Method and system for detecting oil state of gearbox of wind power generation equipment

Abstract

The invention discloses a wind power generation equipment gearbox oil state detection method and system, and the method comprises the steps: obtaining an oil sample, carrying out the quantitative analysis of solid, liquid and gas pollutants through an oil pollution degree analysis model, and processing the parameters such as kinematic viscosity through an oil physicochemical property analysis model; and after the tensors of the two results are fused, feature decoupling and deep feature mining are performed, and finally, the oil state is judged by a state evaluation decision model. The system is correspondingly provided with six units such as a sample acquisition unit and a pollution analysis unit to realize the functions. According to the scheme, the defects that traditional detection is single and correlation analysis is lacked are overcome, a certain index is not independently detected any more, pollution and physicochemical property information is comprehensively integrated, the relation between the pollution and the physicochemical property information is deeply analyzed through tensor fusion, feature decoupling and other technical means, accurate evaluation of the oil state of the gearbox is achieved, and the detection accuracy is improved. And misjudgment caused by one-sided detection or no consideration of mutual relations is effectively avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of gear box detection, in particular to a wind power generation equipment gear box oil state detection method and system. BACKGROUND

[0002] In the field of wind power generation, the gear box as a core transmission component, its stable operation is directly related to the efficiency and reliability of the entire wind power generation system. The accurate detection of the gear box oil state is the key link to ensure the normal operation of the gear box. With the development of wind power generation equipment towards large-scale and intelligent direction, the traditional oil state detection method has been difficult to meet the actual demand.

[0003] The prior art has many deficiencies in oil state detection. On the one hand, the detection means is relatively single, mostly only for detecting one or several indicators of oil, such as only focusing on the content of solid particle contaminants in the oil, or only analyzing the kinematic viscosity of the oil, which cannot comprehensively reflect the overall state of the oil. This one-sided detection method is easy to misjudge the oil state of the gear box and miss the best maintenance and processing opportunity. On the other hand, the existing detection method lacks in-depth analysis of the mutual relationship between oil pollution and physicochemical properties. In fact, the contaminants in the oil will affect its physicochemical properties, and the change of physicochemical properties will also exacerbate the influence of contaminants, both of which interact to affect the operation of the gear box. But the prior art fails to combine pollution degree analysis and physicochemical property analysis, making it difficult to accurately assess the actual state of the gear box oil, and unable to provide strong protection for the efficient operation of the wind power generation equipment. SUMMARY

[0004] In order to overcome the shortcomings and deficiencies of the prior art, the present application provides a wind power generation equipment gear box oil state detection method and system.

[0005] The technical solution adopted by the present application is a wind power generation equipment gear box oil state detection method, which comprises:

[0006] Step S1: obtaining a gear box oil sample, based on an oil pollution degree analysis model, quantitatively analyzing solid particle contaminants, liquid contaminants and gas contaminants in the oil sample, and through constructing a multi-dimensional pollution factor matrix, extracting and encoding the particle size distribution and concentration information of different types of contaminants;

[0007] Step S2: using an oil physicochemical property analysis model, the kinematic viscosity, acid value, water content and oxidation stability parameters of the gear box oil are established, and a physicochemical property parameter space is established, and through nonlinear mapping, the measured parameter values are converted into corresponding feature vectors;

[0008] Step S3: Tensor fusion is performed on the multi-dimensional pollution factor matrix obtained in step S1 and the physicochemical property eigenvector obtained in step S2 to form a composite feature tensor containing pollution information and physicochemical property information;

[0009] Step S4: Based on the composite feature tensor, a preset feature decoupling algorithm is used to separate the pollution-related features and the physicochemical property-related features in the composite feature tensor to obtain a pollution feature sub-tensor and a physicochemical property feature sub-tensor;

[0010] Step S5: Deep feature mining is performed on the pollution feature sub-tensor and the physicochemical property feature sub-tensor, respectively, to extract high-order pollution features and high-order physicochemical property features by constructing a multi-level feature extraction network;

[0011] Step S6: The extracted high-order pollution features and high-order physicochemical property features are input into a state evaluation decision model, and a preset decision rule is used to comprehensively determine the gear box oil state, and an output gear box oil state detection result is output.

[0012] Further, the multi-dimensional pollution factor matrix of the oil pollution degree analysis model construction in step S1 is calculated by the following formula:

[0013]

[0014] wherein, M cF represents a multi-dimensional pollution factor matrix; n is the number of pollutant types; a i is the weight coefficient of the i-th type of pollutant, which is determined according to the anti-pollution performance parameters of the gear box oil; D i is the particle size distribution vector of the i-th type of pollutant, which contains the proportion information of pollutants in different particle size intervals; C i is the concentration scalar of the i-th type of pollutant; f i is the feature encoding function of the i-th type of pollutant, and the encoding rule is adjusted according to the lubrication characteristics of the gear box oil.

[0015] Further, the physicochemical property parameter space conversion formula of the oil physicochemical property analysis model established in step S2 is as follows:

[0016] V PP = β·ReLU(γ·[V, A, W, O] T )

[0017] wherein, V PPrepresents the converted physicochemical property feature vector; β is a scale scaling matrix, the element value of which is set according to the specification parameters of the gearbox oil; γ is a weight matrix, which is determined according to the performance standard parameters of the gearbox oil; V is the measured value of the kinematic viscosity of the gearbox oil; A is the measured value of the acid value; W is the measured value of the moisture content; O is the measured value of the oxidation stability index; ReLU is a linear rectification activation function, which is used to enhance the nonlinear expression ability of the feature vector.

[0018] Further, in the step S3, the tensor fusion process adopts a combination of tensor product and weighted summation, and the specific fusion formula is:

[0019]

[0020] wherein, T CFP represents the composite feature tensor; m is the dimension of the physicochemical property feature vector; ω j is the weight coefficient of the jth dimension physicochemical property feature vector in the fusion process, which is determined by the use condition parameters of the gearbox oil; V PPj is the jth dimension component of the physicochemical property feature vector V PP ; and represents the tensor product operation.

[0021] Further, in the step S4, the feature decoupling algorithm is based on the principle of tensor decomposition, and the composite feature tensor is decomposed by the following formula:

[0022]

[0023] wherein, T CF represents the pollution feature sub-tensor; T PP represents the physicochemical property feature sub-tensor; ° represents the Hadamard product operation of the tensor; E is the decomposition error tensor, and the norm of the error tensor is minimized through an iterative optimization algorithm, and the parameter adjustment in the optimization process is based on the stability parameters of the gearbox oil.

[0024] Further, in the step S5, the multi-level feature extraction network comprises a plurality of feature extraction layers, and the feature extraction formula of each layer is:

[0025] F l = σ(θ l · F l-1 + b l )

[0026] wherein, F l is the feature tensor extracted by the lth layer; F l-1 is the output feature tensor of the (l-1)th layer; θ l is the weight tensor of the lth layer, and the parameter value is initialized according to the wear characteristic parameters of the gearbox oil; b lis the bias tensor of the l-th layer; σ is an activation function, and a suitable type of activation function is selected according to the performance change trend of the gearbox oil.

[0027] Further, the step S6, the state evaluation decision model adopts a combination of decision tree and rule base, and the decision process is carried out through the following rules:

[0028] If H CF >τ CF and H PP >τ PP , it is determined that the state of the gearbox oil is seriously abnormal;

[0029] If H CF >τ CF and H PP ≤τ PP , it is determined that the state of the gearbox oil is contaminated abnormal;

[0030] If H CF ≤τ CF and H PP >τ PP , it is determined that the state of the gearbox oil is abnormal in physicochemical properties;

[0031] If H CF ≤τ CF and H PP ≤τ PP , it is determined that the state of the gearbox oil is normal;

[0032] Wherein, H CF is the comprehensive measurement value of the extracted high-order pollution characteristics, which is calculated according to the cleanliness standard parameters of the gearbox oil; H PP is the comprehensive measurement value of the high-order physicochemical property characteristics, which is determined according to the quality standard parameters of the gearbox oil; τ CF and τ PP are the determination thresholds of the pollution characteristics and the physicochemical property characteristics, respectively, which are set by statistical analysis of a large number of historical detection data of gearbox oil samples and in combination with the performance limit parameters of the gearbox oil.

[0033] Further, before the step S1, a step of determining the collection position of the gearbox oil sample is further included, and the collection position is determined by the following method:

[0034] According to the internal flow field distribution characteristics of the gearbox and in combination with the flowability parameters of the gearbox oil, an oil sample collection position optimization model is established, which determines the best collection position by calculating the representativeness index R I of the oil sample at different positions, and the calculation formula is as follows:

[0035]

[0036] wherein p is the number of candidate collection positions; δ k is the weight coefficient of the kth candidate collection position, which is determined by the structural parameters of the gearbox; Corr(S k , S ref ) is the correlation measure value of the oil sample at the kth candidate collection position and the reference sample, which is constructed according to the standard performance parameters of the gearbox oil; by comparing the representative indexes R I of each candidate collection position, the position with the largest representative index is selected as the oil sample collection position.

[0037] Further, after step S6, a confidence evaluation step of the detection result is further included, which is calculated by the following formula:

[0038]

[0039] wherein C E represents the confidence of the detection result; r is the number of reference indicators for evaluating the confidence; λ q is the weight coefficient of the qth reference indicator, which is set according to the detection accuracy requirement parameters of the gearbox oil; Conf(R q ) is the confidence value corresponding to the qth reference indicator, the reference indicator includes the repeatability of the detection data and the consistency with the historical detection data, and the confidence value is calculated according to the stability parameters of the gearbox oil and the performance parameters of the detection equipment.

[0040] The wind power equipment gearbox oil state detection system comprises:

[0041] An oil sample acquisition and pollution quantification analysis unit is configured to acquire a gearbox oil sample, and quantitatively analyze solid particle pollutants, liquid pollutants and gas pollutants in the oil sample based on an oil pollution quantification analysis model to construct a multi-dimensional pollution factor matrix.

[0042] A physicochemical property parameter analysis and feature vector conversion unit is configured to analyze the kinematic viscosity, acid value, water content and oxidation stability parameters of the gearbox oil, establish a physicochemical property parameter space based on an oil physicochemical property analysis model, and convert the measured parameter values into corresponding feature vectors.

[0043] A pollution-physicochemical feature tensor fusion unit is configured to fuse the multi-dimensional pollution factor matrix and the physicochemical property feature vectors to form a composite feature tensor containing pollution information and physicochemical property information.

[0044] A feature decoupling and sub-tensor separation unit is configured to separate the pollution-related features and the physicochemical property-related features in the composite feature tensor based on a preset feature decoupling algorithm to obtain a pollution feature sub-tensor and a physicochemical property feature sub-tensor.

[0045] The multi-level feature extraction and high-order feature mining unit is used for deep feature mining of the pollution feature sub-tensor and the physicochemical property feature sub-tensor respectively, and extracting high-order pollution features and high-order physicochemical property features.

[0046] The oil state comprehensive judgment and result output unit is used for comprehensively judging the gear box oil state based on the extracted high-order pollution features and high-order physicochemical property features through a preset decision rule, and outputting a gear box oil state detection result.

[0047] Beneficial effects: The wind power equipment gear box oil state detection method and system proposed by the application are not limited to single index analysis in the detection means, but simultaneously quantitatively analyze solid particles, liquid and gas pollutants in the oil sample, and combine multiple physicochemical property parameters such as kinematic viscosity and acid value to construct a multi-dimensional pollution factor matrix and a physicochemical property feature vector, form a composite feature tensor through tensor fusion, and realize comprehensive detection of the oil state. In terms of analysis logic, the system does not consider pollution and physicochemical properties in isolation, but uses feature decoupling algorithm to separate pollution and physicochemical property features in the composite feature tensor, mines high-order features through a multi-level feature extraction network, and comprehensively judges based on a decision tree and a rule base, fully considers the interaction relationship between the two, accurately evaluates the oil state, and further includes sample collection position optimization and detection result confidence evaluation, etc. to ensure the representativeness and reliability of the detection result. The method and system comprehensively improve the accuracy and effectiveness of the gear box oil state detection, provide a strong guarantee for the stable operation of the wind power equipment, can timely find potential problems of the oil, and avoid equipment failure caused by misjudgment or incomplete detection. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The method step flowchart of the application;

[0049] Figure 2 The system unit composition diagram of the application. DETAILED DESCRIPTION

[0050] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict, and the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0051] As shown in the figure, the wind power equipment gear box oil state detection method comprises: Figure 1

[0052] ​Step S1: Obtain a gear box oil sample, quantitatively analyze solid particle pollutants, liquid pollutants and gas pollutants in the oil sample based on an oil pollution degree analysis model, and extract and encode the particle size distribution and concentration information of different types of pollutants by constructing a multi-dimensional pollution factor matrix;

[0053] Specifically, the main task of step S1 is to obtain a gear box oil sample and quantitatively analyze the pollutants in the sample based on an oil pollution degree analysis model. In an embodiment, when obtaining the oil sample, a dedicated sampler is usually used to extract from a specific sampling port of the wind power equipment gear box under a stable running state. The sampling port position is optimized to ensure that the collected sample is representative. After obtaining the sample, a particle counter is used to detect solid particle pollutants. The number of particles in different particle size intervals is accurately measured by the laser scattering principle, and the particle size distribution and concentration are determined. For liquid pollutants, a chromatographic analyzer is used to identify the types of liquid pollutants and determine their mixing proportions based on the separation characteristics of different substances in the chromatographic column. The gas analyzer is used to analyze the dissolved gas components and content in the oil by using the gas sensor. Finally, the detection results are constructed into a multi-dimensional pollution factor matrix according to the established rules to complete the feature extraction and coding of the pollution information.

[0054] Accurate quantification of pollution information is the core basis for judging the pollution degree of the gear box oil. During the long-term operation of the wind power equipment, dust, sand and other impurities from the outside world may invade the oil, and the friction and wear of the internal components of the gear box may also produce metal particles and other pollutants. These different types of pollutants have different harmful ways and degrees of harm to the gear box components. By comprehensive and detailed analysis and coding, the specific condition of the oil pollution can be mastered, and the potential threat to the normal operation of the gear box can be evaluated, thereby providing important data support for subsequent detection and maintenance decision-making. If this step is ignored and the pollution information is not accurately obtained, the oil pollution problem may not be discovered in time, the gear box components may be prematurely worn and degraded, and even equipment failure may occur, affecting the stable operation of the wind power equipment.

[0055] Step S2: Use an oil physicochemical property analysis model to analyze the kinematic viscosity, acid value, water content and oxidation stability parameters of the gear box oil, establish a physicochemical property parameter space, and convert the measured parameter values into corresponding feature vectors by nonlinear mapping.

[0056] Specifically, step S2 uses an oil physicochemical property analysis model to analyze the kinematic viscosity, acid value, water content, oxidation stability, and other important parameters of the gearbox oil. In specific implementation, the kinematic viscosity is measured using a capillary viscometer, which measures the flow time of the oil through the capillary tube at a specified temperature based on the Poiseuille law, and then calculates the kinematic viscosity value; the acid value is determined using the potentiometric titration method, which reacts with the acidic substances in the oil using a suitable titrant, and determines the titration endpoint by measuring the change in electrode potential to obtain the acid value; the water content is detected using the Karl Fischer titration method, which reacts with water using Karl Fischer reagent, and determines the water content by measuring the consumption of reagent; the oxidation stability is determined by the rotating bomb method, which records the time required to reach the specified pressure drop value under specific temperature and pressure conditions to evaluate the oxidation stability of the oil. After obtaining the measured values of each parameter, a physicochemical property parameter space is established through a specially designed software algorithm, and these parameters are converted into corresponding feature vectors using nonlinear mapping to realize the digital expression of the physicochemical properties of the oil.

[0057] The physicochemical properties of the oil directly affect its lubrication, corrosion prevention, and other properties, and thus determine the reliability of the gearbox. The appropriateness of the kinematic viscosity directly relates to whether the components of the gearbox can be well lubricated, and excessive viscosity will increase the frictional resistance between components, leading to increased energy consumption, while low viscosity will not form an effective lubricating oil film, exacerbating component wear; high acid value means that the oil may have oxidized or been contaminated, which can corrode the gearbox components; high water content can emulsify the oil and damage its lubrication performance; poor oxidation stability can accelerate the deterioration of the oil properties. By converting these interrelated and interdependent physicochemical property parameters into feature vectors, the comprehensive performance status of the oil can be systematically analyzed, and the trends of each parameter can be captured to provide key evidence for accurately determining whether the oil is suitable for continued use. If this step is skipped, the internal performance changes of the oil cannot be thoroughly understood, the normal operation of the gearbox cannot be guaranteed, and equipment failures may occur due to poor oil performance, increasing maintenance costs and downtime.

[0058] Step S3: Tensor fusion of the multi-dimensional pollution factor matrix obtained in step S1 and the physicochemical property feature vector obtained in step S2 to form a composite feature tensor containing pollution information and physicochemical property information;

[0059] Specifically, step S3 is to perform tensor fusion on the multi-dimensional pollution factor matrix obtained in step S1 and the physicochemical property eigenvector obtained in step S2 to form a composite feature tensor. In the specific implementation process, the fusion operation is realized by using a specially written tensor operation program on the data processing platform of a high-performance computer. When performing tensor product operation, the elements of each dimension of the multi-dimensional pollution factor matrix and the physicochemical property eigenvector are operated according to the tensor operation rules to obtain a preliminary fusion result. Then, according to the operating condition parameters of the gearbox oil, such as operating temperature, load size, and rotating speed, the corresponding weight coefficients are assigned to the physicochemical property eigenvectors of different dimensions through a pre-established weight distribution model. The weight distribution model is constructed based on a large amount of experimental data and actual operating experience, and can accurately reflect the influence degree of each physicochemical property parameter on the oil condition under different operating conditions. Finally, the preliminary fusion result is weighted and summed to obtain a composite feature tensor containing pollution information and physicochemical property information, and the high integration of oil condition data is realized.

[0060] This step plays a key role in accurately evaluating the oil condition. The pollution condition and physicochemical property of oil are closely related and interact with each other. The presence of pollutants may accelerate the oxidation process of oil, leading to changes in physicochemical properties such as acid value and kinematic viscosity. Changes in physicochemical properties, such as reduced viscosity, may also affect the suspension and dispersion state of pollutants in oil, thereby affecting the degree of wear of gearbox components caused by pollutants. Through tensor fusion, these complex interrelationships can be presented in data form, allowing subsequent analysis to be conducted from a holistic perspective and avoiding isolated consideration of the pollution or physicochemical properties of oil. This comprehensive data integration lays a foundation for more accurate understanding of the actual condition of oil, provides a more reliable and comprehensive basis for the maintenance and management of the gearbox, and helps maintenance personnel to develop more scientific and reasonable maintenance strategies to ensure the stable operation of wind power equipment.

[0061] Step S4: based on the composite feature tensor, using a pre-set feature decoupling algorithm, separating the pollution-related features and physicochemical property-related features in the composite feature tensor to obtain a pollution feature sub-tensor and a physicochemical property feature sub-tensor;

[0062] Specifically, step S4 separates the pollution-related features and the physicochemical property-related features in the composite feature tensor obtained in step S3 using a preset feature decoupling algorithm to obtain a pollution feature sub-tensor and a physicochemical property feature sub-tensor. In an embodiment, the feature decoupling algorithm is implemented by computer programming, and the algorithm is based on the principle of tensor decomposition and uses an iterative optimization calculation method. First, the initial parameters in the algorithm are reasonably set according to the stability parameters of the gearbox oil, such as the oxidation stability and emulsion stability of the oil. Then, the related parameters in the decomposition process are adjusted through multiple iterations, and in each iteration, the norm of the decomposition error tensor is calculated, and the parameters are optimized with the goal of minimizing the norm. In the specific calculation process, mathematical methods such as matrix operations and vector operations are used to gradually decompose the composite feature tensor into two relatively independent sub-tensors. During the iteration process, the accuracy and reliability of the decomposition results are monitored in real time, and when the norm of the decomposition error tensor meets the preset accuracy requirement, the iteration is stopped, and the final pollution feature sub-tensor and physicochemical property feature sub-tensor are obtained.

[0063] Although the composite feature tensor integrates rich oil state information, separating the pollution features and the physicochemical property features can more clearly understand the changes and interaction mechanisms of the two in the actual analysis and evaluation process. For example, when analyzing the pollution feature sub-tensor, the change trend of different types of pollutants and their influence on the wear of gearbox components can be studied, so as to determine whether oil filtration or replacement is needed; the analysis of the physicochemical property feature sub-tensor helps to determine whether the lubrication performance and corrosion resistance of the oil meet the requirements of the gearbox operation, and whether additives or other treatment methods are needed to improve the physicochemical properties. This separation operation improves the accuracy and accuracy of the analysis, avoids the difficulty of analysis caused by the mixing of pollution and physicochemical property information, and provides clear and explicit data support for subsequent more accurate evaluation of oil state and maintenance decision-making.

[0064] Step S5: Deep feature mining is performed on the pollution feature sub-tensor and the physicochemical property feature sub-tensor respectively, and high-order pollution features and high-order physicochemical property features are extracted by constructing a multi-level feature extraction network.

[0065] Specifically, step S5 performs deep feature mining on the pollution feature sub-tensor and the physicochemical property feature sub-tensor respectively, and is implemented by constructing a multi-level feature extraction network. In actual construction of the multi-level feature extraction network, based on a deep learning framework such as TensorFlow or PyTorch, according to the wear characteristics parameters of the gearbox oil such as the gear material characteristics, the wear rate during operation, etc., the weight tensor of each layer in the network is initialized and set. At the same time, according to the performance change trend of the oil under different working conditions, a suitable activation function is selected, such as the ReLU function, the Sigmoid function, etc. During network operation, each layer receives the feature tensor output by the previous layer as input, multiplies the input feature tensor with the weight tensor of the layer through matrix multiplication operation, adds the bias tensor to obtain the preliminary calculation result, and then inputs the preliminary calculation result into the activation function for nonlinear transformation to obtain the output feature tensor of the layer. Through multiple layers of such calculation and activation operations, high-order pollution features and high-order physicochemical property features are gradually extracted from the pollution feature sub-tensor and the physicochemical property feature sub-tensor. In the network training process, a large amount of historical oil detection data is used as training samples, and the weight parameters of each layer in the network are continuously adjusted through the back propagation algorithm to minimize the error between the predicted result and the actual result, thereby improving the accuracy and effectiveness of the network feature extraction.

[0066] It is difficult to comprehensively and accurately evaluate the actual state of the oil under complex working conditions by only obtaining the basic pollution and physicochemical property information of the oil. By mining high-order features through the multi-level feature extraction network, the subtle trends and potential laws of the oil state change can be captured. For example, in the early wear stage of the gearbox, the high-order pollution features can find abnormal changes in the number and particle size distribution of some tiny particle pollutants, which may not be obvious in basic analysis, but indicate potential wear problems; for the oxidation process of the oil, the high-order physicochemical property features can more accurately reflect the degree and speed of performance degradation, providing a basis for taking measures to prevent further deterioration of the oil performance. These high-order features provide more valuable information for subsequent state evaluation, making the detection result more forward-looking and reliable, which helps the maintenance personnel to discover potential problems of the oil state in time, to develop maintenance plans in advance, to reduce the probability of equipment failure, and to ensure the stable and efficient operation of the wind power equipment.

[0067] Step S6: input the extracted high-order pollution features and high-order physicochemical property features into a state evaluation decision model, comprehensively judge the state of the gearbox oil through a preset decision rule, and output the gearbox oil state detection result.

[0068] Specifically, step S6 inputs the high-order contamination features and high-order physicochemical property features extracted in step S5 into a state evaluation decision model, comprehensively determines the gear box oil state through a preset decision rule, and outputs a gear box oil state detection result. In implementation, the state evaluation decision model is constructed based on a computer software system and adopts a structure combining a decision tree and a rule base. The decision rule is set based on statistical analysis of historical detection data of a large number of gear box oil samples, and in combination with cleanliness standard parameters, quality standard parameters and performance limit parameters of the gear box oil. In specific construction of the decision tree, key parameters of the high-order contamination features and the high-order physicochemical property features are taken as node division bases, the relationship between features and oil states in the historical data is learned continuously to determine an optimal node division mode, and a hierarchical decision tree structure is formed. The rule base stores specific rules for oil state determination in various cases, such as a rule for determining that the gear box oil state is seriously abnormal when the comprehensive measurement value of the high-order contamination features is higher than a contamination feature determination threshold and the comprehensive measurement value of the high-order physicochemical property features is also higher than a physicochemical property feature determination threshold; conversely, if both are lower than the threshold values, the corresponding rule is used to determine that the gear box oil state is normal. After the high-order contamination features and the high-order physicochemical property features are input, the decision model first classifies preliminarily through the decision tree, then accurately determines in combination with the rules in the rule base, and finally outputs the gear box oil state detection result.

[0069] The detection result output in this step is directly related to the maintenance decision and operation safety of the gear box. By combining the high-order features with the scientifically set decision rule, the oil state can be objectively and accurately evaluated. Based on the detection result, the operation and maintenance personnel can take corresponding measures in time, such as immediately arranging replacement of the oil when the detection result shows that the oil state is seriously abnormal to avoid gear box failure caused by oil problems, or performing filtration treatment or strengthening the monitoring frequency when the oil state is slightly abnormal. This step makes the entire detection process form a complete closed loop, from sample acquisition, contamination analysis, physicochemical property analysis, feature fusion and decoupling, deep feature mining to final state evaluation and result output, thereby providing comprehensive and reliable technical support for the management of the gear box oil state and helping to improve the operation efficiency of the wind power equipment, reduce the maintenance cost and ensure the stable operation of the wind power system.

[0070] Preferably, the multi-dimensional pollution factor matrix of the oil pollution degree analysis model is calculated by the following formula in step S1.

[0071]

[0072] wherein, M CF represents the multi-dimensional pollution factor matrix; n is the number of types of contaminants; α iis the weight coefficient of the ith type of contaminant, which is determined according to the anti-pollution performance parameters of the gear box oil; D i is the particle size distribution vector of the ith type of contaminant, which contains the proportion information of the contaminants in different particle size intervals; C i is the concentration scalar of the ith type of contaminant; f i is the feature coding function for the ith type of contaminant, which adjusts the coding rules according to the lubrication characteristic parameters of the gear box oil.

[0073] Specifically, the technology of constructing the multi-dimensional pollution factor matrix in step S1 is refined. In the implementation process, different types of contaminants in the gear box oil, such as solid particles, liquid impurities, and gas components, need to be considered. For each type of contaminant, its key information needs to be obtained, such as the proportion distribution of solid particles of different sizes, and the content of liquid and gas contaminants. At the same time, according to the anti-pollution performance of the gear box oil, different weights are given to each type of contaminant. The oil with excellent anti-pollution performance has relatively small influence on some contaminants, and the weight is set to be low. In the implementation, professional detection equipment such as particle counter, chromatographic analyzer and gas analyzer is used to accurately measure the data of each type of contaminant. Then, according to the pre-set rules and processes, these data are integrated to construct a multi-dimensional pollution factor matrix. This matrix can comprehensively and systematically present the pollution status of the oil, and provides a key data basis for subsequent in-depth analysis of the oil status and judgment of whether it affects the normal operation of the gear box.

[0074] Preferably, the physicochemical property analysis model of step S2 is established, and the physicochemical property parameter space conversion formula is as follows:

[0075] V PP =β·ReLU(γ·[V,A,W,O] T )

[0076] Wherein, V PP represents the converted physicochemical property feature vector; β is a scale scaling matrix, the element value of which is set according to the specification parameters of the gear box oil; γ is a weight matrix, which is determined according to the performance standard parameters of the gear box oil; V is the measured value of the kinematic viscosity of the gear box oil; A is the measured value of the acid value; W is the measured value of the water content; O is the measured value of the oxidation stability index; ReLU is a linear rectification activation function, which is used to enhance the nonlinear expression ability of the feature vector.

[0077] Specifically, the conversion of the physicochemical property parameters of the oil in step S2 includes multiple important indicators such as kinematic viscosity, acid value, moisture content, and oxidation stability. Different physicochemical property indicators reflect different performance characteristics of the oil, and are interrelated and affect each other. In the implementation process, special detection instruments are used, such as a capillary viscometer to measure the kinematic viscosity and a potentiometric titrator to determine the acid value, to obtain accurate parameter values. Then, according to the specifications and performance requirements of the oil, the measured parameters are processed. Through specific algorithms and procedures, these parameters are converted into feature vectors. In the conversion process, the numerical range of the feature vector is adjusted according to the specifications of the oil, and the importance of each parameter is determined according to the performance standards, so as to highlight the key physicochemical property indicators. The final feature vector presents the originally dispersed physicochemical property parameters in a more convenient and processed digital form, facilitating subsequent systematic research on the physicochemical property state of the oil and its changing trend.

[0078] Preferably, in step S3, the tensor fusion process adopts a combination of tensor product and weighted summation, and the specific fusion formula is:

[0079]

[0080] wherein T CFP represents the composite feature tensor; m is the dimension of the physicochemical property feature vector; ω j is the weight coefficient of the j-th dimension of the physicochemical property feature vector in the fusion process, which is determined by the operating condition parameters of the gearbox oil; V PPj is the j-th component of the physicochemical property feature vector V PP ; and represents the tensor product operation.

[0081] Specifically, after completing the separate analysis of the oil pollution information and the physicochemical property information in step S3, the two parts of information need to be deeply fused. When fusing, it is considered that in different operating conditions of the gearbox oil, the influence of each physicochemical property parameter on the oil state is different, for example, in high temperature and high load conditions, the importance of certain physicochemical property indicators will be significantly improved. Therefore, according to the actual operating conditions, appropriate weights are given to physicochemical property features of different dimensions. In the implementation, the multi-dimensional pollution factor matrix and the physicochemical property feature vector obtained in the previous steps are first obtained, and then the pollution information matrix and the physicochemical property feature vector are operated by using a high-performance computer and professional data processing software according to specific fusion rules. Through such operation, a composite feature tensor containing pollution information, physicochemical property information, and their mutual relationship is formed. The composite feature tensor integrates the comprehensive information of the oil state, and provides comprehensive data support for more accurate evaluation of the oil state and the development of reasonable maintenance strategies.

[0082] Preferably, the step S4, the feature decoupling algorithm is based on the principle of tensor decomposition, the composite feature tensor is decomposed by the following formula:

[0083]

[0084] Wherein, T CF represents the pollution feature sub-tensor; T PP represents the physicochemical property feature sub-tensor; ° represents the Hadamard product operation of tensor; E is the decomposition error tensor, the norm of the error tensor is minimized by the iterative optimization algorithm, and the parameter adjustment in the optimization process is based on the stability parameters of the gearbox oil.

[0085] Specifically, since the fused composite feature tensor contains mixed information of oil pollution and physicochemical properties, in order to more clearly and accurately analyze the influence of the two on the oil state, it is necessary to separate the pollution-related features and physicochemical property-related features in the composite feature tensor. When implementing this step, the initial parameters of the feature decoupling algorithm are set according to the stability-related characteristics of the gearbox oil, such as antioxidant stability, emulsion resistance stability, etc. Then, through computer programming, the parameters in the decomposition process are adjusted through multiple iterations. In each iteration, the accuracy of the decomposition result is monitored to ensure that the decomposition error is as small as possible. When the preset accuracy requirement is reached, the iteration is stopped, and the pollution feature sub-tensor and the physicochemical property feature sub-tensor are successfully obtained. The separated sub-tensors enable subsequent in-depth analysis of the pollution features and physicochemical property features, which helps to more accurately judge the oil state and provide more targeted basis for gearbox maintenance.

[0086] Preferably, the step S5, the multi-level feature extraction network contains multiple feature extraction layers, and the feature extraction formula of each layer is:

[0087] F l = σ(θ l · F l-1 + b l )

[0088] Wherein, F l is the feature tensor extracted by the lth layer; F l-1 is the output feature tensor of the (l-1)th layer; θ l is the weight tensor of the lth layer, and the parameter value is initialized according to the wear characteristic parameters of the gearbox oil; b l is the bias tensor of the lth layer; σ is the activation function, and the appropriate activation function type is selected according to the performance change trend of the gearbox oil.

[0089] Specifically, in order to mine more valuable and more essential information of oil state from the pollution feature sub-tensor and the physicochemical property feature sub-tensor, a multi-level feature extraction network is constructed. When constructing the network, according to the gear material characteristics of the gearbox oil, the wear rate and other wear characteristics parameters in the running process, the weight tensor of each layer of the network is initialized and set. At the same time, combined with the performance change trend of the oil under different working conditions, a suitable activation function is selected. In the implementation process, the pollution and physicochemical property feature sub-tensors decoupled in the foregoing are input into the network, and each layer of the network will process the input feature tensor, and through matrix multiplication, bias addition and activation function operation and other operations, high-order pollution features and high-order physicochemical property features are gradually extracted. With the increase of the network level, the extracted features are more and more abstract, and can more and more reflect the deep change rule of the oil state. These high-order features provide more in-depth and valuable data support for accurately evaluating the oil state and predicting the performance change trend of the oil.

[0090] Preferably, in the step S6, the state evaluation decision model adopts a combination of decision tree and rule base, and the decision process is carried out through the following rules:

[0091] if H CF >τ CF and H PP >τ PP , it is determined that the gearbox oil state is seriously abnormal;

[0092] if H CF >τ CF and H PP ≤τ PP , it is determined that the gearbox oil state is pollution abnormal;

[0093] if H CF ≤τ CF and H PP >τ PP , it is determined that the gearbox oil state is physicochemical property abnormal;

[0094] if H CF ≤τ CF and H PP ≤τ PP , it is determined that the gearbox oil state is normal;

[0095] wherein H CF is a comprehensive measurement value of the extracted high-order pollution feature, which is calculated according to the cleanliness standard parameter of the gearbox oil; H PP is a comprehensive measurement value of the high-order physicochemical property feature, which is determined according to the quality standard parameter of the gearbox oil; τ CF and τ PPThe determination threshold of the pollution characteristic and the physicochemical property characteristic respectively, the threshold is set by statistical analysis on a large number of historical detection data of gear box oil samples, and combining with the performance limit parameters of the gear box oil.

[0096] Specifically, after obtaining the high-order pollution characteristic and the high-order physicochemical property characteristic, it is necessary to determine the state of the gear box oil according to certain standards and rules. In actual operation, according to the cleanliness standard, quality standard and performance limit of the gear box oil and other related parameters, the comprehensive measurement value of the high-order pollution characteristic and the high-order physicochemical property characteristic is calculated to quantitatively reflect the degree of pollution and physicochemical property characteristic. At the same time, through systematic statistical analysis on a large number of historical detection data of gear box oil samples, combined with the performance limit parameters of the oil, the determination threshold of the pollution and physicochemical property characteristic is set. After comparing the comprehensive measurement value with the determination threshold, the oil state is determined according to the combination of the pre-set decision tree and rule base. If a certain condition is met, the oil state is determined as serious abnormality, pollution abnormality, physicochemical property abnormality or normal. These clear determination rules provide clear basis for the operation and maintenance personnel to accurately determine the oil state and timely make reasonable maintenance decisions.

[0097] Preferably, before step S1, it further includes a step of determining the collection position of the gear box oil sample, and the collection position is determined by the following method:

[0098] According to the internal flow field distribution characteristics of the gear box, combined with the flowability parameters of the gear box oil, an oil sample collection position optimization model is established, the model determines the best collection position by calculating the representativeness index R I of the oil sample at different positions, and the calculation formula is:

[0099]

[0100] Wherein, p is the number of candidate collection positions; δ k is the weight coefficient of the kth candidate collection position, which is determined by the structural parameters of the gear box; Corr(S k , S ref ) is the correlation measurement value of the oil sample at the kth candidate collection position and the reference sample, and the reference sample is constructed according to the standard performance parameters of the gear box oil; by comparing the representativeness index R I of each candidate collection position, the position with the largest representativeness index is selected as the oil sample collection position.

[0101] Specifically, in order to ensure that the collected oil sample can truly and comprehensively reflect the overall state of the oil in the gearbox, it is necessary to scientifically and reasonably determine the sample collection position. In actual operation, first, the flow field distribution in the gearbox is studied to understand the flow law of the oil in the gearbox, and the flowability characteristics of the gearbox oil are combined. Then, a plurality of candidate collection positions are determined in the gearbox, and the relevant parameters of the oil sample are measured for each candidate position. According to the standard performance parameters of the gearbox oil, a reference sample is constructed. The correlation between each candidate position sample and the reference sample is calculated, and the weight coefficient of each candidate position is determined in combination with the structural parameters of the gearbox, and then the representative index of each position is calculated. Finally, the representative indexes of each candidate position are compared, and the position with the largest index is selected as the final sample collection position. The sample collected in this way can represent the oil state in the gearbox to the greatest extent and provide a reliable data basis for subsequent accurate detection and analysis.

[0102] Preferably, after step S6, there is also a confidence evaluation step for the detection result, which is calculated by the following formula:

[0103]

[0104] wherein C E represents the confidence of the detection result; r is the number of reference indexes for evaluating the confidence; λ q is the weight coefficient of the qth reference index, which is set according to the detection accuracy requirement parameters of the gearbox oil; Conf(R q ) is the confidence value corresponding to the qth reference index, and the reference index includes the repeatability of the detection data and the consistency with the historical detection data, and the confidence value is calculated according to the stability parameters of the gearbox oil and the performance parameters of the detection equipment.

[0105] Specifically, after completing the detection and determination of the oil state of the gearbox, in order to enable the maintenance personnel to better understand the reliability of the detection result, it is necessary to evaluate the confidence of the detection result. In actual operation, first, the reference indexes for evaluation are determined, such as the repeatability of the detection data and the consistency with the historical detection data. Then, according to the specific requirements of the detection accuracy of the gearbox oil, the corresponding weight coefficient is set for each reference index, and the higher the detection accuracy requirement, the greater the weight of some key indexes. Then, according to the stability of the gearbox oil itself and the performance parameters of the detection equipment, the confidence value of each reference index is calculated. Finally, these data are substituted into the pre-set evaluation formula to calculate the confidence of the detection result. Through this confidence value, the maintenance personnel can intuitively judge the reliability of the detection result, so as to more scientifically make reasonable maintenance decisions according to the detection result and improve the accuracy and effectiveness of the maintenance work.

[0106] AsFigure 2 The wind power equipment gearbox oil liquid state detection system shown comprises:

[0107] The oil sample acquisition and pollution quantification analysis unit is configured to acquire a gearbox oil sample and quantitatively analyze solid particle pollutants, liquid pollutants and gas pollutants in the oil sample based on an oil pollution analysis model to construct a multi-dimensional pollution factor matrix.

[0108] The physicochemical property parameter analysis and feature vector conversion unit is configured to analyze the kinematic viscosity, acid value, moisture content and oxidation stability parameters of the gearbox oil, establish a physicochemical property parameter space based on an oil physicochemical property analysis model, and convert the measured parameter values into corresponding feature vectors.

[0109] The pollution-physicochemical feature tensor fusion unit is configured to fuse the multi-dimensional pollution factor matrix and the physicochemical property feature vectors to form a composite feature tensor containing pollution information and physicochemical property information.

[0110] The feature decoupling and sub-tensor separation unit is configured to separate the pollution-related features and the physicochemical property-related features in the composite feature tensor based on a preset feature decoupling algorithm to obtain a pollution feature sub-tensor and a physicochemical property feature sub-tensor.

[0111] The multi-level feature extraction and high-order feature mining unit is configured to perform deep feature mining on the pollution feature sub-tensor and the physicochemical property feature sub-tensor respectively to extract high-order pollution features and high-order physicochemical property features.

[0112] The oil state comprehensive judgment and result output unit is configured to comprehensively judge the gearbox oil state based on the extracted high-order pollution features and high-order physicochemical property features through a preset decision rule and output a gearbox oil liquid state detection result.

[0113] The output end of the oil sample acquisition and pollution quantification analysis unit is connected to the first input end of the pollution-physicochemical feature tensor fusion unit, the output end of the physicochemical property parameter analysis and feature vector conversion unit is connected to the second input end of the pollution-physicochemical feature tensor fusion unit, the output end of the pollution-physicochemical feature tensor fusion unit is connected to the input end of the feature decoupling and sub-tensor separation unit, the two output ends of the feature decoupling and sub-tensor separation unit are respectively connected to the two input ends of the multi-level feature extraction and high-order feature mining unit, and the output end of the multi-level feature extraction and high-order feature mining unit is connected to the input end of the oil state comprehensive judgment and result output unit.

[0114] The present application precisely grasps the oil state through multi-dimensional deep analysis and innovative model construction. The advantages thereof are described below from the aspects of comprehensive detection, relationship analysis and detection optimization.

[0115] Firstly, in view of the problem of single detection means in the prior art, the detection method and system realize comprehensive detection in multiple dimensions. Instead of detecting only one or several indicators of the oil, the solid particle pollutants, liquid pollutants and gas pollutants in the oil sample are quantitatively analyzed, and multiple physicochemical property parameters such as the kinematic viscosity, acid value, water content and oxidation stability are comprehensively considered. By constructing a multi-dimensional pollution factor matrix and a physicochemical property characteristic vector, and performing tensor fusion, a composite feature tensor containing rich information is formed, which comprehensively covers the oil pollution and physicochemical property state, avoids misjudgment caused by one-sided detection, and provides a more comprehensive and accurate data basis for the oil state evaluation of the gearbox.

[0116] Secondly, in order to solve the problem that the prior art lacks in-depth analysis of the mutual relationship between oil pollution and physicochemical properties, the feature decoupling algorithm is used to separate the pollution-related features and the physicochemical property-related features in the composite feature tensor, and then high-order features are mined through a multi-level feature extraction network. Based on the state evaluation decision model combining the decision tree and the rule base, the mutual interaction relationship is fully considered, and the oil state of the gearbox is comprehensively determined. This method breaks the limitation of traditional isolated analysis, can more accurately evaluate the actual state of the oil, timely discovers potential problems, and effectively avoids the situation of missing the maintenance opportunity due to not considering the interaction.

[0117] In addition, the detection method and system also optimize the sample collection and result evaluation links. By combining the internal flow field distribution of the gearbox and the oil flowability parameters, a sample collection position optimization model is established to determine the best collection position and ensure that the collected sample is representative. By evaluating the confidence of the detection results, considering reference indexes such as the repeatability of the detection data and the consistency with historical data, combining the stability of the gearbox oil and the performance parameters of the detection equipment, the reliability of the detection results is evaluated, and the accuracy and effectiveness of the detection are further improved, providing a solid guarantee for the stable operation of the wind power equipment.

[0118] In the description of the present application, it should be noted that, unless otherwise specified and limited, the terms "arrangement", "installation", "connection", "connection", "fixing" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be connected inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0119] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement of parts shown. It is to be understood that all related terms not specifically defined in the specification shall be interpreted in accordance with United States Patent Office interpretation of Dictionary of Electrical and Electronic Terms, Third Edition, as updated and revised for the Patent and Trademark Office, and published in 2001.

Claims

1. A method for detecting the condition of gearbox oil in wind power generation equipment, characterized in that, The method includes: Step S1: Obtain gearbox oil samples. Based on the oil contamination analysis model, perform quantitative analysis on solid particulate contaminants, liquid contaminants and gaseous contaminants in the oil samples. By constructing a multidimensional contamination factor matrix, extract and encode the particle size distribution and concentration information of different types of contaminants. Step S2: Using the oil physicochemical property analysis model, establish a physicochemical property parameter space for the kinematic viscosity, acid value, water content, and oxidation stability parameters of the gearbox oil. Through nonlinear mapping, convert the measured parameter values ​​into corresponding feature vectors. Step S3: Perform tensor fusion between the multidimensional pollution factor matrix obtained in step S1 and the physicochemical property feature vector obtained in step S2 to form a composite feature tensor containing pollution information and physicochemical property information; Step S4: Based on the composite feature tensor, using a preset feature decoupling algorithm, separate the pollution-related features and physicochemical property-related features in the composite feature tensor to obtain the pollution feature sub-tensor and the physicochemical property feature sub-tensor; Step S5: Perform deep feature mining on the pollution feature sub-tensor and the physicochemical property feature sub-tensor respectively, and extract high-order pollution features and high-order physicochemical property features by constructing a multi-level feature extraction network; Step S6: Input the extracted high-order contamination features and high-order physicochemical properties into the state assessment decision model, and make a comprehensive judgment on the gearbox oil state through the preset decision rules, and output the gearbox oil state detection results.

2. The method for detecting the oil condition of a wind power generation equipment gearbox according to claim 1, characterized in that, In step S1, the multidimensional contamination factor matrix constructed by the oil contamination analysis model is calculated using the following formula: Among them, M CF This represents a multidimensional pollution factor matrix; n is the number of pollutant types; α i D represents the weighting coefficient for the i-th type of contaminant, which is determined based on the anti-fouling performance parameters of the gearbox oil. i C represents the particle size distribution vector of the i-th type of pollutant, containing information on the proportion of pollutants in different particle size ranges; i f is the concentration scalar value of the i-th type of pollutant; i The coding rules are adjusted based on the lubrication characteristic parameters of the gearbox oil to form the feature coding function for the i-th type of pollutant.

3. The method for detecting the oil condition of a wind power generation equipment gearbox according to claim 1, characterized in that, In step S2, the spatial transformation formula for the physicochemical property parameters established by the oil physicochemical property analysis model is as follows: V PP Nβ·ReLU(γ·[V,A,W,O] T ) Among them, V PP β represents the transformed physicochemical property feature vector; β is the scaling matrix, with element values ​​set according to the gearbox oil specifications; γ is the weight matrix, determined based on the gearbox oil performance standard parameters; V is the measured kinematic viscosity of the gearbox oil; A is the measured acid value; W is the measured water content; O is the measured oxidation stability index; ReLU is the linear rectification activation function, used to enhance the nonlinear expressive power of the feature vector.

4. The method for detecting the oil condition of a wind power generation equipment gearbox according to claim 1, characterized in that, In step S3, the tensor fusion process employs a combination of tensor product and weighted summation. The specific fusion formula is as follows: Among them, T CFP Represents the composite feature tensor; m is the dimension of the physicochemical property feature vector; ω j V represents the weighting coefficient of the j-th dimension physicochemical property eigenvector during the fusion process; this coefficient is determined by the operating parameters of the gearbox oil. PPj V is the eigenvector of physicochemical properties. PP The j-th dimension component; This represents the tensor product operation.

5. The method for detecting the oil condition of a wind power generation equipment gearbox according to claim 1, characterized in that, In step S4, the feature decoupling algorithm is based on the principle of tensor decomposition and decomposes the composite feature tensor using the following formula: Among them, T CF T represents the pollution characteristic tensor; PP Characteristic tensors representing physicochemical properties; The Hadamard product operation represents the tensor; E is the decomposed error tensor, and the norm of this error tensor is minimized through an iterative optimization algorithm. The parameter adjustments during the optimization process are based on the stability parameters of the gearbox oil.

6. The method for detecting the oil condition of a wind power generation equipment gearbox according to claim 1, characterized in that, In step S5, the multi-level feature extraction network contains multiple feature extraction layers, and the feature extraction formula for each layer is: F l =σ(θ l ·F l-1 +b l ) Among them, F l F is the feature tensor extracted from the l-th layer; l-1 θ is the output feature tensor of the (l-1)th layer; l b is the weight tensor of the l-th layer, and its parameter values ​​are initialized based on the wear characteristics of the gearbox oil; l σ is the bias tensor of the l-th layer; σ is the activation function, and the appropriate activation function type is selected according to the performance change trend of the gearbox oil.

7. The method for detecting the oil condition of a wind power generator gearbox according to claim 1, characterized in that, In step S6, the state assessment decision model adopts a combination of decision tree and rule base, and the decision-making process is carried out according to the following rules: If H CF >τ Cf And H PP >τ PP If so, the gearbox oil condition is determined to be seriously abnormal; If H CF >τ CF And H PP ≤τ PP If so, the gearbox oil condition is determined to be abnormally contaminated; If H CF ≤τ CF And H PP >τ PP If so, the gearbox oil condition is determined to be abnormal in terms of physical and chemical properties; If H CF ≤τ CF And H PP ≤τ PP If so, the gearbox oil condition is determined to be normal; Among them, H CF The comprehensive metric value for the extracted high-order contamination characteristics was calculated based on the standard parameters for gearbox oil cleanliness; H PP τ is a comprehensive measure of higher-order physicochemical properties, determined based on the quality standard parameters of gearbox oil; CF and τ PP These are the threshold values ​​for judging pollution characteristics and physicochemical properties, respectively. These threshold values ​​are set by statistically analyzing historical test data from a large number of gearbox oil samples and combining them with the performance limit parameters of the gearbox oil.

8. The method for detecting the oil condition of a wind power generator gearbox according to claim 1, characterized in that, Before step S1, the method further includes a step of determining the sampling location of the gearbox oil sample, wherein the sampling location is determined by the following method: Based on the internal flow field distribution characteristics of the gearbox and the fluidity parameters of the gearbox oil, an optimization model for oil sample collection location is established. This model calculates the representativeness index R of oil samples at different locations. I The optimal sampling location is determined using the following formula: Where p is the number of candidate acquisition locations; δ k The weighting coefficient for the j-th candidate acquisition position is determined by the structural parameters of the gearbox; Corr(S) k S ref _ represents the correlation metric between the oil sample at the k-th candidate sampling location and the reference sample, which is constructed based on the standard performance parameters of gearbox oil; the correlation is measured by comparing the representativeness index R of each candidate sampling location. I The location with the highest representative index was selected as the oil sample collection location.

9. The method for detecting the oil condition of a wind power generator gearbox according to claim 1, characterized in that, Following step S6, a confidence assessment step for the detection results is also included, wherein the confidence assessment is calculated using the following formula: Among them, C E The confidence level of the test result is represented by r; r is the number of reference indicators used to assess the confidence level; λ q This is the weighting coefficient for the q-th reference indicator, which is set according to the required testing accuracy of the gearbox oil; Conf(R) q ) represents the confidence level value corresponding to the q-th reference indicator. The reference indicators include the repeatability of the test data and the consistency with historical test data. The confidence level value is calculated based on the stability parameters of the gearbox oil and the performance parameters of the test equipment.

10. A gearbox oil condition monitoring system for wind power generation equipment, characterized in that, include: The oil sample acquisition and contamination quantification analysis unit is used to acquire gearbox oil samples and quantify the solid particulate contaminants, liquid contaminants and gaseous contaminants in the oil samples based on the oil contamination analysis model to construct a multidimensional contamination factor matrix. The physicochemical property parameter analysis and feature vector conversion unit is used to establish a physicochemical property parameter space based on the oil physicochemical property analysis model for kinematic viscosity, acid value, water content, and oxidation stability parameters of gearbox oil, and convert the measured parameter values ​​into corresponding feature vectors. The pollution-physicochemical feature tensor fusion unit is used to fuse the multidimensional pollution factor matrix with the physicochemical property feature vector to form a composite feature tensor containing pollution information and physicochemical property information. The feature decoupling and sub-tensor separation unit is used to separate the pollution-related features and physicochemical property-related features in the composite feature tensor based on a preset feature decoupling algorithm, so as to obtain the pollution feature sub-tensor and the physicochemical property feature sub-tensor. The multi-level feature extraction and high-order feature mining unit is used to perform deep feature mining on the pollution feature sub-tensor and the physicochemical property feature sub-tensor respectively, and extract high-order pollution features and high-order physicochemical property features; The oil condition comprehensive judgment and result output unit is used to comprehensively judge the gearbox oil condition based on the extracted high-order contamination characteristics and high-order physicochemical properties, and output the gearbox oil condition detection results through preset decision rules.

Images