A wind turbine main shaft fault detection method, electronic device and program product

Abstract

The application provides a wind turbine main shaft fault detection method, an electronic device and a program product. In the main shaft fault detection, a radar sensor is introduced, the micro-Doppler effect is used to directly detect the surface state of the main shaft, the radar can capture the surface roughness change caused by the slight deformation or wear of the surface when the main shaft rotates, and even if the wear amount is extremely small, it can also be accurately identified through the phase / frequency change of the reflected wave. Combined with the eccentricity feature, the abnormal balance state of the main shaft can be further reflected, a more sensitive detection means is provided for early faults, and the fault detection rate is improved. In the embodiment, a multi-modal fusion of a radar and a temperature sensor is adopted, the temperature rise rate extracted by the temperature sensor can reflect the abnormality of the friction state of the main shaft, is complementary to the surface roughness and the eccentricity detected by the radar, and also improves the fault detection rate.

Description

Technical Field

[0001] This application relates to the field of wind turbine technology, and more specifically, to a method, electronic equipment and software product for detecting faults in the main shaft of a wind turbine. Background Technology

[0002] The wind turbine main shaft is the core component of the unit's transmission system, playing a crucial role in transmitting the kinetic energy of the wind turbine. It transmits the enormous torque captured by the turbine blades to the gearbox (in geared units) or directly to the generator (in direct-drive units). Simultaneously, it must withstand axial thrust, bending moments, and other complex dynamic loads from components such as the blades and hub, and transfer these non-torsional loads to the support structure of the main frame. As the central hub of the rotating system, the main shaft supports the entire rotor system. Its structure typically connects to the hub via flanges, mounts bearings on the journals, and connects to downstream equipment via couplings at the shaft ends. Therefore, it must possess extremely high strength and fatigue resistance. The main shaft faces various failure risks during long-term operation. Common failure modes include mechanical damage (such as bending, misalignment, axial movement, and journal wear), bearing failure (such as roller breakage, cage damage, wear due to lubrication failure, corrosion, fracture, and aging), and internal cracks caused by long-term exposure to alternating loads (fluctuating loads). If these cracks are not detected in time, they may continue to propagate, eventually leading to catastrophic fracture failure.

[0003] The necessity of fault detection for the main shaft is reflected in the following aspects: First, once a serious failure occurs in the main shaft or its bearings, the entire nacelle often needs to be lifted off the tower for return to the factory for repair, resulting in extremely high replacement costs (up to hundreds of thousands of yuan or even higher) and long repair cycles. Second, unplanned downtime caused by failures will result in huge power generation losses, directly affecting the economic benefits of the wind farm. Third, and most seriously, a sudden fracture failure of the main shaft poses a significant safety risk of causing the entire wind turbine to collapse, threatening the lives and property of personnel. In addition, main shaft failures may also cause collateral damage to the bearing mounting positions of the main frame and even adjacent critical components such as the gearbox, further expanding the scope of losses. Traditional periodic maintenance methods require shutdown and inspection by professionals at high altitudes (over 100 meters), which is not only dangerous and costly, but also makes it difficult to detect early and minute damage changes inside the main shaft or bearings, exhibiting significant lag and limitations.

[0004] Existing technologies have developed solutions for real-time monitoring of spindle operation using MEMS vibration sensors. However, traditional MEMS vibration sensors are not sensitive enough to surface micro-wear of less than 0.5 mm, resulting in a high rate of missed early fault detection. Summary of the Invention

[0005] The purpose of this application is to provide a method, electronic device and program product for detecting faults in the main shaft of a wind turbine, in order to solve the problem that the existing technology uses MEMS vibration sensors to monitor the operating status of the main shaft in real time, resulting in a high rate of missed detection of early faults.

[0006] This application provides a method for detecting faults in the main shaft of a wind turbine, comprising:

[0007] The rotating spindle is monitored in real time using radar and temperature sensors to obtain radar and temperature information;

[0008] The radar information was preprocessed and features were extracted to obtain the surface roughness and eccentricity;[1]

[0009] Temperature information is preprocessed and features are extracted to obtain the temperature rise rate;

[0010] By fusing surface roughness, eccentricity, and temperature rise rate, multidimensional features are obtained.

[0011] By inputting multidimensional features into the trained fault detection model, fault detection results are obtained.

[0012] Surface roughness refers to the error in the microscopic geometry of a machined surface, typically described by the height, spacing, or shape characteristics of micro-irregularities. In mechanical engineering, it reflects the degree of peak-valley undulations on the surface and is an important indicator for evaluating the surface quality of parts. When radar waves irradiate the spindle surface, the microscopic peak-valley structure alters the reflection path of the radar waves, producing a micro-Doppler frequency shift related to surface roughness (the Doppler effect caused by minute surface deformations). By analyzing the spectral characteristics of radar signals, surface roughness can be quantitatively extracted, thereby detecting early micro-wear.

[0013] Eccentricity refers to the radial distance between the actual axis of rotation of a rotating component (such as a spindle) and its geometric axis (ideal axis). Simply put, it reflects the degree of "wobbling" during spindle rotation. Radar can indirectly calculate eccentricity by measuring the phase or frequency changes of the reflected wave as the spindle rotates. For example, when the spindle radially wobbles, the radar wave reflection path will periodically shift; by analyzing the periodicity and amplitude of this shift, the eccentricity can be quantitatively extracted.

[0014] In the above technical solution, a radar sensor is introduced into the spindle fault detection. Utilizing the micro-Doppler effect (Doppler frequency shift caused by minute target motion), the surface condition of the spindle is directly detected. The radar can capture minute deformations or surface roughness changes caused by wear during spindle rotation. Even with extremely small wear (e.g., at the micrometer level), it can be accurately identified through phase / frequency changes in the reflected wave. Combined with the eccentricity (radial runout during spindle rotation) characteristics, it can further reflect abnormal spindle balance (e.g., eccentricity caused by bearing wear), providing a more sensitive detection method for early faults and improving the fault detection rate. This embodiment employs multi-modal fusion of radar and a temperature sensor. The temperature rise rate (temperature change rate) extracted by the temperature sensor can reflect abnormalities in the spindle friction state (e.g., early wear leading to increased friction and a significantly increased temperature rise rate), complementing the surface roughness and eccentricity detected by the radar, thus further improving the fault detection rate.

[0015] In some alternative implementations, the radar is installed inside the wind turbine housing, facing the main shaft, and the radar is oriented at a set angle to the central axis of the main shaft, the set angle being less than 70 degrees and greater than 20 degrees.

[0016] In the above technical solution, the incident angle range of 20°-70° allows the radar wave to be incident on the main shaft surface in an "oblique" manner. At this time, the tiny wear or cracks on the surface when the main shaft rotates will change the reflection path of the radar wave, generating a richer Doppler frequency shift signal (i.e., micro-Doppler effect), thereby improving the detection sensitivity of micron-level surface deformation.

[0017] In some alternative implementations, it also includes:

[0018] Vibration information is obtained by using vibration sensors to detect the rotating spindle in real time.

[0019] Vibration information is preprocessed and features are extracted to obtain envelope entropy and harmonic energy ratio.

[0020] Envelope entropy: By demodulating the envelope of the vibration signal, the periodic impact intensity of the signal is quantified. Early faults (such as pitting corrosion on the inner ring of a bearing) will generate periodic impact vibrations, and envelope entropy can effectively capture such low-amplitude, highly periodic signals.

[0021] Harmonic energy ratio: By analyzing the energy proportion of the fundamental frequency and its harmonics in a vibration signal, faults such as gear wear and shaft misalignment can be identified. For example, gear wear can lead to a significant increase in the energy of certain harmonic components. The harmonic energy ratio can quantify this change and improve the early identification capability of gear faults.

[0022] In the above technical solution, vibration characteristics (envelope entropy, harmonic energy ratio) reflect the dynamic impact of the spindle rotation (such as periodic vibration caused by bearing failure) and changes in harmonic components (such as abnormal harmonic energy distribution caused by gear wear), which can accurately identify faults in key components such as bearings and gears. Radar characteristics (surface roughness, eccentricity) reflect abnormalities in the spindle surface deformation and balance state (such as journal wear, installation errors). Temperature characteristics (temperature rise rate) reflect the dynamic process of frictional heat generation (such as local overheating caused by poor lubrication). The combination of these three features can cover the mechanical, thermal, and vibrational dimensions of spindle faults, avoiding missed detections due to the limitations of a single sensor.

[0023] In some alternative implementations, surface roughness, eccentricity, and temperature rise rate are feature-fused, including:

[0024] Based on the rotational speed, determine the weight settings for radar information, temperature information, and vibration information;

[0025] Based on the weight settings, surface roughness, eccentricity, temperature rise rate, envelope entropy, and harmonic energy ratio are fused to obtain multidimensional features.

[0026] In some alternative implementations, the weight of the radar information is: w_uwb = 1 / (1 + k);

[0027] The weight of the vibration information: w_vib = a1×k;

[0028] Weight of temperature information: w_temp = a2;

[0029] Where k is the speed adaptive weight; when the speed rpm is greater than the speed set value, k = a3 × rpm; when the speed rpm is less than or equal to the speed set value, k = a4; a1, a2, a3 and a4 are fixed coefficient values.

[0030] In the above technical solution, at low speeds, the weights of radar information, vibration information, and temperature information remain unchanged; at high speeds, as the speed increases, the weight of radar information decreases, the weight of vibration information increases, and the weight of temperature information remains unchanged. The weight adjustment strategy in this embodiment adapts to the differences in fault characteristic sensitivity at different speeds, optimizes sensor data fusion, balances computational resources and detection accuracy, and ultimately reduces the early fault false alarm rate, improving system robustness and practicality.

[0031] In some alternative implementations, surface roughness, eccentricity, and temperature rise rate are feature-fused, including:

[0032] The weight settings for radar information, temperature information, and vibration information are determined based on the rotational speed and current temperature.

[0033] Based on the weight settings, surface roughness, eccentricity, temperature rise rate, envelope entropy, and harmonic energy ratio are fused to obtain multidimensional features.

[0034] In some alternative implementations, the weight of the radar information is: w_uwb = 1 / (1 + k);

[0035] When the current temperature is less than or equal to the temperature setpoint, the weight of the vibration information is: w_vib = a1×k; when the current temperature is greater than the temperature setpoint, the weight of the vibration information is: w_vib = a5.

[0036] Weight of temperature information: w_temp = a2;

[0037] Where k is the speed adaptive weight; when the speed rpm is greater than the speed set value, k = a3 × rpm; when the speed rpm is less than or equal to the speed set value, k = a4; a1, a2, a3, a4 and a5 are fixed coefficient values.

[0038] In the above technical solution, the weight of vibration information is relatively high in low-temperature scenarios, and the weight of vibration information increases with the increase of rotational speed. In high-temperature scenarios, the weight of vibration information remains at a relatively low level. This is because high temperatures may be caused by lubrication failure, increased friction, or thermal expansion, in which case the vibration signal may be distorted due to thermal deformation.

[0039] In some alternative implementations, temperature and vibration sensors are mounted on the main shaft bearing housing of the wind turbine.

[0040] In some alternative implementations, the fault detection model is a time-series sensitive model.

[0041] In the aforementioned technical solutions, time-series sensitive models (such as LSTM, GRU, and Transformer) can memorize the historical operating states of the spindle and capture the gradual evolution of fault characteristics. For example, early bearing wear can lead to a gradual increase in the envelope entropy of the vibration signal and a slow increase in the temperature rise rate. The model can learn these temporal patterns and issue an early warning before the fault becomes obvious (such as before the wear reaches a threshold). Spindle faults may be caused by small anomalies accumulated over a long period of time (such as slow wear caused by poor lubrication). Time-series models can effectively capture dependencies across long time windows through gating mechanisms (such as the forget gate and input gate of LSTM) or self-attention mechanisms (such as Transformer), avoiding short-term noise from masking long-term trends.

[0042] In some alternative implementations, the output of the fault detection model includes at least one of the following:

[0043] The probability of surface wear, the probability of bearing pitting, the probability of shaft bending, and the probability of lubrication failure[2].

[0044] Among them, surface wear is caused by long-term friction, poor lubrication or load fluctuations, which leads to the gradual loss of surface material, forming micron-level scratches or pits.

[0045] Pitting in bearings is caused by localized stress concentration, fatigue, or lubrication failure, which leads to the shedding of surface metal and the formation of tiny pits (pits).

[0046] Shaft bending is caused by manufacturing defects, installation errors, or long-term asymmetrical loads leading to elastic deformation of the shaft, resulting in increased eccentricity during rotation.

[0047] Lubrication failures include: aging, leakage, or contamination of lubricating oil leading to insufficient oil film thickness, resulting in direct metal-to-metal contact.

[0048] An electronic device provided in this application includes a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the machine-readable instructions, when executed by the processor, perform any of the methods described above.

[0049] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of any of the methods described above. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 A flowchart illustrating the steps of a wind turbine main shaft fault detection method provided in this application embodiment;

[0052] Figure 2 This is a schematic diagram of a possible structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0053] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0054] Please refer to Figure 1 , Figure 1 A flowchart of a method for detecting faults in the main shaft of a wind turbine provided in this application includes:

[0055] Step S1: Use radar and temperature sensors to detect the rotating spindle in real time and obtain radar information and temperature information;

[0056] The radar can be a UWB radar module, such as the NXP SR040 UWB chip, with an operating frequency of 6.5GHz±1.5GHz and a detection accuracy of ±0.1mm (surface wear).

[0057] The temperature sensor can be a PT1000 platinum resistance thermometer, with a measurement range of -40℃ to 150℃ and an accuracy of ±0.5℃. Temperature detection can also be achieved using an infrared thermal imaging array, enabling non-contact detection of the spindle's temperature information.

[0058] Step S2: Preprocess and extract features from the radar information to obtain surface roughness and eccentricity;

[0059] The preprocessing of radar information includes pulse compression: compressing wide pulses into narrow pulses.

[0060] Surface roughness refers to the error in the microscopic geometry of a machined surface, usually described by the height, spacing, or shape characteristics of micro-irregularities. In mechanical engineering, it reflects the degree of peak-valley undulations on the surface and is an important indicator for evaluating the surface quality of parts. When radar waves irradiate the surface of a spindle, the microscopic peak-valley structure alters the reflection path of the radar waves, producing a micro-Doppler frequency shift related to surface roughness (the Doppler effect caused by minute surface deformations). By analyzing the spectral characteristics of radar signals, surface roughness can be quantitatively extracted, thereby detecting early micro-wear.

[0061] Eccentricity refers to the radial distance between the actual axis of rotation of a rotating component (such as a spindle) and its geometric axis (ideal axis). Simply put, it reflects the degree of "wobbling" during spindle rotation. Radar can indirectly calculate eccentricity by measuring the phase or frequency changes of the reflected wave as the spindle rotates. For example, when the spindle radially wobbles, the radar wave reflection path will periodically shift; by analyzing the periodicity and amplitude of this shift, the eccentricity can be quantitatively extracted.

[0062] Step S3: Preprocess and extract features from the temperature information to obtain the temperature rise rate;

[0063] The preprocessing of temperature information includes moving average: calculating the average value of values ​​within a continuous window in the data sequence to eliminate short-term fluctuations and highlight long-term trends.

[0064] Temperature rise rate refers to the rate at which the temperature of a device or component changes over time during operation, usually expressed as the temperature increase per unit time (e.g., °C / min or °C / hour). In wind turbine main shaft fault detection, temperature rise rate is one of the key monitoring parameters. Its core function is to identify potential faults (such as lubrication failure, bearing wear, or abnormal thermal expansion) at an early stage by observing the dynamic trend of temperature changes.

[0065] Step S4: The surface roughness, eccentricity, and temperature rise rate are fused to obtain multidimensional features;

[0066] In this process, corresponding weights are set for surface roughness, eccentricity, and temperature rise rate, and feature fusion is performed based on these weights to obtain multidimensional features.

[0067] Step S5: Input the multidimensional features into the trained fault detection model to obtain the fault detection results.

[0068] Among them, the fault detection model can adopt the lightweight edge model TinyFaultNet, which has a small model size, reduced cost, and low power consumption, and can be applied to edge device MCUs.

[0069] In this embodiment, a radar sensor is introduced into the spindle fault detection. Utilizing the micro-Doppler effect (Doppler frequency shift caused by minute target motion), the surface condition of the spindle is directly detected. The radar can capture minute deformations or surface roughness changes caused by wear during spindle rotation. Even with extremely small wear (e.g., at the micrometer level), it can accurately identify the problem through phase / frequency changes in the reflected wave. Combined with the eccentricity (radial runout during spindle rotation) characteristics, it can further reflect abnormal spindle balance (e.g., eccentricity caused by bearing wear), providing a more sensitive detection method for early faults and improving the fault detection rate. This embodiment employs multi-modal fusion of radar and a temperature sensor. The temperature rise rate (temperature change rate) extracted by the temperature sensor reflects abnormalities in the spindle friction state (e.g., early wear leading to increased friction and a significantly increased temperature rise rate), complementing the surface roughness and eccentricity detected by the radar, thus further improving the fault detection rate.

[0070] Specifically, extracting surface roughness and eccentricity from radar information features involves first extracting the following features from the radar echo signal, and then determining the surface roughness and eccentricity based on these features:

[0071] 1) Vibration characteristics:

[0072] Vibration amplitude: Monitors the magnitude of vibration of the shaft in the radial (X, Y direction) and axial (Z direction). Abnormally increased amplitude usually indicates problems such as imbalance, misalignment, bearing wear, looseness, or blade damage.

[0073] Vibration frequency: Precisely measures the fundamental frequency (rotational frequency) of shaft vibration and its harmonic and subharmonic components. Specific faults (such as bearing defects, gear meshing problems, imbalance, misalignment) will produce characteristic frequency components (such as the bearing's pass frequency, the gear's meshing frequency, and its sidebands). UWB can capture these spectral characteristics.

[0074] Vibration modes: Analyze the time-domain waveform and trajectory of vibration (such as shaft center trajectory). Specific vibration modes (such as beat vibration, impact vibration, chaos) are characteristic of certain faults (such as friction, severe loosening, crack propagation).

[0075] 2) Displacement and Deformation:

[0076] Shaft center position / trajectory: Real-time monitoring of the average position of the shaft's rotation center and its trajectory (shaft center trajectory) during one rotation. Changes in trajectory shape (such as ellipse, banana shape, scattered pattern) can directly reflect problems such as misalignment, shaft bending, excessive bearing clearance, and oil film oscillation.

[0077] Dynamic eccentricity: measures the instantaneous offset of the rotating shaft relative to the theoretical center during rotation.

[0078] Axial displacement: Monitors the movement of the shaft along the axial direction. Excessive axial displacement may be caused by thrust bearing wear, uneven thermal expansion, coupling problems, or misalignment.

[0079] Bending / Deflection: Detects the degree of bending deformation of the shaft during operation (especially during start-up, shutdown, and variable load). Sustained or excessive bending may be caused by gravity, thermal stress, manufacturing residual stress, or damage (such as cracks).

[0080] 3) Rotational speed and rotational stability:

[0081] Speed ​​fluctuation: Accurately measures the actual speed of the shaft and its fluctuations. Abnormal speed fluctuations or modulation components in the speed signal may be related to load changes, control system problems, drivetrain failures (such as gear damage), or electrical faults.

[0082] Torsional vibration: Indirectly inferred by analyzing minute tangential displacement changes or specific spectral components (such as torsional frequency) at specific points on the shaft (e.g., areas with distinctive structural features). Severe torsional vibration is extremely harmful to gears, couplings, and the shaft itself.

[0083] In some alternative implementations, the radar is installed inside the wind turbine housing, facing the main shaft, and the radar is oriented at a set angle to the central axis of the main shaft, the set angle being less than 70 degrees and greater than 20 degrees.

[0084] In this embodiment, the incident angle range of 20°-70° allows the radar wave to be incident on the spindle surface in an "oblique" manner. At this time, the minute wear or cracks on the surface during spindle rotation will change the reflection path of the radar wave, generating a richer Doppler frequency shift signal (i.e., micro-Doppler effect), thereby improving the detection sensitivity of micron-level surface deformation.

[0085] In some alternative implementations, it also includes:

[0086] Vibration information is obtained by using vibration sensors to detect the rotating spindle in real time.

[0087] Vibration information is preprocessed and features are extracted to obtain envelope entropy and harmonic energy ratio.

[0088] The vibration sensor can be a MEMS triaxial vibration sensor, such as the ADI ADXL1002, with a range of ±50g and a sampling rate of 20kHz.

[0089] Envelope entropy: By demodulating the envelope of a vibration signal, the periodic impact intensity of the signal is quantified. Early faults (such as pitting corrosion on the inner ring of a bearing) will generate periodic impact vibrations, and envelope entropy can effectively capture such low-amplitude, highly periodic signals.

[0090] Harmonic energy ratio: By analyzing the energy proportion of the fundamental frequency and its harmonics in a vibration signal, faults such as gear wear and shaft misalignment can be identified. For example, gear wear can lead to a significant increase in the energy of certain harmonic components. The harmonic energy ratio can quantify this change and improve the early identification capability of gear faults.

[0091] In this embodiment, vibration characteristics (envelope entropy, harmonic energy ratio) reflect the dynamic impact of the spindle rotation (such as periodic vibration caused by bearing failure) and changes in harmonic components (such as abnormal harmonic energy distribution caused by gear wear), which can accurately identify faults in key components such as bearings and gears. Radar characteristics (surface roughness, eccentricity) reflect abnormalities in the spindle surface deformation and balance state (such as journal wear, installation errors). Temperature characteristics (temperature rise rate) reflect the dynamic process of frictional heat generation (such as local overheating caused by poor lubrication). The combination of these three features can cover the mechanical, thermal, and vibrational dimensions of spindle faults, avoiding missed detections due to the limitations of a single sensor.

[0092] For offshore wind power scenarios, typhoon-induced vibration information includes main shaft fault characteristics and typhoon vortex-induced vibration characteristics. The typhoon vortex-induced vibration characteristics will affect the accuracy of fault diagnosis. Since the frequency range of typhoon vortex-induced vibration is typically 0.1–5Hz (low-frequency dominant), while the frequency range of main shaft fault vibration is typically 50 Hz–2 kHz (high-frequency harmonics), this embodiment employs the following methods to reduce typhoon interference:

[0093] FFT spectrum analysis was performed to extract the low-frequency components of 0.1–5 Hz by calculating the spectrum of the vibration signal.

[0094] Energy percentage calculations are performed. If the low-frequency energy percentage is greater than 70%, it is determined to be typhoon interference.

[0095] Notch filtering is performed in the 0.1–5 Hz frequency band to preserve high-frequency fault characteristics.

[0096] Dynamic threshold adjustments can be made, for example, by increasing the weight of radar information (without typhoon impact) and temperature information, or by decreasing the weight of vibration information.

[0097] In some alternative implementations, surface roughness, eccentricity, and temperature rise rate are feature-fused, including:

[0098] Based on the rotational speed, determine the weight settings for radar information, temperature information, and vibration information;

[0099] Based on the weight settings, surface roughness, eccentricity, temperature rise rate, envelope entropy, and harmonic energy ratio are fused to obtain multidimensional features.

[0100] In some alternative implementations, the weight of the radar information is: w_uwb = 1 / (1 + k);

[0101] The weight of the vibration information: w_vib = a1×k;

[0102] Weight of temperature information: w_temp = a2;

[0103] Where k is the speed adaptive weight; when the speed rpm is greater than the speed set value, k = a3 × rpm; when the speed rpm is less than or equal to the speed set value, k = a4; a1, a2, a3 and a4 are fixed coefficient values.

[0104] In one specific embodiment, the rotational speed is set to 100 revolutions per minute, a1=0.8, a2=0.2, a3=0.02, and a4=0.5.

[0105] In this embodiment, at low speeds, the weights of radar information, vibration information, and temperature information remain unchanged; at high speeds, as the speed increases, the weight of radar information decreases, the weight of vibration information increases, and the weight of temperature information remains unchanged. This weight adjustment strategy adapts to the differences in fault characteristic sensitivity at different speeds, optimizes sensor data fusion, balances computational resources and detection accuracy, ultimately reducing the early fault detection rate and improving system robustness and practicality.

[0106] In some alternative implementations, surface roughness, eccentricity, and temperature rise rate are feature-fused, including:

[0107] The weight settings for radar information, temperature information, and vibration information are determined based on the rotational speed and current temperature.

[0108] Based on the weight settings, surface roughness, eccentricity, temperature rise rate, envelope entropy, and harmonic energy ratio are fused to obtain multidimensional features.

[0109] In some alternative implementations, the weight of the radar information is: w_uwb = 1 / (1 + k);

[0110] When the current temperature is less than or equal to the temperature setpoint, the weight of the vibration information is: w_vib = a1 × k; when the current temperature is greater than the temperature setpoint, the weight of the vibration information is: w_vib = a5.

[0111] Weight of temperature information: w_temp = a2;

[0112] Where k is the speed adaptive weight; when the speed rpm is greater than the speed set value, k = a3 × rpm; when the speed rpm is less than or equal to the speed set value, k = a4; a1, a2, a3, a4 and a5 are fixed coefficient values.

[0113] In one specific embodiment, a1=0.8, a2=0.2, a3=0.02, a4=0.5, a5=0.7, the temperature setting is 70 degrees, and the rotation speed setting is 100 revolutions per minute.

[0114] In this embodiment, the vibration information has a higher weight in low-temperature scenarios, and the weight of the vibration information increases with the increase of rotational speed. In high-temperature scenarios, the weight of the vibration information remains at a lower weight because high temperature may be caused by lubrication failure, increased friction, or thermal expansion, in which case the vibration signal may be distorted due to thermal deformation.

[0115] In some alternative implementations, temperature and vibration sensors are mounted on the main shaft bearing housing of the wind turbine.

[0116] In some alternative implementations, the fault detection model is a time-series sensitive model.

[0117] In this embodiment, time-series sensitive models (such as LSTM, GRU, and Transformer) can memorize the historical operating state of the spindle and capture the gradual evolution of fault characteristics. For example, early bearing wear can lead to a gradual increase in the envelope entropy of the vibration signal and a slow increase in the temperature rise rate. The model can learn these temporal patterns and issue an early warning before the fault becomes obvious (such as before the wear reaches a threshold). Spindle faults may be caused by small anomalies accumulated over a long period of time (such as slow wear caused by poor lubrication). Time-series models can effectively capture dependencies across long time windows through gating mechanisms (such as the forget gate and input gate of LSTM) or self-attention mechanisms (such as Transformer), avoiding short-term noise from masking long-term trends.

[0118] In some alternative implementations, the output of the fault detection model includes at least one of the following:

[0119] The probability of surface wear, bearing pitting, shaft bending, and lubrication failure.

[0120] Among them, surface wear is caused by long-term friction, poor lubrication or load fluctuations, which leads to the gradual loss of surface material, forming micron-level scratches or pits.

[0121] Pitting in bearings is caused by localized stress concentration, fatigue, or lubrication failure, which leads to the shedding of surface metal and the formation of tiny pits (pits).

[0122] Shaft bending is caused by manufacturing defects, installation errors, or long-term asymmetrical loads leading to elastic deformation of the shaft, resulting in increased eccentricity during rotation.

[0123] Lubrication failures include: aging, leakage, or contamination of lubricating oil leading to insufficient oil film thickness, resulting in direct metal-to-metal contact.

[0124] In some alternative implementations, the spindle health status can be comprehensively assessed based on the probability of surface wear, bearing pitting, shaft bending, and lubrication failure. The future maintenance cycle is then determined based on the spindle health status to avoid downtime.

[0125] Figure 2 This illustration shows a possible structure of an electronic device provided in an embodiment of this application. (Refer to...) Figure 2 The electronic device includes a processor, memory, and a communication interface, which are interconnected and communicate with each other via a communication bus and / or other forms of connection mechanism (not shown).

[0126] The memory includes one or more (only one is shown in the figure), which can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The processor and other possible components can access the memory to read and / or write data to it.

[0127] The processor comprises one or more (only one is shown in the figure), which can be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Microcontroller Unit (MCU), a Network Processor (NP), or other conventional processors; it can also be a special-purpose processor, including a Neural-network Processing Unit (NPU), a Graphics Processing Unit (GPU), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. Furthermore, when there are multiple processors, some can be general-purpose processors, and others can be special-purpose processors.

[0128] The communication interface includes one or more (only one is shown in the figure), which can be used to communicate directly or indirectly with other devices to exchange data. The communication interface may include interfaces for wired and / or wireless communication.

[0129] One or more computer program instructions may be stored in the memory, and the processor may read and execute these computer program instructions to implement the methods provided in the embodiments of this application.

[0130] Understandable. Figure 2 The structure shown is for illustrative purposes only; the electronic device may also include structures that are more complex than those shown. Figure 2 The more or fewer components shown, or having the same Figure 2 The different structures shown. Figure 2 The components shown can be implemented using hardware, software, or a combination thereof. Electronic devices may be physical devices, such as PCs, laptops, tablets, mobile phones, servers, embedded devices, etc., or they may be virtual devices, such as virtual machines, virtualized containers, etc. Furthermore, electronic devices are not limited to a single device; they can also be a combination of multiple devices or a cluster of a large number of devices.

[0131] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of any of the methods described above.

[0132] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0133] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0134] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0135] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0136] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for detecting faults in the main shaft of a wind turbine generator, characterized in that, include: The rotating spindle is monitored in real time using UWB radar and temperature sensors to obtain radar and temperature information. The radar information is preprocessed and features are extracted to obtain the surface roughness and eccentricity characterized by radar echo features; The surface roughness is used to characterize the microscopic unevenness of the spindle surface, and the eccentricity is used to characterize the radial offset of the actual axis of rotation relative to the geometric center axis when the spindle rotates. The temperature information is preprocessed and features are extracted to obtain the temperature rise rate; The surface roughness, eccentricity, and temperature rise rate are fused to obtain multidimensional features; The multidimensional features are input into the trained fault detection model to obtain the fault detection results; The radar is installed inside the wind turbine housing, facing the main shaft, and the radar is set at a set angle with the central axis of the main shaft, the set angle being less than 70 degrees and greater than 20 degrees.

2. The method as described in claim 1, characterized in that, Also includes: Vibration information is obtained by using vibration sensors to detect the rotating spindle in real time. The vibration information is preprocessed and features are extracted to obtain the envelope entropy and harmonic energy ratio.

3. The method as described in claim 2, characterized in that, The feature fusion of the surface roughness, eccentricity, and temperature rise rate includes: The weight settings for the radar information, temperature information, and vibration information are determined based on the rotational speed. Based on the weight settings, the surface roughness, eccentricity, temperature rise rate, envelope entropy, and harmonic energy ratio are fused to obtain the multidimensional features.

4. The method as described in claim 3, characterized in that, The weight of the radar information is: w_uwb = 1 / (1 +k); The weight of the vibration information: w_vib = a1×k; The weight of the temperature information: w_temp = a2; Where k is the speed adaptive weight; when the speed rpm is greater than the speed set value, k = a3 × rpm; when the speed rpm is less than or equal to the speed set value, k = a4; a1, a2, a3 and a4 are fixed coefficient values.

5. The method as described in claim 2, characterized in that, The feature fusion of the surface roughness, eccentricity, and temperature rise rate includes: The weight settings for the radar information, temperature information, and vibration information are determined based on the rotational speed and current temperature. Based on the weight settings, the surface roughness, eccentricity, temperature rise rate, envelope entropy, and harmonic energy ratio are fused to obtain the multidimensional features.

6. The method as described in claim 5, characterized in that, The weight of the radar information is: w_uwb = 1 / (1 +k); When the current temperature is less than or equal to the temperature setpoint, the weight of the vibration information is: w_vib = a1 × k; when the current temperature is greater than the temperature setpoint, the weight of the vibration information is: w_vib = a5. The weight of the temperature information: w_temp = a2; Where k is the speed adaptive weight; when the speed rpm is greater than the speed set value, k = a3 × rpm; when the speed rpm is less than or equal to the speed set value, k = a4; a1, a2, a3, a4 and a5 are fixed coefficient values.

7. The method as described in claim 1, characterized in that, The temperature sensor and vibration sensor are mounted on the main shaft bearing housing of the wind turbine.

8. The method as described in claim 1, characterized in that, The fault detection model is a time-series sensitive model.

9. The method as described in claim 1, characterized in that, The output of the fault detection model includes at least one of the following: The probability of surface wear, bearing pitting, shaft bending, and lubrication failure.

10. An electronic device, characterized in that, include: A processor and a memory, the memory storing machine-readable instructions executable by the processor, which, when executed by the processor, perform the method as described in any one of claims 1-9.

11. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-9.

Images