Fault diagnosis and early warning method and system for fan transmission mechanism based on double-signal characteristics

Abstract

The invention provides a fan transmission mechanism fault diagnosis early warning method and system based on double-signal characteristics, and belongs to the technical field of fan transmission mechanism fault diagnosis. Comprising the following steps: synchronously acquiring an impact pulse signal and a vibration signal in the operation process of a fan transmission mechanism; performing preprocessing and characteristic parameter calculation on the obtained signal to obtain an impact characteristic parameter and a vibration characteristic parameter; based on the fault feature frequency of the fan transmission mechanism, a fault feature model is constructed in combination with an equipment fault library, and the fault feature model is adopted for fault diagnosis; and when the fault diagnosis result is that the operation state is abnormal, generating early warning information and executing corresponding early warning operation according to a multi-stage alarm mechanism and a delay alarm algorithm. According to the invention, the early warning of the abnormal operation state of the fan transmission mechanism can be realized while the accurate fault detection of the whole shaft system range of the fan transmission mechanism is ensured.

Description

Technical Field

[0001] This invention belongs to the field of fault diagnosis technology for wind turbine transmission mechanisms, and particularly relates to a fault diagnosis and early warning method and system for wind turbine transmission mechanisms based on dual-signal characteristics. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In the wind power industry, the wind turbine drive mechanism, as the core power transmission component of a wind turbine generator set, directly determines the unit's power generation efficiency and operating costs through its operational stability and continuous operation capability. With the continuous expansion of wind power installed capacity and the increasing years of unit operation, wind turbine drive mechanisms are prone to failure due to long-term high-load and variable-condition operation. Therefore, efficient fault diagnosis and early warning for wind turbine drive mechanisms are of great significance.

[0004] However, existing fault diagnosis or early warning methods for wind turbine drive mechanisms generally suffer from the following technical defects: Currently, fault diagnosis methods for wind turbine drive mechanisms mainly rely on single vibration signal analysis techniques. These methods, through time-domain or frequency-domain analysis of collected vibration signals, can identify some obvious fault characteristics, especially effective in the middle and later stages of a fault. However, because vibration signals are easily affected by low-frequency interference from normal equipment operation and the external environment, these methods are insensitive to early, weak fault impact characteristics, making early fault warning difficult. Furthermore, relying solely on vibration signals for diagnosis under complex operating conditions (such as speed fluctuations and load changes) is prone to misjudgment due to signal confusion, failing to meet the requirements for accurate fault type identification and precise fault location.

[0005] To improve diagnostic performance, some existing technologies have attempted to introduce impact pulse detection to form a simple signal combination with vibration analysis. However, these methods mostly remain at the level of simple data superposition, resulting in short early warning cycles for early faults and insufficient accuracy and reliability of diagnostic conclusions. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, the present invention provides a fault diagnosis and early warning method and system for wind turbine transmission mechanism based on dual signal features, which can achieve early warning of abnormal operating status of wind turbine transmission mechanism while ensuring accurate fault detection across the entire shaft system.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for fault diagnosis and early warning of wind turbine drive mechanism based on dual signal characteristics.

[0008] A fault diagnosis and early warning method for wind turbine drive mechanisms based on dual-signal features includes: Composite data acquisition is performed on the wind turbine transmission mechanism to simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and the obtained impact pulse signal and vibration signal are preprocessed. The characteristic parameters of the preprocessed impact pulse and vibration signal are calculated to obtain the impact characteristic parameters and vibration characteristic parameters. The fault characteristic frequency is determined based on the natural frequency of the key components in the wind turbine transmission mechanism, and a fault characteristic model is constructed by combining it with a preset equipment fault database. The fault characteristic model is then used to diagnose the fault by analyzing the impact characteristic parameters and vibration characteristic parameters. When the fault diagnosis result indicates an abnormal operating status, an early warning message containing the fault type and fault point is generated, and the corresponding early warning operation is executed according to the preset multi-level alarm mechanism and delayed alarm algorithm.

[0009] Furthermore, the preprocessing includes analog-to-digital conversion, envelope filter processing, rectification, smoothing filter fixing, low-pass filtering, sampling, symptom enhancement, and interference suppression operations.

[0010] Furthermore, after extracting the vibration feature parameters corresponding to the vibration signal, the vibration feature parameters are mapped to a preset numerical range to ensure that feature parameters with different dimensions and different numerical ranges can be compared and analyzed in a unified manner.

[0011] Furthermore, the construction of the fault feature model includes: based on the fault feature frequency, and combined with an equipment fault library containing data on multiple types of bearings, gearboxes, and equipment failures, establishing a mapping relationship between impact feature parameters, vibration feature parameters, fault types, and fault points.

[0012] Furthermore, the construction of the fault feature model also includes: First, for rolling bearing faults under non-stationary speeds, the order bispectral analysis method is used to demodulate the envelope of the rolling bearing fault signal, and after estimating the fault characteristic frequency, the envelope signal is resampled at equal angles; then, the non-stationary time domain signal is converted into a stationary angular domain signal, and the resampled signal is subjected to bispectral analysis and diagonal slices are taken to establish the correlation between rolling bearing faults and feature parameters.

[0013] Furthermore, the construction of the fault feature model also includes: superimposing and comparing the real-time monitoring data of the wind turbine transmission mechanism with historical data in the equipment fault database.

[0014] Furthermore, fault diagnosis is performed using a fault characteristic model, including: using a multivariate state estimation method, taking the health data of the wind turbine drive mechanism during normal operation as a benchmark, comparing the correlation between the impact characteristic parameters and vibration characteristic parameters to be analyzed and the health data, and determining whether the operating state of the wind turbine drive mechanism is abnormal through residual analysis.

[0015] Furthermore, the multivariate state estimation method includes: constructing a process memory matrix based on historical observation vectors during normal operation of the wind turbine drive mechanism, the process memory matrix covering the entire dynamic process of normal operation of the wind turbine drive mechanism; subsequently, comparing the real-time extracted feature parameters with the vectors in the process memory matrix, and estimating the difference between the real-time state and the normal state through weight vectors.

[0016] Furthermore, the multi-level alarm mechanism includes three levels: normal, warning, and alarm, with different levels corresponding to different early warning response strategies; the delayed alarm algorithm is used to perform delayed verification of abnormal signals to eliminate false alarms caused by instantaneous interference.

[0017] The second aspect of the present invention provides a fault diagnosis and early warning system for wind turbine drive mechanisms based on dual-signal features.

[0018] A fault diagnosis and early warning system for wind turbine transmission mechanisms based on dual-signal characteristics includes: The data acquisition and conditioning module is configured to: perform composite data acquisition on the wind turbine transmission mechanism, simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and preprocess the obtained impact pulse signal and vibration signal; The host computer software monitoring module is configured to: calculate the characteristic parameters of the pre-processed impact pulse and vibration signal to obtain the impact characteristic parameters and vibration characteristic parameters; The model building module is configured to: determine the fault characteristic frequency based on the inherent frequency of the key components in the wind turbine transmission mechanism, and build a fault characteristic model in combination with a preset equipment fault library. The fault diagnosis module is configured to: use the established fault characteristic model to perform fault diagnosis by analyzing impact characteristic parameters and vibration characteristic parameters; The early warning module is configured to generate early warning information including the fault type and fault point when the fault diagnosis result indicates an abnormal operating status, and execute the corresponding early warning operation according to the preset multi-level alarm mechanism and time-delay alarm algorithm.

[0019] The above one or more technical solutions have the following beneficial effects: This invention employs a composite data acquisition method for the wind turbine drive mechanism, simultaneously acquiring impact pulse signals and vibration signals during the operation of the wind turbine drive mechanism. The obtained impact pulse signals and vibration signals are then preprocessed and characteristic parameters are calculated to obtain impact characteristic parameters and vibration characteristic parameters. By adopting composite monitoring and collaborative analysis, the limitations of single vibration signal analysis in being insensitive to early, subtle faults and susceptible to interference can be effectively overcome. Compared to existing technologies, this invention not only keenly captures subtle impact characteristics characterizing early damage, enabling early fault warning, but also effectively eliminates interference through signal complementarity and verification, significantly improving the accuracy of fault type identification and fault location under complex operating conditions.

[0020] This invention achieves deep fusion and intelligent diagnosis of composite monitoring signals by constructing a fault feature model based on the equipment's inherent frequency and a historical fault database, surpassing simple signal superposition. The fault feature model integrates multi-parameter data and is trained and optimized based on equipment mechanisms and massive historical data, enabling earlier and more accurate identification of fault development trends and patterns. This allows the invention to significantly extend the early warning cycle and greatly improve the accuracy and reliability of fault diagnosis conclusions, providing a solid basis for predictive maintenance decisions.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1 This is a flowchart of the fault diagnosis and early warning method for wind turbine transmission mechanism based on dual signal features in Embodiment 1 of the present invention.

[0024] Figure 2 This is a flowchart of the preprocessing of impact pulse signals and vibration signals in Embodiment 1 of the present invention.

[0025] Figure 3 This is a schematic diagram showing the relationship between some functional modules in Embodiment 1 of the present invention.

[0026] Figure 4 This is a schematic diagram of the display interface corresponding to the monitoring function module in Embodiment 1 of the present invention.

[0027] Figure 5 This is a schematic diagram of the FFT spectrum of a rotor imbalance fault vibration signal in Embodiment 1 of the present invention. Detailed Implementation

[0028] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0030] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0031] Example 1 This embodiment discloses a fault diagnosis and early warning method for wind turbine transmission mechanisms based on dual-signal features.

[0032] like Figure 1 As shown, the fault diagnosis and early warning method for wind turbine drive mechanism based on dual-signal features includes: Step S1: Perform composite data acquisition on the wind turbine transmission mechanism, simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and preprocess the obtained impact pulse signal and vibration signal; Step S2: Calculate the characteristic parameters of the preprocessed impact pulse and vibration signal to obtain the impact characteristic parameters and vibration characteristic parameters; Step S3: Determine the fault characteristic frequency based on the natural frequency of the key components in the wind turbine transmission mechanism, and construct a fault characteristic model in conjunction with a preset equipment fault database; use the constructed fault characteristic model to perform fault diagnosis by analyzing impact characteristic parameters and vibration characteristic parameters. Step S4: When the fault diagnosis result indicates an abnormal operating status, generate early warning information including the fault type and fault point, and execute the corresponding early warning operation according to the preset multi-level alarm mechanism and delayed alarm algorithm.

[0033] Based on the above process, this invention can achieve early warning of abnormal operating conditions of the wind turbine drive mechanism while ensuring accurate fault detection across the entire shaft system. To facilitate understanding of the technical solution of this invention, the specific implementation methods are further explained and described below.

[0034] In step S1, composite data acquisition is performed on the wind turbine transmission mechanism to simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and the obtained impact pulse signal and vibration signal are preprocessed.

[0035] Impact pulse sensors and vibration sensors are installed at key parts of the wind turbine drive mechanism. These key parts include the bearing housing, gearbox input and output shafts, and the free end and input end of the generator.

[0036] Furthermore, the impact pulse sensor employs a built-in metal rod structure, and the vibration sensor utilizes a piezoelectric crystal accelerometer structure. The signals acquired by the impact pulse sensor and vibration sensor are rectified and filtered in a signal acquisition box installed inside the wind turbine nacelle to remove interference signals from the useful signals and improve the clarity of the acquired signals.

[0037] like Figure 2 As shown, the obtained impact pulse signal and vibration signal are preprocessed, including: 1) Analog-to-digital conversion: A 24-bit analog-to-digital converter (ADC) is used with a sampling frequency of 102.4KHz, which ensures high-precision data acquisition and more realistic acquisition of the data spectrum.

[0038] 2) Envelope Filter Processing: This process removes high-energy low-frequency signals generated by imbalance and misalignment, while retaining low-energy high-frequency signals from the wind turbine drive mechanism. As the wind turbine drive mechanism deteriorates, energy shifts from higher to lower frequencies. In the early stages of failure, the increase in mechanical noise floor is more pronounced at higher frequencies (5–10 kHz), while the development of later-stage failures is more evident at lower frequencies (<5 kHz). By applying (digital) filters at different frequencies, early signs of damage can be detected, and the damage process can be tracked.

[0039] 3) Rectification: Converting the negative part of a signal into a positive part, a process accomplished through digital processing.

[0040] 4) Fixed Smoothing Filter: A fixed smoothing filter maintains a constant number of samples per revolution. This is achieved by continuously tracking the rotational speed (in parallel with vibration data acquisition) and adjusting the sampling time interval. Even if the rotational speed changes significantly during data acquisition, the number of samples per revolution remains constant, as the smoothing filter adjusts the number of samples according to the changes in rotational speed.

[0041] 5) Low-pass filter: The cutoff frequency of the low-pass filter is determined by the operating conditions of the fan, and is generally selected as 20Hz.

[0042] 6) Sampling: When the low-pass filter is 40kHz, 102,400 samples are sampled per second, and all are selected; when the low-pass filter is 20kHz, 51,200 samples are sampled per second, and 1 is selected from every 2 samples; when the low-pass filter is 5kHz, 12,800 samples are sampled per second, and 1 is selected from every 8 samples; when the low-pass filter is 10Hz, 25.6 samples are sampled per second, and 1 is selected from every 4,000 samples.

[0043] 7) Symptom enhancement: Symptom enhancement is achieved by enhancing repetitive signals and suppressing random signals. In terms of magnitude, time synchronization averaging is used, while the spectrum is averaged using FFT averaging.

[0044] 8) Interference suppression: Interference suppression is performed as the device rotates. One strong pulse is processed every 10 revolutions. If the default measurement time of 50 revolutions is used, 5 strong pulses are processed.

[0045] 9) Time-domain signal gain: After rectification, filtering, and other processing steps, the time-domain signal waveform is enhanced. For example, after processing the 10 acquired signals, several weak signals will be removed, thus better reflecting the authenticity of the acquired signals.

[0046] 10) High-resolution peak: This reflects the true amplitude of the impact and avoids the influence of interference from single events. The high-resolution peak is calculated by extrapolating the histogram from the slope on the right side of the histogram.

[0047] In step S2, the characteristic parameters of the preprocessed impact pulse and vibration signal are calculated to obtain the impact characteristic parameters and vibration characteristic parameters, and then time domain analysis and frequency domain analysis are performed.

[0048] Time-domain analysis was performed on the impact pulse signal and vibration signal respectively to obtain time-domain characteristic parameters. Frequency-domain analysis was also performed on the aforementioned impact signal and vibration signal respectively to obtain frequency-domain characteristic parameters. The calculation methods for each parameter are defined, including: a) Valid values. The calculation method for valid values ​​is as follows: ; in, Indicates the effective value of the vibration signal. Indicates the vibration signal at Amplitude at any given moment; Indicates the number of sampling points.

[0049] b) Peak-to-peak value. The peak-to-peak value is calculated as follows: ; in, Peak-to-peak value The maximum value of the vibration signal. This represents the minimum value of the vibration signal.

[0050] c) Kurtosis. The kurtosis is calculated as follows: ; in, As a kurtosis index; The mean; The standard deviation is, and ; Indicates sample size. Indicates the first One observation value.

[0051] d) Impulse Index. The impulse index is calculated as follows: ; in, For pulse indicators, This is the absolute average.

[0052] e) Skewness. The skewness is calculated as follows: ; in, For valid values, This indicates the skewness value.

[0053] The pre-processed impact pulses and vibration signals are then subjected to secondary processing: root mean square (RMS) calculations are performed on the multiple sets of impact pulse signals and multiple sets of vibration signals to obtain the feature calculation values ​​corresponding to each set of signals, in order to suppress the influence of sudden impacts on feature parameter extraction.

[0054] In the specific implementation process, multiple consecutive sets of data are collected each time, with each set containing 1024 sampling points. By calculating the root mean square value (i.e., effective value) of multiple sets (e.g., 10 sets) of data and taking the average, sudden impact signals can be effectively smoothed, highlighting regular fault characteristics. The results are then stored as feature calculation values. Feature extraction is performed on the preprocessed impact pulse signal and vibration signal. This feature extraction process can be achieved through time-domain feature extraction and frequency-domain feature extraction. 1) Calculation of time-domain characteristic parameters.

[0055] The impact pulse is the impact value of carpet value (background noise level) and high-frequency energy.

[0056] The time-domain parameters of the preprocessed impact pulse signal and vibration signal are calculated separately. The time-domain parameters of the impact pulse signal include peak value (maximum value of a single impact), carpet value (background noise level), and pulse index. The time-domain characteristic parameters of the vibration signal include peak-to-peak value, kurtosis, and skewness. Further, the time-domain characteristic parameters of the impact pulse signal, such as peak value and carpet value, can be specifically expressed as follows: ; ; in, Indicates an impact pulse signal. This indicates the carpet value of the impact pulse signal; Indicates the signal type. Indicates carrier wave, Indicates actual power. This represents 1 milliwatt.

[0057] 2) Calculation of frequency domain characteristic parameters.

[0058] The vibration signal is obtained by performing a Fourier transform (FFT) on the vibration signal. The amplitudes corresponding to the rotor's first and second harmonics are then searched within the FFT spectrum. The calculation of these characteristic parameters and the FFT spectrum are performed by the system software, and the relevant characteristic parameters are stored in a database. When equipment malfunctions or its operating condition deteriorates, auxiliary analysis is performed by querying the relevant characteristic parameters in the database. Furthermore, the relationship between the integrated data acquisition and conditioning module and the monitoring function module within the system is as follows: Figure 3 As shown, the display interface corresponding to the monitoring function module is as follows: Figure 4 As shown, the data in the data acquisition and conditioning module is sent to the monitoring function module for time-domain waveform display, FFT spectrum calculation, time-domain characteristic parameter calculation, dimensionless index calculation, and frequency domain parameter calculation, and is displayed in real time.

[0059] 3) Time-domain analysis, including: filtering, probability density analysis, autocorrelation function analysis, cross-correlation function analysis, and bar plot display.

[0060] a) Filtering. Filtering methods include low-pass filtering, high-pass filtering, band-pass filtering, and band-stop filtering.

[0061] b) Probability Density. The probability density of a signal represents the probability that its amplitude falls within a specified interval. When the signal changes, its probability density waveform also changes. Therefore, signal identification can be achieved by analyzing the probability density waveform. If the original signal is... Then the probability density of the signal for: ; in, The maximum value in the signal. It is the minimum value in the signal; The number of equal segments to divide the amplitude interval. The number of data points that fall within the specified amplitude range.

[0062] c) Autocorrelation function. Autocorrelation analysis is a commonly used method in time-domain analysis. The autocorrelation function of a signal can highlight the periodic components and suppress the non-periodic components. Therefore, the autocorrelation function of a signal can also be used as a method to extract the periodic components of a signal. If the original signal is... Then the autocorrelation function of the signal for: ; in, Indicates the periodic time. Indicates the magnitude of change over time.

[0063] d) Cross-correlation function. Cross-correlation analysis is used to determine the correlation between two signals. When the two signals have the same frequency, the cross-correlation function is a periodic component of the same frequency; when the two signals have different frequencies, the cross-correlation function is zero, meaning the two signals are uncorrelated. If the two original signals are respectively... , To represent, the cross-correlation function of the signal for: .

[0064] e) Bar plot. When a vibration signal changes, its time-domain characteristic parameters (mean, peak value, peak-to-peak value, RMS value, etc.) also change. By analyzing the time-domain characteristic parameters of the signal, the degree of signal change can be estimated. A bar plot can visually reflect the time-domain characteristic parameters of the signal, including the mean, peak value, peak-to-peak value, RMS value, and root mean square error.

[0065] 4) Frequency domain analysis, including: FFT spectrum, self-power spectrum, cross-power spectrum, cepstrum, envelope spectrum, waterfall plot and order spectrum.

[0066] a) FFT Spectrum. The FFT spectrum of a signal reveals its various frequency components. For mechanical equipment, when abnormal operating conditions occur, the frequency components of the vibration signal will change. Therefore, amplitude spectrum analysis can effectively determine the operating status of the equipment. For example, the FFT spectrum of a rotor imbalance fault vibration signal at a rotor speed of 3000 r / min would look like this: Figure 5 As shown, when the rotor speed is 3000 r / min, an imbalance fault signal occurs, and its vibration amplitude changes accordingly; conversely, the vibration amplitude remains stable.

[0067] b) Autopower Spectrum. The autopower spectrum reflects the energy of each frequency component in the signal, and can be determined from the signal's FFT spectrum. If X ( f If is the FFT spectrum of the signal, then the power spectrum of the signal can be expressed as: ; in, This represents the power spectrum of the original signal.

[0068] c) Cross-power spectrum. The cross-power spectrum is a frequency domain description of the correlation between two signals, describing the amplitude and phase relationship between them. Similar to the self-power spectrum, the cross-power spectrum can also be determined from the FFT spectra of the two signals. If the FFT spectra of the two signals are respectively... and Then the cross-power spectrum can be expressed as: ; Among them, the interactive power spectrum Indicates two signals and In frequency The cross power spectral density at a certain point is used to analyze the linear correlation and phase relationship between two signals in the frequency domain; and Corresponding to and The discrete Fourier transform; This indicates the complex conjugate operation.

[0069] d) Cepstrum. Cepstrum analysis is also a commonly used method for analyzing mechanical vibration signals. It separates sideband signals, transforming periodic components that are difficult to distinguish in the power spectrum into discrete line spectra in the cepstrum diagram. Cepstrum analysis is widely used in analyzing gear fault signals with many sideband components. Its calculation method is as follows: ; in, The cepstrum obtained after transformation This represents the inverse Fourier transform.

[0070] e) Envelope Spectrum. When mechanical equipment malfunctions, low-frequency and high-frequency components often modulate each other. For example, when localized damage occurs on the surface of a rolling bearing component, periodic pulse forces are generated during operation, exciting the high-frequency inherent vibrations of the bearing components, resulting in modulation. To extract the low-frequency components from the modulated signal, demodulation is required. In practice, bandpass filtering is typically used to select a frequency band of interest, and demodulation is performed on the signal in that band. Hilbert envelope demodulation is a commonly used demodulation method, and the amplitude spectrum of the demodulated signal is the Hilbert envelope spectrum. The calculation method for the Hilbert envelope spectrum is as follows: Original signal The result obtained after Hilbert transform is: ; in, This represents the parameter obtained after performing Hilbert operations on the original signal.

[0071] Using the original signal x (t ) and Hilbert transform The analytical signal can be obtained: ; in, This represents the transformed analytic signal. It represents the imaginary unit 1.

[0072] Furthermore, analyze the signal. The magnitude function can be expressed as: ; in, That is, the envelope of the real signal, which is calculated. The envelope spectrum of a signal can be obtained from its amplitude spectrum or power spectrum.

[0073] f) Waterfall Plot. The waterfall plot is a commonly used method for analyzing vibration signals during the acceleration and deceleration process of rotating machinery. It is obtained by calculating the spectrum of vibration signals collected at different speeds or times and plotting it as a three-dimensional spectrum. The waterfall plot clearly shows the changes in the frequency and amplitude of each vibration component over time or speed.

[0074] g) Order Spectrum. Vibration signals from rotating machinery exhibiting changes in rotational speed are typical non-stationary signals, unsuitable for conventional spectral analysis. Order ratio analysis is a novel non-stationary signal processing method developed in recent years. It transforms time-domain non-stationary signals into angular-domain stationary signals through equal-angle sampling. Analyzing the angular-domain stationary signals allows for the extraction of non-stationary characteristics from the vibration signal. Traditional order ratio analysis methods are divided into hardware-based order ratio tracking and computational order ratio tracking. Hardware-based order ratio tracking achieves equal-angle sampling of the rotor using hardware to obtain angular-domain stationary signals. Computational order ratio tracking simultaneously samples the original vibration signal and rotational speed pulse signal at equal time intervals. The rotational speed pulse signal is used to obtain the rotor's rotational speed information, which determines the timing of equal-angle sampling. Interpolation sampling of the original vibration signal is then performed at these times to achieve equal-angle sampling. The computational order ratio tracking method is used to perform order ratio analysis on vibration signals during acceleration and deceleration. The computational order ratio tracking method is as follows: The calculation of order tracking assumes that the reference axis rotation speed undergoes uniform acceleration within a short time interval. Under this premise, the rotation angle of the reference axis... It can be represented as: ; in, , and There are three undetermined coefficients.

[0075] Determine the rotation increment corresponding to the bond phase pulse If the reference axis generates one key phase pulse per revolution, then The undetermined coefficients in the above formula can be obtained by fitting the arrival times of three consecutive pulses. , and get: ; Once the undetermined coefficients are obtained, the sampling time at equal angles can be determined by the following formula: ; in, The sampling interval is equal in angle. At the equal in angle sampling time... By performing interpolation on the original signal, equal-angle sampled data can be obtained. This is determined by the equal-angle sampling interval. The sampling order ratio can be obtained. The order spectrum of the vibration signal during acceleration and deceleration can be obtained by performing amplitude spectrum analysis on the sampled data at equal angles.

[0076] 5) Shaft Center Trajectory Analysis. Shaft center trajectory analysis involves using two sensors with a 90-degree angle between them to measure the rotor's vibration in the X and Y directions. The rotor's shaft center trajectory is obtained by plotting the X (Y) direction vibration data on the x-axis and the Y (X) direction vibration data on the y-axis. The rotor's shaft center trajectory typically changes when its operating state changes. For example, when the rotor experiences an imbalance fault, the shaft center trajectory is elliptical; when the rotor is misaligned, the shaft center trajectory is banana-shaped.

[0077] 6) Trend Analysis. The electromechanical equipment condition monitoring system periodically stores relevant vibration characteristic parameters. By plotting the time history curves of these parameters, the vibration trend of the relevant electromechanical equipment can be analyzed. Taking the effective value of vibration acceleration as an example, if the effective value of vibration acceleration of a certain electromechanical equipment does not change much over a certain period of time (such as one year), it indicates that the equipment is operating stably. If the effective value of vibration acceleration of a certain electromechanical equipment shows an increasing trend, it indicates that the operating condition of the equipment is gradually deteriorating, and the changes in the condition of the equipment need to be closely monitored.

[0078] In step S3, the fault characteristic frequency is determined based on the natural frequency of the key components in the wind turbine drive mechanism, and a fault characteristic model is constructed in conjunction with the preset equipment fault database. The fault characteristic model is then used to diagnose the fault by analyzing the impact characteristic parameters and vibration characteristic parameters.

[0079] Based on the fault symptom domain, the fault early warning method for the transmission device involves matching the fault symptom domain with the standard parameter domain to determine if the transmission device has a fault trend. If a fault trend is determined, fault location is performed based on the fault symptom domain and the fault cause domain. The fault cause domain uses the target characteristic parameters of component failure from Research Content 1, requiring at least one of them to meet the requirements of time-domain characteristic parameters, frequency-domain characteristic parameters, relative values ​​of measurement results, shaft center characteristic parameters, and temperature values ​​of temperature signals. In the fault location process, firstly, the membership degree of each target element in the fault symptom domain is determined, and the membership degree of the target elements constitutes the fault symptom matrix. Then, the membership degree of the target characteristic parameters in the fault cause domain is determined, and the membership degree of the target characteristic parameters constitutes the fault cause matrix. Based on the fuzzy fault matrix and the fuzzy relation matrix, the fuzzy cause matrix is ​​determined. Based on the fuzzy cause matrix, the failure probability of the target faulty component is determined. Finally, fault location is performed based on the fault symptom matrix, the fault cause matrix, and the fuzzy relation matrix. The implementation method is as follows: Fuzzy fault diagnosis methods utilize membership functions and fuzzy relation matrices from fuzzy set theory to describe the relationship between faults and features. The fault diagnosis process based on fuzzy theory includes: There are two domains of discourse: the domain of fault symptoms. U and the domain of fault causes V Fault Symptoms Domain U Represented as: ; Among them, each element The corresponding membership degree The fuzzy fault vector is formed as The cause of the malfunction is Each element membership degree The fuzzy cause vector is .

[0080] There exists a fuzzy relationship between the two universes, which can be represented by a fuzzy relationship matrix, i.e.: ; If the fuzzy relation matrix is ​​known and fuzzy fault symptom vector Then the fuzzy cause vector can be obtained. The calculation formula is as follows: ; ; Based on the calculated fault cause vector, the probability of each fault occurring can be calculated. Generally, the fault cause corresponding to the highest probability is the actual fault cause of the equipment. As fuzzy fault diagnosis theory suggests, the fault symptom vector should be clearly defined when constructing a fuzzy fault diagnosis system. Fuzzy cause vector and fuzzy relation matrix .

[0081] Faults in rotating equipment typically include rotor imbalance, misalignment, looseness, and bearing failure. For example, when a rotor is unbalanced, the amplitude of the first rotational frequency in the vibration signal is relatively large; when a rotor is misaligned, the amplitude of the second rotational frequency in the vibration signal spectrum is relatively large; when the rotor support is loose, higher harmonics of the rotational frequency will appear in the vibration signal spectrum, sometimes accompanied by fractional harmonics; when a bearing failure occurs, there are usually more impacts in the vibration signal, the kurtosis value of the vibration signal will increase, and a resonance band will appear in the high-frequency part of the spectrum. Therefore, the kurtosis value and the energy of the resonance band can be considered as two main criteria for evaluating bearing failure.

[0082] By performing FFT analysis on the vibration data and identifying the amplitude of characteristic frequencies in the FFT spectrum using a computer, the fault symptom vector of the generator can be determined. ; in, Each corresponds to the membership degree of its respective characteristic frequency.

[0083] To place the values ​​of each membership degree between [0,1], fuzzy membership functions can be used to fuzzify the amplitudes corresponding to the above frequencies. The fuzzy membership function is as follows: ; in, x The amplitude corresponding to the above frequencies, and the coefficient. a It is usually set to 0.

[0084] Fuzzy cause vector Y Fault symptom vectors And fuzzy relation matrix Multiplying them together yields the result, i.e. According to the fuzzy cause vector Y This allows you to calculate the probability of each fault occurring.

[0085] In step S4, when the fault diagnosis result indicates an abnormal operating status, an early warning message containing the fault type and fault point is generated, and the corresponding early warning operation is executed according to the preset multi-level alarm mechanism and delayed alarm algorithm.

[0086] When the fault diagnosis result indicates an abnormal operating status, an early warning message is generated, including the fault type and fault location. In the specific implementation process, the fault diagnosis module outputs structured data, including the device ID and measurement point location, fault type (such as "bearing outer ring damage", "gear meshing failure"), fault location (such as "wind turbine drive shaft rear bearing"), timestamp, and severity level; then, the above structured data is encapsulated into JSON or XML format and sent to the early warning module through a message middleware (such as RabbitMQ).

[0087] After generating an early warning message, the corresponding warning operation is executed according to the preset multi-level alarm mechanism and delayed alarm algorithm. The multi-level alarm mechanism includes three levels: normal, warning, and alarm. When the level is normal, no operation is performed, and the status is displayed in green. When the level is warning, the status is displayed in yellow, a log is recorded, and an email notification is sent to maintenance personnel. When the level is alarm, the status is displayed in red, an audible alarm is triggered, and an SMS and APP push are sent to management personnel.

[0088] The delayed alarm algorithm is used to verify abnormal signals over time, eliminating false alarms caused by transient interference. In implementation, the initial state is normal; when an anomaly is detected, it enters a waiting state and starts a delay timer (e.g., 10 minutes); while in the waiting state, if the anomaly persists, the anomaly time is accumulated; if the anomaly disappears, it returns to the normal state; when the accumulated anomaly time exceeds a threshold (e.g., 8 minutes), it switches to alarm state. Through delayed verification, short-term interference (such as external shocks and electrical noise) is filtered out, ensuring alarm accuracy.

[0089] Example 2 This embodiment discloses a fault diagnosis and early warning system for wind turbine transmission mechanisms based on dual-signal characteristics.

[0090] A fault diagnosis and early warning system for wind turbine transmission mechanisms based on dual-signal characteristics includes: The data acquisition and conditioning module is configured to: perform composite data acquisition on the wind turbine transmission mechanism, simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and preprocess the obtained impact pulse signal and vibration signal; The host computer software monitoring module is configured to: calculate the characteristic parameters of the pre-processed impact pulse and vibration signal to obtain the impact characteristic parameters and vibration characteristic parameters; The model building module is configured to: determine the fault characteristic frequency based on the inherent frequency of the key components in the wind turbine transmission mechanism, and build a fault characteristic model in combination with a preset equipment fault library. The fault diagnosis module is configured to: use the established fault characteristic model to perform fault diagnosis by analyzing impact characteristic parameters and vibration characteristic parameters; The early warning module is configured to generate early warning information including the fault type and fault point when the fault diagnosis result indicates an abnormal operating status, and execute the corresponding early warning operation according to the preset multi-level alarm mechanism and time-delay alarm algorithm. Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0091] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A fault diagnosis and early warning method for wind turbine transmission mechanisms based on dual-signal characteristics, characterized in that, include: Composite data acquisition is performed on the wind turbine transmission mechanism to simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and the obtained impact pulse signal and vibration signal are preprocessed. The characteristic parameters of the preprocessed impact pulse and vibration signal are calculated to obtain the impact characteristic parameters and vibration characteristic parameters. The fault characteristic frequency is determined based on the natural frequency of the key components in the wind turbine transmission mechanism, and a fault characteristic model is constructed by combining it with a preset equipment fault database. The fault characteristic model was established, and fault diagnosis was performed by analyzing the impact characteristic parameters and vibration characteristic parameters. When the fault diagnosis result indicates an abnormal operating status, an early warning message containing the fault type and fault point is generated, and the corresponding early warning operation is executed according to the preset multi-level alarm mechanism and delayed alarm algorithm.

2. The method for fault diagnosis and early warning of wind turbine transmission mechanism based on dual-signal features as described in claim 1, characterized in that, The preprocessing includes analog-to-digital conversion, envelope filter processing, rectification, smoothing filter fixing, low-pass filtering, sampling, symptom enhancement, and interference suppression operations.

3. The fault diagnosis and early warning method for wind turbine transmission mechanism based on dual-signal features as described in claim 1, characterized in that, After extracting the vibration feature parameters corresponding to the vibration signal, the vibration feature parameters are mapped to a preset numerical range to ensure that feature parameters with different dimensions and different numerical ranges can be compared and analyzed in a unified manner.

4. The method for fault diagnosis and early warning of wind turbine transmission mechanism based on dual-signal features as described in claim 1, characterized in that, The construction of the fault feature model includes: based on the fault feature frequency, and combined with an equipment fault library containing data on multiple types of bearings, gearboxes, and equipment failures, establishing a mapping relationship between impact feature parameters, vibration feature parameters, fault types, and fault points.

5. The fault diagnosis and early warning method for wind turbine transmission mechanism based on dual-signal features as described in claim 4, characterized in that, The construction of the fault feature model also includes: First, for rolling bearing faults under non-stationary speeds, the order bispectral analysis method is used to demodulate the envelope of the rolling bearing fault signal, and after estimating the fault feature frequency, the envelope signal is resampled at equal angles; then, the non-stationary time domain signal is converted into a stationary angular domain signal, and the resampled signal is subjected to bispectral analysis and diagonal slices are taken to establish the correlation between rolling bearing faults and feature parameters.

6. The method for fault diagnosis and early warning of wind turbine transmission mechanism based on dual-signal features as described in claim 4, characterized in that, The construction of the fault feature model also includes: superimposing and comparing the real-time monitoring data of the wind turbine transmission mechanism with historical data in the equipment fault database.

7. The method for fault diagnosis and early warning of wind turbine transmission mechanism based on dual-signal features as described in claim 1, characterized in that, Fault diagnosis is performed using a fault characteristic model, including: using a multivariate state estimation method, taking the health data of the wind turbine drive mechanism during normal operation as a benchmark, comparing the correlation between the impact characteristic parameters and vibration characteristic parameters to be analyzed and the health data, and determining whether the operating state of the wind turbine drive mechanism is abnormal through residual analysis.

8. The method for fault diagnosis and early warning of wind turbine transmission mechanism based on dual-signal features as described in claim 7, characterized in that, The multivariate state estimation method includes: constructing a process memory matrix based on historical observation vectors during normal operation of the wind turbine drive mechanism, the process memory matrix covering the entire dynamic process of normal operation of the wind turbine drive mechanism; subsequently, comparing the real-time extracted feature parameters with the vectors in the process memory matrix, and estimating the difference between the real-time state and the normal state through weight vectors.

9. The method for fault diagnosis and early warning of wind turbine transmission mechanism based on dual-signal features as described in claim 1, characterized in that, The multi-level alarm mechanism includes three levels: normal, warning, and alarm, with different levels corresponding to different early warning response strategies; the delayed alarm algorithm is used to perform delayed verification of abnormal signals to eliminate false alarms caused by transient interference.

10. A fault diagnosis and early warning system for wind turbine transmission mechanisms based on dual-signal characteristics, characterized in that, include: The data acquisition and conditioning module is configured to: perform composite data acquisition on the wind turbine transmission mechanism, simultaneously acquire the impact pulse signal and vibration signal during the operation of the wind turbine transmission mechanism, and preprocess the obtained impact pulse signal and vibration signal; The host computer software monitoring module is configured to: calculate the characteristic parameters of the pre-processed impact pulse and vibration signal to obtain the impact characteristic parameters and vibration characteristic parameters; The model building module is configured to: determine the fault characteristic frequency based on the inherent frequency of the key components in the wind turbine transmission mechanism, and build a fault characteristic model in combination with a preset equipment fault library. The fault diagnosis module is configured to: use the established fault characteristic model to perform fault diagnosis by analyzing impact characteristic parameters and vibration characteristic parameters; The early warning module is configured to generate early warning information including the fault type and fault point when the fault diagnosis result indicates an abnormal operating status, and execute the corresponding early warning operation according to the preset multi-level alarm mechanism and time-delay alarm algorithm.

Images