Wind turbine generator gearbox fault diagnosis method and device and computer equipment

Abstract

The invention relates to the technical field of wind power generation, and discloses a wind turbine generator gearbox fault diagnosis method and device and computer equipment, and the method comprises the steps: collecting multi-source data of a wind turbine generator gearbox; respectively extracting a time domain feature, a frequency domain feature, a temperature feature and a working condition feature of the multi-source data; performing weighted feature fusion on the time domain feature, the frequency domain feature, the temperature feature and the working condition feature to obtain a real-time fusion feature; obtaining historical health data of the wind turbine generator gearbox, establishing a health degree reference of the wind turbine generator gearbox based on the historical health data, and constructing a health degree model based on the real-time fusion features and the health degree reference; and performing fault diagnosis on the wind turbine generator gearbox by adopting the health degree model. According to the method, multi-source data of the gearbox of the wind turbine generator is combined, and gearbox health state evaluation and fault type identification are realized through dynamic feature fusion and deep learning modeling.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and particularly to a fault diagnosis method, device and computer equipment for a wind turbine gearbox. Background Art

[0002] The existing fault diagnosis technologies for wind turbine gearboxes mainly include the following two categories: 1. Single sensor analysis method: ① Vibration signal analysis: Detect gear wear and bearing damage through time-frequency domain features (such as root mean square (RMS), kurtosis, and spectral energy), but it is sensitive to environmental noise and difficult to distinguish fault modes under complex working conditions.

[0003] ② Temperature monitoring method: Judge lubrication failure or abnormal friction based on the temperature change trend, but the response is lagging and it cannot capture instantaneous impact faults.

[0004] ③ SCADA (Supervisory Control and Data Acquisition) parameter threshold method: Set fixed threshold alarms using parameters such as power and speed, but lack multi-parameter coupling analysis and have a high false alarm rate.

[0005] 2. Simple data fusion method: Adopt static weight to fuse features such as vibration and temperature, but do not consider the influence of dynamic changes in working conditions on feature sensitivity, resulting in insufficient robustness of the fusion index.

[0006] Based on this, there is an urgent need for a fault diagnosis method for wind turbine gearboxes based on multi-source data fusion. Summary of the Invention

[0007] In view of this, the present invention provides a fault diagnosis method, device and computer equipment for a wind turbine gearbox to solve the problem of low accuracy in fault diagnosis of wind turbine gearboxes due to the lack of multi-source data fusion in the prior art.

[0008] In a first aspect, the present invention provides a fault diagnosis method for a wind turbine gearbox, the method comprising: Collect multi-source data of the wind turbine gearbox; Extract the time domain features, frequency domain features, temperature features and working condition features of the multi-source data respectively; Perform weighted feature fusion on the time domain features, frequency domain features, temperature features and working condition features to obtain real-time fusion features; Obtain the historical health data of the wind turbine gearbox, establish a health benchmark for the wind turbine gearbox based on the historical health data, and construct a health model based on the real-time fusion features and the health benchmark; Fault diagnosis of the gearbox of a wind turbine is carried out using a health model.

[0009] A fault diagnosis method for the gearbox of a wind turbine provided by the present invention collects multi-source data of the gearbox of the wind turbine, avoiding the limitations of a single data source, providing a rich and comprehensive data basis for fault diagnosis, extracting features from four dimensions of time domain, frequency domain, temperature and working conditions respectively, constructing a complete fault-sensitive feature system, realizing dynamic weighted feature fusion, automatically allocating weights according to the concentration degree of feature distribution, highlighting the contribution degree of fault-sensitive features, establishing a health benchmark using historical health data, and constructing a health model based on real-time fusion features and the health benchmark, which can accurately evaluate the health state of the gearbox, provide a scientific basis for fault diagnosis, and solve the problem of low accuracy of fault diagnosis of the gearbox of a wind turbine due to the lack of multi-source data fusion in the prior art.

[0010] In an optional implementation manner, the multi-source data includes vibration signals, temperature signals and working condition parameters; Collecting multi-source data of the gearbox of a wind turbine includes: Arranging three-axis vibration sensors on the high-speed shaft, low-speed shaft and planet carrier of the gearbox of the wind turbine, and using the three-axis vibration sensors to collect the vibration signals of the gearbox of the wind turbine; Arranging temperature sensors on the bearing seat and the wall of the gearbox near the meshing gears of the wind turbine, and using the temperature sensors to collect the temperature signals of the gearbox of the wind turbine; Using a SCADA monitoring system to collect the working condition parameters of the wind turbine, and the working condition parameters include rotational speed, power and torque; Arranging a torque sensor at the coupling of the low-speed shaft of the gearbox of the wind turbine, and determining an adaptive sampling frequency adjustment strategy based on the rotational speed and torque using the torque sensor.

[0011] A fault diagnosis method for the gearbox of a wind turbine provided by the present invention arranges various types of sensors on the high-speed shaft, low-speed shaft and planet carrier of the gearbox, synchronously collects multi-source data such as vibration and temperature, combines with SCADA working condition parameters, and comprehensively covers the operation state information of the gearbox. Different types of data complement each other, avoiding the limitations of a single data source, providing a rich and comprehensive data basis for fault diagnosis, proposing a sensor arrangement scheme, combining micro sensors to solve the problem of space limitation of the planet carrier, and designing a dynamic sampling frequency adjustment strategy based on rotational speed and torque to balance data accuracy and storage cost.

[0012] In an optional implementation manner, before respectively extracting the time-domain features, frequency-domain features, temperature features and working condition features of the multi-source data, the method further includes: The CEEMDAN algorithm is used to reduce the noise of the vibration signal, and the preset formula is used to standardize the vibration signal, temperature signal and working condition parameters, where: When the CEEMDAN algorithm is used to reduce the noise of vibration signals, a preset speed synchronous noise template is used to suppress the interference components related to the vibration signal and the gear meshing frequency, and the physical and statistical double criteria are used to screen the intrinsic mode functions corresponding to the vibration signals.

[0013] The present invention provides a wind turbine gearbox fault diagnosis method, which introduces a speed synchronization noise template, directionally suppresses interference components related to the gear meshing frequency, and uses physical and statistical double criteria (correlation coefficient + kurtosis) to screen effective IMFs to improve the impact signal retention capability.

[0014] In an optional implementation, time domain features, frequency domain features, temperature features, and operating condition features of multi-source data are extracted separately, including: Extract the time domain features and frequency domain features of the vibration signal; the time domain features include the root mean square value and kurtosis, and the frequency domain features include the spectrum energy integral and the sideband energy ratio; Extracting temperature features of the temperature signal, the temperature features include temperature rise rate and temperature distribution uniformity index; The operating characteristics of the operating parameters are extracted, and the operating characteristics include load-speed coupling factor and power fluctuation entropy.

[0015] The present invention provides a wind turbine gearbox fault diagnosis method, which extracts features from four dimensions: time domain, frequency domain, temperature and operating conditions, and constructs a complete fault-sensitive feature system. Time domain features (root mean square value, kurtosis) are sensitive to uniform wear and local impact of the gearbox; frequency domain features (spectral energy integral, sideband energy ratio) can accurately capture specific frequency faults and modulation phenomena; temperature features (temperature rise rate, temperature distribution uniformity index) can effectively reflect lubrication status and cooling system faults; operating condition features (load-speed coupling factor, power fluctuation entropy) can characterize mechanical transmission efficiency and power stability. This multi-dimensional feature extraction method can deeply mine the fault information contained in the data and improve the sensitivity and accuracy of fault diagnosis.

[0016] In an optional implementation, weighted feature fusion is performed on time domain features, frequency domain features, temperature features, and operating condition features to obtain real-time fusion features, including: Normalize the time domain features, frequency domain features, temperature features and operating condition features respectively; Calculate the characteristic information entropy corresponding to the normalized time domain characteristics, frequency domain characteristics, temperature characteristics and working condition characteristics, and perform entropy weight method weight calculation based on the characteristic information entropy of each characteristic to obtain the weight corresponding to each characteristic; Perform weighted feature linear fusion on time-domain features, frequency-domain features, temperature features, operating conditions features, and their corresponding weights to obtain real-time fusion features.

[0017] A fault diagnosis method for a wind turbine gearbox provided by the present invention realizes dynamic weighted feature fusion based on the information entropy theory, automatically assigns weights according to the concentration degree of feature distribution, and highlights the contribution degree of features sensitive to faults. The weights are adjusted in real time according to the data changes, which can adapt to the changes in the importance of features under different operating conditions and fault scenarios. Compared with the fixed-weight fusion method, it can more accurately reflect the actual operating state of the gearbox. At the same time, a comprehensive fault-sensitive index is generated through normalization processing and weighted linear fusion, reducing the influence of dimension differences and enhancing the scientificity and practicability of feature fusion.

[0018] In an alternative embodiment, a health benchmark of the wind turbine gearbox is established based on historical health data, and a health model is constructed based on the real-time fusion features and the health benchmark, including: Use historical health data to train a preset gated recurrent unit model to obtain the health benchmark of the wind turbine gearbox; Calculate the health score based on the real-time fusion features, the health benchmark, and the preset health state tolerance threshold, and calculate the change trend of the health score within a preset time period; Use the health score and its change trend, as well as the preset expert rule base as the fault type decision rule, and construct a health model based on the fault type decision rule.

[0019] A fault diagnosis method for a wind turbine gearbox provided by the present invention uses historical health data to train a GRU model to establish a health benchmark. The GRU model can effectively process long sequence data and capture the time change law of the gearbox operating state. Set strict training constraints, including appropriate loss functions, sufficient training data volume, and overfitting control strategies, to ensure the reliability and generalization ability of the model. Through real-time health score calculation, compare the real-time fusion features with the health benchmark, and combine the dynamically adjusted threshold to accurately evaluate the health state of the gearbox and provide a scientific basis for fault diagnosis.

[0020] In an alternative embodiment, the health model is used to diagnose faults in the wind turbine gearbox, including: Obtain the real-time multi-source data of the wind turbine gearbox and input the real-time multi-source data of the wind turbine gearbox into the health model to obtain the predicted health score; When the predicted health score is less than the preset alarm threshold, trigger an alarm and prompt the existence of a fault risk; Dynamically update the real-time feature weights based on the weight learning rate, historical fusion features, and real-time fusion features, and calculate the real-time feature weight vector; Construct a standard library of fault types for the gearbox of a wind turbine generator set, calculate the cosine similarity between the real-time feature weight vector and the standard weight patterns of each fault type in the standard library of fault types, and take the fault type with the highest cosine similarity as the fault diagnosis result.

[0021] A fault diagnosis method for the gearbox of a wind turbine generator set provided by the present invention can collect data in real time and calculate the health score, which can timely detect abnormal changes in the operating state of the gearbox and achieve early warning of faults. Combining threshold alarm and fault mode matching based on cosine similarity, and using an expert rule library for fault type decision-making, it not only considers the absolute value of the health score but also pays attention to its change trend. At the same time, by comparing with the standard weight pattern, the accuracy and reliability of fault diagnosis are improved. In an optional implementation manner, the fault diagnosis method for the gearbox of a wind turbine generator set further includes: Updating the fault type decision rule and the health model by using the newly diagnosed fault type.

[0022] A fault diagnosis method for the gearbox of a wind turbine generator set provided by the present invention. The online self-optimization mechanism of the model can update the model parameters and decision rules by using the newly diagnosed fault data, form a diagnostic closed loop, continuously improve the diagnostic accuracy, and adapt to changes in equipment performance and newly emerging fault types.

[0023] In a second aspect, the present invention provides a fault diagnosis device for the gearbox of a wind turbine generator set, and the device includes: A data acquisition module for acquiring multi-source data of the gearbox of a wind turbine generator set; A feature extraction module for respectively extracting the time-domain features, frequency-domain features, temperature features, and operating condition features of the multi-source data; A feature fusion module for performing weighted feature fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features to obtain real-time fusion features; A health model construction model for obtaining the historical health data of the gearbox of a wind turbine generator set, establishing a health benchmark for the gearbox of a wind turbine generator set based on the historical health data, and constructing a health model based on the real-time fusion features and the health benchmark; A fault diagnosis module for performing fault diagnosis on the gearbox of a wind turbine generator set by using the health model.

[0024] In a third aspect, the present invention provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the fault diagnosis method for the gearbox of a wind turbine generator set according to the first aspect or any corresponding implementation manner thereof.

[0025] Fourthly, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute the wind turbine gearbox fault diagnosis method according to the first aspect or any corresponding embodiment thereof as described above.

[0026] Fifthly, the present invention provides a computer program product, including computer instructions, and the computer instructions are used to cause a computer to execute the wind turbine gearbox fault diagnosis method according to the first aspect or any corresponding embodiment thereof as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0028] Figure 1 is a flowchart of the wind turbine gearbox fault diagnosis method according to an embodiment of the present invention; Figure 2 is a flowchart of another wind turbine gearbox fault diagnosis method according to an embodiment of the present invention; Figure 3 is a flowchart of yet another wind turbine gearbox fault diagnosis method according to an embodiment of the present invention; Figure 4 is a structural block diagram of the wind turbine gearbox fault diagnosis device according to an embodiment of the present invention; Figure 5 is a schematic hardware structure diagram of the computer device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] In order to make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0030] As the core transmission component of the wind turbine, the failure rate of the wind power gearbox accounts for more than 30% of the total failures of the whole machine. The disadvantages of the prior art include: ① Poor data synchronization: The traditional method fails to achieve strict time alignment of vibration, temperature, and SCADA data, resulting in the loss of feature correlation.

[0031] ② Insufficient noise suppression: Modal aliasing problems exist in vibration signal denoising methods (such as traditional EMD), affecting the extraction of high-frequency impact components.

[0032] ③ Rigid feature fusion: Fixed weight allocation cannot adapt to changes in feature importance under variable speed and variable load conditions, reducing diagnostic sensitivity.

[0033] ④ Limitations in health modeling: Health assessment based on static thresholds or shallow models is difficult to capture the non-linear time-series characteristics of the gearbox degradation process.

[0034] This embodiment provides a fault diagnosis method for a wind turbine gearbox. Through high-precision synchronous acquisition and noise reduction of multi-source heterogeneous data (vibration, temperature, torque, speed, power), dynamic extraction and fusion of fault-sensitive features under complex working conditions, and real-time quantitative assessment of the health status of the gearbox, the effect of accurate identification of early faults is achieved.

[0035] According to an embodiment of the present invention, an embodiment of a fault diagnosis method for a wind turbine gearbox is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0036] In this embodiment, a fault diagnosis method for a wind turbine gearbox is provided, which can be used for wind turbines. Figure 1 It is a flowchart of a fault diagnosis method for a wind turbine gearbox according to an embodiment of the present invention. As Figure 1 shown, the process includes the following steps: Step S101, collect multi-source data of the wind turbine gearbox.

[0037] Specifically, three-axis vibration sensors, temperature sensors, and torque sensors are arranged on the high-speed shaft, low-speed shaft, and planet carrier of the wind turbine gearbox, and the sampling frequency is adaptively adjusted (1 kHz - 50 kHz). Through vibration acceleration sensors, infrared temperature sensors, torque sensors, and the SCADA system, vibration signals, temperature signals, torque signals, and SCADA operating conditions parameters of the wind turbine are obtained.

[0038] Step S102, extract the time-domain features, frequency-domain features, temperature features, and operating conditions features of the multi-source data respectively.

[0039] Specifically, time-domain features and frequency-domain features are extracted from the vibration signal, temperature features are extracted from the temperature signal, and operating conditions features are extracted from the SCADA operating conditions parameters.

[0040] Step S103: Perform weighted feature fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features to obtain real-time fusion features.

[0041] Specifically, weighted feature fusion refers to the process in the fault diagnosis of wind turbine gearboxes of extracting time-domain, frequency-domain, temperature, and operating condition features from multi-source data such as vibration, temperature, and operating conditions, assigning corresponding weights according to the importance of each feature for fault diagnosis, and then performing a linear combination to generate a comprehensive fault sensitivity index.

[0042] That is, after normalizing the time-domain features, frequency-domain features, temperature features, and operating condition features, the weights of each feature are dynamically calculated based on the information entropy theory to highlight the contribution degree of the fault-sensitive features, and the normalized features are linearly fused according to the weights to generate a comprehensive fault sensitivity index, that is, the real-time fusion features are obtained.

[0043] Step S104: Obtain the historical health data of the wind turbine gearbox, establish a health benchmark for the wind turbine gearbox based on the historical health data, and construct a health model based on the real-time fusion features and the health benchmark.

[0044] Specifically, the health benchmark is a key reference standard for measuring the operating state of equipment in the fault diagnosis of wind turbine gearboxes, providing a quantitative basis for fault diagnosis. Specifically, the health benchmark refers to the quantitative standard of the normal operating state established based on the historical healthy operation data of the equipment in the fault diagnosis system of the wind turbine gearbox, which is used to measure the deviation degree of the current equipment operating state from the ideal healthy state.

[0045] Train a Gated Recurrent Unit (GRU) model based on the historical health data to establish a health benchmark for the normal operating state of the gearbox. Compare the real-time fusion features with the health benchmark, calculate the health score, and finally construct a health model based on the health score.

[0046] Step S105: Use the health model to perform fault diagnosis on the wind turbine gearbox.

[0047] Specifically, input the multi-source data of the wind turbine gearbox collected in real time into the health model to obtain the health score, diagnose whether there is a fault based on the relationship between the health score and the preset score threshold. If there is a fault, use the fault mode matching to diagnose the fault type, so as to achieve the purpose of accurately evaluating the health state of the gearbox.

[0048] The fault diagnosis method for the gearbox of a wind turbine provided in this embodiment collects multi-source data of the gearbox of the wind turbine, avoiding the limitations of a single data source, providing a rich and comprehensive data basis for fault diagnosis, extracting features from four dimensions: time domain, frequency domain, temperature, and working condition, constructing a complete fault-sensitive feature system, realizing dynamic weighted feature fusion, automatically allocating weights according to the concentration degree of feature distribution, highlighting the contribution degree of fault-sensitive features, establishing a health benchmark using historical health data, and constructing a health model based on real-time fusion features and the health benchmark, which can accurately evaluate the health state of the gearbox, provide a scientific basis for fault diagnosis, and solve the problem of low accuracy in fault diagnosis of the gearbox of a wind turbine due to the lack of multi-source data fusion in the prior art.

[0049] In this embodiment, a fault diagnosis method for the gearbox of a wind turbine is provided, which can be used for a wind turbine. Figure 2 It is a flowchart of the fault diagnosis method for the gearbox of a wind turbine according to an embodiment of the present invention, as Figure 2 shown, and this process includes the following steps: Step S201, collect multi-source data of the gearbox of the wind turbine.

[0050] Specifically, the multi-source data includes vibration signals, temperature signals, and working condition parameters; three-axis vibration sensors (vibration acceleration sensors) are arranged on the high-speed shaft, low-speed shaft, and planet carrier of the gearbox of the wind turbine, temperature sensors (infrared temperature sensors) are arranged on the bearing seat and the wall of the gearbox near the meshing gears of the wind turbine, a torque sensor is arranged at the coupling of the low-speed shaft of the gearbox of the wind turbine, and the sampling frequency is adaptively adjusted (1 kHz - 50 kHz). Through the vibration acceleration sensors, infrared temperature sensors, torque sensors, and SCADA monitoring system, the original data of the gearbox of the wind turbine is obtained, including: Vibration signals: , sampling frequency = 12.8 kHz; Temperature signal: T(t), sampling interval = 1 second; SCADA working condition parameters: collect the rotational speed R (rpm), power P (MW), and load torque L (kN·m); All the above data are marked with a unified timestamp , time window length Δt = 10 minutes.

[0051] The sensor selection and parameter configuration are shown in Table 1 below: Table 1 Sensor Selection and Parameter Configuration

[0052] The above step S201 includes: Step S2011, arrange triaxial vibration sensors on the high-speed shaft, low-speed shaft and planet carrier of the wind turbine gearbox, and use the triaxial vibration sensors to collect the vibration signals of the wind turbine gearbox.

[0053] Specifically, the installation of the triaxial vibration sensors includes: High-speed shaft: Installation method: magnetic base (neodymium iron boron permanent magnet) + thread reinforcement (M6 stainless steel bolt).

[0054] Axial alignment: X-axis (radial), Y-axis (axial), Z-axis (tangential), aligned with the gear rotation direction.

[0055] Anti-resonance design: attach a rubber damping pad (Shore hardness 70A) to the base of the triaxial vibration sensor to suppress high-frequency resonance.

[0056] Planet carrier: Solution for limited space: miniature triaxial sensor (size 15×10×10mm), fixed by epoxy resin adhesive, with built-in temperature compensation module.

[0057] Step S2012, arrange temperature sensors on the bearing housing and the wall of the gearbox near the meshing gears of the wind turbine gearbox, and use the temperature sensors to collect the temperature signals of the wind turbine gearbox.

[0058] Specifically, the installation of the temperature sensors includes: Use armored PT100 temperature sensors; Installation position on the bearing housing: Install at the position near the rolling elements of the bearing housing. It is recommended to install on the side or bottom of the bearing housing, and evenly apply a layer of thermal conductive silicone grease (thermal conductivity ); Gear meshing surface: Select on the wall of the gearbox near the gear meshing area, and use thermal conductive adhesive (thermal conductivity ) to closely paste the temperature sensor to the wall of the gearbox.

[0059] Step S2013, use the SCADA monitoring system to collect the operating condition parameters of the wind turbine, and the operating condition parameters include speed, power and torque.

[0060] Specifically, the speed is represented by and the torque is represented by .

[0061] Step S2014, arrange a torque sensor at the low-speed shaft coupling of the wind turbine gearbox, and based on the speed and torque, use the torque sensor to determine the adaptive sampling frequency adjustment strategy.

[0062] Specifically, the integration of the torque sensor: Non-contact design: adopt the magnetoelastic torque measurement technology to avoid the signal interference of the traditional slip ring structure.

[0063] Dynamic compensation: Built-in strain gauge temperature drift compensation circuit (temperature drift < 0.005% FS / ℃).

[0064] Adopt a torque sensor to determine the adaptive sampling frequency adjustment strategy, including: Benchmark rule: Low-speed shaft (rotation speed < 30 rpm): Vibration sampling rate 1 kHz. High-speed shaft (rotation speed ≥ 1000 rpm): Vibration sampling rate 50 kHz.

[0065] Dynamic adjustment algorithm: Based on the rotational speed and torque Calculate the critical frequency, and the formula is as follows: (1); Among them, is the number of teeth, is the adjustment coefficient, is the maximum torque, and the sampling frequency is adjusted dynamically (meeting the Nyquist theorem, that is, the Nyquist sampling theorem). Triggered sampling: When the short-time energy of the vibration signal suddenly increases by > 30%, it automatically switches to the 50 kHz high-speed sampling mode and lasts for 200 ms.

[0066] Step S202, use the CEEMDAN algorithm to denoise the vibration signal, and at the same time use a preset formula to standardize the vibration signal, temperature signal, and working condition parameters, where: when using the CEEMDAN algorithm to denoise the vibration signal, use a preset rotational speed synchronous noise template to suppress the interference components related to the gear meshing frequency, and use physical statistics dual criteria to screen the intrinsic mode functions corresponding to the vibration signal.

[0067] Specifically, the process of improving the CEEMDAN algorithm for vibration signal denoising is as follows:

[0068] Step 1: Signal initialization: Input the vibration signal , set the total average number (usually ), the noise amplitude coefficient (default ), where is the standard deviation. is the standard deviation.

[0069] Step 2: Iterative decomposition: For the th iteration ( ), directional noise injection: Generate a rotational speed synchronous noise signal , the formula is as follows: (2); Fundamental frequency: (unit: Hz); Gearbox meshing frequency ; Among them, is the rotational speed synchronous noise signal, which is a function of time ; , are both summation indices, used to represent the order of harmonics; is a positive integer, representing the highest order of harmonics, used to determine the upper limit of summation, that is, considering the highest order harmonic component in the signal; , is the amplitude coefficient (determined by historical data regression), , are both random initial phases.

[0070] Generate a perturbation signal containing a rotational speed noise template: (3); EMD decomposition (Empirical Mode Decomposition): Perform EMD decomposition on to obtain the first layer of intrinsic mode functions .

[0071] Residual calculation: ; Loop decomposition: Repeat the above process for the residual ( ) until the IMF (Intrinsic Mode Function, hereinafter referred to as IMF) termination condition (std < 0.3) is met.

[0072] Step 3: Ensemble average: Take the average of the iteration results to obtain the final IMF component: ( ) (4).

[0073] Step 4: Effective IMF screening, that is, screening IMF based on physical and statistical dual criteria: Correlation coefficient criterion: (Keep ); Kurtosis criterion: (Keep ); Among them, is the d th intrinsic mode function and the original vibration signal The correlation coefficient between them ranges from [-1, +1]; , , E[ ], respectively represent covariance, standard deviation, expected value, and mean; ; .

[0074] Step 5: Signal reconstruction: Add the filtered IMFs to obtain the denoised signal : where, is the effective IMF index set.

[0075] Step 6. Key parameter optimization: Noise amplitude coefficient (Dynamic adjustment strategy): (5); where, is the frequency corresponding to the maximum allowable speed of the gearbox.

[0076] Adaptive selection of the number of iterations according to signal complexity : (6); Use the improved adaptive complete ensemble empirical mode decomposition to remove high-frequency noise from the vibration signal.

[0077] Denoised vibration signal: =CEEMDAN( , K , ε ); where, K is the number of intrinsic mode functions (IMFs) for vibration signal decomposition, automatically determined by the number of spectral peaks of the signal; ε is the noise amplitude coefficient, with an initial value =0.2, dynamically adjusted according to the signal-to-noise ratio: (7).

[0078] Use the preset formula to standardize the vibration signal, temperature signal, and operating conditions parameters to eliminate the dimension. The preset formula is as follows: Normalized temperature , , , : (8); Equation (8) is applied to all input data: ∈{ , T, , L, P}; is the signal mean value (the data normalization benchmark, calculated based on a sliding time window and recalculated every 10 minutes), and is calculated in real time as the arithmetic mean of the data within the time window Δ t : = (9); wherein, is the signal standard deviation (the data normalization benchmark, calculated based on a sliding time window and recalculated every 10 minutes), and is calculated in real time as: = (10); wherein, represents the multi-source data of the sampling, n represents the total number of sampling points, and k is the data of the k-th sampling point.

[0079] Step S203: Extract the time-domain features, frequency-domain features, temperature features, and working condition features of the multi-source data respectively.

[0080] Specifically, the above step S203 includes: Step S2031: Extract the time-domain features and frequency-domain features of the vibration signal; the time-domain features include the root mean square value and kurtosis, and the frequency-domain features include the spectral energy integral and the sideband energy ratio.

[0081] Specifically, the calculation of the time-domain features: 1. Root Mean Square (RMS): The average energy of the vibration signal, sensitive to uniform wear (such as tooth surface wear), and the formula is as follows: (11); wherein, is the data of the k -th sampling point of the vibration signal after noise reduction, is the time window = 10 minutes, and the number of sampling points within it is × .

[0082] 2. Kurtosis: The intensity of the impact component of the vibration signal, sensitive to local impact (such as bearing spalling), and the formula is as follows: (12); wherein, is the mean value of the vibration signal, ; is the standard deviation of the vibration signal, .

[0083] Calculation of frequency-domain features: 1. Spectrum energy integration, the formula is as follows: (13); Among them, i = 1, 2, …, M, is the Fourier transform result of the vibration signal; is the center frequency of the i th characteristic frequency band, which is determined by the physical parameters of the gearbox: = ( is the number of teeth of the gear); is the bandwidth, by default = 0.5 × (half octave); M is the number of monitored meshing frequency orders, usually M = 3 (fundamental frequency, 2 times frequency, 3 times frequency); is the vibration energy representing a specific frequency band, which is sensitive to specific frequency components (such as broken teeth, eccentricity faults).

[0084] 2. Sideband energy ratio: Reflects the degree of modulation fault and is sensitive to modulation phenomena (such as misalignment of the shafting), the formula is as follows: (14); Among them, ; Q is the number of sidebands, by default Q = 2; Physical meaning: An increase in the sideband energy ratio indicates a modulation phenomenon (such as shaft misalignment, gear eccentricity).

[0085] Step S2032, extract the temperature features of the temperature signal, and the temperature features include the temperature rise rate and the temperature distribution uniformity index.

[0086] Specifically, the temperature rise rate (Temperature rate) represents the temperature rise rate after restoring the dimension, which is directly related to the lubrication state and friction loss of the gearbox and is used to detect abnormal heating (such as frictional heating caused by lack of oil in the bearing), the formula is as follows: (15); Among them, is the original temperature value before standardization (unit: °C); is the time interval, by default = 5 minutes (set according to the thermal inertia characteristics of the gearbox); , : The mean and standard deviation of the temperature signal.

[0087] The Temperature Distribution Uniformity Index can detect local overheating caused by the failure of the cooling system. The formula is as follows: (16); Among them, the value of formula (16) ∈ [0, 1]. Approaching 0 indicates uneven temperature distribution (such as local overheating); is the temperature fluctuation range in the healthy state, which is obtained by statistical analysis of historical data (for example = 2.5).

[0088] Step S2033: Extract the operating characteristics of the operating parameters. The operating characteristics include the load-speed coupling factor and the power fluctuation entropy.

[0089] Specifically, for SCADA operating characteristic extraction: Extract the operating coupling characteristics from the standardized SCADA operating parameters (speed , load , power ), which characterize abnormal operating states. The extracted operating coupling characteristics include: 1. The Load-Speed Coupling Factor reflects the mechanical transmission efficiency and is strongly correlated with gear wear and shaft alignment. The formula is as follows: (17); Among them, is a constant for preventing zero repair, = 0.1 MW; Physical meaning: It represents the mechanical power transmission efficiency. An abnormal increase indicates slippage of the transmission chain or gear wear.

[0090] 2. The Power Fluctuation Entropy quantifies the power stability and is sensitive to grid anomalies or intermittent gear failures. The formula is as follows: (18); Among them, , is the power value binning interval (divide P ∈ [0, 8] MW into Y = 10 intervals); n is the number of power sampling points within the time window Δt = 10 minutes; An increase indicates an enhanced randomness of power fluctuations (such as grid disturbances or intermittent gear jamming); Anti-interference design is adopted for feature extraction: In the calculation of the temperature rise rate, the dimension is restored through , which can avoid the ambiguity of physical meaning caused by standardization. Power binning (Y = 10) balances the calculation efficiency and feature resolution.

[0091] Step S204: weighted feature fusion is performed on the time domain features, frequency domain features, temperature features and working condition features to obtain real-time fusion features. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0092] Step S205, obtain the historical health data of the wind turbine gearbox, establish the health benchmark of the wind turbine gearbox based on the historical health data, and build a health model based on the real-time fusion feature and the health benchmark. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0093] Step S206: Use the health model to perform fault diagnosis on the wind turbine gearbox. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0094] The wind turbine gearbox fault diagnosis method provided in this embodiment arranges various types of sensors on the high-speed shaft, low-speed shaft and planetary frame of the gearbox, synchronously collects multi-source data such as vibration and temperature, and combines SCADA working condition parameters to fully cover the gearbox operating status information. Different types of data complement each other, avoid the limitations of a single data source, and provide a rich and comprehensive data basis for fault diagnosis. The proposed sensor arrangement scheme combines micro sensors to solve the problem of planetary frame space limitations, and designs a dynamic sampling frequency adjustment strategy based on speed and torque to balance data accuracy and storage costs. The speed synchronization noise template is introduced to directionally suppress the interference components related to the gear meshing frequency, and the physical statistical dual criteria (correlation coefficient + kurtosis) are used to screen effective IMFs to improve the impact signal retention capability. Features are extracted from four dimensions: time domain, frequency domain, temperature and working conditions, respectively, to construct a complete fault sensitive feature system. This multi-dimensional feature extraction method can deeply mine the fault information contained in the data and improve the sensitivity and accuracy of fault diagnosis.

[0095] In this embodiment, a wind turbine gearbox fault diagnosis method is provided, which can be used for wind turbines. Figure 3 FIG. 1 is a flow chart of a method for diagnosing a fault in a wind turbine gearbox according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps: Step S301, collect multi-source data of wind turbine gearbox. Figure 1 Step S201 of the illustrated embodiment will not be described in detail here.

[0096] Step S302, respectively extract the time domain features, frequency domain features, temperature features and working condition features of the multi-source data. Figure 1 Step S203 of the illustrated embodiment will not be described in detail here.

[0097] Step S303: Perform weighted feature fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features to obtain real-time fusion features.

[0098] Specifically, the above-mentioned step S303 includes: Step S3031: Normalize the time-domain features, frequency-domain features, temperature features, and operating condition features respectively.

[0099] Specifically, the following formula is used to normalize the time-domain features, frequency-domain features, temperature features, and operating condition features: (19); Where, , is the historical mean of the feature under healthy conditions (obtained by statistical analysis of the gearbox fault-free operation data, updated quarterly); is the historical standard deviation of the feature under healthy conditions (statistical range: at least 6 months of data, updated quarterly); are the root mean square value, kurtosis, 2x frequency spectrum energy integral, 3x frequency spectrum energy integral, and sideband energy ratio respectively; are the temperature rise rate, temperature distribution uniformity index, load speed coupling factor, and power fluctuation entropy respectively; is the feature corresponding to the vibration signal, are the temperature features and operating condition features.

[0100] Constraint condition: If (no fluctuation of the feature), force = 1 to avoid division-by-zero errors.

[0101] Step S3032: Calculate the feature information entropy corresponding to the normalized time-domain features, frequency-domain features, temperature features, and operating condition features, and calculate the entropy weight method weights based on the feature information entropy of each feature to obtain the weights corresponding to each feature.

[0102] Specifically, the weights of each feature are dynamically calculated based on the information entropy theory to highlight the contribution degree of the fault-sensitive features. The following formula is used for feature information entropy calculation: (20); Where, , is to evenly divide the value range of the feature into Y = 10 intervals; is the number of samples in the statistical window (default = 1440, corresponding to 24-hour data).

[0103] Physical meaning: The smaller it is, the more concentrated the distribution of the feature (more sensitive to faults).

[0104] The entropy weight method is used to calculate the feature fusion weight, and the formula is as follows: (21); where n’ is the total number of features, and the constraint condition is: ; Physical meaning: It is adjusted in real time with the update of ; The weight is negatively correlated with the entropy . The lower the entropy (the stronger the ability of the feature to distinguish faults), the higher the weight.

[0105] Step S3033: Perform weighted feature linear fusion on the time-domain features, frequency-domain features, temperature features, and operating conditions features and their corresponding weights to obtain real-time fusion features.

[0106] Specifically, for weighted feature fusion: linearly fuse the normalized features according to the weights to generate a comprehensive fault sensitivity index, and the formula is as follows: (22); where is a dimensionless comprehensive index, that is, the real-time fusion feature. The value range is usually [-3, 3]. Exceeding the threshold indicates a fault. The wind turbine gearbox fault diagnosis method provided in this embodiment realizes dynamic weighted feature fusion based on the information entropy theory, automatically assigns weights according to the concentration degree of the feature distribution, and highlights the contribution degree of the features sensitive to faults. The weights are adjusted in real time with the change of data, and can adapt to the change of the importance of features under different working conditions and fault scenarios. Compared with the fixed-weight fusion method, it can more accurately reflect the actual operating state of the gearbox. At the same time, a comprehensive fault sensitivity index is generated through normalization processing and weighted linear fusion, reducing the influence of dimension difference and enhancing the scientificity and practicability of feature fusion.

[0107] Step S304: Obtain the historical health data of the wind turbine gearbox, establish a health benchmark for the wind turbine gearbox based on the historical health data, and construct a health model based on the real-time fusion features and the health benchmark.

[0108] Specifically, the above step S304 includes: Step S3041: Train a preset gated recurrent unit model with historical health data to obtain a health benchmark for the wind turbine gearbox.

[0109] Specifically, the preset gated recurrent unit model is the GRU (Gated Recurrent Unit) model. The gated recurrent unit model is trained based on historical health data to establish a health benchmark under the normal operation state of the gearbox. The formula for the health benchmark is as follows: (23); Where: is the fused feature sequence in the historical health data (time window t ∈ ); Θ is the set of GRU model parameters, including the input gate weight , the reset gate weight , the hidden layer weight , the bias term b , etc.; Θ = { , , , b}, which is obtained by training with historical health data; it is updated once every 100 new sets of fault data.

[0110] The training constraint uses a loss function. The formula for the loss function is as follows: (24); Training data volume: N ≥ samples (covering different seasons and working conditions).

[0111] Step S3042: Calculate the health score based on the real-time fused feature, the health benchmark, and the preset health state tolerance threshold, and calculate the change trend of the health score within a preset time period.

[0112] Specifically, compare the real-time fused feature with the health benchmark to calculate the health score.

[0113] (25); Among them, the preset time period is 10 minutes, that is, the health score is calculated every 10 minutes; Threshold is the health state tolerance threshold, calculated as the 95th percentile of the historical health data residuals (re-counted monthly), and its formula is: (26); Scoring range: ∈ (-∞, 1], the health state is set to , if it is less than 0.8, it is an unhealthy state.

[0114] In this embodiment, new data is collected every 10 minutes and the health score is calculated , forming a time series Through the sliding window technique, the most recent m ratings (e.g., m = 6, corresponding to 1-hour data) are tracked in real time, providing a data basis for trend analysis.

[0115] Step S3043: Use the health score, its change trend, and a preset expert rule base as the fault type decision rule, and construct a health model based on the fault type decision rule.

[0116] Specifically, the expert rule base can adopt a hierarchical structure, including a basic rule layer, a composite rule layer, and an extended rule layer. The basic rule layer defines single-index threshold rules, such as "trigger a primary warning when (St < 0.6)"; the composite rule layer combines the logical relationships of multiple indicators, such as "if (St < 0.6) and { }< -0.1) and the load speed coupling factor exceeds the normal range, it is determined as a serious fault"; the extended rule layer dynamically adds rules according to new fault types, such as adding "when the oil temperature rises abnormally and the vibration signal shows low-frequency fluctuations for the gearbox oil leakage fault, start the oil detection process".

[0117] Match the real-time health score, change trend characteristics with the expert rule base. Adopt a forward reasoning strategy to retrieve the rules that meet the conditions one by one from the rule base. For example, if it is detected that the health score drops suddenly, the power fluctuation entropy increases abnormally, and the sideband energy ratio rises, and the "local gear damage" rule is matched, it is initially diagnosed as a local gear damage fault. For the case of multi-rule matching conflicts, introduce evidence theory or fuzzy logic to resolve the conflicts, and determine the final fault type by synthesizing the support degrees of multiple rules. Use the trained and rule-matched GPU model as the health model.

[0118] A fault diagnosis method for a wind turbine gearbox provided in this embodiment uses historical health data to train a GRU model to establish a health benchmark. The GRU model can effectively process long sequence data and capture the time variation law of the gearbox operating state. Set strict training constraints, including appropriate loss functions, sufficient training data volume, and overfitting control strategies, to ensure the reliability and generalization ability of the model. Through the calculation of the real-time health score, compare the real-time fusion features with the health benchmark, and combine with the dynamically adjusted threshold to accurately evaluate the health state of the gearbox and provide a scientific basis for fault diagnosis.

[0119] Step S305: Use the health model to diagnose the faults of the wind turbine gearbox.

[0120] Specifically, the above step S305 includes: Step S3051: Obtain the multi-source data of the real-time wind turbine gearbox, and input the multi-source data of the real-time wind turbine gearbox into the health model to obtain the predicted health score.

[0121] Specifically, for the detailed description, refer to Steps S301 to S304, which will not be elaborated here.

[0122] Step S3052: When the predicted health score is less than the preset alarm threshold, an alarm is triggered, and a fault risk is prompted.

[0123] Specifically, the alarm trigger condition judgment formula is as follows: < and (27); where, is the alarm threshold, with a default = 0.6; is the health decline rate threshold, with a default = -0.1 / minute.

[0124] Step S3053: Dynamically update the real-time feature weights based on the weight learning rate, historical fusion features, and real-time fusion features, and calculate the real-time feature weight vector.

[0125] Specifically, adjust the weights according to the real-time diagnosis result feedback to improve the model adaptability. The update rule is expressed as: (28); where, , is the weight learning rate, with an initial value = 0.01, which is dynamically adjusted according to the convergence situation (for example, adjusted once every diagnosis cycle (10 minutes)): ; is the fusion feature value of manually marked or historically diagnosed faults (i.e., historical fusion features).

[0126] Step S3054: Construct a fault type standard library for the wind turbine gearbox faults, calculate the cosine similarity between the real-time feature weight vector and each fault type standard weight pattern in the fault type standard library, and take the fault type with the highest cosine similarity as the fault diagnosis result.

[0127] Specifically, the fault type representation formula is: .

[0128] where, = is the dynamic weight vector; is the standard weight pattern of the k th type of fault (preset by expert experience and historical fault data analysis; new fault types are expanded every quarter; such as the bearing fault mode = [0.15, 0.25, 0.1,...]).

[0129] The cosine similarity calculation formula is as follows: The wind turbine gearbox fault diagnosis method provided in this embodiment can collect data in real time and calculate the health score, which can timely detect the abnormal changes in the operating state of the gearbox and realize the early warning of faults. Combining threshold alarm and fault mode matching based on cosine similarity, and using the expert rule base for fault type decision-making, it not only considers the absolute value of the health score but also pays attention to its change trend. At the same time, by comparing with the standard weight mode, the accuracy and reliability of fault diagnosis are improved.

[0130] Step S306: Update the fault type decision rule and the health model by using the newly diagnosed fault type.

[0131] Specifically, fine-tune the parameters of the GRU model, and the formula is as follows: (29); (30); where M is the amount of new fault data; is the new parameter set of the GRU model, is the old parameter set of the GRU model, is a function with as the parameter and is a gradient vector.

[0132] Dynamic correction of the health state tolerance threshold: (31); where = 0.9: Threshold update smoothing coefficient.

[0133] The wind turbine gearbox fault diagnosis method provided in this embodiment, the online self-optimization mechanism of the model can update the model parameters and decision rules by using the newly diagnosed fault data, form a diagnostic closed loop, continuously improve the diagnostic accuracy, and adapt to the changes in equipment performance and newly emerging fault types.

[0134] The wind turbine gearbox fault diagnosis method provided in this embodiment has the following beneficial effects: 1. Improvement of diagnostic accuracy: The multi-source data fusion increases the fault detection accuracy rate to 98.2% (compared with 85.3% of single vibration analysis).

[0135] 2. Enhancement of anti-interference ability: After improving CEEMDAN noise reduction, the signal-to-noise ratio (SNR) is increased by 6 - 8 dB, effectively retaining the fault impact components.

[0136] 3. Dynamic adaptive optimization: The adaptive sampling strategy reduces the data volume by 30%, and at the same time ensures the high-speed capture of 50 kHz for transient impacts (> 30% energy sudden increase); weight Along with the feature information entropy Dynamic adjustment, automatically focusing on key features at different life stages of the gearbox (early wear, severe faults). The learning rate The adaptive mechanism avoids overfitting or slow convergence of the model.

[0137] 4. Early warning ability: Health score The dual mechanisms of health score and weight pattern matching can trigger an early warning 24 - 48 hours before a fault occurs, with a false alarm rate < 2%; The falling rate threshold Filters instantaneous interference.

[0138] 5. Noise reduction adaptability: Through the dynamic adjustment of ε, the fault impact components can still be retained under strong noise conditions (such as sudden changes in wind speed).

[0139] 6. Enhanced fault sensitivity: Suppress high - entropy and low - efficiency features (such as environmental temperature interference) through the entropy weight method, and strengthen the decision weights of low - entropy features (such as vibration kurtosis).

[0140] 7. Model evolution ability: Online update of GRU parameters Θ and threshold Threshold enables the system to adapt to the performance degradation of the gearbox. The fault mode library Is extensible and supports new - type fault diagnosis. In this embodiment, a fault diagnosis device for a wind turbine gearbox is also provided. This device is used to implement the above - mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that realizes a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0141] This embodiment provides a fault diagnosis device for a wind turbine gearbox, as Figure 4 shown, including: A data acquisition module 401, which is used to acquire multi - source data of the wind turbine gearbox.

[0142] A feature extraction module 402, which is used to extract the time - domain features, frequency - domain features, temperature features, and operating condition features of the multi - source data respectively.

[0143] A feature fusion module 403, which is used to perform weighted feature fusion on the time - domain features, frequency - domain features, temperature features, and operating condition features to obtain real - time fusion features.

[0144] A health model construction model 404, which is used to obtain the historical health data of the wind turbine gearbox, establish a health benchmark for the wind turbine gearbox based on the historical health data, and construct a health model based on the real - time fusion features and the health benchmark.

[0145] The fault diagnosis module 405 is used to perform fault diagnosis on the gearbox of the wind turbine by using a health model.

[0146] In some alternative embodiments, the multi-source data includes vibration signals, temperature signals, and operating conditions parameters; the data acquisition module 401 includes: A vibration signal acquisition unit, which is used to arrange three-axis vibration sensors on the high-speed shaft, low-speed shaft, and planetary carrier of the gearbox of the wind turbine, and collect the vibration signals of the gearbox of the wind turbine by using the three-axis vibration sensors.

[0147] A temperature signal acquisition unit, which is used to arrange temperature sensors on the bearing seat and the wall of the gearbox near the meshing gears of the gearbox of the wind turbine, and collect the temperature signals of the gearbox of the wind turbine by using the temperature sensors.

[0148] An operating conditions parameters acquisition unit, which is used to collect the operating conditions parameters of the gearbox of the wind turbine by using a SCADA monitoring system, and the operating conditions parameters include rotational speed, power, and torque.

[0149] An adaptive sampling frequency adjustment unit, which is used to arrange a torque sensor at the low-speed shaft coupling of the gearbox of the wind turbine, and determine an adaptive sampling frequency adjustment strategy based on the rotational speed and torque by using the torque sensor.

[0150] In some alternative embodiments, the wind turbine gearbox fault diagnosis device further includes: A noise reduction and standardization processing module, which is used to perform noise reduction processing on the vibration signals by using the CEEMDAN algorithm, and at the same time perform standardization processing on the vibration signals, temperature signals, and operating conditions parameters by using a preset formula, where: when performing noise reduction processing on the vibration signals by using the CEEMDAN algorithm, a preset rotational speed synchronous noise template is used to suppress the interference components related to the gear meshing frequency of the vibration signals, and the intrinsic mode functions corresponding to the vibration signals are screened by using a physical statistics double criterion.

[0151] In some alternative embodiments, the feature extraction module 402 includes: A time-frequency domain feature extraction unit, which is used to extract the time-domain features and frequency-domain features of the vibration signals; the time-domain features include root mean square value and kurtosis, and the frequency-domain features include spectral energy integral and sideband energy ratio.

[0152] A temperature feature extraction unit, which is used to extract the temperature features of the temperature signals, and the temperature features include temperature rise rate and temperature distribution uniformity index.

[0153] An operating conditions feature extraction unit, which is used to extract the operating conditions features of the operating conditions parameters, and the operating conditions features include load rotational speed coupling factor and power fluctuation entropy.

[0154] In some alternative embodiments, the feature fusion module 403 includes: A normalization processing unit for performing normalization processing on time-domain features, frequency-domain features, temperature features, and operating condition features respectively.

[0155] A weight calculation unit for calculating the feature information entropy corresponding to the time-domain features, frequency-domain features, temperature features, and operating condition features after normalization processing, and performing entropy weight method weight calculation based on the feature information entropy of each feature to obtain the weights corresponding to each feature.

[0156] A feature fusion unit for performing weighted feature linear fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features and their corresponding weights to obtain real-time fusion features.

[0157] In some alternative embodiments, the health model construction model 404 includes: A health benchmark acquisition unit for training a preset gated recurrent unit model with historical health data to obtain the health benchmark of the wind turbine gearbox.

[0158] A health score calculation unit for calculating the health score based on the real-time fusion features, health benchmark, and preset health state tolerance threshold, and calculating the change trend of the health score within a preset time period.

[0159] A health model construction unit for using the health score and its change trend, as well as a preset expert rule base as a fault type decision rule, and constructing a health model based on the fault type decision rule.

[0160] In some alternative embodiments, the fault diagnosis module 405 includes: Obtain the real-time multi-source data of the wind turbine gearbox and input the real-time multi-source data of the wind turbine gearbox into the health model to obtain the predicted health score.

[0161] When the predicted health score is less than the preset alarm threshold, an alarm is triggered and a fault risk is prompted.

[0162] Dynamically update the real-time feature weights based on the weight learning rate, historical fusion features, and real-time fusion features, and calculate the real-time feature weight vector.

[0163] Construct a fault type standard library for the faults of the wind turbine gearbox, calculate the cosine similarity between the real-time feature weight vector and each fault type standard weight pattern in the fault type standard library, and use the fault type with the highest cosine similarity as the fault diagnosis result.

[0164] In some alternative embodiments, the wind turbine gearbox fault diagnosis device further includes: An update module for updating the fault type decision rule and the health model using the newly diagnosed fault type.

[0165] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be repeated here.

[0166] The wind turbine gearbox fault diagnosis device in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0167] An embodiment of the present invention also provides a computer device having the above-mentioned Figure 4 shown wind turbine gearbox fault diagnosis device.

[0168] Please refer to Figure 5 , Figure 5 which is a schematic structural diagram of a computer device provided by an optional embodiment of the present invention. As Figure 5 shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor system). Figure 5 One processor 10 is taken as an example in

[0169] The processor 10 can be a central processor, a network processor, or a combination thereof. Among them, the processor 10 can further include a hardware chip. The above-mentioned hardware chip can be an application specific integrated circuit, a programmable logic device, or a combination thereof. The above-mentioned programmable logic device can be a complex programmable logic device, a field programmable gate array, a general array logic, or any combination thereof.

[0170] Among them, the memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0171] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories may be connected to the computer device through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0172] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk, or a solid-state drive; the memory 20 may further include a combination of the above types of memories.

[0173] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 may be connected through a bus or other means. Figure 5 Taking connection through a bus as an example.

[0174] The input device 30 may receive input digital or character information, and generate key signal inputs related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (such as an LED), and a haptic feedback device (such as a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device may be a touch screen.

[0175] Embodiments of the present invention also provide a computer-readable storage medium. The method according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code that is originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored as such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiments is implemented.

[0176] A part of the present invention can be applied as a computer program product, for example, computer program instructions. When executed by a computer, through the operation of the computer, the methods and / or technical solutions according to the present invention can be called or provided. Those skilled in the art should understand that the forms of existence of computer program instructions in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.

[0177] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A fault diagnosis method for a wind turbine gearbox, characterized in that, The method includes: Collecting multi-source data of the gearbox of a wind turbine; Separately extracting the time-domain features, frequency-domain features, temperature features, and operating condition features of the multi-source data; Performing weighted feature fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features to obtain real-time fusion features; Obtaining the historical health data of the gearbox of the wind turbine, establishing a health benchmark for the gearbox of the wind turbine based on the historical health data, and constructing a health model based on the real-time fusion features and the health benchmark; Using the health model to perform fault diagnosis on the gearbox of the wind turbine.

2. The method according to claim 1, wherein The multi-source data includes vibration signals, temperature signals, and operating condition parameters; The collecting of the multi-source data of the gearbox of the wind turbine includes: Arranging three-axis vibration sensors on the high-speed shaft, low-speed shaft, and planet carrier of the gearbox of the wind turbine, and using the three-axis vibration sensors to collect the vibration signals of the gearbox of the wind turbine; Arranging temperature sensors on the bearing seat and the wall of the gearbox near the meshing gears of the gearbox of the wind turbine, and using the temperature sensors to collect the temperature signals of the gearbox of the wind turbine; Using a SCADA monitoring system to collect the operating condition parameters of the wind turbine, and the operating condition parameters include rotational speed, power, and torque; Arranging a torque sensor at the low-speed shaft coupling of the gearbox of the wind turbine, and based on the rotational speed and torque, using the torque sensor to determine an adaptive sampling frequency adjustment strategy.

3. The method according to claim 2, wherein Before separately extracting the time-domain features, frequency-domain features, temperature features, and operating condition features of the multi-source data, the method further includes: Using the CEEMDAN algorithm to perform noise reduction processing on the vibration signals, and at the same time using a preset formula to perform normalization processing on the vibration signals, temperature signals, and operating condition parameters, where: When using the CEEMDAN algorithm to perform noise reduction processing on the vibration signals, using a preset rotational speed synchronous noise template to suppress the interference components related to the gear meshing frequency of the vibration signals, and using a physical statistics double criterion to screen the intrinsic mode functions corresponding to the vibration signals.

4. The method according to claim 2, wherein The separately extracting the time-domain features, frequency-domain features, temperature features, and operating condition features of the multi-source data includes: Extracting the time-domain features and frequency-domain features of the vibration signals; the time-domain features include root mean square value and kurtosis, and the frequency-domain features include spectral energy integral and sideband energy ratio; Extracting the temperature features of the temperature signals, and the temperature features include temperature rise rate and temperature distribution uniformity index; Extracting the operating condition features of the operating condition parameters, and the operating condition features include load rotational speed coupling factor and power fluctuation entropy.

5. The method according to claim 1, characterized in that, Performing weighted feature fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features to obtain real-time fusion features, including: Performing normalization processing on the time-domain features, frequency-domain features, temperature features, and operating condition features respectively; Calculating the feature information entropy corresponding to the time-domain features, frequency-domain features, temperature features, and operating condition features after normalization processing, and performing entropy weight method weight calculation based on the feature information entropy of each feature to obtain the weights corresponding to each feature; Performing weighted feature linear fusion on the time-domain features, frequency-domain features, temperature features, and operating condition features and the corresponding weights to obtain real-time fusion features.

6. The method according to claim 1, wherein Based on the historical health data, a health benchmark of the wind turbine gearbox is established, and a health model is constructed based on the real-time fusion features and the health benchmark, including: Using the historical health data to train a preset gated recurrent unit model to obtain a health benchmark of the wind turbine gearbox; Calculating a health score based on the real-time fusion features, the health benchmark, and a preset health status tolerance threshold, and calculating the change trend of the health score within a preset time period; Taking the health score and its change trend, as well as a preset expert rule base, as fault type decision rules, and constructing a health model based on the fault type decision rules.

7. The method according to claim 1, wherein The fault diagnosis of the wind turbine gearbox using the health model includes: Obtaining multi-source data of the real-time wind turbine gearbox and inputting the multi-source data of the real-time wind turbine gearbox into the health model to obtain a predicted health score; When the predicted health score is less than a preset alarm threshold, an alarm is triggered and a fault risk is prompted; Dynamically updating the real-time feature weights based on the weight learning rate, historical fusion features, and real-time fusion features, and calculating the real-time feature weight vector; Constructing a fault type standard library for the faults of the wind turbine gearbox, calculating the cosine similarity between the real-time feature weight vector and each fault type standard weight pattern in the fault type standard library, and taking the fault type with the highest cosine similarity as the fault diagnosis result.

8. The method according to claim 6, characterized in that, The method further includes: Updating the fault type decision rules and the health model using the newly diagnosed fault type.

9. A fault diagnosis device for a wind turbine gearbox, characterized in that, The device includes: A data acquisition module for acquiring multi-source data of the wind turbine gearbox; A feature extraction module for respectively extracting the time-domain features, frequency-domain features, temperature features, and working condition features of the multi-source data; A feature fusion module for performing weighted feature fusion on the time-domain features, frequency-domain features, temperature features, and working condition features to obtain real-time fusion features; A health model construction model for obtaining the historical health data of the wind turbine gearbox, establishing a health benchmark of the wind turbine gearbox based on the historical health data, and constructing a health model based on the real-time fusion features and the health benchmark; A fault diagnosis module for performing fault diagnosis on the wind turbine gearbox using the health model.

10. A computer device, characterized in that, Including: A memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the wind turbine gearbox fault diagnosis method according to any one of claims 1 to 8.

Images