Fan bearing composite fault diagnosis method, device, electronic equipment and system

Abstract

The application relates to a fan bearing composite fault diagnosis method, device, electronic equipment and system, and belongs to the technical field of wind turbine equipment state monitoring and fault diagnosis, wherein the fan bearing composite fault diagnosis method comprises the following steps: converting the pretreated vibration signal obtained by processing the original vibration signal of the fan bearing based on a rotating speed signal into an angular domain stationary signal, and simultaneously obtaining a calibrated fault characteristic order through calibration; adopting an improved hippo optimization algorithm (IHOA) and an enhanced angular domain feature mode decomposition (EAFMD) to obtain an optimal parameter combination; combining the optimal parameter combination to decompose the angular domain stationary signal into multiple modal components by using the EAFMD, and determining a best modal sequence through an adaptive modal screening mechanism, and then extracting fault features from each optimal modal component to realize fan bearing composite fault diagnosis. The application can improve the diagnosis precision of the fan bearing composite fault under variable rotating speed and strong noise conditions.

Description

Technical Field

[0001] This invention relates to the field of wind turbine equipment condition monitoring and fault diagnosis technology, and in particular to a method, device, electronic equipment and system for diagnosing complex faults in wind turbine bearings. Background Technology

[0002] The efficiency of wind energy utilization is highly dependent on the operational stability of wind turbines. Rolling bearings are critical transmission components of wind turbines, and their health directly determines the reliability and power generation efficiency of the entire system. However, wind turbine rolling bearings typically operate under harsh conditions such as high loads, variable speeds, and high noise levels for extended periods. These harsh conditions accelerate bearing fatigue degradation, leading to failures.

[0003] When a rolling bearing fails, it typically generates a series of periodic impact pulses. Under varying speeds and strong noise interference, the weak composite fault signal not only exhibits strong non-stationarity but is also often severely submerged by background noise, posing a significant challenge to feature extraction. Current mainstream signal processing methods mainly include frequency band segmentation, blind deconvolution, and signal decomposition. Frequency band segmentation and blind deconvolution are essentially limited to feature extraction within a single frequency band. However, composite faults often induce resonance in multiple frequency bands simultaneously. This characteristic leads these methods to over-rely on prior knowledge for iterative optimization, making it difficult to extract complete fault features. Signal decomposition breaks down the original vibration signal into several independent modal components. The most representative signal decomposition algorithms include empirical mode decomposition (EMD), ensemble empirical mode decomposition (EMD), and variational mode decomposition (VMD). EMD suffers from mode aliasing and endpoint effects; while ensemble EMD suppresses aliasing, it inevitably results in residual white noise; and VMD fails to effectively utilize the prior information of the periodic impact of the fault signal. In recent years, Eigenmode Decomposition (EMD) has achieved adaptive separation of fault features without a preset period by constructing a relevant kurtosis objective function, effectively capturing periodic shocks. However, because its evaluation index strictly depends on time-domain periodicity, this method is prone to failure under variable speed conditions.

[0004] Traditional technologies suffer from the problem of difficulty in accurately separating and extracting complex fault features under variable speed and high noise conditions, due to the dependence of filter parameters on prior settings and the failure of time-domain evaluation indicators, which leads to a decrease in fault diagnosis accuracy. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, device, electronic equipment and system for diagnosing complex faults in wind turbine bearings, in order to solve the problem that the complex fault features are difficult to separate and extract accurately under variable speed and high noise conditions due to the dependence of filter parameters on prior settings and the failure of time-domain evaluation indicators, which leads to a decrease in fault diagnosis accuracy.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides a method for diagnosing complex faults in wind turbine bearings, comprising: The original vibration signal and rotational speed signal of the wind turbine bearing are acquired, and the original vibration signal is preprocessed to obtain the preprocessed vibration signal. Based on the rotational speed signal, the preprocessed vibration signal is converted into an angular domain stationary signal, and the fault characteristic order is calibrated to obtain the calibrated fault characteristic order. With maximizing the average Hoyer exponent as the optimization objective, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) to obtain the target parameter combination. The average Hoyer exponent is the arithmetic mean of the Hoyer exponents of each periodic signal in the angular domain. The target parameter combination includes the optimal number of filters and the optimal filter length. Based on the target parameter combination, the angular domain stationary signal is decomposed using the EAFMD to obtain multiple modal components; Adaptive mode extraction of composite faults is performed on the multiple modal components to generate target mode sequences corresponding to each bearing fault type; Based on the target mode sequence, order matched filters are introduced to suppress residual background noise, and the fault characteristic order and harmonic components of each fault type are extracted to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

[0007] In one possible implementation, the preprocessing of the original vibration signal includes: Perform a fast Fourier transform on the original vibration signal to obtain a complex spectrum, which includes an amplitude spectrum and a phase spectrum; A whitening filter is constructed, and the amplitude spectrum is whitened using the whitening filter to obtain the whitened frequency domain signal; Based on the original phase spectrum and the whitened amplitude spectrum, the time-domain signal is reconstructed by inverse fast Fourier transform, and the energy of the reconstructed signal is restored.

[0008] In one possible implementation, converting the preprocessed vibration signal into an angular-domain stationary signal based on the rotational speed signal specifically includes: Based on the instantaneous rotational speed values ​​corresponding to each sampling point in the rotational speed signal, a discrete phase accumulation sequence is established by integrating the accumulated rotational angles; The preprocessed vibration signal is resampled at equal angles based on the discrete phase accumulation sequence to convert the preprocessed vibration signal into a stationary signal in the angular domain.

[0009] In one possible implementation, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) to obtain a combination of target parameters, including: Initialize the improved Hippo Optimization Algorithm (IHOA) with its population size, maximum number of iterations, problem dimension, and search boundaries for the number and length of filters. Based on the parameter combination of the number of filters and the filter length corresponding to the current individual, the filter bank is initialized using the maximum entropy principle, and the coefficients of the filter bank are updated using the differential gradient model. Calculate the average Hoyer index value for each modal component, and define the maximum average Hoyer index value as the fitness value of the current individual; The fitness value of the current individual is compared with the individual's historical best value and the global historical best value to update the local best position of the individual and the global best position of the population. Tent chaotic inertial weights and progressive refraction mirror learning are used to update the position of the hippopotamus individual, respectively. Determine if the termination condition is met. If not, increment the iteration count and re-execute the filter bank coefficient update step. If the condition is met, output the target parameter combination, which includes the optimal number of filters and the optimal filter length.

[0010] In one possible implementation, the updating of the hippopotamus individual's position using Tent chaotic inertial weights and progressive refraction mirror learning includes: Using the current individual's position and fitness value as input, chaotic variables are generated using Tent mapping. The amplitude of the chaotic variables is modulated by a decay factor to construct adaptive inertial weights. The adaptive inertial weights are substituted into the current individual's position update process to generate the updated individual position. The fitness value is recalculated based on the updated individual position. The recalculated fitness value is compared with the individual's historical best value and the global historical best value to update the individual's local best position and the population's global best position. Taking the global optimal position and fitness value under the current iteration as input, a mirror-reverse solution is generated based on the law of refraction of light. The fitness value of the mirror-reverse solution is calculated and compared with the fitness value of the global optimal position. If the fitness value of the mirror-reverse solution is greater than the fitness value of the global optimal position, the mirror-reverse solution is updated to the new global optimal position; otherwise, the current global optimal position remains unchanged.

[0011] In one possible implementation, the step of decomposing the angular-domain stationary signal using the EAFMD based on the target parameter combination to obtain multiple modal components includes: Initialize the filter bank based on the optimal number of filters and the optimal filter length in the target parameter combination; A differential gradient model of the filter bank is constructed, and the coefficients of the filter bank are iteratively updated through a multi-scale step size adjustment strategy until the convergence condition is met. In each iteration, the angular domain stationary signal is filtered according to the current filter coefficients to separate each modal component, and the average Hoyer exponent of each modal component is calculated. The iterative updates cause the filter bank to converge in the direction of maximizing the average Hoyer exponent, thereby decomposing the angular domain stationary signal into multiple modal components.

[0012] In one possible implementation, the adaptive mode extraction of the multiple modal components for composite faults to generate target mode sequences corresponding to each bearing fault type includes: The modal components are initially screened using the average Hoyer index, eliminating redundant components below the threshold and retaining valid components. Construct the envelope order spectrum of the effective components, extract the order corresponding to the maximum energy peak in each envelope order spectrum, match and classify the order corresponding to the maximum energy peak with the calibrated fault feature order, and classify each effective modal component into the corresponding bearing fault type according to the matching and classification results. Calculate the envelope order harmonic signal-to-noise ratio (SNR) of all candidate modal components corresponding to each type of fault, and select the modal component with the largest SNR from the candidate components of each type of fault. The mode components with the highest signal-to-noise ratio selected under various fault conditions are combined to generate target mode sequences corresponding to each bearing fault type.

[0013] Secondly, the present invention also provides a wind turbine bearing composite fault diagnosis device, comprising: The signal acquisition and preprocessing module is used to acquire the original vibration signal and speed signal of the wind turbine bearing, and to preprocess the original vibration signal to obtain the preprocessed vibration signal. The angular domain conversion and order calibration module is used to convert the preprocessed vibration signal into an angular domain stationary signal based on the rotational speed signal, and to obtain the calibrated fault characteristic order by calibrating the fault characteristic order. The parameter adaptive optimization module is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) with the goal of maximizing the average Hoyer exponent. The target parameter combination is obtained by using the improved Hippo Optimization Algorithm (IHOA) to optimize the average Hoyer exponent of each periodic signal in the angular domain. The target parameter combination includes the optimal number of filters and the optimal filter length. The signal decomposition module is used to decompose the angular domain stationary signal according to the target parameter combination using the EAFMD to obtain multiple modal components; The adaptive mode extraction module is used to perform adaptive mode extraction of the multiple modal components for composite faults, and generate target mode sequences corresponding to each bearing fault type. The feature enhancement and diagnosis module is used to introduce order matched filters based on the target mode sequence to suppress residual background noise, extract the fault feature order and harmonic components of each fault type, so as to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

[0014] Thirdly, the present invention also provides an electronic device, including a signal acquisition unit, a memory, and a processor, wherein the signal acquisition unit is used to acquire the original vibration signal and rotation speed signal of the wind turbine bearing; The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the wind turbine bearing composite fault diagnosis method described in any of the above implementations.

[0015] Fourthly, the present invention also provides a wind turbine bearing composite fault diagnosis system, which includes an electronic device, a data acquisition device, and a result output device as described above. The data acquisition device and the result output device are respectively connected to the electronic device. The data acquisition device is used to acquire the original vibration signal and speed signal of the wind turbine bearing, and the result output device is used to output and display the composite fault diagnosis result of the wind turbine bearing.

[0016] The beneficial effects of this invention are as follows: The wind turbine bearing composite fault diagnosis method provided by this invention acquires the original vibration signal and speed signal of the wind turbine bearing, and preprocesses the original vibration signal to obtain the preprocessed vibration signal; based on the speed signal, the preprocessed vibration signal is converted into a stationary signal in the angular domain, and the fault feature order is calibrated to obtain the calibrated fault feature order; with maximizing the average Hoyer exponent as the optimization objective, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Feature Mode Decomposition (EAFMD) to obtain the target parameter combination; according to the target parameter combination, the stationary signal in the angular domain is decomposed using EAFMD to obtain multiple modal components; adaptive mode extraction of composite faults is performed on the multiple modal components to generate target mode sequences corresponding to each bearing fault type; based on the target mode sequences, order matched filters are introduced to suppress residual background noise, and the fault feature order and its harmonic components of each fault type are extracted to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions. Thus, the technical challenge of accurately separating and extracting composite fault features and reducing fault diagnosis accuracy caused by the dependence of filter parameters on prior settings and the failure of time-domain evaluation indicators under variable speed and high noise conditions was successfully solved, realizing adaptive diagnosis of composite faults of wind turbine bearings under variable speed and high noise conditions. Attached Figure Description

[0017] Figure 1 This is a flowchart of an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 2 A schematic diagram of the overall process for diagnosing composite faults in wind turbine bearings, as provided in this invention, is shown in an embodiment of the method for diagnosing composite faults in wind turbine bearings. Figure 3 A schematic diagram of the IHOA-optimized EAFMD core parameter flow of an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 4 A schematic diagram of the rotational speed curve of the experimental signal in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 5 The time-domain waveform of a noisy composite fault signal is shown in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention. Figure 6 An envelope order diagram of a noisy composite fault signal in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 7 A schematic diagram of the fitness convergence curve of an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 8 The optimal mode time-domain diagram of outer ring fault in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 9 The optimal mode envelope spectrum order diagram of an outer ring fault in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 10 The optimal mode time-domain diagram of inner ring fault in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 11 The optimal mode envelope spectrum order diagram of inner ring faults in an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention; Figure 12 A functional block diagram of a wind turbine bearing composite fault diagnosis device provided in an embodiment of this application; Figure 13 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0020] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] Before demonstrating the embodiments, the following terms will be explained.

[0023] CPW: Cepstrum Pre-whitening (CPW) is a preprocessing method that whitens a signal in the cepstrum domain to reduce deterministic interference components.

[0024] FCO: Fault Characteristic Order (FCO) refers to the characteristic order value of a bearing fault in the angular domain, which depends on the bearing's geometric parameters.

[0025] HI: Hoyer Index (HI) is a normalized index that measures the sparsity of a sequence.

[0026] AHI: Average Hoyer Index (AHI) is the arithmetic mean of the Hoyer indices of the signals in each period within the angular domain, used to measure the sparsity of the global signal.

[0027] HOA: Hippopotamus Optimization Algorithm (HOA), a metaheuristic optimization algorithm that simulates the defense and escape strategies of a hippopotamus population.

[0028] IHOA: Improved Hippopotamus Optimization Algorithm (IHOA) is an optimization algorithm that integrates three improvement strategies on the basis of HOA: Latin hypercube sampling, Tent chaotic inertial weights, and progressive refraction mirror learning.

[0029] LHS: Latin Hypercube Sampling (LHS) is a sampling strategy that ensures the initial population is uniformly distributed across all dimensions of the solution space through a stratified interval mechanism.

[0030] TCIW: Tent Chaotic Inertia Weight (TCIW) is an adaptive step size adjustment strategy built using the ergodicity and randomness of the Tent mapping.

[0031] PRML: Progressive Refractive Mirror Learning (PRML) is a strategy that uses the law of light refraction to generate potential reverse solutions to the current optimal solution in order to expand the search boundary.

[0032] EAFMD: Enhanced Angle-domain Feature Mode Decomposition (EAFMD) is a method for adaptive signal decomposition in the angular domain using the average Hoyer exponent as the objective function.

[0033] EOHNR: Envelope Order Spectrum Harmonic-to-Noise Ratio (EOHNR) is the ratio of the total energy of the fault characteristic order and its harmonics in the envelope order spectrum to the energy of the local background noise.

[0034] OMF: Order Matching Filter (OMF) is a filter that suppresses residual background noise by constructing a passband function to match the order of fault characteristics and their harmonic components.

[0035] This invention provides a method, apparatus, electronic device, and system for diagnosing complex faults in wind turbine bearings, which will be described below.

[0036] Figure 1 This is a schematic flowchart of an embodiment of the wind turbine bearing composite fault diagnosis method provided by the present invention. The execution subject of the wind turbine bearing composite fault diagnosis method can be a computer, a portable smart device, or a cloud server, etc. The specific execution subject is the storage medium in the computer, and this embodiment does not limit it.

[0037] like Figure 1 As shown, the method for diagnosing complex faults in wind turbine bearings includes: S101. Obtain the original vibration signal and rotational speed signal of the wind turbine bearing, and preprocess the original vibration signal to obtain the preprocessed vibration signal.

[0038] Before proceeding with the explanation, this embodiment will first describe the technical principles involved in the wind turbine bearing composite fault diagnosis method provided by the present invention. Under complex operating conditions, the frequency harmonics and electromagnetic interference in the original vibration signal can easily mask fault characteristics. In the cepstral domain, deterministic components are mapped as high-frequency discrete spikes, while bearing fault impacts focus on the low-frequency region. Therefore, cepstral pre-whitening (CPW) is used to whiten the amplitude spectrum to weaken high-energy harmonics.

[0039] However, traditional CPWs tend to amplify background noise while suppressing high-energy harmonics. Therefore, a generalized CPW is constructed by introducing a whitening intensity adjustment factor, achieving a dynamic balance between harmonic suppression and noise control. The overall process for composite fault diagnosis in this embodiment can be referred to... Figure 2 As shown.

[0040] In some embodiments, step S101 includes: performing a fast Fourier transform on the original vibration signal to obtain a complex spectrum, the complex spectrum including an amplitude spectrum and a phase spectrum; constructing a whitening filter and using the whitening filter to whiten the amplitude spectrum to obtain a whitened frequency domain signal; reconstructing the time domain signal based on the original phase spectrum and the whitened amplitude spectrum by an inverse fast Fourier transform, and restoring the energy of the reconstructed signal.

[0041] Specifically, the complex spectrum X(f) is obtained by performing a Fast Fourier Transform (FFT) on the original vibration signal x(t), and its expression is as follows:

[0042] Where A(f) is the amplitude spectrum and φ(f) is the phase spectrum containing the time-domain characteristics of the fault impact.

[0043] Construct a whitening filter, and obtain the frequency domain expression therefrom. as follows:

[0044] in, The whitening intensity coefficient is... This is a regularization parameter used to ensure the stability of the values.

[0045] The time-domain signal is reconstructed based on the original phase and the equalized amplitude spectrum through inverse fast Fourier transform (IFFT).

[0046]

[0047] Since whitening alters the signal's amplitude and energy level, energy restoration is necessary. The reconstructed time-domain signal... The expression is as follows:

[0048] in, , x(t) and The standard deviation.

[0049] S102. Based on the rotational speed signal, the preprocessed vibration signal is converted into a stationary signal in the angular domain, and the fault characteristic order is obtained by calibrating the fault characteristic order.

[0050] It should be noted that frequency modulation caused by variable speed operation makes the signal non-stationary, and the resulting spectral ambiguity renders traditional frequency domain analysis ineffective. Therefore, a time-angular domain mapping relationship is constructed using the speed integral method to convert the time-domain non-stationary signal into an angular-domain stationary signal.

[0051] In some embodiments, step S102 includes: establishing a discrete phase accumulation sequence by accumulating rotation angle integrals based on the instantaneous rotation speed values ​​corresponding to each sampling point in the rotation speed signal; and performing equal-angle resampling on the preprocessed vibration signal based on the discrete phase accumulation sequence to convert the preprocessed vibration signal into an angularly stationary signal.

[0052] Let the instantaneous rotational speed at time t be v(t), and its cumulative rotational angle θ(t) be defined as:

[0053] Since the collected vibration signal is a discrete sequence, the above continuous integral needs to be numerically discretized by introducing the sampling frequency. Its discrete phase accumulation formula can be expressed as:

[0054] Where fs and N are the sampling frequency and the number of sampling points, respectively, and v(n) is the instantaneous rotational speed (in r / s) corresponding to the nth sampling point.

[0055] The effective periodicity (FCO) depends on the bearing's geometric parameters. However, due to random slip and assembly errors, it often deviates from the theoretical value. If the theoretical value is directly used to divide the period, the accumulated phase error will weaken the stationarity of the characteristic. Therefore, a coarse-to-fine hierarchical search strategy is proposed to obtain the optimal FCO. The optimization model is defined as follows.

[0056]

[0057] Where AHI is the average Hoyer exponent operator, Ω is the search interval for optimization, and FCO is the search interval for optimization. opt This represents the final optimal fault characteristic order.

[0058] Using FCO opt The periodic index sequence of the segmented signal is defined as follows:

[0059] Where k is the period number, n k This is the starting index for the k-th fault cycle.

[0060] S103. With maximizing the average Hoyer index as the optimization objective, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Feature Mode Decomposition (EAFMD) to obtain the target parameter combination.

[0061] It should be noted that in the time domain, varying rotational speed disrupts the statistical stationarity of fault shocks, causing traditional time-domain metrics to become invalid. However, in the angular domain, the sparsity of the signal remains unchanged. The Hoyer index (HI) is a normalized metric for measuring the sparsity of a sequence, defined as follows:

[0062] Where L is the number of sampling points in the m-th period, u(n) is the instantaneous amplitude of the signal mode, and ||u||1 and ||u||2 represent the L1 and L2 norms of the sequence, respectively.

[0063] To measure the sparsity of the global signal, the average Hoyer exponent (AHI) is used as the final objective function. AHI is defined as the arithmetic mean of the periods HI in each angular domain, and its calculation formula is as follows:

[0064] Among them, S m Let M be the signal sequence of the m-th period, where ||sm||1 and ||sm||2 correspond to the L1 and L2 norms of the signal, respectively, and M is the total number of periods.

[0065] EAFMD updates the filter bank coefficients using AHI as the objective function, thereby achieving the separation of signal modes. Let the signal sequence be Z = {z(1), z(2), ..., z(n)}, with length N. The core theory of EAFMD can be expressed as solving the following constrained optimization problem:

[0066] Among them, c k (n) represents the k-th mode. ωk represents the k-th FIR filter coefficient of length D, c k,m This is the segmented signal of the k-th decomposed mode within this period.

[0067] To solve this nonlinear optimization problem, a multi-scale step size adjustment strategy is adopted, and the following difference gradient model is constructed:

[0068] By introducing the unit norm constraint to ensure numerical stability, the coefficient iterative update equation is derived:

[0069] Where s is the iteration index, βs is the decay step size, Δs is the standard normal random perturbation, and H(.) is the unit step function.

[0070] In some embodiments, step S103 includes: initializing the population size, maximum number of iterations, problem dimension, number of filters K, and search boundary with length L of IHOA; initializing the filter bank according to the parameter combination corresponding to the current individual using the maximum entropy principle, and then updating the coefficients of the filter bank using the differential gradient model; calculating the AHI value of each modal component, and defining the maximum value as the fitness of the current individual; evaluating the current fitness value and comparing it with the individual and the global historical best value, and then updating the local optimum and global optimum position of the individual; updating the hippopotamus position and enhancing population diversity using TCIW and PRML; determining whether the termination condition is met, if not, incrementing the iteration count by 1 and returning to the position update step, otherwise outputting the global optimum parameter combination [K_best, L_best].

[0071] It should be noted that the position of the hippopotamus individual is updated using Tent chaotic inertial weights and progressive refraction mirror learning, respectively. This includes: taking the current position and fitness value of the individual as input, generating chaotic variables using Tent mapping, modulating the amplitude of the chaotic variables through a decay factor to construct adaptive inertial weights, substituting the adaptive inertial weights into the current individual's position update process to generate the updated individual position, and recalculating the fitness value based on the updated individual position. The recalculated fitness value is compared with the individual's historical best value and the global historical best value to update the individual's local best position and the population's global best position. Taking the global best position and fitness value under the current iteration as input, generating a mirror inversion solution based on the law of light refraction, calculating the fitness value of the mirror inversion solution, and comparing the fitness value of the mirror inversion solution with the fitness value of the global best position. If the fitness value of the mirror inversion solution is greater than the fitness value of the global best position, then the mirror inversion solution is updated to the new global best position; otherwise, the current global best position remains unchanged.

[0072] It should be understood that HOA achieves global optimization by simulating the defense and escape strategies of the hippo population. However, this algorithm is prone to getting trapped in local optima due to the decay of population diversity in the later stages of iteration. To address this, a multi-strategy improved HOA is proposed. The improved hippo optimization algorithm uses a Latin hypercube sampling strategy to initialize the population distribution in the initialization phase, introduces Tent chaotic inertial weights to dynamically adjust the step size in the position update phase, and employs progressive refraction mirror learning to expand the search boundary in the later stages of iteration. For example: 1) In the initialization phase, LHS is used to optimize the initial population distribution, enhancing the ergodicity of the solution space. 2) In the position update phase, Tent chaotic inertial weights are introduced to dynamically adjust the step size, suppressing premature convergence. 3) In the later stages of iteration, PRML is used to expand the search boundary, improving the ability to escape local optima.

[0073] It should be noted that the complete process of IHOA adaptively optimizing the core parameters of EAFMD is as follows: Figure 3 As shown, the specific steps are as follows: a. Initialize the population size, maximum number of iterations, problem dimension, number of filters K, and search boundary with length L for IHOA. b. Initialize the filter bank using the maximum entropy principle based on the parameter combination corresponding to the current individual. Then update the coefficients of the filter bank using the differential gradient model. c. Calculate the AHI value of each modal component and define the maximum value as the fitness of the current individual. d. Evaluate the current fitness value and compare it with the individual and the global historical best value. Then update the local optimum and the global optimum position of the individual. e. Update the position of male hippos and enhance population diversity using TCIW and PRML. Determine whether the termination condition is met. If not, increment the iteration count by 1 and return to step e. Otherwise, go to step f. f. Output the global optimal parameter combination [Kbest, Lbest].

[0074] Specifically, the Latin hypercube sampling is as follows: LHS ensures that the initial population is uniformly distributed across all dimensions of the solution space through a stratified interval mechanism. Let the population size be S and the search space dimension be V. The mathematical model for initializing the hippopotamus population can be expressed as:

[0075] Among them, X ij Let be the initial position of the i-th individual in the j-th dimension, where i ∈ {1, 2, ..., S} and j ∈ {1, 2, ..., V}. B j with U j Let P represent the lower and upper bounds of the j-th dimension search space, respectively, where η is a random number following U(0,1). j For a random permutation of the set {1,2,…,S}, P j (i) represents the i-th element of the permutation.

[0076] Furthermore, the Tent chaotic inertia weights are as follows: TCIW utilizes the ergodicity and randomness of the Tent mapping to construct an adaptive step size adjustment strategy. Let... The iterative equation for the chaotic variables generated by this mapping can be expressed as:

[0077] Where τ is the current iteration number.

[0078] A linear decay factor is introduced to modulate the amplitude of the chaotic variable, constructing an adaptive weight ψ(τ). Its mathematical model can be expressed as:

[0079] Where, ψmin and ψ max These are the lower and upper limits of the inertia weight, respectively, T. max This represents the maximum number of iterations.

[0080] Substituting into the position update equation, the final iterative formula is:

[0081] Among them, X i (τ) and Xnew i (τ) represents the position of individual i before and after the update, X best Let r1 be the current global optimal position, r1 be a random number within (0,1), and I1 be a random integer in the set {1,2}.

[0082] Furthermore, the progressive refraction mirror image learning is as follows: The potential reverse position of the current optimal solution is generated based on the law of refraction of light. Let ρ1 and ρ2 represent the refractive indices of the two media, θ1 and θ2 be the angle of incidence and the angle of refraction, and h1 and h2 be the propagation distances. From geometric relationships, we can obtain:

[0083] Among them, X j * represents the mirror-image inverse solution generated through the refraction mechanism.

[0084] Let μ = ρ1 / ρ2 and h1 = h2, then we can obtain:

[0085] After simplification, the inverse solution can be expressed as:

[0086] Among them, X best,j This represents the j-th dimension component of the current global optimal solution.

[0087] To meet the dynamic search requirements of the algorithm at different iteration stages, a PRML mechanism is designed. Its mathematical model can be expressed as:

[0088] As shown in the above equation, in the early stages of iteration, μ ≥ 1 guides the algorithm to perform a refined search near the center of the search space. As the iteration progresses, μ tends to 1, and this mechanism degenerates into standard backward learning.

[0089]

[0090] Through this incremental optimization mechanism, the algorithm can expand its search range to escape local optima. The updated global optimum can be represented as:

[0091] S104. Based on the target parameter combination, the angular domain stationary signal is decomposed using the EAFMD to obtain multiple modal components.

[0092] It should be noted that, based on the optimal parameter combination [K_best, L_best] obtained from the above steps, the signal is decomposed into a series of independent modal components using EAFMD, providing initial components for subsequent adaptive mode extraction.

[0093] In some embodiments, step S104 includes: initializing a filter bank according to the optimal number of filters and the optimal filter length in the target parameter combination; constructing a differential gradient model of the filter bank; iteratively updating the coefficients of the filter bank through a multi-scale step size adjustment strategy until the convergence condition is met; wherein, in each iteration, the angular domain stationary signal is filtered according to the current filter coefficients to separate each modal component, and the average Hoyer exponent of each modal component is calculated; the iterative update causes the filter bank to converge in the direction of maximizing the average Hoyer exponent, so as to decompose the angular domain stationary signal into multiple modal components.

[0094] S105. Perform adaptive mode extraction of composite faults on the multiple modal components to generate target mode sequences corresponding to each bearing fault type.

[0095] It should be noted that adaptively extracting effective fault components from the modal components obtained from signal decomposition is the key to achieving composite fault separation. To this end, an adaptive modal extraction mechanism is proposed.

[0096] In some embodiments, step S105 includes: using AHI to perform preliminary screening of modal components, removing redundant components below a threshold, and retaining effective components; constructing the envelope order spectrum of the effective components, extracting the order corresponding to the maximum energy peak in each envelope order spectrum, matching and classifying the order corresponding to the maximum energy peak with the calibrated fault characteristic order, and classifying each effective modal component into the corresponding bearing fault type according to the matching and classification results; introducing the envelope order harmonic signal-to-noise ratio (EOHNR) as an optimization criterion, calculating the EOHNR values ​​of all candidate modal components corresponding to each type of fault, and selecting the modal component with the largest signal-to-noise ratio from the candidate components of each type of fault; combining the modal components with the largest signal-to-noise ratio selected under each type of fault to generate a target modal sequence corresponding to each bearing fault type.

[0097] Specifically, AHI is used for preliminary screening of modal components. By setting an AHI threshold, redundant components below the threshold are eliminated, thus retaining the effective components.

[0098] Construct the envelope order spectrum of the effective components and extract the order of the maximum energy peak, ôq. If the absolute deviation of ôq from FCO is within the tolerance δ, then the mode is classified into the corresponding fault type Ωi. The mathematical definition of this classification process is as follows:

[0099] Envelope order harmonic signal-to-noise ratio (EOHNR) is introduced as the final screening criterion. This index is defined as the ratio of the energy of the fault characteristic harmonics to the energy of the local background noise in the order spectrum. It not only highlights the periodic impact characteristics of the fault harmonics but also quantitatively characterizes their intensity. By calculating the EOHNR values ​​of each candidate mode component and selecting the maximum value, the optimal mode sequence is adaptively generated. Its mathematical definition is as follows:

[0100] Where Etotal is the total energy of FCO and its harmonics, and Enoise is the corresponding noise energy.

[0101] S106. Based on the target mode sequence, order matched filters are introduced to suppress residual background noise, and the fault characteristic order and harmonic components of each fault type are extracted to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

[0102] It should be noted that, to further suppress residual noise and enhance fault characteristics, an order matched filter (OMF) is introduced. This is achieved by constructing a passband function to accurately match the FCO and its harmonic components; its mathematical definition is:

[0103] Where r is the harmonic index, σ is the tolerance bandwidth, and c is the noise suppression coefficient.

[0104] Furthermore, this embodiment is illustrated with an experimental example. The signal used in the experiment is the composite fault vibration signal of a variable-speed bearing acquired by a fault test platform. The sampling frequency for signal acquisition was set to 25.6 kHz, the sampling duration was 10.24 s, and the number of sampling points was 262,144. The composite fault vibration signal of the first 1 second was extracted for subsequent analysis. Based on the specific parameters of the bearing under test, the theoretical fault characteristic orders of the inner and outer rings were calculated to be 5.4256 and 3.5744, respectively. The speed change curve during the experiment is shown below. Figure 4 As shown.

[0105] To simulate strong noise conditions, Gaussian white noise with a signal-to-noise ratio of -5dB was added to the composite fault vibration signal. Figure 5 and Figure 6The time-domain waveform and envelope order spectrum of the noisy composite fault signal are shown below. Subsequently, the signal is adaptively decomposed using IHOA-EAFMD. The initialization parameters for the experimental analysis are consistent with those for the simulation analysis. The fitness convergence curve during the optimization process is shown below. Figure 7 As shown.

[0106] As shown by the convergence curve, the algorithm converges in the 8th iteration, and the optimal parameter combination is [7, 95]. Using this parameter combination, EAFMD is performed, and finally, 7 independent modal components are decomposed.

[0107] Based on the above decomposition results, an adaptive modal recognition mechanism is used to select the optimal feature modes of each fault, and OMF is used for feature enhancement. Figure 8 and Figure 9 The figures show the time-domain plot of the optimal mode of outer ring fault and the order plot of the envelope spectrum after processing by the proposed method. Figure 10 and Figure 11 These are the time-domain plot and envelope spectrum order plot of the optimal mode of inner-circle fault after processing by the proposed method. Figures 8-11 As shown, the FCO and its harmonic order spectral peaks of each fault are clearly visible. This proves that the proposed method has the ability to accurately extract the characteristics of complex faults under variable speed and high noise conditions.

[0108] This embodiment acquires the original vibration and rotational speed signals of the wind turbine bearing and preprocesses the original vibration signal to reduce deterministic interference components. Based on the rotational speed signal, the preprocessed vibration signal is converted into a stationary signal in the angular domain, and the fault feature order is calibrated through a coarse-to-fine hierarchical search strategy to overcome the problem of spectral ambiguity under variable speed conditions. Furthermore, with maximizing the average Hoyer exponent as the optimization objective, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD). The average Hoyer exponent serves as a normalized measure of the sparsity of signals in each period within the angular domain, effectively overcoming the limitation of traditional time-domain evaluation indicators failing under variable speed conditions. The improved Hippo Optimization Algorithm utilizes Latin hypercube sampling and Tent chaotic inertial weighting. The fusion of three strategies—repetition, progressive refraction, and mirror learning—significantly enhances global exploration and local development capabilities, reducing the risk of traditional Hippo optimization algorithms getting trapped in local optima. Based on the optimal number and length of filters obtained through optimization, EAFMD is used to decompose the angular domain stationary signal into multiple modal components. Subsequently, adaptive mode extraction of composite faults is performed on multiple modal components. Through a three-level screening mechanism—initial screening using the average Hoyer exponent, envelope order spectrum matching and classification, and envelope order harmonic signal-to-noise ratio optimization—target mode sequences corresponding to each bearing fault type are adaptively generated, solving the problem of the number of modes depending on prior parameter settings during signal decomposition. Finally, order-matched filters are introduced based on the target mode sequences to suppress residual background noise and extract the fault characteristic order and its harmonic components for each fault type. This successfully solved the technical challenge of accurately separating and extracting composite fault features and reducing fault diagnosis accuracy under variable speed and high noise conditions, which is caused by the dependence of filter parameters on prior settings and the failure of time-domain evaluation indicators. It achieved high-precision and adaptive diagnosis of composite faults in wind turbine bearings under variable speed and high noise conditions, significantly improving the reliability and robustness of wind power equipment condition monitoring.

[0109] Based on the same idea as the wind turbine bearing composite fault diagnosis method in the above embodiments, this application also provides a wind turbine bearing composite fault diagnosis device 1200, which can be used to perform the above-described wind turbine bearing composite fault diagnosis method. Figure 12 As shown, the module comprises a signal acquisition and preprocessing module 1201, an angular domain conversion and order calibration module 1202, a parameter adaptive optimization module 1203, a signal decomposition module 1204, a modality adaptive extraction module 1205, and a feature enhancement and diagnosis module 1206. In some embodiments, the above modules can be programmable software instructions stored in memory and executable by a processor. It is understood that in other embodiments, the above modules can also be program instructions or firmware embedded in the processor.

[0110] The signal acquisition and preprocessing module 1201 is used to acquire the original vibration signal and speed signal of the wind turbine bearing, and preprocess the original vibration signal to obtain the preprocessed vibration signal. The angular domain conversion and order calibration module 1202 is used to convert the preprocessed vibration signal into an angular domain stationary signal based on the rotation speed signal, and to obtain the calibrated fault characteristic order by calibrating the fault characteristic order. The parameter adaptive optimization module 1203 is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) with the goal of maximizing the average Hoyer exponent and using the improved Hippo Optimization Algorithm (IHOA) to obtain the target parameter combination. The average Hoyer exponent is the arithmetic mean of the Hoyer exponents of each periodic signal in the angular domain. The target parameter combination includes the optimal number of filters and the optimal filter length. The signal decomposition module 1204 is used to decompose the angular domain stationary signal according to the target parameter combination using the EAFMD to obtain multiple modal components; The adaptive mode extraction module 1205 is used to perform adaptive mode extraction of composite faults on the multiple modal components and generate target mode sequences corresponding to each bearing fault type. The feature enhancement and diagnosis module 1206 is used to introduce order matched filters based on the target mode sequence to suppress residual background noise, extract the fault feature order and harmonic components of each fault type, so as to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

[0111] Please refer to Figure 13 , Figure 13 This is a schematic diagram of an embodiment of the electronic device of this application. In this embodiment of the invention, the electronic device 1300 includes a processor 1301, a memory 1302, a display 1303, and a signal acquisition device 1304. Figure 13 Only some components of the electronic device 1300 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0112] In some embodiments, processor 1301 may be a central processing unit (CPU), microprocessor or other data processing chip, used to run program code stored in memory 1302 or process data, such as the wind turbine bearing composite fault diagnosis method of the present invention.

[0113] In some embodiments, display 1303 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 1303 is used to display information from electronic device 1300 and to display visual user applications. Components 1301-1303 of electronic device 1300 communicate with each other via a system bus.

[0114] In one embodiment, when the processor 1301 executes the wind turbine bearing composite fault diagnosis program in the memory 1302, the following steps can be implemented: The original vibration signal and rotational speed signal of the wind turbine bearing are acquired, and the original vibration signal is preprocessed to obtain the preprocessed vibration signal. Based on the rotational speed signal, the preprocessed vibration signal is converted into an angular domain stationary signal, and the fault characteristic order is calibrated to obtain the calibrated fault characteristic order. With maximizing the average Hoyer exponent as the optimization objective, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) to obtain the target parameter combination. The average Hoyer exponent is the arithmetic mean of the Hoyer exponents of each periodic signal in the angular domain. The target parameter combination includes the optimal number of filters and the optimal filter length. Based on the target parameter combination, the angular domain stationary signal is decomposed using the EAFMD to obtain multiple modal components; Adaptive mode extraction of composite faults is performed on the multiple modal components to generate target mode sequences corresponding to each bearing fault type; Based on the target mode sequence, order matched filters are introduced to suppress residual background noise, and the fault characteristic order and harmonic components of each fault type are extracted to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

[0115] It should be understood that when the processor 1301 executes the aero-optical effect sequence image restoration program in the memory 1302, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0116] Furthermore, the embodiments of the present invention do not specifically limit the type of the electronic device 1300 mentioned. The electronic device 1300 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 1300 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0117] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0118] The above provides a detailed description of the wind turbine bearing composite fault diagnosis method, device, electronic equipment, and system provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for diagnosing a combined fault of a fan bearing, characterized by, include: The original vibration signal and rotational speed signal of the wind turbine bearing are acquired, and the original vibration signal is preprocessed to obtain the preprocessed vibration signal. Based on the rotational speed signal, the preprocessed vibration signal is converted into an angular domain stationary signal, and the fault characteristic order is calibrated to obtain the calibrated fault characteristic order. With maximizing the average Hoyer exponent as the optimization objective, the improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) to obtain the target parameter combination. The average Hoyer exponent is the arithmetic mean of the Hoyer exponents of each periodic signal in the angular domain. The target parameter combination includes the optimal number of filters and the optimal filter length. Based on the target parameter combination, the angular domain stationary signal is decomposed using the EAFMD to obtain multiple modal components; Adaptive mode extraction of composite faults is performed on the multiple modal components to generate target mode sequences corresponding to each bearing fault type; Based on the target mode sequence, order matched filters are introduced to suppress residual background noise, and the fault characteristic order and harmonic components of each fault type are extracted to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

2. The fan bearing compound fault diagnosis method according to claim 1, characterized in that, The preprocessing of the original vibration signal includes: Perform a fast Fourier transform on the original vibration signal to obtain a complex spectrum, which includes an amplitude spectrum and a phase spectrum; A whitening filter is constructed, and the amplitude spectrum is whitened using the whitening filter to obtain the whitened frequency domain signal; Based on the original phase spectrum and the whitened amplitude spectrum, the time-domain signal is reconstructed by inverse fast Fourier transform, and the energy of the reconstructed signal is restored.

3. The fan bearing compound fault diagnosis method according to claim 1, characterized in that, The process of converting the preprocessed vibration signal into a stationary signal in the angular domain based on the rotational speed signal specifically includes: Based on the instantaneous rotational speed values ​​corresponding to each sampling point in the rotational speed signal, a discrete phase accumulation sequence is established by integrating the accumulated rotational angles; The preprocessed vibration signal is resampled at equal angles based on the discrete phase accumulation sequence to convert the preprocessed vibration signal into a stationary signal in the angular domain.

4. The fan bearing compound fault diagnosis method according to claim 1, characterized in that, The improved Hippo Optimization Algorithm (IHOA) is used to adaptively optimize the core parameters of Enhanced Angular Domain Feature Mode Decomposition (EAFMD) to obtain a combination of target parameters, including: Initialize the improved Hippo Optimization Algorithm (IHOA) with its population size, maximum number of iterations, problem dimension, and search boundaries for the number and length of filters. Based on the parameter combination of the number of filters and the filter length corresponding to the current individual, the filter bank is initialized using the maximum entropy principle, and the coefficients of the filter bank are updated using the differential gradient model. Calculate the average Hoyer index value for each modal component, and define the maximum average Hoyer index value as the fitness value of the current individual; The fitness value of the current individual is compared with the individual's historical best value and the global historical best value to update the local best position of the individual and the global best position of the population. Tent chaotic inertial weights and progressive refraction mirror learning are used to update the position of the hippopotamus individual, respectively. Determine if the termination condition is met. If not, increment the iteration count and re-execute the filter bank coefficient update step. If the condition is met, output the target parameter combination, which includes the optimal number of filters and the optimal filter length.

5. The fan bearing compound fault diagnosis method according to claim 4, characterized in that, The method of updating the position of individual hippos using Tent chaotic inertial weights and progressive refraction mirror learning includes: Using the current individual's position and fitness value as input, chaotic variables are generated using Tent mapping. The amplitude of the chaotic variables is modulated by a decay factor to construct adaptive inertial weights. The adaptive inertial weights are substituted into the current individual's position update process to generate the updated individual position. The fitness value is recalculated based on the updated individual position. The recalculated fitness value is compared with the individual's historical best value and the global historical best value to update the individual's local best position and the population's global best position. Taking the global optimal position and fitness value under the current iteration as input, a mirror-reverse solution is generated based on the law of refraction of light to calculate the fitness value of the mirror-reverse solution. The fitness value of the mirror-reverse solution is compared with the fitness value of the global optimal position. If the fitness value of the mirror-reverse solution is greater than the fitness value of the global optimal position, the mirror-reverse solution is updated to the new global optimal position; otherwise, the current global optimal position remains unchanged.

6. The fan bearing compound fault diagnosis method according to claim 1, characterized in that, The step involves decomposing the angular-domain stationary signal using the EAFMD based on the target parameter combination to obtain multiple modal components, including: Initialize the filter bank based on the optimal number of filters and the optimal filter length in the target parameter combination; A differential gradient model of the filter bank is constructed, and the coefficients of the filter bank are iteratively updated through a multi-scale step size adjustment strategy until the convergence condition is met. In each iteration, the angular domain stationary signal is filtered according to the current filter coefficients to separate each modal component, and the average Hoyer exponent of each modal component is calculated. The iterative updates cause the filter bank to converge in the direction of maximizing the average Hoyer exponent, thereby decomposing the angular domain stationary signal into multiple modal components.

7. The fan bearing compound fault diagnosis method according to claim 1, characterized in that, The adaptive mode extraction of composite faults from the multiple modal components to generate target mode sequences corresponding to each bearing fault type includes: The modal components are initially screened using the average Hoyer index, eliminating redundant components below the threshold and retaining valid components. Construct the envelope order spectrum of the effective components, extract the order corresponding to the maximum energy peak in each envelope order spectrum, match and classify the order corresponding to the maximum energy peak with the calibrated fault feature order, and classify each effective modal component into the corresponding bearing fault type according to the matching and classification results. Calculate the envelope order harmonic signal-to-noise ratio (SNR) of all candidate modal components corresponding to each type of fault, and select the modal component with the largest SNR from the candidate components of each type of fault. The mode components with the highest signal-to-noise ratio selected under various fault conditions are combined to generate target mode sequences corresponding to each bearing fault type.

8. A fan bearing compound fault diagnosis device, characterized by, include: The signal acquisition and preprocessing module is used to acquire the original vibration signal and speed signal of the wind turbine bearing, and to preprocess the original vibration signal to obtain the preprocessed vibration signal. The angular domain conversion and order calibration module is used to convert the preprocessed vibration signal into an angular domain stationary signal based on the rotational speed signal, and to obtain the calibrated fault characteristic order by calibrating the fault characteristic order. The parameter adaptive optimization module is used to adaptively optimize the core parameters of Enhanced Angular Domain Eigenmode Decomposition (EAFMD) with the goal of maximizing the average Hoyer exponent. The target parameter combination is obtained by using the improved Hippo Optimization Algorithm (IHOA) to optimize the average Hoyer exponent of each periodic signal in the angular domain. The target parameter combination includes the optimal number of filters and the optimal filter length. The signal decomposition module is used to decompose the angular domain stationary signal according to the target parameter combination using the EAFMD to obtain multiple modal components; The adaptive mode extraction module is used to perform adaptive mode extraction of the multiple modal components for composite faults, and generate target mode sequences corresponding to each bearing fault type. The feature enhancement and diagnosis module is used to introduce order matched filters based on the target mode sequence to suppress residual background noise, extract the fault feature order and harmonic components of each fault type, so as to realize the diagnosis of composite faults of wind turbine bearings under variable speed and strong noise conditions.

9. An electronic device, comprising: It includes a signal acquisition unit, a memory, and a processor, wherein the signal acquisition unit is used to acquire the original vibration signal and rotational speed signal of the wind turbine bearing; The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the wind turbine bearing composite fault diagnosis method according to any one of claims 1 to 7.

10. A fan bearing compound fault diagnosis system, characterized in that, The wind turbine bearing composite fault diagnosis system includes the electronic device, data acquisition device, and result output device as described in claim 9. The data acquisition device and the result output device are respectively connected to the electronic device. The data acquisition device is used to acquire the original vibration signal and speed signal of the wind turbine bearing. The result output device is used to output and display the extended wind turbine bearing composite fault diagnosis results.

Images