Early warning method for electric corrosion of generator bearing of wind turbine generator

Abstract

An early warning method for electric corrosion of a generator bearing of a wind turbine generator relates to the technical field of generators and specifically comprises the following steps: respectively acquiring vibration data and ultrasonic data of the generator bearing under the operation condition of the wind turbine generator; performing feature extraction on the vibration data and the ultrasonic data; constructing a wind turbine generator bearing electrocorrosion early warning model; substituting the extracted feature data into a wind turbine generator bearing electrocorrosion early warning model to obtain early warning information; the method is used for solving the problems that in the operation site of the wind driven generator, the electric corrosion fault of the bearing is high in cost, inaccurate and not timely, and hidden dangers are brought to safe operation of the wind driven generator.

Description

Technical Field

[0001] The present invention relates to the technical field of generators, and in particular to an early warning method for electrical corrosion of a generator bearing of a wind turbine generator set. Background Art

[0002] Wind energy, a renewable, pollution-free energy source, has become the world's fastest-growing clean energy source in recent years. In wind turbines, generator bearings are critical components that influence system efficiency and lifespan. However, due to the long-term high-speed rotation and electromagnetic field conditions in generator bearings, electrical corrosion is a common fault. High-frequency discharges from the converter switching power supply in wind turbine systems, which disrupt the bearing lubricant film, are the root cause of bearing electrical corrosion failures, seriously impacting the safe and stable operation of wind turbines.

[0003] Currently, there is a lack of reliable on-site detection methods for electrical corrosion in wind turbine generator bearings. Severe electrical corrosion in the generator bearings is typically observed during maintenance after a wind turbine failure. Regularly removing the wind turbine generator bearings for inspection would require cumbersome procedures and maintenance downtime, resulting in significant costs. In terms of online monitoring, shaft voltage and shaft current are key parameters contributing to electrical corrosion in generator bearings, but these parameters are difficult to capture during real-time operation. Vibration analysis is the most commonly used method for detecting wind turbine generator bearing failures. However, it struggles to detect early failures, identify the severity of damage, or accurately determine whether electrical corrosion is the cause of abnormal vibration data. Using single detection methods, such as commonly used vibration analysis, it is impossible to provide accurate, real-time early warning of electrical corrosion in wind turbine bearings during operation.

[0004] Chinese patent document CN118747305A discloses a "Method and Related Device for Diagnosing Electrical Corrosion Faults in Wind Turbine Bearings." This method constructs a wind turbine bearing model by acquiring ultrasonic signal data generated by the wind turbine bearing during operation and combining it with bearing structural information. The ultrasonic signal data generated by the wind turbine bearing under test is then input into the generator bearing model to obtain a generator bearing model diagnosis result. This technical solution relies on ultrasonic analysis technology, resulting in a single fault identification method and a failure to provide accurate early warning information for generator bearing electrical corrosion faults.

[0005] Therefore, we propose an early warning method that can accurately determine whether the generator bearing has an electrical corrosion fault state. Summary of the Invention

[0006] The purpose of the present invention is to provide an early warning method for electrical corrosion of bearings of wind turbine generators, which is used to solve the problem that bearing electrical corrosion failures at the wind turbine operation site are expensive, inaccurate, and untimely, posing hidden dangers to the safe operation of the wind turbine.

[0007] The present invention is achieved through the following technical solutions:

[0008] An early warning method for electrical corrosion of a wind turbine generator bearing, specifically comprising:

[0009] Under the condition of wind turbine operation, the vibration data and ultrasonic data of the generator bearing are obtained respectively;

[0010] performing feature extraction on the vibration data and the ultrasonic data;

[0011] Construct an early warning model for electrical corrosion of wind turbine generator bearings;

[0012] The extracted feature data is brought into the wind turbine generator bearing electrical corrosion early warning model to obtain early warning information.

[0013] Furthermore, the frequency range of the vibration acquisition is 0-10 kHz, and the frequency range of the ultrasonic acquisition is 20-500 kHz.

[0014] Furthermore, an acceleration sensor is used to collect vibration data of the generator bearing in a rigid contact manner, and an ultrasonic sensor is used to monitor ultrasonic signals generated by oil film discharge in the generator bearing.

[0015] Furthermore, the feature extraction step of the vibration signal specifically includes:

[0016] Perform CEEMDAN decomposition on the vibration signal to obtain several intrinsic mode functions;

[0017] Screen out the IMF components containing the characteristic frequencies of bearing electrical corrosion faults;

[0018] Perform synchronous compression transformation on the filtered IMF components to generate high-resolution time-frequency diagrams;

[0019] The amplitude of the fault characteristic frequency and the amplitude of the fault characteristic harmonic in the time-frequency diagram are extracted as vibration characteristic quantities.

[0020] Furthermore, the feature extraction step of the ultrasonic signal specifically includes:

[0021] Perform CEEMDAN decomposition on the ultrasonic signal to obtain several eigenmode functions;

[0022] Screen out the IMF components containing the characteristic frequencies of bearing electrical corrosion faults;

[0023] Perform synchronous compression transformation on the filtered IMF components to generate high-resolution time-frequency diagrams;

[0024] The frequency and amplitude of irregular impact in the time domain waveform are counted, and the root mean square and peak factor of the ultrasonic time domain discrete signal are calculated as ultrasonic feature quantities.

[0025] Furthermore, the ensemble average formula of the CEEMDAN decomposition is:

[0026]

[0027] Where x is the original vibration signal; w (n) is the adaptive white noise added for the nth time; β is the control parameter of the noise amplitude; N is the number of noise auxiliary signals; EMD(·) k is the kth IMF component obtained after performing empirical mode decomposition on the noisy signal.

[0028] Furthermore, the screening process of the IMF components is as follows:

[0029] Calculate the power spectral density of each IMF component;

[0030] Identify IMF components that match the fault characteristic frequencies associated with bearing components;

[0031] The IMF components dominated by high-frequency noise are eliminated, and the effective IMF components reflecting fault information are retained.

[0032] Furthermore, the calculation formula of the synchronous compression transformation is:

[0033]

[0034] Where, T(t,ω) is the short-time Fourier transform result of the IMF component; is the instantaneous frequency estimate; g(η) is the normalization factor; Δη is the frequency bandwidth.

[0035] Furthermore, the formula for extracting the fault characteristic frequency amplitude is:

[0036]

[0037] Where SST(τ,f) is the time-frequency energy distribution after synchronous compression transformation; f is the characteristic frequency of the bearing fault; and A(f) is the amplitude corresponding to the fault frequency.

[0038] Furthermore, the wind turbine generator bearing electrical corrosion early warning model specifically includes:

[0039] Three-level thresholds for vibration and ultrasonic waves are preset respectively;

[0040] Dynamically update the threshold value according to the operation status;

[0041] Assign weights to each threshold according to the type and importance of the feature;

[0042] The total warning score is calculated using comprehensive weighting.

[0043] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0044] The present invention discloses an early warning method for electrical corrosion of a wind turbine generator bearing. By analyzing vibration data and ultrasonic data, the health of the generator bearing can be monitored in real time, and early warning information of bearing electrical corrosion faults can be obtained on site, thereby reducing the hidden dangers of bearing safety failures and minimizing maintenance costs and downtime. In addition, by combining the two fault detection technologies of vibration analysis and ultrasonic analysis, fault information of electrical corrosion of the wind turbine generator bearing can be effectively extracted, greatly improving the accuracy of identifying bearing electrical corrosion faults from the perspective of obtaining detection data sources.

[0045] By using feature extraction methods, the characteristic quantities of vibration signals and ultrasonic signals are fused and processed, which can fully and multi-dimensionally reflect the fault characteristics of bearing electrical corrosion and predict the specific components and severity of the fault.

[0046] In addition, the combined algorithm of CEEMDAN decomposition and synchronous compression transform can effectively suppress modal aliasing problems and improve time-frequency resolution, making the warning results more accurate.

[0047] By setting the frequency receiving range of the vibration sensor to 0-10kHz and the frequency receiving range of the ultrasonic sensor to 20-500kHz, the low-frequency and high-frequency ranges of different development stages of bearing electrical corrosion faults can be effectively covered. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic flow chart of a method of the present invention is shown;

[0049] Figure 2 Schematic diagram of the system structure of the present invention;

[0050] Figure 3 Schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0052] Example 1

[0053] like Figure 1 The present invention provides an early warning method for electrical corrosion of a wind turbine generator bearing, specifically comprising:

[0054] Under the condition of wind turbine operation, the vibration data and ultrasonic data of the generator bearing are obtained respectively;

[0055] It can monitor the health of generator bearings in real time, obtain early warning information of bearing electrical corrosion failure on site, reduce the hidden dangers of bearing safety failure, and minimize maintenance costs and downtime;

[0056] In addition, accelerometers are used to collect vibration data from wind turbine generator bearings using a rigid contact method, with a frequency range of 0 to 10 kHz. Due to the strong coupling and interference between the vibration responses of wind turbine components, the original vibration signal often contains a large amount of noise, which can overwhelm fault information related to bearing electrical corrosion. To suppress the random noise components in the detection signal, a wind turbine bearing fault diagnosis method based on stochastic resonance (SR) is employed to enhance fault signature information and improve fault identification. In nonlinear systems, the signal-to-noise ratio of the output signal is increased by varying the noise intensity, generating a resonance mechanism similar to dynamics, thereby achieving signal noise reduction. The SR method has the advantage of utilizing noise to enhance weak signal characteristics, thus compensating for the drawback of traditional noise reduction methods, where removing noise also results in the removal of useful information.

[0057] Ultrasonic sensors are also used to monitor ultrasonic signals generated by oil film discharge in the generator bearings, with a frequency range of 20 to 500 kHz. Early failures caused by bearing electrical corrosion have characteristic frequency components that are primarily high-frequency. Ultrasonic analysis overcomes the difficulty in detecting the weak, high-frequency fault signals generated by bearing electrical corrosion. When electrical corrosion occurs in a bearing, the oil film on the bearing surface breaks down. The high-frequency discharge of the bearing lubricating oil film caused by the converter switching power supply changes the amplitude and frequency of the ultrasonic waves. These changes in the ultrasonic characteristics enable early warning and diagnosis of bearing electrical corrosion failures.

[0058] performing feature extraction on the vibration data and the ultrasonic data;

[0059] Before feature extraction, the vibration data and ultrasonic data are denoised to ensure that the corresponding features extracted can accurately represent the working conditions of the generator bearings;

[0060] Construct an early warning model for electrical corrosion of wind turbine generator bearings;

[0061] The extracted feature data is fed into the wind turbine generator bearing electrical corrosion early warning model to obtain early warning information. Furthermore, the wind turbine generator bearing electrical corrosion early warning model can provide certain guidance based on the early warning information. For example, when a minor fault is determined to be in the early stage, the model uses the feature quantity with a higher early warning level, such as the amplitude at a certain fault frequency, to infer the parts inside the bearing that may be worn, and then notify professionals to replace the grease and improve the bearing lubrication environment. When the fault is determined to be in the development stage, the model infers the faulty parts corresponding to the feature quantity with a higher early warning level, and notifies professionals to closely monitor changes in parameters such as bearing temperature and motor current. When the fault is determined to be in the early stage of serious failure, professionals are notified to stop the wind turbine and dismantle it for maintenance.

[0062] Example 2

[0063] As an embodiment, the feature extraction step of the vibration signal specifically includes:

[0064] The vibration signal is decomposed by CEEMDAN, which is an improved ensemble empirical mode decomposition, to obtain several intrinsic mode functions.

[0065] Screen out the IMF components containing the characteristic frequencies of bearing electrical corrosion faults;

[0066] Perform synchronous compression transformation on the filtered IMF components to generate high-resolution time-frequency diagrams;

[0067] The amplitude of the fault characteristic frequency and the amplitude of the fault characteristic harmonic in the time-frequency diagram are extracted as vibration characteristic quantities.

[0068] In addition, the feature extraction step of the ultrasonic signal specifically includes:

[0069] The ultrasonic signal is decomposed by CEEMDAN to obtain several intrinsic mode functions, which are used to suppress modal aliasing and extract effective IMF components;

[0070] Screen out the IMF components containing the characteristic frequencies of bearing electrical corrosion faults;

[0071] Perform synchronous compression transformation on the filtered IMF components to generate high-resolution time-frequency diagrams;

[0072] Count the frequency and amplitude of irregular shocks in the time domain waveform;

[0073] The root mean square and peak factor of the ultrasonic discrete time domain signal are calculated as ultrasonic feature quantities.

[0074] In addition, in the feature extraction step of vibration signals and ultrasonic signals, the ensemble average formula of CEEMDAN decomposition is:

[0075]

[0076] Where x is the original vibration signal; w (n) is the adaptive white noise added for the nth time; β is the control parameter of the noise amplitude, usually 0.2 to 0.3; N is the number of noise auxiliary signals; EMD(·) k is the kth IMF component obtained after performing empirical mode decomposition on the noisy signal;

[0077] During the decomposition process of CEEMDAN, white noise of a specific amplitude is added to the original signal to generate multiple groups of noise auxiliary signals. Empirical mode decomposition (EMD) is performed on each group of signals to extract the IMFs of various orders. The noise effect is eliminated by ensemble averaging to obtain the final IMF components. This method can effectively suppress the modal aliasing problem in traditional EMD and improve the stability and accuracy of signal decomposition.

[0078] In the feature extraction step of the vibration signal and the ultrasonic signal, the screening process of the IMF component is as follows:

[0079] Calculate the power spectral density of each IMF component;

[0080] Identify IMF components that match the fault characteristic frequencies associated with bearing components;

[0081] The IMF components dominated by high-frequency noise are eliminated, and the effective IMF components reflecting fault information are retained.

[0082] In the feature extraction step of vibration and ultrasonic signals, the filtered IMF components are subjected to synchronous compression transform (SST) to generate high-resolution time-frequency diagrams. The SST time-frequency redistribution formula is used to improve the time-frequency resolution and focus on transient fault characteristics:

[0083]

[0084] Where T(t,ω) is the short-time Fourier transform (STFT) result of the IMF component; The instantaneous frequency is estimated using the phase derivative; g(η) is the normalization factor; and Δη is the frequency bandwidth, used for energy redistribution. In this method, a short-time Fourier transform is performed on the IMF components to obtain the initial time-frequency distribution. Synchronous compression is then performed along the frequency direction on the time-frequency plane to concentrate the energy on the instantaneous frequency curve, thereby improving the time-frequency resolution. This method can accurately capture transient fault characteristics in nonstationary vibration signals, significantly outperforming traditional envelope analysis, especially in strong noise environments.

[0085] In addition, the amplitude of the fault characteristic frequency and its harmonics are extracted from the time-frequency diagram as the vibration characteristic quantity. The extraction formula of the fault characteristic frequency amplitude is:

[0086]

[0087] Where SST(τ,f) is the time-frequency energy distribution after synchronous compression transformation; f is the characteristic frequency of the bearing fault, such as BPFO and BPFI; A(f) is the amplitude corresponding to the fault frequency f, which is used to quantify the severity of electrical corrosion. In the time-frequency diagram after synchronous compression transformation, the energy distribution area corresponding to the fault characteristic frequency is located along the frequency axis. Within the time window [t1, t2], the maximum energy amplitude A(f) at frequency f is extracted as the core feature quantity, reflecting the fundamental frequency energy intensity of bearing electrical corrosion. For the harmonic components of the fault frequency, the first two to three harmonics of the fault frequency, such as 2f and 3f, are extracted. Repeat the above steps to extract the harmonic amplitudes A(2f) and A(3f). Higher-order harmonics, such as the 4th harmonic and above, are often significantly affected by noise and have lower diagnostic value, so it is recommended to only include the first three harmonics. The energy peak corresponding to the fault characteristic frequency is located in the time-frequency diagram, and the peak amplitude and the distribution characteristics of its harmonic components are statistically analyzed to quantify the severity of bearing electrical corrosion.

[0088] For ultrasonic signals, the frequency and amplitude of irregular impacts occurring in the time domain waveform are statistically analyzed. The impact amplitude directly quantifies the transient impact intensity caused by electrical corrosion, while the impact frequency reflects the intensity of impact events. Generally speaking, the higher the impact frequency and the larger the amplitude, the more severe the bearing failure caused by electrical corrosion. In addition, static analysis is performed on the ultrasonic time domain discrete signal x[n] to calculate the root mean square (RMS):

[0089]

[0090] Where N is the total number of sampling points. RMS is used to evaluate the strength and stability of the signal, reflecting the average energy level of the signal. It is used to evaluate the persistence of electrical corrosion. An increase in the RMS value usually means an increase in the degree of electrical corrosion of the bearing.

[0091] Next, calculate the crest factor:

[0092]

[0093] Where max{|x[n]|} is the maximum absolute value of the signal, and the crest factor is defined as the ratio of the maximum absolute value of the signal to its root mean square value. The crest factor is used to evaluate the peak and impact characteristics of the signal, reflecting the relative strength of the peak in the signal. It is strongly correlated with early failures. An increase in the crest factor usually means that there are more impact or transient signals in the bearing, which may be a sign of electrical corrosion.

[0094] In summary, the characteristic quantities reflecting the electrical corrosion fault of the wind turbine bearing are extracted, including the fundamental amplitude V1 of the main fault characteristic frequency of the vibration signal, the amplitude V2 of its 2nd harmonic and the amplitude V3 of its 3rd harmonic, the amplitude U1 and frequency U2 of the impact of the ultrasonic time domain signal, and the root mean square U3 and peak factor U4 obtained after static analysis.

[0095] Example 3

[0096] As an embodiment, the wind turbine generator bearing electrical corrosion early warning model specifically includes:

[0097] Three-level thresholds for vibration and ultrasound are preset to display different warning levels;

[0098] The vibration characteristic threshold is set to V1 (1) , V2 (2) , V3 (3) , and based on the threshold of the vibration signal characteristic quantity, collect the vibration data of the same type of wind turbine bearings in the normal, slightly corroded, developed and severely corroded stages; calculate the distribution range of the fault frequency and its harmonic amplitude in each stage, and use the mean ± 3σ as the threshold boundary;

[0099] The ultrasonic feature threshold is set to For the ultrasonic signal characteristic value threshold, the threshold of the impact amplitude U1 is taken from the P95, P99, and P99.9 quantiles under normal working conditions in the historical data; the threshold of the impact frequency U2 is set at 1.5, 3.0, and 5.0 times the number of impacts N0 under normal working conditions; the threshold of the RMS value U3 is set according to the offset of 120%, 150%, and 200% of the baseline value; the threshold of the peak factor U4 is set according to the normal distribution statistics, and the normal working condition data is collected to calculate the mean μ and standard deviation σ, and the thresholds are set to μ+2σ, μ+3σ, and μ+4σ.

[0100] Dynamically update the threshold value according to the operation status;

[0101] The dynamic update is triggered automatically every three months, or when the cumulative running time of the wind turbine reaches 2000 hours, the threshold update is forced to be updated. During the update, the vibration and ultrasonic data of the wind turbine to be tested for the last three months are extracted and recalculated according to the above rules.

[0102] In addition, considering environmental adaptability correction, when the temperature sensor detects that the ambient temperature is >50℃ or <-20℃, or the humidity is >80%, the trigger threshold is temporarily adjusted:

[0103]

[0104] Where T is temperature; H is humidity; ΔT is the difference from the critical temperature; and H is the difference from the critical humidity.

[0105] Assign weights to each threshold according to the type and importance of the feature;

[0106] Among them, the vibration feature has a high contribution in the fault development stage, and the fundamental frequency amplitude V1 directly reflects the fault energy intensity, so its weight is 0.40; the 2nd frequency V2 is sensitive to early faults and has a weight of 0.15; the 3rd frequency V3 is more susceptible to noise interference and has a weight of 0.05; the ultrasonic impact feature has a high warning sensitivity for serious faults. The impact amplitude U1 quantifies the transient discharge intensity and has a weight of 0.20; the impact frequency U2 reflects the event density and has a weight of 0.10. The ultrasonic static feature is mainly used to assist in confirming the persistence of the fault. The RMS value U3 has a weight of 0.07, and the Crest Factor U4 has a weight of 0.03, because they are less sensitive to early faults. The total warning score is calculated by comprehensive weighting:

[0107]

[0108] The superscript 'warn' represents the warning level of the corresponding feature, ranging from 1 to 4. The above are recommended weights, which can be adjusted based on the importance of the corresponding feature to fault development, warning sensitivity, and persistence.

[0109] The overall early warning score is calculated using comprehensive weighting;

[0110] The warning level of electrical corrosion of the generator bearing under test is determined based on the total warning score, including normal operation stage, minor fault stage, fault development stage, and serious fault stage, as shown in the following table.

[0111] S<1.5 Normal operation stage 1.5≤S<2.5 Slight corrosion stage <![CDATA[2.5≤U i <3.5]]> Development corrosion stage <![CDATA[U i ≥3.5]]> Severe corrosion stage

[0112] Example 4

[0113] As attached Figure 2 The wind turbine generator bearing electrical corrosion early warning system shown includes a vibration sensor, an ultrasonic sensor, a data processing unit and a model judgment unit;

[0114] The vibration sensor and ultrasonic sensor are used to obtain vibration signals and ultrasonic signals of the generator bearing respectively;

[0115] The data processing unit is used to extract characteristic data from the vibration signal and the ultrasonic signal;

[0116] The model judgment unit is preset with thresholds of various characteristic data, so that the electrical corrosion condition of the generator bearing of the wind turbine generator set can be judged after comprehensive calculation.

[0117] Example 5

[0118] As attached Figure 3 An electronic device as shown is characterized by comprising:

[0119] Processor, memory, communication interface;

[0120] The memory is used to store executable instructions of the processor;

[0121] Wherein, the processor is configured to execute the above-mentioned early warning method for electrical corrosion of the generator bearing of the wind turbine generator set by executing the executable instructions.

[0122] A readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the above-mentioned early warning method for electrical corrosion of a generator bearing of a wind turbine generator is implemented.

[0123] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A wind turbine generator bearing electrical corrosion early warning method, characterized in that: Specifically include: Under the condition of wind turbine operation, the vibration data and ultrasonic data of the generator bearing are obtained respectively; performing feature extraction on the vibration data and the ultrasonic data; Construct an early warning model for electrical corrosion of wind turbine generator bearings; The extracted feature data is brought into the wind turbine generator bearing electrical corrosion early warning model to obtain early warning information.

2. The early warning method for electrical corrosion of a wind turbine generator bearing according to claim 1, characterized in that: The frequency range of the vibration acquisition is 0-10 kHz, and the frequency range of the ultrasonic acquisition is 20-500 kHz.

3. The early warning method for electrical corrosion of a wind turbine generator bearing according to claim 1, characterized in that: An acceleration sensor is used to collect vibration data of the generator bearing in a rigid contact manner, and an ultrasonic sensor is used to monitor ultrasonic signals generated by oil film discharge in the generator bearing.

4. The early warning method for electrical corrosion of a wind turbine generator bearing according to claim 1, characterized in that: The feature extraction step of the vibration signal specifically includes: Perform CEEMDAN decomposition on the vibration signal to obtain several intrinsic mode functions; Screen out the IMF components containing the characteristic frequencies of bearing electrical corrosion faults; Perform synchronous compression transformation on the filtered IMF components to generate high-resolution time-frequency diagrams; The amplitude of the fault characteristic frequency and the amplitude of the fault characteristic harmonic in the time-frequency diagram are extracted as vibration characteristic quantities.

5. The early warning method for electrical corrosion of a wind turbine generator bearing according to claim 1, characterized in that: The feature extraction step of the ultrasonic signal specifically includes: Perform CEEMDAN decomposition on the ultrasonic signal to obtain several eigenmode functions; Screen out the IMF components containing the characteristic frequencies of bearing electrical corrosion faults; Perform synchronous compression transformation on the filtered IMF components to generate high-resolution time-frequency diagrams; The frequency and amplitude of irregular impact in the time domain waveform are counted, and the root mean square and peak factor of the ultrasonic time domain discrete signal are calculated as ultrasonic feature quantities.

6. The early warning method for electrical corrosion of a wind turbine generator bearing according to any one of claims 4 or 5, characterized in that: The ensemble average formula of the CEEMDAN decomposition is: Where x is the original vibration signal; w (n) is the adaptive white noise added for the nth time; β is the control parameter of the noise amplitude; N is the number of noise auxiliary signals; EMD(·) k is the kth IMF component obtained after performing empirical mode decomposition on the noisy signal.

7. The early warning method for electrical corrosion of a wind turbine generator bearing according to any one of claims 4 or 5, characterized in that: The screening process of the IMF components is: Calculate the power spectral density of each IMF component; Identify IMF components that match the fault characteristic frequencies associated with bearing components; The IMF components dominated by high-frequency noise are eliminated, and the effective IMF components reflecting fault information are retained.

8. The early warning method for electrical corrosion of a wind turbine generator bearing according to any one of claims 4 or 5, characterized in that: The calculation formula of the synchronous compression transformation is: Where, T(t,ω) is the short-time Fourier transform result of the IMF component; is the instantaneous frequency estimate; g(η) is the normalization factor; Δη is the frequency bandwidth.

9. The wind turbine generator bearing electrical corrosion early warning method according to claim 4, characterized in that: The extraction formula of the fault characteristic frequency amplitude is: Where SST(τ,f) is the time-frequency energy distribution after synchronous compression transformation; f is the characteristic frequency of the bearing fault; and A(f) is the amplitude corresponding to the fault frequency.

10. The early warning method for electrical corrosion of a wind turbine generator bearing according to claim 1, characterized in that: The wind turbine generator bearing electrical corrosion early warning model specifically includes: Three-level thresholds for vibration and ultrasonic waves are preset respectively; Dynamically update the threshold value according to the operation status; Assign weights to each threshold according to the type and importance of the feature; The total warning score is calculated using comprehensive weighting.

Images