.

[0111] Figure 2 This is the principle diagram of blind source separation. m Wind turbine gearbox raw signal x Depend on n Dimensionally independent unknown source signal s and the mixing matrix A Composition, relying on signal amplitude characteristic parameters and separation matrix W , separated from the original signal n dimensional sparse feature parameters y .

[0112] Figure 3 This is a schematic diagram of sparse feature component extraction. The training set data in the original signal of the wind turbine gearbox is used to construct a separation matrix that optimizes the sparse feature extraction performance, including constructing a virtual channel using the complex local average decomposition algorithm. The complex product function components are obtained after centering and whitening to obtain a separation matrix with excellent sparse feature extraction performance. The test set data is subjected to sparse feature decomposition to obtain sparse feature parameters that can effectively characterize the state of the original signal.

[0113] This embodiment uses the invented method to sparsely decompose the vibration signal of a certain wind turbine gearbox. The signal comes from a planetary gearbox fault simulation system driven by a motor. The system consists of a motor, a parallel shaft gearbox, a planetary gear, a low-speed bearing, a high-speed bearing and a magnetic brake. It can simulate several different gearbox fault conditions. In this embodiment, four working conditions (normal, sun gear fault, planetary gear fault, and ring gear fault) are simulated. The sampling frequency is 24kHz. 200 groups of data are collected for each type, of which 160 groups are used as training data and 40 groups are used as test data. Data collection is performed on four known types of signals to obtain sparse parameters of fault characteristics corresponding to the original signal, and compared with the original vibration signal, short-time Fourier transform and physical modeling analysis method. This example uses a fixed parameter support vector machine as a gearbox fault classifier to evaluate the diagnostic accuracy of four different signal processing methods. The results are shown in Table 1. From the results in Table 1, it can be seen that the sparse decomposition method proposed in the present invention has a 100% diagnostic accuracy for gearbox signals in both normal and sun gear fault states, and a 98.75% combined diagnostic accuracy for the four signal states. All diagnostic results are superior to other signal processing methods.

[0114] Table 1 Diagnostic results of different signal processing methods

[0115] .

[0116] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0117] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0118] In another embodiment provided by the present application, a computer program product including instructions is also provided. When the computer program product is run on a computer, the computer executes any of the methods for modeling wireless channels in concrete in the above embodiments.

[0119] It is understandable that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above method.

[0120] The embodiment of the present application also provides an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus.

[0121] Memory, used to store computer programs;

[0122] The processor is used to implement the above-mentioned wireless channel modeling method in concrete when executing the program stored in the memory.

[0123] The communication bus mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus may be divided into an address bus, a data bus, a control bus, etc.

[0124] The communication interface is used for communication between the above electronic device and other devices.

[0125] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0126] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0127] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0128] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0129] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0130] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sparse decomposition method for wind turbine gearbox signal, characterized in that: The following steps are included: Constructing a sparse decomposition structure of a wind turbine gearbox signal, wherein the sparse decomposition structure includes a separation matrix, a complex fast independent component analysis, a complex local average decomposition, and constructing a virtual channel; The amplitude characteristic parameters of the wind turbine gearbox under different working conditions are measured, and the input and output vector dimensions of the sparse characteristic parameters are determined based on the measured data; The amplitude of the wind turbine gearbox signal and the corresponding time code are used as the input of complex local average decomposition to construct a virtual channel. The characteristic parameters of the virtual channel are used as the input of complex fast independent component analysis to construct an iterative separation rule. The separation matrix is ​​further constructed according to the iterative separation rule to obtain an output with a one-to-one correspondence between the original signal and the sparse characteristic parameters. The sparse decomposition structure is trained using historical measured data. The trained model is used to input the given original signal of the wind turbine gearbox to obtain the corresponding sparse feature parameters. Construct a sparse decomposition structure of the wind turbine gearbox signal, including, The blind source separation algorithm is used to separate the sparse characteristic parameters from the original signal based on the signal amplitude characteristic parameters when the aliasing of the original signal of the wind turbine gearbox is not detected. The expression is: in It is obtained after separation dimensional sparse feature parameters, is the constructed separation matrix, It is collected dimensional original signal, is the mixing matrix, yes Dimensionally independent unknown source signal; Constructing the sparse decomposition structure of the wind turbine gearbox signal also includes complex fast independent component analysis to estimate the separation matrix , the specific expression of maximizing the objective function in the estimation process is as follows: in represents the mean, represents any non-quadratic function, yes dimensional vector, representing the complex separation matrix In a row, Represents the centralization and whitening , wherein the steps include: in After centralization samples of time steps, is the length of the observed signal used for analysis, is the whitening matrix, is the covariance matrix The eigenvalue diagonal matrix of is the matrix consisting of the corresponding eigenvectors; The sparse decomposition structure of the wind turbine gearbox signal is constructed by assuming that the amplitude of the source signal is 1, that is, the amplitude of the separated signal is also 1. The Lagrangian relaxation method is used to reconstruct the objective function as follows: in is the Lagrange multiplier; the Newton iteration method maximizes the reconstruction of the objective function The expression is as follows: in Indicates During the iteration , and They are The first and second derivatives of ; Normalization should be performed in the next iteration to ensure the uniqueness of each new row component. The normalization formula is as follows: ; Complex local average decomposition and construction of virtual channels specifically include using complex local average decomposition to construct virtual channels to solve the underdetermined blind source separation problem where the number of input signals is less than the number of unknown source signals, the amplitude of the wind turbine gearbox signal and the corresponding time coding The components in each projection direction are respectively projected to the directions of 0 and π / 2, and the components in each projection direction are processed to obtain the local mean function and the local envelope function, and the frequency modulation signal and envelope signal of the real part and the imaginary part are obtained. The envelope signal and the pure normalized frequency modulation signal are multiplied to obtain the product function components, including: S21: Projected to the directions of 0 and π / 2 respectively: in represents the real part of a complex value; S22: Separate deduction and Local mean function , and the local envelope function , , construct the wind turbine gearbox signal amplitude and corresponding time encoding The complex local mean function as follows: in is the number of product functions, is the number of iterations in the decomposition; S23: Calculate the local envelope demodulation function as follows: S24: Calculate frequency modulated rotation signal and as follows: Repeat S22 to S24, and and Bring in and , until and Equal to one or and is a pure normalized frequency modulated signal; S25: Calculate the product of all local envelope functions generated during the iteration process and obtain and The envelope signal is as follows: in is the maximum number of iterations; S26: Seek No. Substituting the components of the complex product function, it is calculated as follows: S27: Calculate the residual signal : check Does it contain any rotation mode? If so, As new , and repeat steps S21 to S27; otherwise, terminate the process.

2. The wind turbine gearbox signal sparse decomposition method according to claim 1, characterized in that: The invention also includes dividing the wind turbine gearbox signal amplitude and the corresponding time-coded data obtained by historical actual measurement into a training set and a test set according to a ratio of 4:1, using the training set to train the fault identification of the original data and the decomposed complex product function component, and using the test set to obtain the sparse decomposition structure of the wind turbine gearbox signal with excellent final identification effect, wherein the loss function used in the training process is: In the formula and Respectively represent The true output category of the training samples and the identification category of the complex product function component, Indicates error.

3. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method according to claim 1 or 2.

4. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to claim 1 or 2.