Wind power bearing intelligent diagnosis method and system
Through the combination of the Elk Optimization Algorithm Optimization Algorithm and the CNN Network Model, the problem of noise interference in wind power bearing fault diagnosis is solved, and high-precision and efficient fault identification are achieved.
Patent Information
- Application Number
- CN202510327382.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2025-07-04
AI Technical Summary
Traditional wind power bearing fault diagnosis methods are difficult to accurately identify fault characteristics under noise interference, resulting in low diagnostic efficiency.
The parameters of the variable modal decomposition algorithm (VMD) are optimized by using the Elk Optimization Algorithm (EHO), combined with the CNN network model, feature extraction and diagnosis of wind power bearing failure data are performed, weak defect signals are decomposed by the optimization of the VMD algorithm, and feature components are screened using energy entropy and network model is trained.
It improves the accuracy and efficiency of wind power bearing fault diagnosis, can effectively filter out noise interference, and provides efficient and reliable intelligent diagnosis solutions.
Smart Images
Figure CN120253231A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of mechanical fault diagnosis, and particularly to an intelligent diagnosis method and system for wind power bearings. Background Art
[0002] The wind power bearing is one of the core components of a wind turbine generator set. Bearing failures will directly affect the operation safety and efficiency of the generator. Due to the complex operating environment of the wind turbine, the vibration signal of the bearing is often submerged by noise, making it difficult for traditional fault diagnosis methods to accurately identify the bearing fault characteristics.
[0003] In recent years, vibration signal analysis has been widely used in bearing fault diagnosis. Among them, the variational mode decomposition algorithm has gradually become a research hotspot because of its excellent performance in dealing with non-stationary and non-linear signals. However, the performance of the variational mode decomposition algorithm depends on the reasonable setting of parameters.
[0004] To overcome these defects, the present application proposes an intelligent diagnosis method and system for wind power bearings. Summary of the Invention
[0005] The purpose of the present application is to provide an intelligent diagnosis method and system for wind power bearings, aiming to solve the problem of low fault diagnosis efficiency caused by noise interference.
[0006] To achieve the above purpose, the present application provides the following technical solutions:
[0007] In the first aspect, the present application provides an intelligent diagnosis method for wind power bearings, including:
[0008] Obtain wind power bearing fault data;
[0009] Process the bearing fault data to obtain standardized bearing fault data;
[0010] Extract features from the standardized bearing fault data through a constructed wind power bearing detection algorithm to obtain bearing fault data modal components;
[0011] Input the bearing fault data modal components into a network model for training to obtain an optimal network model;
[0012] Use the optimal network model to diagnose the standardized target bearing fault data and generate a wind power bearing diagnosis result.
[0013] In the second aspect, the present application provides an intelligent diagnosis system for wind power bearings, including:
[0014] An acquisition module: obtain wind power bearing fault data;
[0015] A standardization module: process the bearing fault data to obtain standardized bearing fault data;
[0016] Optimization module: Extract features from the standardized bearing fault data through the constructed wind power bearing detection algorithm to obtain the modal components of the bearing fault data;
[0017] Diagnosis module: Input the modal components of the bearing fault data into the network model for training to obtain the optimal network model; Use the optimal network model to diagnose the standardized target bearing fault data and generate the wind power bearing diagnosis result.
[0018] In a third aspect, the present application provides a computer device, which includes a processor and a memory coupled to the processor. Among them, the memory stores program instructions for implementing a wind power bearing intelligent diagnosis method; The processor is used to execute the program instructions stored in the memory to implement a wind power bearing intelligent diagnosis.
[0019] In a fourth aspect, the present application provides a storage medium storing program instructions that can be run by a processor, and the program instructions are used to execute a wind power bearing intelligent diagnosis method.
[0020] The present application provides a wind power bearing intelligent diagnosis method and system, which has the following beneficial effects:
[0021] In the present application, by acquiring wind power bearing fault data and performing standardization processing, further optimizing the parameters of the variational mode decomposition algorithm (VMD) through the elk optimization algorithm (EHO), using the global optimization algorithm to optimize parameters such as the decomposition number and penalty factor of the VMD algorithm, then applying the optimized VMD algorithm to decompose the acoustic emission signal and vibration signal generated by weak defects, analyzing the energy entropy of each modal function, selecting components that match the fault characteristics according to a specific threshold for feature reconstruction, dividing the reconstructed data into a training set and a test set according to a certain ratio, optimizing the network model weights using the training set, and after multiple iterative training and testing processes, obtaining the network model with the best performance, so as to use the trained network model to output the wind power bearing diagnosis result and realize the intelligent diagnosis process of wind power bearing fault noise reduction. Description of the Drawings
[0022] Figure 1 It is a flowchart of a wind power bearing intelligent diagnosis method according to Embodiment 1 of the present application;
[0023] Figure 2 It is an implementation flowchart of a wind power bearing intelligent diagnosis method according to Embodiment 1 of the present application;
[0024] Figure 3 It is a flowchart of the elk optimization algorithm (EHO) according to Embodiment 1 of the present application;
[0025] Figure 4 Schematic diagram of the preset algorithm in the elk optimization algorithm (EHO) of Embodiment 1 of the present application;
[0026] Figure 5 Schematic diagram of the optimization process of the variational mode decomposition algorithm of Embodiment 1 of the present application;
[0027] Figure 6 Schematic diagram of the framework of the intelligent diagnosis method and network model for wind power bearings in Embodiment 1 of the present application;
[0028] Figure 7 Schematic diagram of the structure of an intelligent diagnosis system for wind power bearings in Embodiment 2 of the present application;
[0029] Figure 8 Schematic diagram of the structure of a computer device in Embodiment 3 of the present application;
[0030] Figure 9 Schematic diagram of the structure of a storage medium in Embodiment 4 of the present application. Detailed implementation manners
[0031] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0032] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.
[0033] Embodiment 1
[0034] Please refer to Figure 1 , which is a schematic diagram of the process of an intelligent diagnosis method for wind power bearings in Embodiment 1 of the present application; the steps include:
[0035] S100: Obtain wind power bearing fault data.
[0036] In this embodiment, bearing fault data reflecting the health status of the wind power bearing is collected, and the bearing fault data includes: defect acoustic emission signals and vibration characteristic signals.
[0037] S200: Process the bearing fault data to obtain the standardized bearing fault data.
[0038] In this embodiment, the bearing fault data is standardized, and its expression is:
[0039]
[0040] In the above formula, X′ z is the standardized bearing fault data; X z is the bearing fault data; N is the number of fault data; μ and σ are the standardization parameters, μ is the mean value, and σ is the standard deviation value.
[0041] S300: Extract features from the standardized bearing fault data through the constructed wind turbine bearing detection algorithm to obtain the modal components of the bearing fault data.
[0042] In this embodiment, the wind turbine bearing detection algorithm is established based on the elk optimization algorithm and the variational mode decomposition algorithm, and the steps include:
[0043] S310: Based on the Hilbert spectral function, optimize the parameters of the variational mode decomposition algorithm through the elk optimization algorithm to obtain the optimal number of modal components and the optimal penalty factor value.
[0044] The elk optimization algorithm (EHO) consists of three different behaviors: the rutting season, the calving season, and the selection season. After subdivision, it is divided into steps S311 to S316, specifically including:
[0045] S311: Initialize the parameters of the elk optimization algorithm and the optimization problem, embed the preset problem into the elk optimization algorithm, and set the objective function, the design buck rate B r , the size of the elk population EHS, and the maximum number of iterations M_Itr.
[0046] The expression of the objective function is:
[0047]
[0048] Among them, f(x) is the objective function, which is used to measure the fitness of each elk or the solution x = x1, 2, …, n , n is the total number of attributes in each solution; the variable x in the elk i is an attribute of this elk indexed by i, and x i ∈[lb i , b i , lb i is the lower bound of the attribute x i , and ub i is the upper bound of the attribute x i .
[0049] The buck rate Br determines the initial buck rate in the elk herd. The other two standard parameters are the elk herd size or population size (EHS) and the maximum number of iterations (M_Itr). This scheme simulates the natural reproduction mechanism of the elk herd. Initializing EHO aims to provide the objective function for evaluating solutions and clarify the solution representation of the search space type.
[0050] S312: Obtain an initial elk herd, which includes male and female elks; assume the elk herd EH is an n-dimensional matrix with a size of n×EHS.
[0051] The expression of the elk herd EH is:
[0052]
[0053] where n is the dimension; in the continuous domain, each solution x j is generated as The fitness value of each solution is calculated using the objective function f(x). Finally, the individual deer in the elk herd EH are sorted in ascending order according to the fitness value, such as f(x 1 ) ≤ … ≤ f(x EHS ).
[0054] S313: Establish an EHO model during the rutting season, divide the population according to the rutting male elk rate, select male elks from the elk herd EH based on a preset fitness value, and allocate female elks to male elks using a preset algorithm.
[0055] During the rutting season, the calculation formula for the male elk population is: Select male elks from the elk herd EH according to the fitness value:
[0056]
[0057] To allocate harems to each male deer in, use roulette wheel selection, each male elk x j in according to its absolute fitness value f(x j ) divided by the sum of the absolute fitness values of all male elks is given its selection probability p j :
[0058]
[0059] S314: During the calving season, each calf reproduces according to the attributes of its male elk father and female elk mother. In calving, the coefficients β and γ are factors that balance the global and local search capabilities of EHO.
[0060] During the calving season, each calf reproduces according to the attributes of its male elk father and female elk mother attributes;
[0061] When the exponent i of the calf x i (t + 1) is the same as that of its male elk, the reproduction of the calf is expressed as:
[0062]
[0063] Among them, α is a random value within the range of [0, 1], which determines the ratio of inheriting attributes from a randomly selected elk herd where k ∈ (1, 2, …, EHS); the higher the value of α, the greater the possibility of random elements participating in the new calves, enhancing diversification.
[0064] When the index of the calf is the same as that of its female elk, the calf x i (t + 1) takes the attributes of the harem x j and the male elk sire as shown in the following formula:
[0065]
[0066] Among them, is the attribute i of the calf j at iteration t + 1; h j is the male elk in the harem j; r is the index of a randomly selected male elk among the current male elks, such that γ and β are random values within the range of [0, 2], which are factors balancing the global and local search capabilities of EHO.
[0067] S315: During the selection season, the male elks, calves, and female elks in the population fuse to form an elk herd EH. The combination of the parents and newborn calves stored in EH forms a matrix, and the EHS optimal individuals are selected through sorting the optimal values as the next generation; this process is called μ + λ - selection, where μ is the parent population and λ is the offspring population.
[0068] During the selection season, all individual elks fuse to form an elk herd EH. The combination of the parents and newborn calves stored in EH forms a matrix Ehtemp, and the elks in EHtemp are sorted in ascending order based on their health values. Finally, the top elks numbered EHS in EH temp will be retained for the next generation, and they will replace the elks in EH, that is, EH j = EH j temp , j = (1, …, EHS). In the evolutionary strategy, this type of selection is called μ + λ - selection, where μ is the parent population and λ is the offspring population.
[0069] S316: Repeat the steps of the repeated estrus season, calving season, and selection season until the preset termination condition is met.
[0070] Repeat steps S313 to S315. The preset termination condition can be the maximum number of iterations, the maximum calculation time, or the reachability of the optimal solution.
[0071] Finally, based on the coefficients β and γ and the μ+λ-selection, the envelope entropy value calculation process is performed on the parameters of the variational mode decomposition algorithm through the variational mode update algorithm, and the optimal number of modal components and the optimal penalty factor value are determined according to the calculation results.
[0072] Among them, the expression of the variational mode update algorithm is:
[0073]
[0074] Among them, is the Wiener filter of the residual; is the Fourier transform of the signal f(t); is the i-th modal update expression; represents the i-th modal expression; is the modal bandwidth corresponding to the frequency ω; is the center frequency of the k-th mode in the n-th iteration process; τ is the fidelity coefficient, n is an integer; k is the initial number of modes; i is different modes or components, and ω is the angular frequency.
[0075] Please refer to Figure 3 , which is the flow schematic diagram of the elk optimization algorithm (EHO) in Embodiment 1 of this application. The optimization process is as follows:
[0076] 1) Initialize the population, and set parameters such as the value range of [,α] and the initial population of EHO;
[0077] 2) Use the natural reproduction mechanism of the elk herd to simulate the reproduction behavior of the natural elk herd;
[0078] 3) Generate an elk herd, calculate the fitness value, and sort it in ascending order according to the fitness at any time;
[0079] 4) During the elk estrus season, divide the family using the buck rate, and the buck herds fight, and the selection probability is given through the fight;
[0080] 5) The elk gives birth to calves. The calves choose the reproduction form according to the male and female parent attributes. The inheritance attribute α determines the ratio of inheriting attributes from the randomly selected elk herd , the exponent r of the randomly selected buck in the buck concentration; γ and β are random values in the range of [0, 2], and they randomly determine the part of the attributes inherited from the previously generated calves;
[0081] 6) During the elk selection season, select all individual deer in the population according to the μ+λ-selection, and retain the excellent individuals for the next generation;
[0082] 7) Repeat 4), 5), 6) until the termination condition is met;
[0083] 5) The variational mode decomposition algorithm decomposes the original signal, calculates the envelope entropy values of each parameter combination [,α] using the variational mode update algorithm, and records the optimal individual position;
[0084] 6) Select the iteration formula according to the size of the convergence factor and iterate and update until the termination condition is met;
[0085] 7) Finally, output the optimal decomposition number K and the quadratic penalty factor α.
[0086] S320: Embed the optimal mode component value and the optimal penalty factor value into the variational mode decomposition algorithm, and combine the elk optimization algorithm to obtain a wind power bearing detection algorithm based on EHO-VMD.
[0087] It should be noted that in the "variational" process of the variational mode decomposition algorithm, the quadratic penalty factor α and the Lagrange multiplier operator λ(t) are used to transform the constrained variational into an unconstrained variational, and then the optimal solution of the variational model is searched iteratively by the alternating direction multiplier method. In order to finally achieve the effective decomposition of the signal adaptively, the main method is to update the bandwidth and frequency center of each mode component multiple times. When the variational mode decomposition algorithm mainly decomposes the original signal, it can decompose the signal into a certain number of mode functions with a central frequency of ω k simultaneously through a preset decomposition number, and each mode u k fluctuates around the central frequency ω k and its bandwidth can be estimated by H 1 Gaussian smoothing.
[0088] When the variational mode decomposition method decomposes the signal, it abandons the sifting principle in the empirical mode decomposition method, but changes the decomposition process into a solution process of a constrained variational problem. The signal is decomposed by constructing and solving the constrained variational model. Assuming that the intrinsic mode functions with different central frequencies and finite bandwidths exist in each mode function, then finding k mode functions u k will be the main purpose of solving the variational problem, and the sum of the bandwidth estimates between each mode is minimized.
[0089] The process of the variational mode decomposition algorithm for modal decomposition of data signals is as follows:
[0090] First, for each mode function, in order to obtain its analytic signal, the unilateral spectrum is obtained by using the Hilbert transform, and the spectrum expression of the analytic signal is:
[0091]
[0092] where δ(t) is the Dirac distribution; {u k} = {u1,…,k} is the set of K BLIMFs decomposed by the variational mode decomposition algorithm; j is the imaginary unit.
[0093] Secondly, by mixing the analytic signals of each mode, a pre-estimated center frequency of is obtained. After shifting the frequency to the baseband, the expression of the mixed analytic signal is:
[0094]
[0095] where is the center frequency.
[0096] Calculate the square value of the L2 norm of the time gradient of the demodulated signal to estimate the bandwidth of the modal component, and its expression is:
[0097]
[0098] Therefore, the construction method of the constrained variational model is:
[0099]
[0100] where {ω k} = {ω1, …, k} is the set of the center frequencies of K modal components, and f represents the input signal.
[0101] Solving the constrained variational model: The method of introducing the quadratic penalty factor α and the Lagrange multiplier λ(t) is adopted to solve the problems that may occur when there is Gaussian noise. The former can ensure the reconstruction accuracy of the signal, while the latter can ensure the strictness of the constraint conditions. The expression of the augmented Lagrangian multiplier L is as follows:
[0102]
[0103] The unconstrained variational problem can be solved by the Alternate Direction Method of Multipliers (ADMM). Using the ADMM algorithm, alternately iterate and update λ n+1 to find the "saddle point" in the extended Lagrangian expression.
[0104] For the update of mode , it can be equivalent to solving a minimization problem:
[0105]
[0106] where λ(t) is the multiplier, For partial derivative calculation.
[0107] Perform Fourier equidistant transformation on the solution formula of the minimization problem of the mode, transform it to the frequency domain for solution, and obtain the update expression of the k-th mode, and its expression is:
[0108]
[0109] where τ is the fidelity coefficient.
[0110] According to the same principle, transform the central frequency solution to the frequency to obtain the update of the central frequency, and its expression is:
[0111]
[0112] where is equivalent to the Wiener filtering of the current remaining amount ; the central frequency is the centroid of the power spectrum of its corresponding mode function ; by performing the inverse Fourier transform on the signal after Wiener filtering, the mode u in the time domain signal is obtained and the real part is taken. k (t) and the real part thereof is obtained.
[0113] Use the alternating direction multiplier algorithm to perform constraints and obtain the optimal solution of the variational model. Finally, k mode components decomposed from the original signal can be obtained.
[0114] It can be understood that based on the elk optimization algorithm, the [,α] parameter combination of the variational mode decomposition algorithm is optimized, and then the optimal parameter combination is found. Using the Hilbert spectrum function as the criterion, the optimal parameters of the variational mode decomposition algorithm are accurately obtained by the elk optimization algorithm, and the number K of mode components and the penalty factor α value of the acoustic emission data of the wind power bearing fault are adaptively determined. At the same time, the elk optimization variational mode decomposition algorithm is used to decompose the acoustic emission signals of the inner and outer rings and the cage faults of the bearing.
[0115] S400: Input the modal components of the bearing fault data into the network model for training to obtain the optimal network model.
[0116] In this embodiment, the bearing fault data after standardization is subjected to modal decomposition processing based on the EHO-VMD wind power bearing detection algorithm to obtain the modal components of the bearing fault data; the average envelope entropy is set, and the modal components of the bearing fault data are screened to obtain the screened modal components of the bearing fault data.
[0117] Perform reconstruction calculation on the screened modal components of the bearing fault data to obtain the reconstructed data.
[0118] Divide the reconstructed data into a training set and a test set according to a preset ratio, and use the training set to optimize the neural network weights, perform iterative training and testing, and obtain the optimal network model;
[0119] Among them, the calculation formula for the average envelope entropy is:
[0120]
[0121] Among them, k and α are the optimal parameter combinations obtained after optimization; is the envelope entropy of the i-th intrinsic mode function IMF after Hilbert demodulation processing;
[0122] The calculation formula for the envelope entropy is:
[0123]
[0124] Among them, N is the number of sampling points; a(i) is the envelope signal; p t is the normalized form of a(i), t represents the time index or time point of the signal. Specifically, t is a sampling point of the signal, which is a position identifier in the time domain and usually represents the state of the signal at a specific moment.
[0125] Subsequently, the energy entropy is used to measure the complexity of the signal energy distribution. The energy entropy is based on the energy distribution of the signal in different frequency or time domain components and reflects the randomness and uncertainty of the signal. The calculation formula for the energy entropy is:
[0126]
[0127] Among them, x i is each order of IMF; p i is the energy proportion of the i-th IMF; H Ei is the energy entropy; E i is the energy of the i-th component;
[0128] The calculation formula for the energy entropy threshold is:
[0129]
[0130] Among them, Q H is the energy entropy threshold, K is an empirical constant, usually taking a value of 2 or 3, representing a multiple of the standard deviation; E Ei is the standard deviation; the threshold set in this way can be used to determine whether the signal exceeds the normal range.
[0131] By selecting the mode functions greater than the threshold for feature reconstruction, the fault feature information can be better retained.
[0132] It can be understood that based on the principle of energy entropy, by calculating the modal components of the bearing fault vibration signal processed by the variational mode decomposition algorithm, the state change of the system can be quantitatively described. According to the principle in the energy entropy algorithm that the average envelope entropy is greater than 3, the modal components with more fault information are selected, effectively improving the performance of the model, and further improving the extraction accuracy of wind power bearing fault data.
[0133] The modal components selected based on the principle of energy entropy are reconstructed and calculated, and their complexity is calculated to comprehensively characterize the wind power bearing fault data and achieve the purpose of intelligent diagnosis. Through the wind power bearing detection algorithm based on the elk optimization algorithm - variational mode decomposition algorithm, the ultimate goal of intelligent diagnosis of wind power bearing fault data is realized.
[0134] S500: Use the optimal network model to diagnose the standardized target bearing fault data and generate a wind power bearing diagnosis result.
[0135] In this embodiment, the optimal network model can be a trained CNN model. The CNN model automatically learns the data features of different fault types, identifies the abnormal patterns during the operation of the motor, and thus completes the intelligent diagnosis of the wind power bearing. Output the intelligent diagnosis result of the wind power bearing to provide technical support for the maintenance and management of wind power equipment.
[0136] In summary, in Embodiment 1 of the present application, by integrating three technologies of EHO, VMD, and the network model, a wind power bearing intelligent diagnosis method is formed, which not only improves the accuracy and efficiency of fault diagnosis, but also demonstrates technological innovation and practicality; specifically, the EHO algorithm is introduced to optimize the VMD parameters, overcoming the problem that the VMD algorithm is sensitive to parameters and improving the signal decomposition effect, thus laying a solid foundation for subsequent feature extraction and fault diagnosis; in addition, careful processing is carried out in feature extraction and reconstruction, and screening is performed by setting the average envelope entropy to ensure the quality of the input data and improve the accuracy of diagnosis.
[0137] Embodiment 2
[0138] Please refer to Figure 7 , which is a schematic structural diagram of a wind power bearing intelligent diagnosis system according to Embodiment 2 of the present application; the specific content includes:
[0139] Acquisition module: Acquire wind power bearing fault data;
[0140] Standardization module: Process the bearing fault data to obtain standardized bearing fault data;
[0141] Optimization module: Feature extraction is performed on the standardized bearing fault data through the constructed wind power bearing detection algorithm to obtain the modal components of the bearing fault data;
[0142] Diagnosis module: The modal components of the bearing fault data are input into the network model for training to obtain the optimal network model; The optimal network model is used to diagnose the standardized target bearing fault data to generate the wind power bearing diagnosis result.
[0143] Embodiment 3
[0144] Please refer to Figure 8 , which is a schematic structural diagram of a computer device according to Embodiment 3 of the present application. The computer device 50 includes a processor 51 and a memory 52 coupled to the processor 51.
[0145] The memory 52 stores program instructions for implementing the above-mentioned intelligent diagnosis method for wind power bearings.
[0146] The processor 51 is used to execute the program instructions stored in the memory 52 to implement an intelligent diagnosis of wind power bearings.
[0147] Among them, the processor 51 can also be called a CPU (Central Processing Unit, central processing unit).
[0148] The processor 51 may be an integrated circuit chip with signal processing capabilities. The processor 51 can also be a general-purpose processor, 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, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.
[0149] Embodiment 4
[0150] Please refer to Figure 9, which is a schematic structural diagram of the storage medium according to Embodiment 4 of the present application. The storage medium of the embodiment of the present application stores a program file 61 that can implement all the above methods. Among them, the program file 61 can be stored in the above storage medium in the form of a software product, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, or devices such as computers, servers, mobile phones, and tablets.
[0151] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, device, article or method including that element.
[0152] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.
[0153] Although the embodiments of the present application have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present application. The scope of the present application is defined by the appended claims and their equivalents.
[0154] Certainly, the present invention can also have other various implementation manners. Based on this implementation manner, other implementation manners obtained by those of ordinary skill in the art without any creative work belong to the scope protected by the present invention.
Claims
1. An intelligent diagnosis method for wind power bearings, characterized in that, Including: Obtain wind power bearing fault data; Process the bearing fault data to obtain standardized bearing fault data; Extract features from the standardized bearing fault data through the constructed wind power bearing detection algorithm to obtain bearing fault data modal components; Input the bearing fault data modal components into the network model for training to obtain the optimal network model; Use the optimal network model to diagnose the standardized target bearing fault data and generate a wind power bearing diagnosis result.
2. The intelligent diagnosis method for a wind power bearing according to claim 1, characterized in that, The wind power bearing detection algorithm is established based on the elk optimization algorithm and the variational mode decomposition algorithm, including: Based on the Hilbert spectrum function, optimize the parameters of the variational mode decomposition algorithm through the elk optimization algorithm to obtain the optimal number of modal components and the optimal penalty factor value; Embed the optimal number of modal components and the optimal penalty factor value into the variational mode decomposition algorithm, and combine the elk optimization algorithm to obtain a wind power bearing detection algorithm based on EHO-VMD.
3. The intelligent diagnosis method for a wind power bearing according to claim 2, characterized in that, In the step of optimizing the parameters of the variational mode decomposition algorithm through the elk optimization algorithm based on the Hilbert spectrum function to obtain the optimal number of modal components and the optimal penalty factor value, it specifically includes the following steps: Initialize the parameters of the elk optimization algorithm and the optimization problem, embed the preset problem into the elk optimization algorithm, set the objective function, design the male elk rate Br, the size EHS of the elk population, and the maximum number of iterations M_Itr; Obtain the initial elk group, which includes male elk and female elk; assume the elk group EH is an n-dimensional matrix with a size of n×EHS; Establish an EHO model during the rutting season, divide the population according to the male elk rate during the rutting period, select male elk from the elk group EH based on the preset fitness value, and assign female elk to male elk using the preset algorithm; During the calving season, each calf reproduces according to the attributes of its male elk father and female elk mother. The coefficients β and γ in calving are factors that balance the global and local search capabilities of EHO; During the selection season, the male elk, calves, and female elk in the population are merged to form the elk group EH. The combination of the parents and newborn calves stored in EH forms a matrix, and the EHS optimal individuals are selected as the next generation through sorting the optimal values; this process is called μ+λ-selection, where μ is the parent population and λ is the offspring population; Repeat the steps of the rutting season, calving season, and selection season until the preset termination condition is met; Based on the coefficients β and γ and μ+λ-selection, calculate the envelope entropy value of the parameters of the variational mode decomposition algorithm through the variational mode update algorithm, and determine the optimal number of modal components and the optimal penalty factor value according to the calculation results.
4. The intelligent diagnosis method for a wind power bearing according to claim 3, characterized in that, The expression of the objective function is: Among them, f(x) is the objective function used to measure the fitness of each elk or the solution x = x1, x2, …, x n , where n is the total number of attributes in each solution; the variable x in the elk i is an attribute of the elk indexed by i, and x i ∈ [lb i , ub i , where lb i is the lower bound of the attribute x i , and ub i is the upper bound of the attribute x i . The expression of the elk group EH is: where n is the dimension; in the continuous domain, each solution x j is generated as 5. The intelligent diagnosis method for a wind power bearing according to claim 4, characterized in that During the estrus season, the calculation formula for the male elk population is as follows: Select male elks from the elk group EH according to the fitness value: each male elk x in j According to its absolute fitness value f(x j ) divided by the sum of the absolute fitness values of all male elks give it a selection probability p j : During the calving season, each young elk is bred according to the attributes of its male elk sire and female elk dam attributes; When the young elk x i (t + 1) has the same exponent i as its male elk, the reproduction of the young elk is expressed as: where α is a random value within the range of [0, 1], which determines the ratio of inheriting attributes from a randomly selected elk herd and k ∈ (1, 2, …, EHS); When the index of the young elk is the same as that of its female elk, then the young elk x i (t + 1) takes the harem x j and the male elk sire The attributes are shown as follows: Among them, is the attribute i of the young elk j at iteration t + 1; h j is the male elk in the harem j; r is the index of a random male deer among the current male deer, such that γ and β are random values within the range of [0, 2], which are factors to balance the global and local search capabilities of EHO.
6. The intelligent diagnosis method for a wind power bearing according to claim 3, wherein, The expression of the variational mode update algorithm is: Among them, is the Wiener filtering of the remaining amount; is the Fourier transform of the signal f(t); is the i-th mode update expression; represents the i-th mode expression; is the mode bandwidth corresponding to the frequency ω; is the center frequency of the k-th mode in the n-th iteration process; τ is the fidelity coefficient, n is an integer; k is the initial mode number; i is different modes or components, and ω is the angular frequency.
7. The intelligent diagnosis method for a wind power bearing according to claim 1, wherein In the step of inputting the bearing fault data modal components into the network model for training to obtain the optimal network model, it specifically includes the following steps: Set the average envelope entropy, screen the bearing fault data modal components to obtain the screened bearing fault data modal components; Reconstruct and calculate the modal components of the filtered bearing fault data to obtain the reconstructed data; Divide the reconstructed data into a training set and a test set according to a preset ratio, use the training set to optimize the neural network weights, perform iterative training and testing, and obtain the optimal network model; Among them, the formula for the average envelope entropy is: Among them, k and α are the optimal parameter combinations obtained after optimization; is the envelope entropy of the i-th intrinsic mode function IMF after Hilbert demodulation processing; The formula for the envelope entropy is: where N is the number of sampling points; a(i) is the envelope signal; p t is the normalized form of a(i), and t is the time index or time point of the signal; Use energy entropy to calculate the complexity of the signal distribution; the formula for the energy entropy is: where x i is each order IMF; p i is the energy proportion of the i-th IMF; H Ei is the energy entropy; E i is the energy of the i-th component; The formula for the energy entropy threshold is: Among them, Q H is the energy entropy threshold, K is the multiple of the standard deviation; E Ei is the standard deviation.
8. An intelligent diagnosis system for wind power bearings, characterized in that, The intelligent wind power bearing diagnosis system includes: An acquisition module: acquire wind power bearing fault data; A normalization module: process the bearing fault data to obtain normalized bearing fault data; An optimization module: extract features from the normalized bearing fault data through a constructed wind power bearing detection algorithm to obtain bearing fault data modal components; A diagnosis module: input the bearing fault data modal components into a network model for training to obtain the optimal network model; use the optimal network model to diagnose the normalized target bearing fault data and generate a wind power bearing diagnosis result.
9. A computer device, characterized in that, The computer device includes a processor and a memory coupled to the processor. Among them, the memory stores program instructions for implementing a wind power bearing intelligent diagnosis method according to any one of claims 1-7; the processor is configured to execute the program instructions stored in the memory to implement a wind power bearing intelligent diagnosis.
10. A storage medium, characterized in that, Store program instructions that can be run by a processor, and the program instructions are used to execute a wind power bearing intelligent diagnosis method according to any one of claims 1-7.