Method for monitoring and diagnosing yaw speed reducer coping with intermittency and variable working conditions

Abstract

The invention relates to the technical field of mechanical fault diagnosis and signal processing analysis, in particular to a yaw speed reducer monitoring and diagnosis method for dealing with intermittency and variable working conditions, which comprises the following steps: S1, acquiring a vibration signal of a yaw speed reducer of a wind turbine generator; s2, according to the amplitude difference of the vibration signals in the action period and the non-action period of the yaw speed reducer; s3, the identified effective vibration signals are amplitude-modulated and frequency-modulated signals, and rotating speed information is estimated by adopting a ridge extraction method based on a cost function; s4, integrating the estimated rotating speed and the identified effective vibration signal; s5, calculating the theoretical fault feature order of each component according to the structural parameters of the speed reducer; according to the method, the adverse effect of intermittent action of the yaw system on data acquisition, storage and monitoring diagnosis is overcome, the feature order corresponding to the fault type is accurately represented, and the adverse effect of the time-varying operation condition of the yaw speed reducer on fault feature extraction and recognition is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault diagnosis technology and signal processing and analysis, and particularly relates to a monitoring and diagnosis method for a yaw reducer that can cope with intermittent and variable working conditions. Background Art

[0002] The wind power yaw reducer is a key component in a wind turbine generator set, which is used to adjust the orientation of the wind turbine generator set's rotor so that it always faces the wind direction to obtain the maximum wind energy. Since wind turbine generator sets are usually installed in outdoor environments and are affected by wind speed and wind direction, the wind power yaw reducer needs to be frequently started and stopped, resulting in intermittent and variable working conditions. This operating characteristic poses a great challenge to the monitoring and diagnosis of the reducer. Especially under complex working conditions, traditional monitoring methods are difficult to accurately judge the working state of the reducer.

[0003] Currently, common monitoring methods mainly include vibration signal-based monitoring, temperature signal-based monitoring, and oil analysis-based monitoring. However, these methods face some challenges when dealing with intermittent and variable working conditions. For example, temperature signals have a large lag, and oil analysis requires regular sampling and cannot achieve real-time monitoring. Although vibration signals can timely characterize the health state, they are easily affected by time-varying working conditions.

[0004] With the development of wind power technology, the capacity and scale of wind turbine generator sets are constantly expanding, posing higher requirements for the reliability and maintenance of wind power yaw reducers. In order to improve the operating reliability of wind power yaw reducers and extend their service life, there is an urgent need for a monitoring and diagnosis method for wind power yaw reducers that can cope with intermittent and variable working conditions to achieve real-time and accurate monitoring of the working state of the reducer and fault diagnosis, and ensure the stable operation of wind turbine generator sets.

[0005] Therefore, the present invention aims to provide a new monitoring and diagnosis method that can effectively cope with the operating characteristics of wind power yaw reducers under intermittent and variable working conditions, improve the monitoring accuracy and fault diagnosis ability of the reducer, and meet the reliable operation requirements of wind turbine generator sets. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a monitoring and diagnosis method for a yaw reducer that can cope with intermittent and variable working conditions, which overcomes the adverse effects of the intermittent operation of the yaw system on data acquisition, storage, monitoring, and diagnosis, accurately characterizes the characteristic orders corresponding to fault types, and avoids the adverse effects of the time-varying operating conditions of the yaw reducer on fault feature extraction and recognition.

[0007] A monitoring and diagnosis method for a yaw reducer that can cope with intermittent and variable working conditions of the present invention includes the following steps:

[0008] S1. Obtain the vibration signal of the yaw reducer of the wind turbine. The vibration signal is from the unit status monitoring system and covers the action and non-action time periods of the yaw system.

[0009] S2. According to the amplitude difference of the vibration signals during the action and non-action periods of the yaw reducer, use sliding median filtering to process the collected signals and identify the effective vibration signals in the action interval.

[0010] S3. The identified effective vibration signal is an amplitude-frequency modulation and frequency modulation signal. Use the ridge extraction method based on the cost function to estimate the rotational speed information and provide signal prior for subsequent cyclic non-stationary analysis.

[0011] S4. Combine the estimated rotational speed and the identified effective vibration signal, calculate the cyclic modulation spectrum, and reveal the time-angle duality characteristics corresponding to the fault pulses.

[0012] S5. Calculate the theoretical fault characteristic orders of each component according to the reducer structure parameters, and compare the theoretical values with the spectral lines in the cyclic modulation spectrum to locate the faulty components.

[0013] Preferably, the sliding median filtering in S2 is used to identify the effective vibration signals. The amplitude of the effective vibration signals is higher than the set threshold, and the signals lower than the threshold are determined as invalid signals and deleted.

[0014] Preferably, in S3, the cost function ridge extraction method is used to estimate the instantaneous frequency of the vibration signal, and the instantaneous shaft rotational speed signal is estimated according to the proportional coefficient between the target component and the shaft rotational speed.

[0015] Preferably, in S4, the cyclic modulation spectrum analysis performs Fourier series expansion on the vibration signal through the angle-time autocorrelation function and extracts the characteristic frequency information corresponding to the reducer fault type.

[0016] Preferably, in S5, the calculation of the fault characteristic orders includes the calculation of the local fault characteristic orders of the sun gear, planet gears, and ring gear, which respectively correspond to different types of gear damages.

[0017] Preferably, the rotational speed estimation method uses short-time Fourier transform to calculate the time-frequency representation of the vibration signal and selects the target component for analysis in the time-frequency domain.

[0018] Preferably, the specific principle of identifying the effective vibration signals in the action interval in S2 is as follows:

[0019] The wind direction sensor of the yaw system is installed at the top of the wind turbine. By measuring the wind direction, it provides wind direction data to the control system. When the wind direction changes and deviates from the set threshold range, the control system issues a yaw command to adjust the orientation of the wind turbine rotor. The yaw drive mechanism drives the yaw reducer through a gear transmission device according to the command of the control system, causing the nacelle of the entire wind turbine to rotate on top of the tower and adjust the orientation of the wind turbine rotor. Wind resources are random, and the action time of the yaw system cannot be predicted. Therefore, the vibration signals collected on the yaw reducer are intermittent, and its effective vibration signals have the characteristics of amplitude modulation and frequency modulation. Identifying effective vibration signals by means of amplitude difference is the basic connotation of overcoming intermittency;

[0020] First, calculate the envelope signal of the collected signal: a(t) = |x(t) + jH[x(t)]|, where H[·] is the Hilbert transform operator. Then, perform sliding median filtering with a window length of N w and a unit sliding step size: a s (t) = MMF[a(t)], where MMF[·] is the sliding median filter. Finally, calculate 3σ as the identification threshold according to a s (t) during the period when the yaw reducer is not operating, and compare a s (t) of the collected signal with the identification threshold. If it is higher than the threshold, it is determined as an effective vibration signal and stored; otherwise, it is determined as an invalid vibration signal and deleted.

[0021] Preferably, in step S3, the specific principle of estimating the rotational speed information by using the ridge extraction method based on the cost function is as follows:

[0022] Under time-varying working conditions, the effective vibration signal of the yaw reducer is an amplitude-modulated and frequency-modulated signal, and the instantaneous frequency of its components characterizes the shaft rotational speed information. First, the time-frequency representation of the effective vibration signal is calculated by using the short-time Fourier transform as follows:

[0023] w(t) is the window function. Then, select a target component on the TFR and use the cost function ridge extraction method to estimate the instantaneous frequency:

[0024] To avoid interference from adjacent components and reduce the parameter search range, construct a time-varying frequency band with the previous moment as the center frequency for the frequency bandwidth to be searched: Finally, solve the instantaneous shaft rotational speed signal according to the proportional coefficient between the target component and the shaft rotational speed:

[0025] Preferably, in step S4, the specific principle of calculating the cyclic modulation spectrum analysis is as follows:

[0026] The fault pulses of the yaw reducer under time-varying speed have time-angle duality. The position of the fault impact is determined by the angular position of the shaft, related to the kinematics of the reducer, synchronized with the machine speed. The resonance frequency and decay time of the fault impact are time-invariant, related to the dynamics of the system, and are the solutions of the time-difference equation. The angle-time cyclic stationary analysis method can analyze signals across the angular domain and time domain simultaneously, facilitating the extraction of subtle reducer fault characteristics under different operating conditions. The angle-time autocorrelation function is periodic, and its Fourier series expansion is:

[0027]

[0028] Among them, the Fourier coefficient is expressed as: The spectral correlation density function is the Fourier transform of the angle-time autocorrelation function: a represents the proportional relationship between the fault characteristic frequency and the rotational frequency, and f is the spectral frequency in the traditional sense.

[0029] Preferably, in S5, the calculation principle of the theoretical fault characteristic order of each component is as follows:

[0030] The yaw reducer is composed of a cascaded planetary gear train. The calculation methods of the fault characteristic orders of various types of gears are as follows:

[0031] If there is local damage to a certain tooth of the sun gear, within one rotation period relative to the planet carrier, the faulty tooth will mesh with all planet gears and generate impacts. Therefore, the local fault characteristic frequency of the sun gear is If there is local damage to a certain tooth of the planet gear, within 1 rotation period relative to the planet carrier, the faulty tooth meshes with both the sun gear and the ring gear, generating 1 impact respectively. The local fault characteristic frequency of the planet gear is: If there is local damage to a certain tooth of the ring gear, within one rotation period relative to the planet carrier, the faulty tooth will mesh with all planet gears and generate impacts. Therefore, the local fault characteristic frequency of the ring gear is According to the model and size parameters of the bearing, the fault characteristic order of the bearing outer ring is: The inner ring fault characteristic order of the inner ring is: The fault characteristic order of the rolling element is: Compare the theoretical fault characteristic order with the dominant a in the spectral correlation density function to identify the faulty component.

[0032] The beneficial effects of the present invention compared with the prior art are as follows: By using sliding median filtering to identify the effective vibration signals of the yaw reducer, the adverse effects of the intermittent operation of the yaw system on data acquisition, storage, monitoring, and diagnosis are overcome. At the same time, by combining the rotational speed estimation method based on the cost function time-frequency ridge extraction method with the angle-time cyclic stationary analysis, the characteristic orders corresponding to the fault types are accurately characterized, avoiding the adverse effects of the time-varying operating conditions of the yaw reducer on fault feature extraction and identification.

[0033] Identifying the effective vibration signals of the yaw reducer and then diagnosing the faults of the yaw reducer under variable operating conditions through angle-time cyclic stationary analysis without a tachometer is the core of the present invention. Through this invention, the fault diagnosis of the yaw reducer can be effectively carried out, reasonably assisting the enterprise's operation and maintenance decision-making. By analyzing the composition of the vibration signals continuously collected from the yaw reducer, a method for accurately identifying the effective vibration signals is designed to overcome the problem of untimely monitoring caused by the intermittent operation of the yaw reducer. Under time-varying operating conditions, the fault pulses in the vibration signals have time-angle duality. A cyclic non-stationary analysis method without a tachometer is designed to overcome the problem of missed diagnosis caused by time-varying operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is the flowchart of the method of the present invention;

[0035] Figure 2 is an example diagram of the time domain of the vibration signals of the yaw reducer collected by the present invention;

[0036] Figure 3 is an example diagram of the effective vibration signals identified after sliding median filtering by the present invention and an example diagram of the instantaneous shaft rotational speed of the estimated effective vibration signals;

[0037] Figure 4 is an example diagram of the angle-time cyclic stationary analysis of the present invention;

[0038] Figure 5 is the architecture diagram of the short-term wind direction prediction model of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. The present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0040] Embodiment

[0041] As Figures 1 to 5 shown, a method for monitoring and diagnosing a yaw reducer to cope with intermittency and variable operating conditions of the present invention adopts the following steps:

[0042] S1. As shown in Figure 1 , the overall time-domain waveform of the collected yaw reducer vibration signal contains effective vibration signals and high-amplitude interference components. Therefore, only by identifying the effective vibration first and then analyzing and storing it can the reliability of the yaw reducer be ensured;

[0043] S2. As shown in Figure 2 , the envelope a(t) = |x(t) + jH[x(t)]| of the collected signal is solved through Hilbert transform. The collected signal segments greater than the threshold are identified as effective vibration signals and stored, while those less than the threshold are identified as invalid vibration signals and discarded by comparing the sliding median filtered signal of the envelope signal with the recognition threshold. Among them, due to the transient high-amplitude characteristics of high-amplitude pulse interference, its amplitude is smoothed during the median filtering process to avoid being misidentified as an effective vibration signal. The recognition threshold is determined through manual analysis and is set to 0.1378 m / s in the present invention 2 ;

[0044] S3. As shown in Figure 4 , first, the short-time fractional Fourier transform time-frequency diagram of a certain segment of effective vibration signal is plotted. Taking the time-frequency ridge line corresponding to the shaft speed as the target component, without loss of generality, there are multiple ridge lines related to the instantaneous shaft speed in the time-frequency diagram, and there is a fixed proportional relationship between them, which is related to the invariant physical parameters of the transmission system. The initial bandwidth is adjusted to 10 Hz, and the extraction result of the time-frequency ridge line is shown as the white dotted line, which is consistent with the trend of the instantaneous frequency in the time-frequency diagram and fits well;

[0045] S4 and S5 use the estimated instantaneous shaft speed as the reference information to calculate the spectral correlation density function By comparing the dominant cyclic order in the angle-time cyclic modulation spectrum with multiple theoretical values calculated in S5, the order 4.93 and its multiples are obvious, which is consistent with the inner ring fault characteristic order of the NSK6203 type bearing. Therefore, it is determined that there is an inner ring fault in the yaw reducer, and corresponding spare parts plans and maintenance plans can be arranged;

[0046] In this embodiment, the core of the present invention is to identify the effective vibration signal of the yaw reducer and then diagnose the fault of the yaw reducer under variable working conditions through angle-time cyclic stationary analysis without a tachometer. Through this invention, the fault diagnosis of the yaw reducer can be effectively carried out, reasonably assisting the enterprise operation and maintenance decision-making. By analyzing the composition of the continuously collected vibration signals of the yaw reducer, a method for accurately identifying effective vibration signals is designed to overcome the problem of untimely monitoring caused by the intermittent operation of the yaw reducer. Under time-varying working conditions, the fault pulses in the vibration signals have time-angle duality, and a cycle non-stationary analysis method without a tachometer is designed to overcome the problem of missed diagnosis caused by time-varying working conditions.

[0047] The main functions achieved by the present invention are:

[0048] 1. Identify the effective vibration signal of the yaw reducer, and then diagnose the faults of the yaw reducer under variable working conditions through angle-time cyclic stationary analysis without a tachometer.

[0049] 2. Design a method to accurately identify the effective vibration signal, overcome the problem of untimely monitoring caused by the intermittent operation of the yaw reducer, and design a cyclic non-stationary analysis method without a tachometer to overcome the problem of missed diagnosis caused by time-varying working conditions.

[0050] 3. The effective vibration signal overcomes the adverse effects of the intermittent operation of the yaw system on data acquisition, storage, monitoring and diagnosis, and avoids the adverse effects of the time-varying operating conditions of the yaw reducer on the extraction and identification of fault characteristics.

[0051] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions, characterized in that, Including the following steps: S1. Obtain the vibration signal of the yaw reducer of the wind turbine. The vibration signal is from the unit status monitoring system and covers the action and non-action time periods of the yaw system; S2. According to the amplitude difference of the vibration signals during the action and non-action periods of the yaw reducer, use sliding median filtering to process the collected signals and identify the effective vibration signals in the action interval; S3. The identified effective vibration signal is an amplitude-frequency and frequency-modulation signal. Use the ridge extraction method based on the cost function to estimate the rotational speed information, providing a signal prior for subsequent cyclic non-stationary analysis; S4. Integrate the estimated rotational speed and the identified effective vibration signal, calculate the cyclic modulation spectrum, and reveal the time-angle duality characteristics corresponding to the fault pulses; S5. Calculate the theoretical fault characteristic orders of each component according to the reducer structure parameters, and compare the theoretical values with the spectral lines in the cyclic modulation spectrum to locate the faulty component.

2. The yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions according to claim 1, characterized in that, The sliding median filtering in S2 is used to identify the effective vibration signals. The amplitude of the effective vibration signal is higher than the set threshold, and the signals lower than the threshold are determined as invalid signals and deleted.

3. A yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions as described in claim 1, characterized in that, In S3, the cost function ridge extraction method is used to estimate the instantaneous frequency of the vibration signal, and the instantaneous shaft rotational speed signal is estimated according to the proportional coefficient between the target component and the shaft rotational speed.

4. A yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions as claimed in claim 1, characterized in that, In S4, the cyclic modulation spectrum analysis performs Fourier series expansion on the vibration signal through the angle-time autocorrelation function and extracts the characteristic frequency information corresponding to the fault type of the reducer.

5. A yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions as described in claim 1, characterized in that, In S5, the calculation of the fault characteristic orders includes the calculation of the local fault characteristic orders of the sun gear, planet gears, and ring gear, corresponding to different types of gear damages respectively.

6. The yaw reducer monitoring and diagnosis method for coping with intermittent and variable working conditions according to claim 1, characterized in that The rotational speed estimation method uses the short-time Fourier transform to calculate the time-frequency representation of the vibration signal and selects the target component for analysis in the time-frequency domain.

7. A yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions according to claim 1, characterized in that, The specific principle of identifying the effective vibration signals in the action interval in S2 is as follows: The wind direction sensor of the yaw system is installed on the top of the wind turbine generator. By measuring the wind direction, it provides wind direction data to the control system. When the wind direction changes and deviates from the set threshold range, the control system issues a yaw command to adjust the orientation of the wind turbine rotor. The yaw drive mechanism drives the yaw reducer through the gear transmission device according to the command of the control system, causing the nacelle of the entire wind turbine to rotate on the top of the tower and adjust the orientation of the wind turbine rotor. Wind resources are random, and the action time of the yaw system cannot be predicted. Therefore, the vibration signals collected on the yaw reducer are intermittent, and its effective vibration signals have amplitude-frequency and frequency-modulation characteristics. Identifying the effective vibration signals by means of the amplitude difference is the basic connotation of overcoming intermittency; First, calculate the envelope signal of the acquired signal: a(t) = |x(t) + jH[x(t)]|, where H[·] is the Hilbert transform operator. Then, perform a sliding median filter with a window length of N w and a unit sliding step size: a s (t) = MMF[a(t)], where MMF[·] is the sliding median filter. Finally, calculate 3σ as the recognition threshold based on a s (t) during the period when the yaw reducer is not operating. Compare a s (t) of the acquired signal with the recognition threshold. If it is higher than the threshold, it is determined as a valid vibration signal and stored; otherwise, it is determined as an invalid vibration signal and deleted.

8. A yaw reducer monitoring and diagnosis method for coping with intermittent and variable working conditions as described in claim 1, characterized in that In S3, the specific principle of using the ridge extraction method based on the cost function to estimate the rotational speed information is as follows: Under time-varying working conditions, the effective vibration signal of the yaw reducer is an amplitude-modulation and frequency-modulation signal, and the instantaneous frequency of its component characterizes the shaft rotational speed information. First, the time-frequency representation of the effective vibration signal is calculated using the short-time Fourier transform as follows: Let \(w(t)\) be the window function. Then, select a target component on the TFR and use the cost function ridge extraction method to estimate the instantaneous frequency: To avoid adjacent component interference and reduce the parameter search range, a time-varying frequency band is constructed with the center frequency at the previous moment for the frequency band to be searched: Finally, the instantaneous shaft speed signal is solved according to the proportional coefficient between the target component and the shaft speed:

9. A yaw reducer monitoring and diagnosis method for coping with intermittency and variable working conditions as claimed in claim 1, characterized in that, In S4, the specific principle of calculating the cyclic modulation spectrum analysis is as follows: The fault pulses of the yaw reducer under time-varying rotational speed have time-angle duality. The position of the fault impact is determined by the angular position of the shaft, related to the kinematics of the reducer, synchronized with the machine speed. The resonance frequency and decay time of the fault impact are time-invariant, related to the dynamics of the system, and are the solutions of the time-difference equation. The angle-time cyclostationary analysis method can analyze signals across the angular domain and time domain simultaneously, facilitating the extraction of subtle reducer fault characteristics under different operating conditions. The angle-time autocorrelation function is periodic, and its Fourier series expansion is as follows: Among them, the Fourier coefficient is expressed as: The spectral correlation density function is the Fourier transform of the angular time autocorrelation function: a represents the proportional relationship between the fault characteristic frequency and the rotational frequency, and f is the spectral frequency in the traditional sense.

10. A yaw reducer monitoring and diagnosis method for coping with intermittent and variable working conditions as described in claim 1, characterized in that In S5, the calculation principle for calculating the theoretical fault characteristic orders of each component is as follows: The yaw reducer is composed of cascaded planetary gear trains. The calculation methods for the fault characteristic orders of various types of gears are as follows: If there is local damage to a certain tooth of the sun gear, then within one rotation period relative to the planet carrier, the faulty tooth will mesh with all planet gears, generating impacts. Therefore, the local fault characteristic frequency of the sun gear is If there is local damage to a certain tooth of a planet gear, then within one rotation period relative to the planet carrier, the faulty tooth meshes with both the sun gear and the ring gear, generating one impact respectively. The local fault characteristic frequency of the planet gear is: If there is local damage to a certain tooth of the ring gear, then within one rotation period relative to the planet carrier, the faulty tooth will mesh with all planet gears, generating impacts. Therefore, the local fault characteristic frequency of the ring gear is According to the model and size parameters of the bearing, the fault characteristic order of the outer ring of the bearing is: The fault characteristic order of the inner ring is: The fault characteristic order of the rolling element is: Compare the theoretical fault characteristic order with the dominant a in the spectral correlation density function to identify the faulty component.

Images