A method and system for detecting the degree of deterioration of fan lubricating oil

Abstract

The present application relates to the technical field of deterioration degree detection, and specifically includes a fan lubricating oil deterioration degree detection method and system, which comprises: obtaining lubricating oil characteristic parameters, determining first and second deterioration sensitive characteristic sets, verifying relevance in combination with historical fault data and lubricating oil replacement records, correcting the deviation of the deterioration degree evaluation function, obtaining the deterioration degree evaluation result, and generating a deterioration degree detection report. The present application solves the technical problem that the deterioration degree evaluation function relies on a single parameter, causing the deterioration cause identification to be ambiguous and the evaluation result to deviate from the actual situation, and the lubricating oil replacement cannot be accurately provided with a basis. The present application realizes the accurate identification of the deterioration cause by distinguishing the deterioration sensitive characteristic sets dominated by contaminants and mechanical wear, verifies the relevance of the deterioration degree evaluation function in combination with the historical fault data and replacement records, corrects the deviation of the deterioration degree evaluation function, improves the accuracy of the deterioration degree evaluation result, and generates a deterioration degree detection report which can provide a scientific basis for the lubricating oil replacement and gear box maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of degradation degree detection, and particularly relates to a degradation degree detection method and system for lubricating oil of a fan. BACKGROUND

[0002] As a core equipment in the fields of new energy power generation and industrial ventilation, the stable operation of the gear box of the fan is directly related to the power generation efficiency, service life and operation safety of the equipment. As a key medium for reducing friction, cooling and cleaning in the gear box, the degradation degree of the lubricating oil will significantly affect the lubricating effect of the gear box. If the degradation is excessive and not replaced in time, it may cause the gear to wear out, the equipment to malfunction and shut down, and even cause safety accidents.

[0003] Currently, the degradation degree detection of the lubricating oil of the fan mainly relies on a single parameter (such as viscosity and acid value) to evaluate the degradation degree, and does not systematically distinguish the different effects of pollutants and mechanical wear on degradation. It is difficult to accurately capture the key causes of degradation, and the degradation degree evaluation function is prone to deviation, resulting in inaccurate degradation degree judgment and difficulty in accurately reflecting the complex process of lubricating oil degradation.

[0004] In summary, the existing technology has the technical problem that the degradation degree evaluation function relies on a single parameter, causing the identification of degradation causes to be ambiguous, the evaluation results to deviate from reality, and the inability to provide accurate basis for lubricating oil replacement, thereby increasing the risk of abnormal operation of the gear box of the fan. SUMMARY

[0005] The present application provides a degradation degree detection method and system for lubricating oil of a fan, which aims to solve the technical problem that the degradation degree evaluation function in the prior art relies on a single parameter, causing the identification of degradation causes to be ambiguous, the evaluation results to deviate from reality, and the inability to provide accurate basis for lubricating oil replacement, thereby increasing the risk of abnormal operation of the gear box of the fan.

[0006] In view of the above problems, the technical scheme of the present application is as follows:

[0007] In a first aspect, the present application provides a degradation degree detection method for lubricating oil of a fan, wherein the method comprises: obtaining lubricating oil characteristic parameters from a lubricating oil sample of a fan gear box, the lubricating oil characteristic parameters including solid particle concentration, spectral data and pressure data; determining a first degradation sensitive feature set under the dominance of pollutants and a second degradation sensitive feature set under the dominance of mechanical wear based on the lubricating oil characteristic parameters; verifying the correlation based on the first degradation sensitive feature set and the second degradation sensitive feature set in combination with historical failure data and lubricating oil replacement records of the fan gear box, correcting the deviation of the degradation degree evaluation function, and obtaining a degradation degree evaluation result; and generating a degradation degree detection report for the lubricating oil sample according to the degradation degree evaluation result.

[0008] Preferably, the solid particle concentration, spectral data in the lubricating oil characteristic parameters are coupled with the equipment operating state of the fan gearbox to analyze the first degradation sensitive feature set under the dominant of pollutants; at the same time, the internal oil flow velocity field of the fan gearbox is analyzed to determine the dynamic response coefficient as the pollution diffusion kinetics index.

[0009] Preferably, the pressure data, spectral data in the lubricating oil characteristic parameters are coupled with the equipment operating state of the fan gearbox to analyze the second degradation sensitive feature set under the dominant of mechanical wear; at the same time, the ratio of the fluctuation amplitude of the pressure data in the meshing period to the peak area of the metal organic compound in the spectral data is obtained, and the ratio after normalization is taken as the oil film degradation stability factor.

[0010] Preferably, based on the first degradation sensitive feature and the second degradation sensitive feature, a degradation feature matrix is constructed; the row vector of the degradation feature matrix corresponds to the lubricating oil sample under multiple time windows, and the column vector includes the particle size distribution entropy in the first degradation sensitive feature set, the spectral pollutant characteristic peak intensity, the pressure fluctuation coefficient in the second degradation sensitive feature set, and the spectral wear metal characteristic peak area.

[0011] Preferably, the pollution diffusion kinetics index is taken as the query vector of the attention network, the oil film degradation stability factor is taken as the key vector of the attention network, and the degradation feature matrix is taken as the value vector of the attention network; through the attention mechanism, the dynamic focusing of the key degradation feature under the weight distribution is carried out.

[0012] Preferably, the attention mechanism adopts a similarity measure based on kernel density estimation instead of inner product operation; the kernel density overlap area of the query vector and the key vector is obtained as a similarity index: the kernel density estimation is carried out on the query vector and the key vector respectively, a Gaussian kernel function is adopted, and the bandwidth is adaptively determined by Silverman rule; the kernel density overlap area of the two kernel density curves corresponding to the query vector and the key vector is calculated.

[0013] Preferably, based on the similarity index and the first similarity threshold and the second similarity threshold; if the first similarity threshold is met, the attention mechanism is inclined to the low-frequency time sequence change component of the particle size distribution entropy and the pressure fluctuation coefficient under the weight distribution; if the second similarity threshold is met, the attention mechanism is inclined to the high-frequency transient mutation component of the spectral pollutant characteristic peak intensity and the spectral wear metal characteristic peak area under the weight distribution.

[0014] Preferably, the low-frequency time sequence change component and the high-frequency transient mutation component are obtained by empirical mode decomposition of each column of the degradation feature matrix, correspond to low-order components and high-order components in a plurality of intrinsic mode functions, and combined with Hilbert transform to extract instantaneous amplitude and frequency characteristics, and then difference weighted and fused.

[0015] Preferably, the key degradation features after difference weighting are input into an improved gated recurrent unit, and noise features are suppressed by the improved gated recurrent unit to construct a degradation degree evaluation function based on the output features of the GRU.

[0016] In a second aspect of the present application, a wind turbine lubricating oil degradation degree detection system is provided, wherein the system comprises: a feature parameter acquisition module configured to acquire lubricating oil feature parameters from a wind turbine gearbox lubricating oil sample, the lubricating oil feature parameters including solid particle concentration, spectral data, and pressure data; a degradation sensitive feature set determination module configured to determine a first degradation sensitive feature set under the dominance of pollutants and a second degradation sensitive feature set under the dominance of mechanical wear based on the lubricating oil feature parameters; a bias correction module configured to correct the bias of a degradation degree evaluation function based on the first degradation sensitive feature set and the second degradation sensitive feature set, and in association with historical failure data and lubricating oil replacement records of the wind turbine gearbox to verify the relevance and correct the bias of the degradation degree evaluation function, thereby obtaining a degradation degree evaluation result; and a degradation degree detection report generation module configured to generate a degradation degree detection report of the lubricating oil sample based on the degradation degree evaluation result.

[0017] In summary, the one or more technical solutions provided in the present application achieve the following technical effects: distinguishing the degradation sensitive feature sets under the dominance of pollutants and mechanical wear, accurately identifying the degradation causes, verifying the relevance and correcting the bias of the degradation degree evaluation function in association with historical failure data and replacement records, improving the accuracy of the degradation degree evaluation result, and generating a degradation degree detection report that can provide a scientific basis for lubricating oil replacement and gearbox maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A flowchart of a wind turbine lubricating oil degradation degree detection method is provided in the present application.

[0019] Figure 2 A structural diagram of a wind turbine lubricating oil degradation degree detection system is provided in the present application.

[0020] Legend: feature parameter acquisition module M100, degradation sensitive feature set determination module M200, bias correction module M300, and degradation degree detection report generation module M400. DETAILED DESCRIPTION

[0021] In Example 1, the present application will be described in detail below with reference to the accompanying drawings, as follows:Figure 1 As shown, the application provides a method for detecting the degradation degree of fan lubricating oil, wherein the method comprises:

[0022] S1: According to the lubricating oil sample of the fan gear box, the lubricating oil characteristic parameters are obtained, including solid particle concentration, spectral data, pressure data; S2: Based on the lubricating oil characteristic parameters, the first degradation sensitive feature set under the dominance of pollutants and the second degradation sensitive feature set under the dominance of mechanical wear are determined.

[0023] Specifically, the lubricating oil sample refers to the data collected from different key parts of the fan gear box, including the oil outlet, gear meshing area, etc., for subsequent feature parameter detection; the solid particle concentration is the number or mass contained in unit volume of lubricating oil, which is a key indicator reflecting the degree of pollution. Common solid particles include dust, impurities and wear debris; spectral data are spectral characteristic information obtained by spectral analysis equipment for detecting lubricating oil samples, which can reflect the composition of pollutants in the oil and gear wear metal elements; pressure data refer to the pressure value and fluctuation of lubricating oil circulating in the gear box, which are collected by pressure sensors installed in the oil circuit and can reflect the running state of the lubricating system, such as oil circuit blockage which can cause abnormal pressure rise; obtaining lubricating oil characteristic parameters means detecting lubricating oil samples by detection equipment to obtain the specific values of the above parameters; the first degradation sensitive feature set under the dominance of pollutants is a more sensitive feature combination to the degradation of lubricating oil caused by pollutants, such as the spectral peak intensity of pollutant elements; the second degradation sensitive feature set under the dominance of mechanical wear is a more sensitive feature combination to the degradation caused by wear debris produced by mechanical wear of gear parts, such as the spectral peak area of wear metal elements and pressure fluctuation frequency; determining the first and second degradation sensitive feature sets means extracting features strongly related to the two types of degradation dominant factors through screening and analysis of characteristic parameters.

[0024] Execution steps: When obtaining lubricating oil characteristic parameters, collect multiple parallel samples from the fan gear box to reduce errors, and use a laser particle counter to detect solid particle concentration; obtain spectral data by atomic emission spectrometer, which can identify the characteristic peaks of elements such as silicon, iron and copper, among which the Si element peak intensity corresponds to external dust pollution, and the Fe and Cu peak intensities correspond to gear wear; use a pressure sensor to collect pressure data and record pressure fluctuations within the meshing period, such as normal range of 2-3 MPa and abnormal range of more than 4 MPa. Through the above steps, raw data is provided for subsequent analysis, which is the basis for distinguishing the causes of degradation.

[0025] In determining the first and second sets of degradation sensitive features, the proportion of 1-5 μm particles in the solid particle concentration and the spectral peak intensity of Si element are analyzed for correlation. If the correlation coefficient is greater than 0.85, the proportion of 1-5 μm particles and the peak intensity of Si element are selected to form the first set of degradation sensitive features, reflecting the degradation dominated by pollutants. The fluctuation amplitude of pressure data and the spectral peak area of Fe element are analyzed to extract the peak area of Fe element and the pressure fluctuation coefficient to form the second set of degradation sensitive features, reflecting the degradation dominated by mechanical wear. The pressure fluctuation coefficient is the fluctuation amplitude / average pressure. Through the above steps, the accurate classification of degradation causes is realized, and targeted features are provided for subsequent evaluation function construction.

[0026] S3: Based on the first set of degradation sensitive features and the second set of degradation sensitive features, the historical failure data and the lubricating oil replacement record of the fan gearbox are associated and verified, the deviation of the degradation degree evaluation function is corrected, and the degradation degree evaluation result is obtained; S4: According to the degradation degree evaluation result, the degradation degree detection report of the lubricating oil sample is generated.

[0027] Specifically, the association verification refers to matching and analyzing the first set of degradation sensitive features and the second set of degradation sensitive features with the historical failure data and the lubricating oil replacement record of the fan gearbox. The historical failure data includes the gearbox failure type, occurrence time, and lubricating oil state at the time of failure in the past period of time. The lubricating oil replacement record includes the replacement time, lubricating oil characteristic parameters at the time of replacement, and running time. The correlation degree of the feature set with the actual degradation result and the failure cause is verified. The degradation degree evaluation function is a mathematical model constructed based on the first set of degradation sensitive features and the second set of degradation sensitive features, which is used for quantitative calculation of the degradation degree of the lubricating oil. The deviation correction refers to adjusting the parameters of the degradation degree evaluation function according to the results of the association verification, reducing the deviation between the output value of the degradation degree evaluation function and the actual degradation state. The degradation degree evaluation result is the quantitative value or grade output by the corrected degradation degree evaluation function, reflecting the current degradation degree of the lubricating oil. The degradation degree detection report is a file generated by comprehensive evaluation results, including the degradation degree, main cause, and recommended maintenance measures.

[0028] Execution steps: Extract the first set of degradation sensitive features, the second set of degradation sensitive features, and the feature data in the lubricating oil replacement record corresponding to the gearbox bearing wear failure in the historical failure data; compare the current first set of degradation sensitive features and the second set of degradation sensitive features with the above historical failure data, and calculate the correlation degree; if there is a deviation between the degradation degree initially output by the degradation degree evaluation function and the actual degradation grade in the same state in the historical failure data, the weight of the peak area of Fe element in the degradation degree evaluation function is adjusted to correct it, so that the corrected evaluation result is closer to the actual degradation grade.

[0029] By the above steps, the error of the modified evaluation result and the actual degradation level is reduced, and the evaluation accuracy is significantly improved. The degradation degree detection report needs to include: the degradation degree evaluation result, the main degradation cause, and the associated historical fault reference. The main degradation cause can be dominated by contaminants or dominated by mechanical wear. Based on the degradation degree detection report, the complex process of lubricating oil degradation can be accurately reflected, and the lubricating oil can be replaced in time to avoid the increase of the abnormal operation risk of the fan gearbox caused by the degradation of the lubricating oil.

[0030] Further, based on the lubricating oil characteristic parameters, a first degradation sensitive feature set under the dominance of contaminants is determined. The method of the present application includes:

[0031] The first degradation sensitive feature set under the dominance of contaminants is determined by coupling analysis of the solid particle concentration and spectral data in the lubricating oil characteristic parameters, combined with the operating state of the fan gearbox. At the same time, the dynamic response coefficient is determined as the contamination diffusion kinetic index by analyzing the differential relationship of the migration and diffusion of contaminants in the internal oil flow velocity field of the fan gearbox.

[0032] Specifically, coupling analysis refers to the correlation analysis of lubricating oil characteristic parameters and equipment operating state parameters of the fan gearbox, such as speed, load, and running time. By establishing mathematical relationships between multiple parameters, the variation law of characteristic parameters under different equipment operating states is revealed. The equipment operating state includes real-time speed, output torque, and cumulative running time of the gearbox, which reflect the working condition of the fan gearbox.

[0033] The internal oil flow velocity field refers to the flow velocity distribution area of the lubricating oil in the fan gearbox under the driving of the gear transmission and the oil pump, and its characteristics can be simulated by fluid simulation. The differential relationship of the migration and diffusion of contaminants is a partial differential equation describing the migration and diffusion law of contaminants in the internal oil flow velocity field with time and space. The dynamic response coefficient is a coefficient that quantifies the response degree of the change of oil flow velocity and the fluctuation of equipment operating state to the diffusion of contaminants. The contamination diffusion kinetic index, i.e., the dynamic response coefficient, is used to characterize the diffusion rate and distribution characteristics of contaminants in the fan gearbox, and reflects the dynamic process of degradation dominated by contaminants.

[0034] Execution step: collect equipment running state parameters, and simultaneously acquire solid particle concentration and spectrum data in the corresponding period; establish a correlation model through coupling analysis, for example, when the rotating speed is increased by 10%, if the proportion of 10-20 μm particles in the solid particle concentration is increased by 15% and the Si element peak intensity is simultaneously increased by 12%, then the proportion of 10-20 μm particles and the Si element peak intensity are included in the first degradation sensitive feature set, and at the same time, in the internal oil flow velocity field analysis, the oil flow velocity distribution is obtained based on fluid simulation, the pollutant concentration change rate at different positions is calculated through the pollutant migration and diffusion differential equation, and then the dynamic response coefficient is determined, specifically, the pollutant migration and diffusion differential equation: wherein c is the pollutant concentration, t is the time, v is the oil flow velocity, and D is the diffusion coefficient; the dynamic response coefficient is taken as the pollution diffusion kinetics index. The coupling analysis strengthens the correlation between the characteristic parameters and the pollutant degradation, so that the first degradation sensitive feature set is more targeted, and the pollution diffusion kinetics index quantifies the dynamic characteristics of the pollutant diffusion, providing a time dimension reference for subsequent degradation evaluation.

[0035] Further, based on the lubricating oil characteristic parameters, a second degradation sensitive feature set under the mechanical wear dominance is determined, and the method of the application comprises:

[0036] The coupling analysis is performed on the pressure data and the spectrum data in the lubricating oil characteristic parameters in combination with the equipment running state of the fan gearbox to determine the second degradation sensitive feature set under the mechanical wear dominance; at the same time, the ratio of the fluctuation amplitude of the pressure data in the meshing period to the peak area of the metal organic compound in the spectrum data is acquired, and the ratio after normalization is taken as an oil film degradation stability factor.

[0037] Specifically, the meshing period refers to the complete time from the start of contact to the disengagement of contact of a pair of gear teeth in the fan gearbox, and the length is related to the gear rotating speed; the fluctuation amplitude of the pressure data in the meshing period refers to the difference between the maximum value and the minimum value of the lubricating oil pressure in one meshing period, reflecting the degree of pressure change when the gears mesh; the peak area of the metal organic compound refers to the integral area of the characteristic peak of the compound formed by the combination of the metal and the organic component in the lubricating oil generated by mechanical wear, which is directly related to the amount of wear products; the normalization processing is to convert the ratio of the pressure fluctuation amplitude to the peak area of the metal organic compound to the numerical range of 0-1, eliminating the dimensional differences under different equipment or detection conditions, so that the results are comparable; the oil film degradation stability factor is the above ratio after normalization, which is used to quantify the stability of the lubricating oil film under the action of mechanical wear; the coupling analysis refers to the correlation of the pressure data, the spectrum data and the equipment running state of the gearbox, and the mathematical relationship among multiple parameters is established to screen the features strongly related to mechanical wear.

[0038] The execution step is: collecting equipment running state parameters, synchronously acquiring pressure data and spectrum data in a corresponding period, such as a pressure fluctuation amplitude of 0.4 MPa in an engagement period, and a ferrous metal organic compound peak area of 3200 area units, wherein the area unit is a numerical value calculated by integrating the spectrum curve in the wavelength range corresponding to the characteristic peak, and reflects the total amount of the substance corresponding to the characteristic peak, wherein the abscissa of the spectrum curve is wavelength / wavenumber, and the ordinate is absorbance / intensity; a corresponding correlation model is established through coupling analysis, for example, when the load increases by 10%, if the pressure fluctuation amplitude increases by 15% and the ferrous metal organic compound peak area increases by 20%, then the pressure fluctuation amplitude and the ferrous metal organic compound peak area are included in the second degradation sensitive feature set dominated by mechanical wear. At the same time, the ratio of the pressure fluctuation amplitude to the metal organic compound peak area is calculated, that is, 0.4 MPa / 3200=1.25*10 -4 ; and then normalized to obtain an oil film degradation stability factor, taking the historical maximum ratio 2*10 -4 of the fan gear box as a reference, and the normalized result is 0.625. In the above steps, the coupling analysis ensures that the second degradation sensitive feature set can accurately reflect the influence of mechanical wear on the degradation of lubricating oil, and effectively supports the accurate differentiation of degradation causes.

[0039] Further, the method of the application further comprises:

[0040] Based on the first degradation sensitive feature and the second degradation sensitive feature, a degradation feature matrix is constructed; the row vector of the degradation feature matrix corresponds to the lubricating oil sample under a plurality of time windows, and the column vector contains the particle size distribution entropy in the first degradation sensitive feature set, the spectrum pollutant characteristic peak intensity, and the pressure fluctuation coefficient in the second degradation sensitive feature set and the spectrum wear metal characteristic peak area.

[0041] Specifically, the degradation feature matrix is a two-dimensional data table formed by organizing the key features in the first degradation sensitive feature set and the second degradation sensitive feature set in time sequence, and is used for system integration of degradation feature information at different time points; the time window refers to a sampling period divided according to a fixed time interval, and each time window corresponds to the collection and detection of a lubricating oil sample; the particle size distribution entropy is an index for describing the uniformity of the particle size distribution of solid particles in the lubricating oil, and the higher the entropy value, the more dispersed the particle size distribution, reflecting the complexity of the source of pollutants; the spectrum pollutant characteristic peak intensity is the peak value height of the corresponding pollutant characteristic peak in the spectrum data, which directly represents the content of the pollutant; the pressure fluctuation coefficient is the ratio of the fluctuation amplitude to the average pressure of the pressure data, which is used to quantify the relative degree of change of the pressure, and reflects the lubrication stability during gear engagement.

[0042] The main material of the gear usually contains iron, and the spectral wear metal characteristic peak area is the integral area of the wear metal characteristic peak corresponding to the iron element in the spectral data. The integral area can more comprehensively reflect the total amount of wear metal than the peak intensity. The size of the integral area is positively correlated with the content of iron filings generated by gear wear in the lubricating oil, that is, the larger the integral area, the more serious the gear wear, and the more iron-based wear debris is generated. The deterioration feature matrix is constructed according to the rule that the row vector corresponds to the time window and the column vector corresponds to the specific feature. The feature parameter values in each time window are filled into the matrix to form a structured data set.

[0043] The execution steps are as follows: determining the time window, detecting and calculating the feature parameters corresponding to the column vector for each time window of the lubricating oil sample: in the first time window, the particle size distribution entropy, the spectral pollutant characteristic peak intensity, the pressure fluctuation coefficient, and the spectral wear metal characteristic peak area corresponding to the iron element are obtained. Similarly, according to the multiple time windows, the deterioration feature matrix is formed; the scattered multi-dimensional and multi-time point feature data is systematically integrated, which provides structured input for focusing on key deterioration features through the attention mechanism in the subsequent process, adopts matrix processing, improves the utilization rate of feature data, and can intuitively reflect the change trend of each feature with time, which provides a data basis for accurately identifying the dynamic change of the deterioration inducement.

[0044] Further, the method of the application further comprises:

[0045] The pollution diffusion dynamics index is used as the query vector of the attention network, the oil film degradation stability factor is used as the key vector of the attention network, and the deterioration feature matrix is used as the value vector of the attention network; the dynamic focusing of the key deterioration feature under the weight distribution is performed through the attention mechanism.

[0046] Specifically, the attention network is an algorithm model simulating human attention mechanism, which can automatically focus on key content from a large amount of information; the query vector is a vector for querying information, taking the pollution diffusion dynamics index as the query basis, and is used to locate key features related to the degradation of the pollutant; the key vector is a vector for matching the query vector, taking the oil film degradation stability factor as the matching basis, and is used to screen key features related to mechanical wear degradation; the value vector is the original data vector to be processed, that is, the degradation feature matrix, which contains specific numerical values of various degradation features; the attention mechanism assigns weights to different features in the value vector by calculating the similarity between the query vector and the key vector, and the higher the weight, the more critical the feature; the weight distribution refers to the importance proportion allocated to each column feature in the degradation feature matrix; the dynamic focusing of the key degradation feature refers to real-time adjustment of the weight of each feature as the pollution diffusion dynamics index, the oil film degradation stability factor and the degradation feature matrix change, so that the model pays more attention to the feature that has the most significant impact on degradation at present.

[0047] The execution steps are as follows: determining the input vector of the attention mechanism: taking the pollution diffusion dynamics index with a time stamp as the query vector, the oil film degradation stability factor under the same time stamp as the key vector, and the degradation feature matrix as the value vector; calculating the similarity between the query vector and the key vector, and assigning weights to each feature in the value vector based on the similarity; multiplying the weight and the numerical value of the corresponding feature in the value vector to obtain the weighted feature combination, and realizing the dynamic focusing on the current key degradation feature; in the above steps, the attention mechanism automatically distinguishes the importance of the features, and solves the problem that the key information is submerged when all features are treated equally, improves the recognition response efficiency of the dominant degradation of the pollutant and the recognition accuracy of the dominant degradation of mechanical wear, and provides more targeted feature input for the construction of the subsequent degradation degree evaluation function.

[0048] Further, the method of the application comprises:

[0049] The attention mechanism uses a similarity measure based on kernel density estimation to replace the inner product operation; the kernel density overlap area of the query vector and the key vector is obtained as a similarity index: kernel density estimation is performed on the query vector and the key vector respectively, a Gaussian kernel function is used, and the bandwidth is adaptively determined by the Silverman rule; the kernel density overlap area of the two kernel density curves corresponding to the query vector and the key vector is calculated.

[0050] Specifically, the kernel density estimation is used to estimate the probability density function of the random variable, and a smooth density curve is fitted by assigning a kernel function weight around the data points to describe the distribution characteristics of the query vector and the key vector; the Gaussian kernel function is a commonly used kernel function in kernel density estimation: wherein K( ) represents the weight value of the kernel function, represents the standardized distance, , The smaller the absolute value of the target point, the closer it is to the target point, and the greater the weight given by the kernel function. The standardization eliminates the influence of the original dimension of the data, effectively capturing the local distribution characteristics of the data; the Silverman rule is a method for adaptively determining the bandwidth in kernel density estimation, and the formula is h = 0.9 min( ,IQR / 1.34) wherein h is the bandwidth, is the standard deviation, IQR is the interquartile range, and n is the sample size.

[0051] The bandwidth is automatically adjusted according to the characteristics of the data to ensure the accuracy of the density estimation, and the bandwidth is used to control the smoothness of the kernel function; the kernel density overlap area refers to the area of the overlapping region between the kernel density curves of the query vector and the key vector, and the larger the area, the more similar the distribution of the two vectors, and vice versa. As an indicator of similarity; the inner product operation is a method for calculating the similarity between query and key in the attention mechanism, and the inner product operation refers to calculating the sum of the products of corresponding elements of two vectors, and the similarity measure based on kernel density estimation replaces the inner product operation. The kernel density overlap area is used to represent the similarity, which can better reflect the similarity of the overall distribution of the vector rather than simply numerical matching.

[0052] ​​The execution step is: kernel density estimation is performed on the query vector and the key vector, if the pollution diffusion dynamics index sequence of the random intercept of the query vector in a period of time is [0.62, 0.65, 0.68, 0.70], and the key vector is the oil film degradation stability factor sequence of the corresponding period, [0.50, 0.53, 0.55, 0.58]; a Gaussian kernel function is used, the bandwidth is calculated by the Silverman rule, and two smooth kernel density curves are fitted respectively; the kernel density overlap area of the two curves is calculated by integration, and the kernel density overlap area of the two kernel density curves corresponding to the query vector and the key vector is calculated. In the above step, preferably, when the query and the key vector are nonlinearly correlated, the similarity of the query and the key is more accurately measured by the kernel density overlap area. Compared with the inner product operation which only focuses on the sum of the numerical product, it can capture the overall distribution characteristics of the vector, reduce the interference of abnormal values on the similarity judgment, and provide a more reliable similarity basis for the weight distribution of the subsequent attention mechanism.

[0053] Further, the method of the application comprises:

[0054] Based on the similarity index and the first similarity threshold value and the second similarity threshold value, if the first similarity threshold value is met, the attention mechanism is inclined to the weight distribution of the low-frequency time sequence change component of the particle size distribution entropy and the pressure fluctuation coefficient, and if the second similarity threshold value is met, the attention mechanism is inclined to the weight distribution of the high-frequency transient mutation component of the spectral pollutant characteristic peak intensity and the spectral wear metal characteristic peak area.

[0055] Specifically, the similarity index is the similarity quantization value of the query vector and the key vector obtained by the kernel density overlap area; the first similarity threshold value and the second similarity threshold value are preset threshold values for judging the degree of similarity, used to divide different similarity intervals; the low-frequency time sequence change component refers to the part of the degradation feature that changes slowly and periodically over time, reflecting the long-term trend of degradation, and is associated with the particle size distribution entropy and the pressure fluctuation coefficient; the high-frequency transient mutation component refers to the part of the degradation feature that changes suddenly and significantly in a short time, reflecting the sudden abnormality of degradation, and is associated with the spectral pollutant characteristic peak intensity and the spectral wear metal characteristic peak area; the weight distribution is inclined to the low-frequency time sequence change component of the particle size distribution entropy and the pressure fluctuation coefficient, and the weight distribution is inclined to the high-frequency transient mutation component of the spectral pollutant characteristic peak intensity and the spectral wear metal characteristic peak area, which means that the attention mechanism increases the weight proportion of the corresponding feature component, so that the model pays more attention to the corresponding part of the information when calculating, and ensures that the weight distribution reflects the similarity and the feature difference degree at the same time, thereby avoiding the dominant influence of the vector angle on the weight distribution.

[0056] The execution steps are: presetting a first similarity threshold value and a second similarity threshold value, and comparing the calculated similarity index with the two threshold values; if the similarity index is 0.7 and the first similarity threshold value is 0.6, the first similarity threshold value is met, indicating that the distribution trend of the pollution diffusion dynamics index and the oil film degradation stability factor is highly similar, at this time, the attention mechanism increases the low-frequency component weight of the particle size distribution entropy and correspondingly reduces the high-frequency component weight, and focuses on the long-term degradation trend; if the similarity index is 0.2 and the second similarity threshold value is 0.3, it is indicated that the distribution trends of the two are significantly different, the pollution diffusion index suddenly rises and the oil film stability factor suddenly drops, and then the high-frequency component weight of the spectral pollutant characteristic peak intensity is increased, the high-frequency component weight of the spectral wear metal characteristic peak area is increased, and the sudden degradation anomaly is focused on capturing; in the above steps, the threshold value is divided to realize the targeted tilt of the weight, so that the model can focus on the key components according to the dynamic relationship of the degradation characteristics, when the first similarity threshold value is met, focusing on the low-frequency component can improve the accuracy of long-term degradation trend prediction; when the second similarity threshold value is met, focusing on the high-frequency component can shorten the identification response time of the sudden degradation event, and provide more accurate feature input for timely warning of potential gearbox faults.

[0057] Further, the method of the application comprises:

[0058] The low-frequency time sequence change component and the high-frequency transient mutation component are obtained by performing signal decomposition on each column of the degradation characteristic matrix through empirical mode decomposition, correspond to low-order components and high-order components in a plurality of intrinsic mode functions respectively, and instantaneous amplitude and frequency characteristics are extracted by combining Hilbert transform for differential weighting fusion.

[0059] Specifically, empirical mode decomposition is an adaptive signal decomposition method that does not need to preset a base function, can decompose a nonlinear and non-stationary time sequence signal into a plurality of intrinsic mode functions with physical meaning, and is used to decompose each column in the degradation characteristic matrix to extract different frequency components; the intrinsic mode function is a component obtained by EMD decomposition, which needs to meet the condition that the number of extreme points and zero-crossing points is equal or differs by at most 1, and the mean value of the upper and lower envelope lines is 0, wherein the low-order IMF corresponds to the low-frequency component changing slowly in the signal, and the high-order IMF corresponds to the high-frequency component changing sharply; the Hilbert transform is a mathematical transform of each IMF, which extracts the instantaneous amplitude and instantaneous frequency by calculating the analytic signal of the signal, so as to quantify the dynamic change characteristics of the feature component; the differential weighting fusion gives different weights according to the importance of the low-frequency time sequence change component and the high-frequency transient mutation component in different degradation stages, and then merges them, for example, increasing the low-frequency component weight in the stable degradation period and increasing the high-frequency component weight in the sudden degradation period, so as to highlight the key information.

[0060] The execution step is: performing empirical mode decomposition on each column feature of the deterioration feature matrix: filtering out components satisfying the IMF condition through the EMD algorithm, if three IMFs are obtained, wherein, the low-order IMF1 has a frequency range of 0.02Hz-0.1Hz, corresponding to a low-frequency time-varying component, reflecting the slow accumulation of the feature over time; the high-order IMF2 has a frequency range of 0.5Hz-2Hz, corresponding to a high-frequency transient mutation component, reflecting sudden changes in a short time; performing Hilbert transform on each IMF to extract the instantaneous amplitude and instantaneous frequency; and differentially weighting and fusing according to the deterioration stage: taking the similarity index meeting the first similarity threshold value as a stable operation period, if the current is in the stable operation period, the weight corresponding to the low-frequency time-varying component is set to exceed 0.5, and the weight corresponding to the high-frequency transient mutation component is set to 0.3; taking the similarity index meeting the second similarity threshold value as an abnormal fluctuation period, if in the abnormal fluctuation period, the weight corresponding to the high-frequency transient mutation component exceeds 0.5, and the weight corresponding to the low-frequency time-varying component is set to 0.4, and a more comprehensive feature vector is formed after fusion.

[0061] In the above steps, the time series information of the original feature is converted into quantifiable dynamic features through signal decomposition and transformation, solving the problem that it is difficult to distinguish between fast and slow changes when traditional features are directly used. Preferably, after EMD decomposition, the low-frequency component is associated with the long-term deterioration trend, and the high-frequency component is associated with the sudden failure; in combination with the instantaneous features of Hilbert transform, when the features after differential weighting and fusion are input into the subsequent model, the time sensitivity of the deterioration degree evaluation is improved, and more fine features are provided for accurate evaluation of the deterioration state of lubricating oil.

[0062] Further, the method of the application comprises:

[0063] The key deterioration features after differential weighting are input into the improved gate recurrent unit; noise features are suppressed through the improved gate recurrent unit, and a deterioration degree evaluation function is constructed based on the output features of the GRU.

[0064] Specifically, the improved gated recurrent unit is an optimization of the conventional gated recurrent unit, which retains the core capability of GRU in processing time series data, and the improvement is reflected in the enhanced filtering capability of the gating mechanism to noise or the memory effect of long time series features, which is suitable for processing time series data of lubricating oil degradation characteristics. The core capability of GRU in processing time series data is to capture time series dependencies through update gate and reset gate. The differentially weighted key degradation features refer to the feature vectors obtained after empirical mode decomposition, Hilbert transform and weight fusion, which contain the weighted results of low-frequency time series change components and high-frequency transient mutation components, and reflect the key information of different degradation stages. Noise features refer to interference information in feature data that is irrelevant to lubricating oil degradation, such as random errors of detection equipment and accidental data fluctuations caused by environmental temperature fluctuations. Suppressing noise features is the process of improving GRU filtering interference data through gating mechanism to highlight effective degradation features. Specifically, the update gate controls the retention proportion of historical information, and the reset gate controls the filtering degree of current information. The construction of degradation degree evaluation function based on GRU output features refers to the mapping of denoised features output by GRU to quantitative degradation degree values, which is usually realized by regression model or neural network to map features to degradation degree.

[0065] The execution steps are as follows: the differentially weighted key degradation features are input into the improved gated recurrent unit, which contains the weighted results of low-frequency time series change components and high-frequency transient mutation components; the improved GRU processes time series features through optimized update gate and reset gate: the update gate dynamically adjusts the retention weight of historical features, and the reset gate strengthens the memory of effective historical features, thereby suppressing noise; after GRU processing, the denoised feature sequence is output, which smooths the jumps caused by noise; the degradation degree evaluation function is constructed based on the output features of GRU; the relationship between the features and the historical labeled degradation degree is fitted through neural network, and the bias of the degradation degree evaluation function is corrected, so that the output value of the degradation degree evaluation function is controlled within a preset threshold, and the accuracy of the degradation degree evaluation result is improved. In the above steps, the improved GRU processes time series features and suppresses noise, providing high-quality input features for the degradation degree evaluation function, solving the problem of evaluation bias caused by noise interference in the original features. Preferably, the fitting goodness of the evaluation function is improved, which lays a precise feature foundation for subsequent correlation verification and bias correction based on historical data, ensuring that the final degradation degree evaluation result can truly reflect the degradation state of the lubricating oil.

[0066] In summary, the beneficial effects of the embodiments of the present application are:

[0067] The degradation degree detection method and system of the fan lubricating oil are provided, and the degradation degree detection method comprises the following steps: obtaining lubricating oil characteristic parameters from a lubricating oil sample of a fan gear box, the lubricating oil characteristic parameters comprising solid particle concentration, spectral data, and pressure data; determining a first degradation sensitive feature set under the dominance of pollutants and a second degradation sensitive feature set under the dominance of mechanical wear based on the lubricating oil characteristic parameters; performing relevance verification on the first degradation sensitive feature set and the second degradation sensitive feature set in combination with historical fault data and lubricating oil replacement records of the fan gear box, correcting the deviation of a degradation degree evaluation function, and obtaining a degradation degree evaluation result; and generating a degradation degree detection report of the lubricating oil sample according to the degradation degree evaluation result. The degradation degree detection method and system of the fan lubricating oil are provided, and the degradation sensitive feature sets dominated by pollutants and mechanical wear are distinguished, the degradation causes are accurately identified, the degradation degree evaluation function is verified in relevance and corrected in deviation in combination with the historical fault data and the replacement records, the accuracy of the degradation degree evaluation result is improved, and the generated degradation degree detection report can provide a scientific basis for lubricating oil replacement and gear box maintenance.

[0068] In the second embodiment, based on the same inventive concept as the degradation degree detection method of the fan lubricating oil in the foregoing embodiments, as shown in the following table, the degradation degree detection system of the fan lubricating oil is provided. Figure 2 As shown in the following table, the degradation degree detection system of the fan lubricating oil is provided.

[0069] The characteristic parameter acquisition module M100 acquires lubricating oil characteristic parameters from a lubricating oil sample of a fan gear box, and the lubricating oil characteristic parameters comprise solid particle concentration, spectral data, and pressure data.

[0070] The degradation sensitive feature set determination module M200 determines a first degradation sensitive feature set under the dominance of pollutants and a second degradation sensitive feature set under the dominance of mechanical wear based on the lubricating oil characteristic parameters.

[0071] The deviation correction module M300 performs relevance verification on the first degradation sensitive feature set and the second degradation sensitive feature set in combination with historical fault data and lubricating oil replacement records of the fan gear box, corrects the deviation of a degradation degree evaluation function, and obtains a degradation degree evaluation result.

[0072] The degradation degree detection report generation module M400 generates a degradation degree detection report of the lubricating oil sample according to the degradation degree evaluation result.

[0073] Further, the degradation sensitive feature set determination module M200 is used to perform the following method.

[0074] The solid particle concentration, the spectrum data in the lubricating oil characteristic parameters are coupled and analyzed in combination with the equipment operation state of the fan gear box to determine a first degradation sensitive feature set under the pollution dominant; meanwhile, the differential relationship of the pollution migration and diffusion is analyzed in the internal oil flow velocity field of the fan gear box to determine a dynamic response coefficient as a pollution diffusion dynamics index.

[0075] Further, the degradation sensitive feature set determination module M200 is configured to perform the following method:

[0076] The pressure data, the spectrum data in the lubricating oil characteristic parameters are coupled and analyzed in combination with the equipment operation state of the fan gear box to determine a second degradation sensitive feature set under the mechanical wear dominant; meanwhile, a ratio of the fluctuation amplitude of the pressure data in the meshing period to the peak area of the metal organic compound in the spectrum data is obtained, and the ratio after normalization is taken as an oil film degradation stability factor.

[0077] Further, the degradation sensitive feature set determination module M200 is further configured to perform the following method:

[0078] Based on the first degradation sensitive feature and the second degradation sensitive feature, a degradation feature matrix is constructed; the row vectors of the degradation feature matrix correspond to lubricating oil samples in multiple time windows, and the column vectors include the particle size distribution entropy in the first degradation sensitive feature set, the spectrum pollution characteristic peak intensity, the pressure fluctuation coefficient in the second degradation sensitive feature set, and the spectrum wear metal characteristic peak area.

[0079] Further, the degradation sensitive feature set determination module M200 is further configured to perform the following method:

[0080] The pollution diffusion dynamics index is taken as a query vector of the attention network, the oil film degradation stability factor is taken as a key vector of the attention network, and the degradation feature matrix is taken as a value vector of the attention network; the dynamic focusing of the key degradation feature under the weight distribution is performed through the attention mechanism.

[0081] Further, the degradation sensitive feature set determination module M200 is further configured to perform the following method:

[0082] The attention mechanism adopts a similarity measurement based on kernel density estimation to replace the inner product operation; the kernel density overlap area of the query vector and the key vector is obtained as a similarity index: the kernel density estimation is performed on the query vector and the key vector respectively, a Gaussian kernel function is adopted, and the bandwidth is adaptively determined through the Silverman rule; the kernel density overlap area of the two kernel density curves corresponding to the query vector and the key vector is calculated.

[0083] Further, the deterioration sensitive feature set determination module M200 is further configured to perform the following method:

[0084] Based on the similarity index and the first similarity threshold value, the second similarity threshold value, if the first similarity threshold value is met, the attention mechanism is inclined to weight distribution to the low-frequency time series change component of the particle size distribution entropy and the pressure fluctuation coefficient, and if the second similarity threshold value is met, the attention mechanism is inclined to weight distribution to the high-frequency transient mutation component of the spectral pollutant characteristic peak intensity and the spectral wear metal characteristic peak area.

[0085] Further, the deterioration sensitive feature set determination module M200 is further configured to perform the following method:

[0086] The low-frequency time series change component and the high-frequency transient mutation component are obtained by signal decomposition of each column of the deterioration feature matrix through empirical mode decomposition, corresponding to low-order components and high-order components in a plurality of intrinsic mode functions, and combined with Hilbert transform to extract instantaneous amplitude and frequency characteristics, and perform differential weighted fusion.

[0087] Further, the bias correction module M300 is further configured to perform the following method:

[0088] The key deterioration features after differential weighting are input into an improved gated recurrent unit, and noise features are suppressed through the improved gated recurrent unit to construct a deterioration degree evaluation function based on the output features of the GRU.

[0089] As described above, any step can be stored in a computer memory without limitation as computer instructions or programs, and can be called and recognized by a computer processor without limitation, and no additional limitation is made here.

[0090] Further, the above technical solution only embodies the preferred technical solution of the technical solution of the embodiments of the present application, and some variations of certain parts made by the person skilled in the art also embody the principles of the novel embodiments of the present application. Obviously, the person skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application.

Claims

1. A method of detecting a degree of deterioration of a lubricating oil for a fan, characterized by, The method comprises: According to the lubricating oil sample of the fan gearbox, the lubricating oil characteristic parameters are obtained, and the lubricating oil characteristic parameters include solid particle concentration, spectral data and pressure data; Based on the lubricating oil characteristic parameters, a first degradation sensitive feature set under the dominance of pollutants and a second degradation sensitive feature set under the dominance of mechanical wear are determined; Based on the first degradation sensitive feature set and the second degradation sensitive feature set, the relevance verification is performed in combination with historical fault data and lubricating oil replacement records of the fan gearbox, the deviation of the degradation degree evaluation function is corrected, and a degradation degree evaluation result is obtained; According to the degradation degree evaluation result, a degradation degree detection report of the lubricating oil sample is generated; Wherein, based on the lubricating oil characteristic parameters, the first degradation sensitive feature set under the dominance of pollutants comprises: Through the solid particle concentration and spectral data in the lubricating oil characteristic parameters, coupling analysis is performed in combination with the equipment operating state of the fan gearbox, and the first degradation sensitive feature set under the dominance of pollutants is determined; At the same time, the differential relationship of the migration and diffusion of pollutants in the internal oil flow velocity field of the fan gearbox is analyzed, and the dynamic response coefficient is determined as the pollution diffusion dynamics index; Wherein, based on the lubricating oil characteristic parameters, the second degradation sensitive feature set under the dominance of mechanical wear comprises: Through the pressure data and spectral data in the lubricating oil characteristic parameters, coupling analysis is performed in combination with the equipment operating state of the fan gearbox, and the second degradation sensitive feature set under the dominance of mechanical wear is determined; At the same time, the ratio of the fluctuation amplitude of the pressure data in the meshing period to the peak area of the metal organic compound in the spectral data is obtained, and the ratio is normalized as an oil film degradation stability factor.

2. A method of detecting the deterioration degree of a fan lubricating oil according to claim 1, characterized by, The method further comprises: Based on the first degradation sensitive feature and the second degradation sensitive feature, a degradation feature matrix is constructed; The row vector of the degradation feature matrix corresponds to the lubricating oil sample in multiple time windows, and the column vector contains the particle size distribution entropy in the first degradation sensitive feature set, the spectral pollutant characteristic peak intensity, the pressure fluctuation coefficient in the second degradation sensitive feature set, and the spectral wear metal characteristic peak area.

3. A method of detecting the deterioration degree of a fan lubricating oil according to claim 2, characterized by, The method further comprises: The pollution diffusion dynamics index is taken as the query vector of the attention network, the oil film degradation stability factor is taken as the key vector of the attention network, and the degradation feature matrix is taken as the value vector of the attention network; Through the attention mechanism, the dynamic focusing of the key degradation feature under the weight distribution is performed.

4. A method of detecting the deterioration degree of a fan lubricating oil according to claim 3, characterized by, The method comprises: The attention mechanism adopts a similarity measure based on kernel density estimation instead of inner product operation; The kernel density overlap area of the query vector and the key vector is obtained as a similarity index: kernel density estimation is performed on the query vector and the key vector respectively, a Gaussian kernel function is adopted, and the bandwidth is adaptively determined by Silverman rule; the kernel density overlap area of the two kernel density curves corresponding to the query vector and the key vector is calculated.

5. A method of detecting the deterioration degree of a fan lubricating oil according to claim 4, characterized by, The method comprises: Based on the similarity index, a first similarity threshold value and a second similarity threshold value; If the first similarity threshold is met, the attention mechanism is inclined to the low-frequency time-varying component of the particle size distribution entropy and the pressure fluctuation coefficient; If the second similarity threshold is met, the attention mechanism is inclined to the high-frequency transient mutation component of the spectral pollutant characteristic peak intensity and the spectral wear metal characteristic peak area.

6. A method of detecting the deterioration degree of a fan lubricating oil according to claim 5, characterized by, The method comprises: The low-frequency time-varying component and the high-frequency transient mutation component are obtained by signal decomposition of each column of the degradation feature matrix through empirical mode decomposition, corresponding to low-order components and high-order components in a plurality of intrinsic mode functions, and combining Hilbert transform to extract instantaneous amplitude and frequency characteristics for differential weighting and fusion.

7. A method of detecting the degree of deterioration of a fan lubricating oil according to claim 6, characterized by, The method comprises: The key degradation features after differential weighting are input into an improved gated recurrent unit; Noise features are suppressed through the improved gated recurrent unit to construct a degradation degree evaluation function based on the output features of the GRU.

8. A system for detecting a degree of deterioration of a lubricating oil of a fan, characterized by comprising: a lubricating oil deterioration degree detection device according to any one of claims 1 to 7. Steps for implementing the method for detecting the degradation degree of a fan lubricating oil according to any one of claims 1-7, the system comprising: A feature parameter acquisition module: acquires lubricating oil feature parameters from a fan gearbox lubricating oil sample, the lubricating oil feature parameters including solid particle concentration, spectral data, and pressure data; A degradation sensitive feature set determination module: determines a first degradation sensitive feature set under pollutant dominance and a second degradation sensitive feature set under mechanical wear dominance based on the lubricating oil feature parameters; A bias correction module: based on the first degradation sensitive feature set and the second degradation sensitive feature set, and in combination with historical failure data and lubricating oil replacement records of the fan gearbox, verifies the relevance, corrects the bias of the degradation degree evaluation function, and obtains a degradation degree evaluation result; A degradation degree detection report generation module: generates a degradation degree detection report of the lubricating oil sample according to the degradation degree evaluation result; Wherein, based on the lubricating oil feature parameters, the first degradation sensitive feature set under pollutant dominance is determined, comprising: Through the solid particle concentration and spectral data in the lubricating oil feature parameters, and in combination with the equipment operating state of the fan gearbox, coupling analysis is performed to determine the first degradation sensitive feature set under pollutant dominance; At the same time, the internal oil flow velocity field of the fan gearbox is analyzed to determine the dynamic response coefficient as the pollutant diffusion dynamics index. Wherein, based on the lubricating oil feature parameters, the second degradation sensitive feature set under mechanical wear dominance is determined, comprising: Through the pressure data and spectral data in the lubricating oil feature parameters, and in combination with the equipment operating state of the fan gearbox, coupling analysis is performed to determine the second degradation sensitive feature set under mechanical wear dominance; At the same time, the ratio of the fluctuation amplitude of the pressure data in the meshing period to the peak area of the metal organic compound in the spectral data is obtained, and the ratio after normalization is taken as the oil film degradation stability factor.

Images