Online monitoring system and method for oil of gearbox of wind turbine generator

Abstract

The invention discloses an on-line monitoring system and method for gearbox oil of a wind turbine generator, belongs to the technical field of gearbox oil monitoring, and solves the problem of how to improve the real-time performance and accuracy of gearbox oil monitoring. According to the invention, multiple sensors are deployed to cooperatively collect oil quality information signals and vibration signals, monitoring is carried out from multiple dimensions of the oil state, and the limitation of monitoring of a single parameter is effectively avoided; by monitoring the oil viscosity, the moisture content and the abrasive particle index of the gear box in real time, the abnormal state caused by lubrication failure or abrasion aggravation can be recognized, and the non-planned shutdown loss can be reduced through a predictive maintenance strategy formulated based on the oil parameter change trend; by monitoring moisture and metal particles in the oil, secondary abrasion caused by lubrication degradation can be effectively inhibited, all-weather state monitoring is conducted on the gear box, safety accidents caused by sudden failures are avoided, and the real-time performance and accuracy of monitoring of the oil in the gear box are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of gearbox oil monitoring technology, and relates to an online monitoring system and method for wind turbine gearbox oil. Background Technology

[0002] To successfully achieve the "dual carbon goals," wind power generation will usher in a period of comprehensive development. However, with this development, some problems are also becoming increasingly apparent, such as gearbox failures, a major source of wind turbine malfunctions. Research indicates that lubrication is the primary factor influencing gearbox failures and even total failure. Therefore, it is necessary to strengthen oil monitoring and fault diagnosis capabilities for wind turbine gearboxes to improve their overall operational status.

[0003] Existing technologies, such as the invention patent with publication number CN102789182A, disclose a novel wind turbine gearbox oil quality monitoring platform based on ARM and Linux, integrating data acquisition, touch screen, SQLite database, RS232 / RS485 communication, wireless communication, Ethernet communication, and A / D functions. The newly added oil quality monitoring typically completes an online monitoring cycle in just a few minutes to tens of minutes, with results automatically entered into the database. It eliminates the need for manual sampling, directly analyzing the oil in the return pipe or gearbox, avoiding problems such as unrepresentative sampling and secondary contamination. This oil quality monitoring addresses the root causes of abnormal gearbox wear, overcoming some significant shortcomings of offline oil monitoring in wind turbine gearboxes. However, in practical use, this method only matches data with data in SQLite, determining the final maintenance report based on the matching result. It struggles to incorporate multimodal information about the gearbox oil, thus failing to comprehensively reflect the gearbox's operating status. Summary of the Invention

[0004] The technical problem to be solved by this invention is how to improve the real-time performance and accuracy of gearbox oil monitoring.

[0005] The present invention solves the above-mentioned technical problems through the following technical solutions: A wind turbine gearbox oil online monitoring system includes a data acquisition module, a processor module, a data analysis module, and an online monitoring module; The data acquisition module is used to collect real-time information related to the gearbox oil, and process and transmit the real-time information related to the gearbox oil to the processor module. The processor module is used to convert the real-time information of the processed gearbox oil into an effective level signal and save it. The data analysis module is used to analyze the information stored in the processor module and determine the wear stage of the gearbox and the cause of abnormal oil quality based on the analysis results. The online monitoring module is used to construct a fault early warning model and calculate the fault early warning threshold range. When the model output result exceeds the fault early warning threshold range, it judges the oil quality information in the gearbox and the normal oil quality information, and at the same time performs a risk assessment of gearbox faults and displays the changes in real-time data.

[0006] Furthermore, the data acquisition module includes: The data acquisition unit is used to deploy various types of sensors to collect real-time oil quality information signals and vibration signals; the oil quality information includes, but is not limited to, viscosity, acid value, alkalinity, water content, iron content, metal abrasive type, size and concentration data; The data processing unit is used to preprocess oil quality information signals and vibration signals.

[0007] Furthermore, the data processing unit specifically comprises: Abnormal data points caused by sensor errors or external interference are identified and removed by statistical methods, and low-pass filtering is used to reduce the noise impact caused by electromagnetic interference or mechanical vibration. The viscosity, acid value, alkalinity, moisture content, and iron content of the gearbox oil quality information are quantitatively calculated. At the same time, the type, size, and concentration data of metal abrasive particles in the gearbox oil quality information are extracted to assess the degree of gear wear. Parameters of different dimensions are standardized to eliminate the impact of magnitude differences on the analysis. Principal component analysis is used to screen key parameters that are sensitive to oil condition in order to reduce data dimensionality.

[0008] Furthermore, the processor module includes: The data conversion unit is used to decode and modulate the preprocessed real-time gearbox oil quality information signal and vibration signal, and convert the effective level signal to obtain the converted gearbox oil quality information and vibration signal. The data storage unit is used to save the converted gearbox oil quality information and vibration signals.

[0009] Furthermore, the data analysis module specifically comprises: Analyze the changing trend of the content of metal abrasive particles in the gearbox oil, and determine the wear stage of the gearbox based on the analysis results; The kinematic viscosity of gearbox oil is analyzed by taking into account its temperature and moisture content, and the cause of abnormal oil quality is determined based on the analysis results.

[0010] Furthermore, the data analysis module uses the following logic to determine the wear stage of the gearbox based on the analysis results: When both flaky and spherical abrasive particles are detected in the oil, and the following conditions are met, it is determined to be in the fatigue wear stage: the average aspect ratio of the flaky abrasive particles is not less than 3:1, the roundness of the spherical abrasive particles is not less than 0.8, the concentration of both types of abrasive particles shows a continuous upward trend within 24 hours and the growth rate exceeds 12%; the load fluctuation amplitude of the synchronously correlated gearbox is less than 8%, and the speed stability deviation is ≤5%. When severe sliding abrasive particles are detected and the following conditions are met, the wear is classified as adhesive wear: the equivalent diameter of the sliding abrasive particles is not less than 50 μm, and the hardness of the abrasive particles is more than 1.2 times that of the gear material; the ratio of the maximum to minimum abrasive particle concentration within 224 hours is greater than 1.15, and the time interval between the peak abrasive particle concentration and the peak gearbox operating temperature does not exceed 30 minutes; the gearbox operating speed stability deviation is ≤3%. When cutting abrasive particles are detected and the following conditions are met, the abrasive wear stage is determined: the aspect ratio of the cutting abrasive particles is not less than 5:1, and high-hardness contaminant particles are present, with the ratio of the concentration of high-hardness contaminant particles to the concentration of cutting abrasive particles not less than 0.6; the concentration of cutting abrasive particles increases by more than 15% within 48 hours, and the effective value of the gearbox vibration acceleration is simultaneously monitored to exceed 2.5 m / s², with the vibration frequency concentrated in the 100-500 Hz frequency band.

[0011] Furthermore, the data analysis module utilizes the following logic to determine the cause of abnormal oil quality based on the analysis results: When the change rate of the kinematic viscosity of the oil exceeds 10% within 1 hour, and the change rates of moisture content, temperature, and dielectric constant within 1 hour are all less than 8%, and multiple incompatible additive components are detected in the oil, the cause of the abnormal oil quality is determined to be the mixing of different types of oil. When the moisture content of the oil increases by more than 5% within 48 hours and the kinematic viscosity decreases by more than 8%, and the humidity of the external environment of the gearbox exceeds 65%, the abnormal oil quality is determined to be caused by moisture intrusion due to desiccant failure or aging of seals. When the acid value of the oil increases by more than 0.3 mg KOH / g within 72 hours, accompanied by a synchronous increase in the content of metal elements, and the gearbox operating temperature is consistently above 90℃, the abnormal oil quality is determined to be caused by overheating and oxidation degradation of the oil.

[0012] Furthermore, the online monitoring module includes: The early warning assessment unit is used to build a fault early warning model and calculate the fault early warning threshold range. When the model output results exceed the fault early warning threshold range, it judges the oil quality information in the gearbox and the normal oil quality information, and at the same time performs a risk assessment of gearbox faults. The real-time data display unit is used to display changes in real-time data, the probability of failure, and corresponding maintenance solutions.

[0013] Furthermore, the early warning assessment unit includes: The fault early warning subunit is used to construct a fault early warning model based on the results of the data analysis module, calculate the fault early warning threshold range, and determine whether the fault value output by the fault early warning model is within the fault early warning threshold range. When the fault value exceeds the fault early warning threshold range, it compares the oil quality information in the gearbox with the normal oil quality information, specifically: Collect historical quality information of gearbox oil within a preset time period to construct a quality information database; the historical quality information specifically refers to the quality information of gearbox oil before the gearbox failure. The historical quality information of the gearbox oil is divided into first historical quality information and second historical quality information. A fault early warning model is constructed based on the first historical quality information, and the fault early warning threshold is determined based on the fault early warning model and the second historical quality information. The abnormal oil quality causes analyzed by the analysis module are input into the fault early warning model. The fault early warning model and fault early warning threshold are used to determine whether the real-time quality information of the gearbox oil meets the fault early warning conditions. If so, an early warning signal is generated. The risk assessment subunit is used to calculate the fault risk assessment coefficient based on the warning results from the fault early warning unit, determine the probability of fault occurrence through the fault risk assessment coefficient, and finally conduct a comprehensive risk assessment of the fault. Specifically: When the quality information of the oil in the gearbox is not greater than or close to the fault warning threshold, the real-time quality information of the gearbox oil is obtained. A comprehensive status information matrix is ​​established based on the current quality information of the gearbox oil, and the fault risk assessment coefficient is determined through the comprehensive status information matrix. The probability of gearbox failure is determined by the failure risk assessment coefficient, and a corresponding maintenance plan is made and recommended based on the probability of failure.

[0014] This invention also provides a method for online monitoring of gearbox oil in wind turbine units, comprising the following steps: S1. Collect real-time information about the gearbox oil, process the real-time information about the gearbox oil, and transmit it to the processor module. S2. Convert the real-time relevant information of the processed gearbox oil into an effective level signal and save it; S3. Analyze the information stored in the processor module, and determine the wear stage of the gearbox and the cause of abnormal oil quality based on the analysis results. S4. Construct a fault early warning model and calculate the fault early warning threshold range. When the model output exceeds the fault early warning threshold range, determine whether the oil quality information in the gearbox is the same as the normal oil quality information. At the same time, conduct a risk assessment of gearbox faults and display the changes in real-time data.

[0015] The advantages of this invention are: (1) This invention uses multiple sensors to collect oil quality information signals and vibration signals in a coordinated manner, and monitors the oil condition from multiple dimensions, effectively avoiding the limitations of monitoring a single parameter. By monitoring the viscosity, moisture content and abrasive index of the gearbox oil in real time, abnormal conditions caused by lubrication failure or increased wear can be identified. For example, the rising trend of iron content can directly reflect the wear process inside the gearbox. Combined with pressure index analysis, potential faults can be warned in advance. The predictive maintenance strategy based on the trend of oil parameter changes can reduce unplanned downtime losses. By monitoring moisture and metal particles in the oil, secondary wear caused by lubrication deterioration can be effectively suppressed, and the gearbox can be monitored around the clock to avoid safety accidents caused by sudden failures. This effectively improves the real-time performance and accuracy of gearbox oil monitoring.

[0016] (2) By monitoring the oil status in the gearbox in real time, the present invention can ensure the stable performance of the gear oil, prevent hydraulic oil contamination caused by incorrect filling or seal damage, and ensure the lubrication effect and reliability of the gearbox.

[0017] By monitoring oil levels, oil change intervals can be accurately calculated, reducing engine damage caused by over-maintenance or missed checks, thereby lowering operating costs. Based on real-time monitoring of gearbox oil quality information, gearbox malfunctions can be predicted, allowing for preventative measures to be taken before malfunctions occur, thus reducing repair costs and losses due to downtime. Based on the predicted malfunction status, more scientific maintenance plans can be developed, reducing over-maintenance or under-maintenance and extending gearbox lifespan.

[0018] By analyzing vibration information, the operating status of the gearbox and the nature and scope of the fault can be determined, thereby identifying the cause of the fault and taking corresponding maintenance measures. It can accurately determine the location and severity of gear faults and help maintenance personnel formulate the best maintenance strategy. Attached Figure Description

[0019] Figure 1 This is a structural diagram of an online monitoring system for wind turbine gearbox oil according to Embodiment 1 of the present invention; Figure 2 This is a flowchart of an online monitoring method for wind turbine gearbox oil according to Embodiment 1 of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments: Example 1 like Figure 1 As shown, specifically, an online monitoring system for wind turbine gearbox oil is disclosed, including a data acquisition module, a processor module, a data analysis module, and an online monitoring module.

[0022] The data acquisition module is used to collect real-time information related to the gearbox oil, process the real-time information related to the gearbox oil, and transmit it to the processor module. The data acquisition module specifically includes: The data acquisition unit is used to deploy various types of sensors to collect real-time oil quality information signals and vibration signals; the oil quality information includes, but is not limited to, viscosity, acid value, alkalinity, moisture content, iron content, metal abrasive type, size and concentration data.

[0023] The data processing unit is used to preprocess the oil quality information signal and vibration signal. The data preprocessing specifically includes: Abnormal data points caused by sensor errors or external interference are identified and removed using statistical methods, and low-pass filtering is used to reduce the noise impact caused by electromagnetic interference or mechanical vibration.

[0024] Quantitative calculations are performed on the viscosity, acid value, base value, moisture content, and iron content in the gearbox oil quality information. Simultaneously, data on the type, size, and concentration of metal abrasive particles are extracted from the gearbox oil quality information to assess the degree of gear wear. Specifically, the extraction of the type, size, and concentration of metal abrasive particles from the gearbox oil quality information includes: (1) By analyzing the characteristics of the preprocessed pulse signal and combining the working principle of the sensor, the type, size and concentration of abrasive particles are deduced in reverse; the characteristics of the pulse signal include amplitude, width, waveform and spectrum; If the pulse signal has a large amplitude and a steep rising edge, it is a ferromagnetic abrasive. If the pulse amplitude is only 1 / 3 to 1 / 2 of the amplitude corresponding to ferromagnetic abrasive and the rising edge is gentle, it is a non-ferromagnetic abrasive. (2) Perform sensor calibration in advance using standard abrasive grains of known size, plot the calibration curve of pulse amplitude and abrasive grain size, extract the peak amplitude of abrasive grain pulse signal, substitute it into the calibration curve, and calculate the equivalent diameter of abrasive grain; for example: the calibration curve of a certain electromagnetic sensor is size = 0.5 × amplitude + 2. If the pulse amplitude is 18mV, then the abrasive grain size is 11μm. (3) Count the effective abrasive pulse signals after preprocessing and count the number of abrasive particles of different sizes; (4) Calculate the abrasive concentration in a unit volume of oil based on the oil flow rate detected by the sensor. The formula is: abrasive concentration (particles / mL) = abrasive count (particles) / detection time (min) × oil flow rate (mL / min). For example, if 200 abrasive particles larger than 5μm are counted in 1 minute and the oil flow rate is 50mL / min, then the concentration is 200 particles / 50mL = 4 particles / m.

[0025] Parameters of different dimensions are standardized to eliminate the impact of magnitude differences on the analysis. Principal component analysis is used to screen key parameters sensitive to oil condition, thereby reducing data dimensionality. Specifically, the screening of key parameters sensitive to oil condition using principal component analysis includes: Based on the preprocessed data, the covariance matrix between each parameter is calculated, the correlation between parameters is analyzed, and the eigenvalues ​​and eigenvectors of the covariance matrix are extracted through eigenvalue decomposition or singular value decomposition. The larger the eigenvalue, the more oil state information is contained in the corresponding principal component. Based on the principle that the eigenvalue is greater than 1 or the cumulative variance contribution rate is ≥85%, the key principal components are determined, and the loading matrix of the key principal components is analyzed. The parameter with the larger the absolute value of the loading coefficient has, the higher its contribution to the principal component, indicating that it is more sensitive to changes in oil condition.

[0026] In this embodiment, if the absolute values ​​of the loading coefficients of acid value, dielectric loss factor, and viscosity in the loading matrix of the first principal component are the largest (e.g., 0.85, 0.82, and 0.78 respectively), then the above three parameters are used as sensitive parameters reflecting the oxidative deterioration of the oil. If the loading coefficients of metal abrasive concentration and particulate contamination are the largest in the loading matrix of the second principal component (e.g., 0.91 and 0.87), then the above two parameters are used as sensitive parameters reflecting oil wear and contamination.

[0027] The processor module is used to convert and save the real-time information related to the processed gearbox oil into effective level signals. The processor module specifically includes: The data conversion unit is used to decode and modulate the preprocessed real-time gearbox oil quality information signal and vibration signal, and convert the effective level signal to obtain the converted gearbox oil quality information and vibration signal, which is represented by the following logic: In this embodiment, the decoding and modulation of the preprocessed real-time gearbox oil quality information signal specifically includes: (1) By using the sensor calibration curve, the preprocessed standard electrical signal is decoded into the actual oil parameter value. The decoded oil parameters are associated with typical gearbox faults to form a correspondence between oil index and fault type. (2) Convert the decoded oil parameters into dimensionless normalized values ​​to eliminate the influence of differences in the magnitude of different parameters. Perform moving average, exponential weighted filtering or differential processing on the time series oil parameters to highlight the changing trend of the parameters rather than instantaneous fluctuations. Generate an oil health index by weighted fusion of multiple oil parameters to comprehensively evaluate the overall state of the oil and avoid misjudgment of a single parameter.

[0028] In this embodiment, the decoding and modulation of the preprocessed real-time gearbox vibration signal specifically includes: (1) Convert the time domain signal into the frequency domain signal by Fourier transform, decode the fault characteristic frequency, accurately locate the faulty component, and for non-stationary vibration signal, decode the time-frequency diagram by wavelet transform to capture the time and frequency distribution of the fault impact signal. (2) Use bandpass filtering or adaptive filtering to retain the frequency band where the fault characteristic frequency is located, filter out irrelevant noise, use the resonance characteristics of the sensor or system to amplify the weak impact signal generated by the fault, and then extract the fault characteristics through envelope detection. For the non-stationary vibration signal of the gearbox, perform angular domain resampling and then synchronous averaging to eliminate the influence of speed fluctuation on meshing frequency and highlight the periodic fault signal of the gear. (3) The abrasive concentration of the oil signal decoding is correlated with the bearing fault frequency amplitude of the vibration signal decoding. The modulated oil health index and vibration health index are fused through the data fusion algorithm to generate the gearbox comprehensive health index, which comprehensively reflects the oil lubrication status and mechanical structure status, and provides a more comprehensive basis for fault early warning.

[0029] The data storage unit is used to save the converted gearbox oil quality information and vibration signals.

[0030] The data analysis module is used to analyze the information stored in the processor module, and determine the wear stage of the gearbox and the cause of abnormal oil quality based on the analysis results. Specifically, the data analysis module is as follows: When both flaky and spherical abrasive particles are detected in the oil and the following conditions are met, it is determined to be in the fatigue wear stage: the average aspect ratio of the flaky abrasive particles is not less than 3:1, the roundness of the spherical abrasive particles is not less than 0.8, the concentration of both types of abrasive particles shows a continuous upward trend within 24 hours and the growth rate exceeds 12%; the load fluctuation of the synchronously associated gearbox is less than 8% and the speed stability deviation is ≤5%; the cause is that under the long-term alternating contact stress, microcracks are generated on the surface of the gearbox friction pair and gradually expand, leading to material spalling. Furthermore, by using the abrasive particle morphology quantification parameters, the abrasive particle differences between fatigue wear and other wear types can be accurately distinguished, avoiding misjudgment. When severe sliding abrasive particles are detected and the following conditions are met, it is determined to be in the adhesive wear stage: the area equivalent diameter of the sliding abrasive particles is not less than 50 μm, and the hardness of the abrasive particles is more than 1.2 times that of the gear material; the ratio of the maximum to minimum abrasive particle concentration within 224 hours is greater than 1.15, and the time interval between the peak abrasive particle concentration and the peak gearbox operating temperature does not exceed 30 minutes; the gearbox operating speed stability deviation is ≤3%; the cause is that the local temperature at the contact point of the friction pair is too high due to load concentration or lubrication failure, causing the metal to melt and adhere, and subsequent relative sliding causes material peeling. By coupling the abrasive particle size, hardness and temperature, the accuracy of adhesive wear judgment is improved. When cutting abrasive particles are detected and the following conditions are met, the abrasive wear stage is determined: the aspect ratio of the cutting abrasive particles is not less than 5:1, and high-hardness contaminant particles are present, with the ratio of the concentration of high-hardness contaminant particles to the concentration of cutting abrasive particles not less than 0.6; the concentration of cutting abrasive particles increases by more than 15% within 48 hours, and the effective value of the gearbox vibration acceleration is simultaneously monitored to exceed 2.5 m / s², with the vibration frequency concentrated in the 100-500 Hz frequency band; the cause is the intrusion of external hard media foreign objects into the contact surface of the friction pair, or the cutting wear caused by hard impurities inside the gear material during relative motion. By using the correlation between the concentration of cutting abrasive particles and contaminant particles and vibration characteristics to assist in the determination, the problem of easy misjudgment by the existing technology that only relies on the ratio of the number of abrasive particles is solved.

[0031] Furthermore, the data analysis module employs the following logic to trace the root cause of oil quality abnormalities: When the change rate of the kinematic viscosity of the oil exceeds 10% within 1 hour, and the change rates of moisture content, temperature, and dielectric constant within 1 hour are all less than 8%, and multiple incompatible additive components are detected in the oil, it is determined that the abnormal oil quality is caused by the mixing of different types of oil. When the moisture content of the oil increases by more than 5% within 48 hours and the kinematic viscosity decreases by more than 8%, and the humidity of the external environment of the gearbox exceeds 65%, the abnormal oil quality is determined to be caused by moisture intrusion due to desiccant failure or aging of seals. When the acid value of the oil increases by more than 0.3 mg KOH / g within 72 hours, accompanied by a synchronous increase in the content of metal elements, and the gearbox operating temperature is consistently above 90°C, the abnormal oil quality is determined to be caused by overheating and oxidation degradation of the oil.

[0032] The online monitoring module is used to construct a fault early warning model and calculate the fault early warning threshold range. When the model output exceeds the fault early warning threshold range, it compares the oil quality information in the gearbox with the normal oil quality information, performs a risk assessment of gearbox faults, and displays the changes in real-time data. The online monitoring module specifically includes: The early warning assessment unit is used to construct a fault early warning model and calculate the fault early warning threshold range. When the model output result exceeds the fault early warning threshold range, it judges the oil quality information in the gearbox and the normal oil quality information, and at the same time performs a risk assessment on the gearbox fault. The early warning assessment unit includes a fault early warning subunit and a risk assessment subunit.

[0033] The fault early warning subunit is used to construct a fault early warning model based on the results of the data analysis module, calculate the fault early warning threshold range, and determine whether the fault value output by the fault early warning model is within the fault early warning threshold range. When the fault value exceeds the fault early warning threshold range, it compares the oil quality information in the gearbox with the normal oil quality information. Specifically, the fault early warning subunit is as follows: (1) Collect historical quality information of gearbox oil within a preset time period and construct a quality information database; in this embodiment, the historical quality information is specifically the quality information of gearbox oil before the gearbox failure. (2) Divide the historical quality information of the gearbox oil into first historical quality information and second historical quality information. Construct a fault early warning model based on the first historical quality information. Determine the fault early warning threshold based on the fault early warning model and the second historical quality information. Specifically, this includes: Historical data of oil parameters that are directly and strongly correlated with gearbox failures are used as the primary historical quality information, including historical changes in metal abrasive concentration, historical levels of particulate contamination, historical growth rate of total acid value, and records of exceeding water content standards. The oil data indirectly related to the fault is used as the second historical quality information, including historical fluctuations in oil viscosity index, historical records of gearbox operating temperature, changes in the concentration of non-critical metal elements, and records of regular oil change cycles. Outliers were removed from the first historical quality information. Data from the first category of the six months prior to the gearbox failure was labeled as pre-failure data, and data from the period without failure was labeled as normal data. Analyze the changing patterns of data before the failure and extract core features, which are metal abrasive concentration, particulate contamination degree and total acid value; and convert these features into identifiable first historical feature parameters, such as 50% Fe growth rate as feature parameter F1. The labeled pre-fault data, normal data, and extracted first historical feature parameters are input into a logistic regression model. The logistic regression model is used to learn the association between the combination of first historical feature parameters and the occurrence of the fault. For example, when "F1 (Fe growth rate ≥ 50%) and particulate pollution level ≥ NAS8" occur simultaneously, the probability of fault is ≥ 70%. The model was validated using first-hand historical data that had not been used for training for nearly a year. The weights of the feature parameters were adjusted (e.g., the weight of the Fe concentration feature was increased to 30%) to ensure that the model accuracy was ≥90%, thus obtaining the fault warning model. Based on the model training results, the warning thresholds for the first historical feature parameters are set: the first-level warning threshold is when a single feature parameter meets the standard (e.g., Fe growth rate ≥ 50%); the second-level warning threshold is when two or more feature parameters meet the standard simultaneously (e.g., Fe growth rate ≥ 50% and total acid value increase ≥ 0.5 mg KOH / g). Configure system trigger rules: when a Level 1 alert is triggered, the system displays a yellow warning; when a Level 2 alert is triggered, an SMS message is automatically sent to maintenance personnel, and the fault risk is simultaneously marked as ≥60%. The early warning model was validated using second historical quality information. If the model triggered an early warning, and the gearbox temperature was higher than normal for three consecutive days in the second historical information, the reliability of the early warning was increased by 20%. (3) Input the abnormal oil quality causes analyzed by the analysis module into the fault warning model. Use the fault warning model and fault warning threshold to determine whether the real-time quality information of the gearbox oil meets the fault warning conditions. If so, generate a warning signal.

[0034] The risk assessment subunit is used to calculate the fault risk assessment coefficient based on the warning results of the fault warning unit, determine the probability of fault occurrence through the fault risk assessment coefficient, and finally conduct a comprehensive risk assessment of the fault. Specifically, the risk assessment subunit is as follows: (1) When the quality information of the oil in the gearbox is not greater than or close to the fault warning threshold, the real-time quality information of the gearbox oil is obtained; (2) Establish a comprehensive status information matrix based on the current quality information of the gearbox oil, and determine the fault risk assessment coefficient through the comprehensive status information matrix, specifically: First, identify the relevant oil parameters that can directly reflect gearbox problems, use the oil parameters as columns of the information matrix, and form a tabular framework for the data collection time. In this embodiment, the relevant oil indicators include contamination-related, aging-related, wear-related, and basic performance indicators. Contamination-related indicators include water content and the amount of impurity particles (e.g., described as clean / small amount / large amount); aging-related indicators include oil color (transparent / slightly yellow / dark brown) and acid value (normal / high / exceeding standard); wear-related indicators include the amount of metal debris (none / trace / obvious) and the content level of key metal elements (iron, copper); basic performance indicators include whether the viscosity is within the acceptable range (normal / high / low) and whether the flash point meets the standard. Secondly, referencing industry standards, determine the risk standards for each indicator, dividing each indicator into three risk ranges: low risk, medium risk, and high risk, and assigning a score from 0 to 10. For example, for moisture content, ≤0.03% is low risk (0-3 points), 0.03%-0.1% is medium risk (4-7 points), and >0.1% is high risk (8-10 points). For example, for metal debris, no debris under a microscope is low risk (0-3 points), a small amount of debris is medium risk (4-7 points), and obvious debris is high risk (8-10 points).

[0035] Next, based on the actual collected data, a corresponding score is assigned to each indicator at each time point, filling the entire tabular framework to obtain a comprehensive status information matrix, specifically: Obtain historical fault data for the gearbox and score them accordingly. For example, metal shavings and particulate contamination each account for 20% (important), viscosity and acid value each account for 15% (relatively important), and moisture and color each account for 10% (generally important), totaling 100%. Identify the relevant oil parameters that exceeded standards before the historical failure, assign high weights to these parameters, and finally form a weight table, for example: metal debris 20%, particulate contamination 20%, viscosity 15%, acid value 15%, moisture 10%, color 10%, flash point 5%, and others 5%; Finally, the state score of each indicator is multiplied by the corresponding weight to obtain a weighted score. For example, 10:05 Iron element 5 points × 30% = 1.5 points, number of particles 4 points × 25% = 1 point. The weighted scores are summed to obtain the risk coefficient, which is then matched with a level. The risk coefficient has a maximum score of 10 points: 0-3 points: low risk (the system displays green, no intervention required); 3-7 points: medium risk (the system displays yellow, reminding users to pay attention to the trend); 7-10 points: high risk (the system triggers a red warning, prompting shutdown for maintenance).

[0036] (3) Determine the probability of gearbox failure through the failure risk assessment coefficient. The higher the risk assessment coefficient, the greater the probability of failure. Make corresponding maintenance plans based on the probability of failure, and recommend maintenance plans (such as oil change and component repair), which specifically include: Gearbox fault data for one year was collected, and fault ranges were determined based on the data. The fault range of 0%-5% (no faults or minor faults) was set, matching the low-risk range. The fault range of 5%-30% (multiple warnings triggered but no faults, faults that occurred, and related fluid indicators showing an upward trend) was set, matching the medium-risk range. The fault range of 30%-90% (serious faults and related fluid indicators exceeding the standard by 2 times) was set, matching the high-risk range. The details are shown in Table 1 below. Table 1. Maintenance Plans Corresponding to Fault Risk Assessment Coefficients

[0037] The real-time data display unit is used to display changes in real-time data, the probability of failure, and corresponding maintenance plans. The real-time data includes real-time monitoring data of gearbox oil temperature, gearbox oil water content, gearbox oil viscosity, gearbox oil dielectric constant, metal abrasive concentration, metal element content, and contaminant particle concentration.

[0038] This invention identifies abnormal conditions caused by lubrication failure or accelerated wear by real-time monitoring of gearbox oil viscosity, moisture content, and abrasive particle indicators. For example, an increasing iron content trend directly reflects the internal wear process of the gearbox. Combined with pressure index analysis, it can provide early warning of potential faults. Predictive maintenance strategies based on oil parameter change trends can reduce unplanned downtime losses. By monitoring moisture and metal particles in the oil, this invention can effectively suppress secondary wear caused by lubrication deterioration, providing 24 / 7 condition monitoring of the gearbox and preventing safety accidents caused by sudden failures.

[0039] This invention also ensures stable gear oil performance by monitoring the oil status in the gearbox in real time, preventing hydraulic oil contamination caused by incorrect filling or seal damage, and ensuring the lubrication effect and reliability of the gearbox. Oil monitoring allows for precise calculation of oil change cycles, reducing engine damage caused by over-maintenance or missed inspections, thereby lowering operating costs. Based on real-time monitoring of gearbox oil quality information, it predicts gearbox faults and allows for proactive measures to prevent sudden gearbox damage, reducing maintenance costs and downtime losses. Based on the predicted fault status, a more scientific maintenance plan can be developed, reducing over-maintenance or under-maintenance and extending gearbox lifespan. Analyzing vibration information allows for assessment of the gearbox's operating status and the nature and scope of faults, thereby identifying the cause of the fault and taking appropriate repair measures. It accurately determines the location and severity of gear faults, helping maintenance personnel develop optimal maintenance strategies.

[0040] Example 2 Specifically, such as Figure 2 As shown, the present invention also provides a method for online monitoring of gearbox oil in wind turbine units, comprising the following steps: S1. Collect real-time information related to the gearbox oil, process the real-time information related to the gearbox oil, and transmit it to the processor module; S1 also includes the following steps: S11. Deploy various types of sensors to collect real-time oil quality information signals and vibration signals; the oil quality information includes, but is not limited to, viscosity, acid value, alkalinity, moisture content, iron content, metal abrasive type, size and concentration data.

[0041] S12. Perform data preprocessing on the oil quality information signal and vibration signal. The data preprocessing specifically includes: Abnormal data points caused by sensor errors or external interference are identified and removed by statistical methods, and low-pass filtering is used to reduce the noise impact caused by electromagnetic interference or mechanical vibration. The viscosity, acid value, alkalinity, moisture content, and iron content of the gearbox oil quality information are quantitatively calculated. At the same time, the type, size, and concentration data of metal abrasive particles in the gearbox oil quality information are extracted to assess the degree of gear wear. Parameters of different dimensions are standardized to eliminate the impact of magnitude differences on the analysis. Principal component analysis is used to screen key parameters that are sensitive to oil condition in order to reduce data dimensionality.

[0042] S2. Convert the real-time relevant information of the processed gearbox oil into an effective level signal and save it; S2 also includes the following steps: S21. Decode and modulate the preprocessed real-time gearbox oil quality information signal and vibration signal, and convert the effective level signal to obtain the converted gearbox oil quality information and vibration signal.

[0043] S22. Save the converted gearbox oil quality information and vibration signal.

[0044] S3. Analyze the information stored in the processor module, and determine the wear stage of the gearbox and the cause of abnormal oil quality based on the analysis results; S3 also includes the following steps: S31. Analyze the changing trend of the number of metal abrasive particles in the gearbox oil, determine the wear stage of the gearbox based on the analysis results, and use the following logic to express the judgment criteria for the wear stage: When both flaky and spherical abrasive particles are detected in the oil, and the following conditions are met, it is determined to be in the fatigue wear stage: the average aspect ratio of the flaky abrasive particles is not less than 3:1, the roundness of the spherical abrasive particles is not less than 0.8, the concentration of both types of abrasive particles shows a continuous upward trend within 24 hours and the growth rate exceeds 12%; the load fluctuation amplitude of the synchronously correlated gearbox is less than 8%, and the speed stability deviation is ≤5%. When severe sliding abrasive particles are detected and the following conditions are met, the wear is classified as adhesive wear: the equivalent diameter of the sliding abrasive particles is not less than 50 μm, and the hardness of the abrasive particles is more than 1.2 times that of the gear material; the ratio of the maximum to minimum abrasive particle concentration within 224 hours is greater than 1.15, and the time interval between the peak abrasive particle concentration and the peak gearbox operating temperature does not exceed 30 minutes; the gearbox operating speed stability deviation is ≤3%. When cutting abrasive particles are detected and the following conditions are met, the abrasive wear stage is determined: the aspect ratio of the cutting abrasive particles is not less than 5:1, and high-hardness contaminant particles are present, with the ratio of the concentration of high-hardness contaminant particles to the concentration of cutting abrasive particles not less than 0.6; the concentration of cutting abrasive particles increases by more than 15% within 48 hours, and the effective value of the gearbox vibration acceleration is simultaneously monitored to exceed 2.5 m / s², with the vibration frequency concentrated in the 100-500 Hz frequency band.

[0045] S32. Analyze the kinematic viscosity of the gearbox oil using its temperature and moisture content. Based on the analysis results, determine the cause of the oil quality abnormality. Use the following logic to represent the judgment criteria for oil quality abnormality: When the change rate of the kinematic viscosity of the oil exceeds 10% within 1 hour, and the change rates of moisture content, temperature, and dielectric constant within 1 hour are all less than 8%, and multiple incompatible additive components are detected in the oil, it is determined that the abnormal oil quality is caused by the mixing of different types of oil. When the moisture content of the oil increases by more than 5% within 48 hours and the kinematic viscosity decreases by more than 8%, and the humidity of the external environment of the gearbox exceeds 65%, the abnormal oil quality is determined to be caused by moisture intrusion due to desiccant failure or aging of seals. When the acid value of the oil increases by more than 0.3 mg KOH / g within 72 hours, accompanied by a synchronous increase in the content of metal elements, and the gearbox operating temperature is consistently above 90°C, the abnormal oil quality is determined to be caused by overheating and oxidation degradation of the oil.

[0046] S4. Construct a fault early warning model and calculate the fault early warning threshold range. When the model output exceeds the fault early warning threshold range, determine the difference between the oil quality information in the gearbox and the normal oil quality information. Simultaneously, perform a risk assessment of the gearbox fault and display the changes in real-time data. S4 includes the following steps: S41. Construct a fault early warning model and calculate the fault early warning threshold range. When the model output exceeds the fault early warning threshold range, determine the quality information of the oil in the gearbox compared to the normal oil quality information, and simultaneously conduct a risk assessment of the gearbox fault. S41 also includes the following steps: S411. Construct a fault early warning model based on the results of the data analysis module, calculate the fault early warning threshold range, and determine whether the fault value output by the fault early warning model is within the fault early warning threshold range. When the fault value exceeds the fault early warning threshold range, compare the oil quality information in the gearbox with the normal oil quality information, specifically: Collect historical quality information of gearbox oil within a preset time period to construct a quality information database; in this embodiment, the historical quality information specifically refers to the quality information of gearbox oil before the gearbox failure. The historical quality information of the gearbox oil is divided into first historical quality information and second historical quality information. A fault early warning model is constructed based on the first historical quality information, and the fault early warning threshold is determined based on the fault early warning model and the second historical quality information. The abnormal oil quality causes analyzed by the analysis module are input into the fault early warning model. The fault early warning model and the fault early warning threshold are used to determine whether the real-time quality information of the gearbox oil meets the fault early warning conditions. If so, an early warning signal is generated.

[0047] Furthermore, the construction of the fault early warning model specifically involves: Genetic algorithms were used to screen variables related to gearbox oil quality from the first historical quality information. The relevant variables included gearbox oil temperature and gearbox oil pressure. The importance of the variables was calculated by combining random forest algorithms, and key features were extracted. Adaptive noise ensemble empirical mode decomposition is used to decompose the time-series characteristics of variables related to gearbox oil quality information, and principal component analysis is combined to reduce dimensionality and reduce redundant information. The relationship between the temporal sequence of gearbox oil quality information changes and related variables of gearbox oil quality information was captured based on a time-recurrent neural network. A neural network algorithm was selected to establish a regression model by utilizing the relationship between the time series of gearbox oil quality information changes and the relevant variables of gearbox oil quality information; Cross-validation was used to evaluate the accuracy of the regression model. Based on the evaluation results and historical failure cases, the warning threshold was adjusted to obtain a fault warning model, which reduced the false alarm rate and predicted the normal range of oil temperature. Multi-source data fusion is performed by combining other gearbox parameters (such as vibration spectrum and lubricating oil particle size) to reduce false alarm rate.

[0048] S412. Calculate the fault risk assessment coefficient based on the warning results from the fault warning unit, determine the probability of fault occurrence through the fault risk assessment coefficient, and finally conduct a comprehensive risk assessment of the fault, specifically as follows: When the quality information of the oil in the gearbox is not greater than or close to the fault warning threshold, the real-time quality information of the gearbox oil is obtained. A comprehensive status information matrix is ​​established based on the current quality information of the gearbox oil, and the fault risk assessment coefficient is determined through the comprehensive status information matrix. The probability of gearbox failure is determined by the failure risk assessment coefficient. Based on the probability of failure, a corresponding maintenance plan is made and a maintenance plan (such as oil change or component repair) is recommended.

[0049] S42. Displays real-time data changes, the probability of failure, and corresponding maintenance solutions.

[0050] Example 3 This invention also provides an online monitoring device for wind turbine gearbox oil, including a micro-water detection element, an oil temperature detection element, an oil pressure detection element, and a metal shavings detection element. The micro-water detection element is installed in the second branch pipe of the wind turbine gearbox oil monitoring circuit to monitor the water content of the oil. The oil temperature detection element is arranged at the bottom of the gearbox oil pan and in the monitoring cavity of the branch pipe of the oil monitoring circuit. This arrangement can cover the normal operating range of the oil and ensure the accuracy of oil temperature monitoring. The oil pressure detection element is set in front of the oil inlet and behind the oil outlet of the lubricating oil monitoring bypass. Specifically, an inlet-side oil pressure detection element is equipped in front of the oil inlet of the online monitoring device, and an outlet-side oil pressure detection element is equipped behind the oil outlet to realize real-time monitoring of pressure fluctuations in the oil circuit system. The metal shavings detection element is installed in the first branch pipe of the wind turbine gearbox oil monitoring circuit to monitor and count metal particles in the oil and can provide feedback on the size and morphological distribution data of the particles.

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or gearbox that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or gearbox.

[0052] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online monitoring system for gearbox oil in a wind turbine generator set, characterized in that, It includes a data acquisition module, a processor module, a data analysis module, and an online monitoring module; The data acquisition module is used to collect real-time information related to the gearbox oil, and process and transmit the real-time information related to the gearbox oil to the processor module. The processor module is used to convert the real-time information of the processed gearbox oil into an effective level signal and save it. The data analysis module is used to analyze the information stored in the processor module and determine the wear stage of the gearbox and the cause of abnormal oil quality based on the analysis results. The online monitoring module is used to construct a fault early warning model and calculate the fault early warning threshold range. When the model output result exceeds the fault early warning threshold range, it judges the oil quality information in the gearbox and the normal oil quality information, and at the same time performs a risk assessment of gearbox faults and displays the changes in real-time data.

2. The online monitoring system for wind turbine gearbox oil according to claim 1, characterized in that, The data acquisition module includes: The data acquisition unit is used to deploy various types of sensors to collect real-time oil quality information signals and vibration signals; the oil quality information includes, but is not limited to, viscosity, acid value, alkalinity, water content, iron content, metal abrasive type, size and concentration data; The data processing unit is used to preprocess oil quality information signals and vibration signals.

3. The online monitoring system for wind turbine gearbox oil according to claim 2, characterized in that, The data processing unit specifically comprises: Abnormal data points caused by sensor errors or external interference are identified and removed by statistical methods, and low-pass filtering is used to reduce the noise impact caused by electromagnetic interference or mechanical vibration. The viscosity, acid value, alkalinity, moisture content, and iron content of the gearbox oil quality information are quantitatively calculated. At the same time, the type, size, and concentration data of metal abrasive particles in the gearbox oil quality information are extracted to assess the degree of gear wear. Parameters of different dimensions are standardized to eliminate the impact of magnitude differences on the analysis. Principal component analysis is used to screen key parameters that are sensitive to oil condition in order to reduce data dimensionality.

4. The online monitoring system for wind turbine gearbox oil according to claim 1, characterized in that, The processor module includes: The data conversion unit is used to decode and modulate the preprocessed real-time gearbox oil quality information signal and vibration signal, and convert the effective level signal to obtain the converted gearbox oil quality information and vibration signal. The data storage unit is used to save the converted gearbox oil quality information and vibration signals.

5. The online monitoring system for wind turbine gearbox oil according to claim 1, characterized in that, The data analysis module specifically includes: Analyze the changing trend of the content of metal abrasive particles in the gearbox oil, and determine the wear stage of the gearbox based on the analysis results; The kinematic viscosity of gearbox oil is analyzed by taking into account its temperature and moisture content, and the cause of abnormal oil quality is determined based on the analysis results.

6. The online monitoring system for wind turbine gearbox oil according to claim 5, characterized in that, The data analysis module uses the following logic to determine the wear stage of the gearbox based on the analysis results: When both flaky and spherical abrasive particles are detected in the oil, and the following conditions are met, it is determined to be in the fatigue wear stage: the average aspect ratio of the flaky abrasive particles is not less than 3:1, the roundness of the spherical abrasive particles is not less than 0.8, the concentration of both types of abrasive particles shows a continuous upward trend within 24 hours and the growth rate exceeds 12%; the load fluctuation amplitude of the synchronously correlated gearbox is less than 8%, and the speed stability deviation is ≤5%. When severe sliding abrasive particles are detected and the following conditions are met, the wear is classified as adhesive wear: the equivalent diameter of the sliding abrasive particles is not less than 50 μm, and the hardness of the abrasive particles is more than 1.2 times that of the gear material; the ratio of the maximum to minimum abrasive particle concentration within 224 hours is greater than 1.15, and the time interval between the peak abrasive particle concentration and the peak gearbox operating temperature does not exceed 30 minutes; the gearbox operating speed stability deviation is ≤3%. When cutting abrasive particles are detected and the following conditions are met, the abrasive wear stage is determined: the aspect ratio of the cutting abrasive particles is not less than 5:1, and high-hardness contaminant particles are present, with the ratio of the concentration of high-hardness contaminant particles to the concentration of cutting abrasive particles not less than 0.6; the concentration of cutting abrasive particles increases by more than 15% within 48 hours, and the effective value of the gearbox vibration acceleration is simultaneously monitored to exceed 2.5 m / s², with the vibration frequency concentrated in the 100-500 Hz frequency band.

7. The online monitoring system for wind turbine gearbox oil according to claim 6, characterized in that, The data analysis module uses the following logic to determine the cause of abnormal oil quality based on the analysis results: When the change rate of the kinematic viscosity of the oil exceeds 10% within 1 hour, and the change rates of moisture content, temperature, and dielectric constant within 1 hour are all less than 8%, and multiple incompatible additive components are detected in the oil, the cause of the abnormal oil quality is determined to be the mixing of different types of oil. When the moisture content of the oil increases by more than 5% within 48 hours and the kinematic viscosity decreases by more than 8%, and the humidity of the external environment of the gearbox exceeds 65%, the abnormal oil quality is determined to be caused by moisture intrusion due to desiccant failure or aging of seals. When the acid value of the oil increases by more than 0.3 mg KOH / g within 72 hours, accompanied by a synchronous increase in the content of metal elements, and the gearbox operating temperature is consistently above 90℃, the abnormal oil quality is determined to be caused by overheating and oxidation degradation of the oil.

8. The online monitoring system for wind turbine gearbox oil according to claim 1, characterized in that, The online monitoring module includes: The early warning assessment unit is used to build a fault early warning model and calculate the fault early warning threshold range. When the model output results exceed the fault early warning threshold range, it judges the oil quality information in the gearbox and the normal oil quality information, and at the same time performs a risk assessment of gearbox faults. The real-time data display unit is used to display changes in real-time data, the probability of failure, and corresponding maintenance solutions.

9. The online monitoring system for wind turbine gearbox oil according to claim 8, characterized in that, The early warning assessment unit includes: The fault early warning subunit is used to construct a fault early warning model based on the results of the data analysis module, calculate the fault early warning threshold range, and determine whether the fault value output by the fault early warning model is within the fault early warning threshold range. When the fault value exceeds the fault early warning threshold range, it compares the oil quality information in the gearbox with the normal oil quality information, specifically: Collect historical quality information of gearbox oil within a preset time period to construct a quality information database; the historical quality information specifically refers to the quality information of gearbox oil before the gearbox failure. The historical quality information of the gearbox oil is divided into first historical quality information and second historical quality information. A fault early warning model is constructed based on the first historical quality information, and the fault early warning threshold is determined based on the fault early warning model and the second historical quality information. The causes of abnormal oil quality analyzed by the analysis module are input into the fault early warning model. The fault early warning model and the fault early warning threshold are used to determine whether the real-time quality information of the gearbox oil meets the fault early warning conditions. If so, an early warning signal is generated. The risk assessment subunit is used to calculate the fault risk assessment coefficient based on the warning results from the fault early warning unit, determine the probability of fault occurrence through the fault risk assessment coefficient, and finally conduct a comprehensive risk assessment of the fault. Specifically: When the quality information of the oil in the gearbox is not greater than or close to the fault warning threshold, the real-time quality information of the gearbox oil is obtained. A comprehensive status information matrix is ​​established based on the current quality information of the gearbox oil, and the fault risk assessment coefficient is determined through the comprehensive status information matrix. The probability of gearbox failure is determined by the failure risk assessment coefficient, and a corresponding maintenance plan is made and recommended based on the probability of failure.

10. A method for online monitoring of gearbox oil in a wind turbine generator set, characterized in that, Includes the following steps: S1. Collect real-time information about the gearbox oil, process the real-time information about the gearbox oil, and transmit it to the processor module. S2. Convert the real-time relevant information of the processed gearbox oil into an effective level signal and save it; S3. Analyze the information stored in the processor module, and determine the wear stage of the gearbox and the cause of abnormal oil quality based on the analysis results. S4. Construct a fault early warning model and calculate the fault early warning threshold range. When the model output exceeds the fault early warning threshold range, determine whether the oil quality information in the gearbox is the same as the normal oil quality information. At the same time, conduct a risk assessment of gearbox faults and display the changes in real-time data.

Images