Wind power gear box wear state evaluation method and system fusing vibration and oil

Abstract

The invention belongs to the technical field of gear box wear state evaluation. The invention provides a vibration and oil fused wind power gear box wear state evaluation method and system, and the method comprises the steps: converting a vibration signal of a wind power gear box into a time-frequency domain image through employing a Markov transfer field, carrying out the resampling of the time-frequency domain image, and obtaining a vibration two-dimensional time-frequency graph with the same pixel as a two-dimensional image of an abrasive particle ring; performing down-sampling on the two-dimensional image of the abrasive particle ring and the two-dimensional time-frequency image of the vibration, fusing the two-dimensional image of the abrasive particle ring after down-sampling and the two-dimensional time-frequency image after down-sampling, and performing up-sampling on a fusion result; and obtaining a wear state evaluation result according to an up-sampling result. According to the method, feature selection deviation caused by subjective factors is avoided, and the accuracy and robustness of wear state recognition are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gearbox wear state assessment, and in particular to a wind power gearbox wear state assessment method and system integrating vibration and oil. Background Art

[0002] The statements in this section merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] In the field of gear system health monitoring, online monitoring of abrasive particles and vibration has become a key means of assessing the health of gearboxes. Abrasive particle monitoring reflects the wear state by analyzing the relative concentration changes of abrasive particles in the oil, and further explores the wear mechanism through abrasive particle morphology analysis. Vibration monitoring, on the other hand, can track the changing trends of tooth surface wear in real time, exhibiting a different sensitivity from abrasive particle monitoring. However, the wear process of a gear system is characterized by distinct stages, and the interaction mechanism between tooth surface wear and dynamic behavior is complex. This leads to a complex nonlinear relationship between the tooth surface wear state and vibration and abrasive particle characteristics, as well as differences in sensitivity between these characteristics. This mechanistic complexity results in significant differences in the sensitivity of a single monitoring method at different wear stages, increasing the uncertainty of wear state assessment. Therefore, relying solely on a single monitoring method makes it difficult to achieve a comprehensive and accurate assessment of the wear state of a gear system's tooth surface.

[0004] Multi-source sensor information fusion technology has significant potential for achieving high-reliability assessment and life prediction for mechanical equipment, providing a new approach for assessing the wear status of gear systems. Compared with single monitoring features, wear particle and vibration features show significant complementarity when assessing the wear status of tooth surfaces. Assessment methods that integrate these two features can effectively improve the accuracy and reliability of assessments. Current mainstream fusion methods rely on manually extracted time-frequency features (mean, kurtosis, etc.) and oil characteristics (concentration, morphology, etc.). This subjective feature selection not only loses detailed information from the original data but also makes it difficult to capture the deep nonlinear correlation between vibration waveforms and wear particle images. Existing feature-layer or decision-layer fusion architectures rely on prior knowledge to define weights and cannot directly utilize the complementary characteristics of multi-source data, resulting in low early wear recognition rates and high false alarm rates. Summary of the Invention

[0005] In order to solve the problems of insufficient cross-modal feature mining, excessive manual intervention and limited fusion levels, the present invention provides a wind turbine gearbox wear state assessment method and system that integrates vibration and oil. The Markov transfer field is used to convert the wind turbine gearbox vibration signal into a time-frequency domain image, and the time-frequency domain image is resampled to obtain a vibration two-dimensional time-frequency map with the same pixels as the wear ring two-dimensional image. The vibration two-dimensional time-frequency map and the wear ring two-dimensional image are directly automatically processed and comprehensively analyzed, thereby avoiding feature selection bias caused by subjective factors. By mining the potential correlation between different data modalities, the details and complex characteristics of the original information can be better retained, so that the correlation and complementarity between features can be fully mined, thereby improving the accuracy and robustness of wear state identification.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for evaluating the wear status of a wind turbine gearbox by integrating vibration and oil.

[0007] A wind turbine gearbox wear condition assessment method integrating vibration and oil includes the following steps: Acquire synchronously collected wind turbine gearbox vibration signals and abrasive ring two-dimensional images used to characterize abrasive concentration changes; The wind turbine gearbox vibration signal is converted into a time-frequency domain image using a Markov transfer field, and the time-frequency domain image is resampled to obtain a vibration two-dimensional time-frequency image having the same pixels as the two-dimensional image of the wear particle ring; Downsampling the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fusing the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsampling the fusion result; The wear status evaluation result is obtained based on the upsampling result.

[0008] In a second aspect, the present invention provides a wind turbine gearbox wear status assessment system that integrates vibration and oil.

[0009] A wind turbine gearbox wear condition assessment system integrating vibration and oil, comprising: An image acquisition unit is configured to: acquire synchronously collected wind turbine gearbox vibration signals and a two-dimensional image of an abrasive ring for characterizing a change in abrasive concentration; The two-dimensional time-frequency diagram generating unit is configured to: convert the wind turbine gearbox vibration signal into a time-frequency domain image using a Markov transfer field, and resample the time-frequency domain image to obtain a vibration two-dimensional time-frequency diagram having the same pixels as the two-dimensional image of the wear particle ring; The feature fusion unit is configured to: downsample the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fuse the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsample the fusion result; The wear status generating unit is configured to obtain a wear status evaluation result according to the upsampling result.

[0010] In a third aspect, the present invention provides a computer device comprising: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein when the computer program is executed by the processor, the method for evaluating the wear state of a wind turbine gearbox integrating vibration and oil as described in the first aspect of the present invention is implemented.

[0011] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by a processor and executing the wind turbine gearbox wear state assessment method integrating vibration and oil as described in the first aspect of the present invention.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The present invention innovatively provides a wind turbine gearbox wear status assessment method that integrates vibration and oil. It converts the one-dimensional vibration time series signal into a two-dimensional time-frequency image through Markov transition field technology, completely retaining the dynamic evolution characteristics of the signal (such as the state transition probability distribution), and avoiding the information loss caused by traditional manual feature extraction.

[0013] The present invention innovatively provides a wind turbine gearbox wear condition assessment method that integrates vibration and oil, constructs a dual-branch U-Net fusion model, directly fuses two-dimensional time-frequency images and original oil images in the encoder, and uses convolutional neural networks to autonomously mine cross-modal correlation features (such as the coupling law between wear impact energy distribution and wear particle morphology changes), significantly improving the feature characterization capability.

[0014] The present invention innovatively provides a wind turbine gearbox wear status assessment method that integrates vibration and oil. From the input of two-dimensional time-frequency images and original oil images to the output of wear status, no human intervention is required, and it supports real-time deployment of wind turbine on-site online monitoring systems.

[0015] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0017] Figure 1 A flow chart of a method for evaluating the wear condition of a wind turbine gearbox by integrating vibration and oil, provided as an exemplary embodiment of the present invention; Figure 2 A schematic diagram of a two-branch neural network provided for an exemplary embodiment of the present invention; Figure 3 A schematic diagram comparing wear assessment results of different signal confusion matrices provided by an exemplary embodiment of the present invention is shown. Figure 3 (A) is a schematic diagram corresponding to multi-source data fusion. Figure 3 (B) is the schematic diagram corresponding to the vibration signal. Figure 3 (C) is the schematic diagram corresponding to the oil wear particles; Figure 4 A schematic diagram of a wind turbine gearbox wear condition assessment system integrating vibration and oil provided by an exemplary embodiment of the present invention; Figure 5 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0020] As described in the background technology, existing fusion methods often require manual extraction of features of different monitoring information to fuse the feature layer and decision layer, which is greatly affected by human subjectivity, resulting in low diagnostic efficiency and poor accuracy. Compared with the fusion of feature layer and decision layer, data-level fusion does not need to rely on manual feature extraction. It directly processes and comprehensively analyzes vibration signals and oil images in the form of raw data, thereby avoiding feature selection bias caused by subjective factors and mining potential correlations between different data modalities. Through data-level fusion, the details and complex characteristics of the original information can be better preserved, so that the correlation and complementarity between features can be fully mined. In addition, data-level fusion can also achieve deep integration of cross-modal information, provide more comprehensive data input for machine learning models, and further improve the accuracy and robustness of fault mode recognition.

[0021] Some researchers have proposed a fault diagnosis method that combines vibration, oil and noise characteristics, aiming to improve the fault diagnosis capability of wind turbine gearboxes under complex working conditions. This patent uses deep learning and DS evidence theory to fuse the non-stationary characteristics of vibration signals, the sound pressure level and spectral characteristics of noise signals, and the type and particle size distribution characteristics of wear particles in the oil to construct a comprehensive evaluation model for fault conditions. However, the above research relies on multiple detection methods, including sensor information such as vibration, noise and oil, and has not conducted in-depth mining and optimization of a single information source. In particular, there is room for improvement in the precision of data fusion. Some researchers have also proposed a multi-source and multi-feature fusion diagnosis method for aircraft engine main bearing faults, which uses multiple sensors (including vibration sensors, temperature sensors and acoustic sensors) to collect aircraft engine main bearing information. Bearing operating status data is processed through signal denoising, feature extraction and machine learning algorithms to detect and locate mechanical equipment faults. However, this study relies more on algorithm processing after feature extraction in the fusion of monitoring information, and does not involve the underlying fusion strategy of multi-sensor data; some researchers have proposed a method and test bench design for gearbox fault diagnosis using multi-source sensor data fusion technology. This study collects the vibration signal, noise signal, temperature signal, displacement signal and lubricating oil data of the gearbox, and combines the multi-sensor data fusion technology based on BP neural network and DS evidence theory to realize the diagnosis of gearbox faults. However, this study mainly relies on feature-level and decision-level fusion, and does not focus on optimizing the synergy of vibration and lubricating oil monitoring information in data-level fusion.

[0022] In view of this, the present invention proposes a wind turbine gearbox wear state assessment method that integrates vibration and oil, such as Figure 1 As shown, the following process is included: S101: Acquire synchronously collected wind turbine gearbox vibration signals and a two-dimensional image of an abrasive ring for characterizing a change in abrasive concentration; S102: converting the wind turbine gearbox vibration signal into a time-frequency domain image using a Markov transfer field, and resampling the time-frequency domain image to obtain a vibration two-dimensional time-frequency image having the same pixels as the two-dimensional image of the wear particle ring; S103: downsampling the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fusing the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsampling the fusion result; S104: Obtain a wear status assessment result according to the upsampling result.

[0023] In S101 of this implementation, specifically, the following steps are included: Through the full-life wear test of wind turbine gearboxes, combined with vibration online sensors and high-throughput oil wear particle image online monitoring sensors, real-time monitoring of gearboxes under different wear conditions is achieved. Vibration sensors and high-throughput oil wear particle image online monitoring sensors are used to synchronously collect the vibration characteristics and oil wear particle information of gearboxes under different wear conditions in real time, and the collected data are marked according to the actual wear status to form a labeled vibration and oil wear particle dataset.

[0024] In S102 of this implementation, specifically, the following steps are included: In a mechanical system, the concentration of wear particles is proportional to the degree of wear of the friction pair. The higher the concentration of wear particles in the lubricating oil, the more severe the wear. A high-throughput oil wear particle image online monitoring sensor can directly acquire a two-dimensional image of the wear particle ring that represents the change in wear particle concentration. At the same time, the synchronously acquired vibration signal is one-dimensional time series data, which needs to be processed and spatially aligned with the wear particle image data to facilitate the subsequent state evaluation of the intelligent network model. The present invention uses the Markov transfer field (MTF) to convert the time series vibration signal into a time-frequency domain image, and further resamples it to obtain a vibration two-dimensional time-frequency map with the same pixels as the wear particle image. Specifically, it includes: Signal discretization: The vibration signal Further discretized into a finite state sequence , by quantizing the function Generate discrete state values: (1); in, is the discrete interval boundary, is the status value, represents the i-th discrete state value, Represents the i-th vibration signal value.

[0025] Transition probability matrix: Use the Markov chain model to calculate the transition probability between each state , describing the dynamic evolution characteristics of the timing signal: (2); Generate MTF time-frequency domain graph: Generate a Markov transition field image (MTF) containing time domain and frequency domain information by performing a logarithmic transformation on the transition probability matrix. Has the following form: (3); in, is a smoothing factor to avoid zero value problems in logarithmic calculations, the matrix The row index of and column indexes They correspond to the horizontal coordinate (X-axis, representing time) and vertical coordinate (Y-axis, representing frequency) of the MTF image pixels respectively.

[0026] The generated Markov transfer field image (MTF) is in the form of a two-dimensional image, which can effectively show the dynamic change characteristics of the vibration signal, facilitating feature extraction and classification by the subsequent intelligent diagnosis model.

[0027] Time-frequency diagram resampling: In order to ensure that the pixels of the Markov transfer field image (MTF) and the oil wear particle image are the same, the Markov transfer field image (MTF) is resampled using an interpolation algorithm to obtain a vibration two-dimensional time-frequency diagram with the same pixels as the oil wear particle image.

[0028] In this implementation, S103 and S104 specifically include: This paper uses the U-Net network architecture to evaluate the wear condition of wind turbine gearboxes. The U-Net architecture consists of an encoder (downsampling part), jump connections, and a decoder (upsampling part), which can effectively extract and fuse information from different modes of vibration and oil. The model architecture is as follows Figure 2 As shown, the model design principles include: Dual-branch independent downsampling: The vibration and oil branches are each downsampled to 5 layers, and feature fusion is performed at the end. Fusion position: channel splicing after downsampling at the 5th layer (at the minimum resolution); Symmetric decoder: Restores resolution through 5 layers of upsampling, with skip connections from the same layer encoder (i.e., the first downsampled layer is skipped with the first upsampled layer, the second downsampled layer is skipped with the second upsampled layer, the third downsampled layer is skipped with the third upsampled layer, and the fourth downsampled layer is skipped with the fourth upsampled layer).

[0029] More specifically, the input includes data from two channels: vibration and oil: Input 1: , represents the vibration two-dimensional time-frequency graph generated by the vibration signal, H represents the height of the image, W represents the width of the image, and R represents the real number domain; Input 2: , represents the oil wear particle image; The two inputs enter their respective independent encoder branches. The encoder branch includes five downsampling layers, each of which contains a 2D convolution layer (Conv2D), a maximum pooling (MaxPool) layer and other structures.

[0030] No. The convolution block expression of the layer is: (4); Where, For the The downsampling layer outputs the feature map; Represents the convolution operation, which is used to extract oil / vibration signal features, etc. Stands for BatchNorm, which means batch normalization; is the ReLU activation function; and For the The convolution kernel weights and biases of the layer; the pooling operation uses maximum pooling to reduce the size of the feature map.

[0031] The features finally extracted and output by each branch are: (5); (6); Among them, 𝐶 indicates that the number of channels increases layer by layer (for example: 16→32→64→128→256), represents the height of the output feature, represents the width of the output feature, Represents the vibration two-dimensional time-frequency diagram The output features after encoding, Representative image of oil wear particles The output features after encoding.

[0032] After extracting key features of vibration and oil through the dual-branch network, deep feature fusion is performed at the minimum resolution to maximize the preservation of the global statistical correlation between vibration and oil and avoid local noise interference caused by early fusion (such as misidentification of oil bubbles as wear). The fused feature map serves as the input of the decoder. The fused feature is expressed as: (7); The decoder performs upsampling and classification reconstruction on the fused feature map. The decoder uses transposed convolution and jump connection to reconstruct the spatial features. The spatial resolution of the oil / vibration image is gradually restored through the deconvolution layer to learn the correlation between the oil / vibration features and the wear state. The deconvolution operation of the layer is: (8); Where, is the transposed convolution, represents transposed convolution, For the The convolution kernel weights and biases of the layer are adjusted, the pooling operation uses maximum pooling to reduce the feature map size, and the output size is restored layer by layer (e.g., 256 → 128 → 64 → 32 → 16).

[0033] The features of the same layer in the encoder are concatenated with the deconvolution results of the current layer of the decoder to enhance the edge and spatial semantic information, which can be expressed as: (9); The output layer adopts a four-level progressive feature compression and classification architecture, and the method flow is as follows: The spatial aggregation layer (feature compression) performs global spatial aggregation on the three-dimensional feature tensor output by the decoder to generate a channel feature vector: (10); Where, Indicates location Place The characteristic value of each channel.

[0034] The fully connected layer (Dense) maps the feature vector processed by the spatial aggregation layer into a hidden representation space through linear transformation, which is used to further learn the association between "wear state" and "image features". It can be expressed as: (11); in, Output vector for the spatial aggregation layer; and is the weight and bias of the fully connected layer; In the wind turbine gearbox wear data scenario, the Dropout layer is used to effectively alleviate the overfitting problem of the model in the training set and improve the generalization ability of identifying different wear stages of wind turbine gearboxes. The process can be expressed as: (12); in, is the drop probability.

[0035] The Softmax layer converts the real number vector output by the fully connected layer (Dense) into a multi-class probability distribution, where each value represents the probability that the input sample belongs to a certain class (wear degree), which can be expressed as: (13); in, Belong to probability of a class (e.g., severe wear); Indicates the number of categories for classification, which is 4 in this invention (running-in, mild, abnormal, severe). Representative The result of the Dropout layer processing of the wear-like state, , Representative The result of the Dropout layer processing of the wear-like state.

[0036] The model training process constructed by the present invention adopts the cross entropy loss function to learn the characteristic distribution of the gearbox's running-in, mild, abnormal, and severe wear states; (14); in, Encode the true label.

[0037] In order to enable the network to find the optimal solution for wear state identification more quickly, the present invention uses Adam for update optimization, which can be expressed as: (15); Where, For the The weight of the step; represents the learning rate; 、 They are the first-order momentum (mean) and second-order momentum (square mean) of the gradient; Represents a very small constant and prevents division by zero.

[0038] After model training is complete, the model's prediction accuracy is evaluated using a validation set of wind turbine gearbox data. Key evaluation metrics include wear feature classification accuracy and the confusion matrix. These metrics ensure the model's accuracy for different gearbox wear stages during actual application, optimize model performance, and provide a basis for wear status classification.

[0039] The wind turbine gearbox wear state assessment method based on multi-source information fusion analysis of oil wear particles and vibration is described in detail. Figure 1 , including the collection of gearbox vibration and oil wear particle signals, signal preprocessing, the construction of intelligent model network, and health status assessment of sample data.

[0040] First, a full-lifecycle wear test of a wind turbine gearbox test bench was conducted. Accelerometer vibration sensors and a high-throughput oil wear particle image online monitoring sensor were used to collect full-lifecycle monitoring signals. The experiment lasted 284 hours, during which vibration and oil online monitoring data were automatically and synchronously collected at 20-minute intervals, resulting in a total of 851 sets of data. During the monitoring process, the vibration signal sampling frequency was 2.56 kHz and the sampling duration was 3 seconds. Oil wear particle ring images were directly generated using the CMOS1 imaging system of the high-throughput oil wear particle image online monitoring sensor. These data cover the complete gear system process, from the run-in period, normal wear, abnormal wear, to severe wear.

[0041] Secondly, based on the sampling frequency and time of the vibration signal, the present invention calculates that each vibration signal generates 7,680 data points within 3 seconds. These 7,680 data points (sampling frequency 2.56 kHz) collected within 3 seconds are discretized into five state intervals (silhouette coefficient 0.81) using K-means clustering, generating a 5×5 Markov transition matrix. To preserve the state transition probability characteristics, a two-step interpolation method is employed: first, bicubic interpolation is used to expand the matrix to 16×16, and then pixel replication is used to scale it up to 64×64 to match the oil image.

[0042] Next, the present invention categorized the 851 processed vibration time-frequency and wear particle map samples into four wear status categories based on the actual wear state of the gearbox. The wear stage was divided into 16 groups: running-in period, 644 groups: mild wear, 116 groups: abnormal wear, and 72 groups: severe wear. To mitigate class imbalance, stratified sampling (75% of each category) was used for the training set: 12 groups: running-in period, 483 groups: mild wear, 87 groups: abnormal wear, and 54 groups: severe wear, for a total of 596 groups. The remaining 255 data sets were used as the test set: running-in period, mild wear, 161 groups: mild wear, 29 groups: abnormal wear, and 18 groups: severe wear, to verify the accuracy and robustness of the proposed method.

[0043] Finally, a dual-channel encoder design (vibration branch / oil branch) is adopted, each branch contains 5 levels of feature extraction layers; each downsampling layer adopts a composite structure of "3×3 convolution → batch normalization → ReLU → 2×2 maximum pooling"; the number of channels increases exponentially (16→32→64→128→256), ensuring the depth of feature extraction while controlling the computational complexity.

[0044] After downsampling at the fifth layer (feature map size 2×2×256), feature concatenation is performed; 1×1 convolution is used for cross-channel information exchange, reducing the concatenated 512-dimensional features to 256 dimensions. The decoder gradually restores the resolution through five layers of transposed convolutions and performs skip connections with the encoder features at the same layer, ultimately outputting a 64×64×16 reconstructed feature.

[0045] The model training uses the Adam optimizer, and the initial learning rate is set to , momentum parameter 、 The cross entropy loss function is used as the loss function to accurately distinguish the four states of running-in, mild, abnormal and severe wear. To prevent overfitting, Dropout regularization (dropout rate 20%) is used during training and an early stopping mechanism is set: when the validation set loss decreases by less than 10 epochs in a row, the training process stops. The training is automatically terminated when the maximum training cycle is limited to 500 times.

[0046] During the model evaluation phase, 255 sets of independent test samples were reserved for performance verification. Quantitative metrics such as overall classification accuracy, precision and recall for each category were calculated, and the confusion matrix was used to comprehensively evaluate the model's recognition capabilities at different wear stages.

[0047] In order to highlight the advantages of the multi-source data fusion method proposed in this invention, the performance of the single-modal model using only vibration signals or oil wear particle images was tested simultaneously. Figure 3 As shown in the confusion matrix, the horizontal axis is the actual wear state of the gearbox, and the vertical axis is the wear state evaluated by the model. Figure 3 (A) is a schematic diagram of multi-source data fusion (accuracy 97%). Figure 3 (B) is the schematic diagram corresponding to the vibration signal (accuracy rate 71%). Figure 3 (C) is a schematic diagram corresponding to oil wear particles (accuracy rate 79%). Figure 3 The number in each grid in the table represents how many times the system evaluated the state as another state when the actual state was a certain state. The numbers on the "diagonal" in the table represent the number of times the system evaluated correctly, and the numbers in other grids represent the number of times the evaluation was wrong.

[0048] Experimental results show that the dual-branch fusion model of the present invention achieves a recognition accuracy of 97% on the test set, which is significantly improved compared with the single-modal model method (about 71% for vibration and about 79% for oil), verifying the effectiveness of the data-level fusion strategy.

[0049] Experimental results show that the method proposed in this invention has better recognition accuracy and robustness than traditional single oil and single vibration signal analysis methods. Its successful application will provide more reliable and efficient technical support for the health monitoring and fault diagnosis of mechanical equipment.

[0050] Figure 4 A wind turbine gearbox wear condition assessment system integrating vibration and oil is shown, comprising: The image acquisition unit 401 is configured to: acquire synchronously collected wind turbine gearbox vibration signals and a two-dimensional image of an abrasive ring for characterizing a change in abrasive concentration; The two-dimensional time-frequency diagram generating unit 402 is configured to: convert the wind turbine gearbox vibration signal into a time-frequency domain image using a Markov transfer field, and resample the time-frequency domain image to obtain a vibration two-dimensional time-frequency diagram having the same pixels as the wear particle ring two-dimensional image; The feature fusion unit 403 is configured to: downsample the wear particle ring two-dimensional image and the vibration two-dimensional time-frequency map respectively, fuse the downsampled wear particle ring two-dimensional image and the downsampled vibration two-dimensional time-frequency map, and upsample the fusion result; The wear state generating unit 404 is configured to: obtain a wear state evaluation result according to the upsampling result It is understandable that each of the above-mentioned units can be separately or completely combined into one or several other units to form a unit, or one (or some) of the units can be further divided into multiple functionally smaller units to form a unit, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In actual applications, the functions of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present application, the system may also include other units. In actual applications, these functions can also be implemented with the assistance of other units and can be implemented by the collaboration of multiple units.

[0051] According to another embodiment of the present application, the system described in this embodiment can be constructed by running a computer program (including program code) capable of executing the steps involved in the corresponding method of the present invention on a general-purpose computing device such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above-mentioned computing device through the computer-readable recording medium and run therein.

[0052] Figure 5 An electronic device is shown, which includes a processor 501, a communication interface 502, and a computer-readable storage medium 503. The processor 501, the communication interface 502, and the computer-readable storage medium 503 may be connected via a bus or other means.

[0053] Among them, the communication interface 502 is used to receive and send data, the computer-readable storage medium 503 can be stored in the memory of the electronic device, the computer-readable storage medium 503 is used to store computer programs, the computer programs include program instructions, and the processor 501 is used to execute the program instructions stored in the computer-readable storage medium 503.

[0054] The processor 501 is the computing core and control core of the electronic device, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions to implement corresponding method processes or corresponding functions.

[0055] The processor 501 is configured to perform the following process: Acquire synchronously collected wind turbine gearbox vibration signals and abrasive ring two-dimensional images used to characterize abrasive concentration changes; The vibration signal of the wind turbine gearbox is converted into a time-frequency domain image by using a Markov transfer field, and the time-frequency domain image is resampled to obtain a vibration two-dimensional time-frequency map having the same pixels as the two-dimensional image of the wear particle ring; downsampling the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fusing the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsampling the fusion result; The wear status evaluation result is obtained based on the upsampling result.

[0056] The present invention also provides a computer-readable storage medium, which is a memory device in an electronic device for storing programs and data. It is understood that the computer-readable storage medium herein may include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The computer-readable storage medium provides storage space that stores the processing system of the electronic device.

[0057] Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by the processor. These instructions may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium herein may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device; alternatively, it may be at least one computer-readable storage medium located remotely from the processor.

[0058] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to implement the following process: Acquire synchronously collected wind turbine gearbox vibration signals and abrasive ring two-dimensional images used to characterize abrasive concentration changes; The vibration signal of the wind turbine gearbox is converted into a time-frequency domain image by using a Markov transfer field, and the time-frequency domain image is resampled to obtain a vibration two-dimensional time-frequency map having the same pixels as the two-dimensional image of the wear particle ring; downsampling the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fusing the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsampling the fusion result; The wear status evaluation result is obtained based on the upsampling result.

[0059] The present invention also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the following process: Acquire synchronously collected wind turbine gearbox vibration signals and abrasive ring two-dimensional images used to characterize abrasive concentration changes; The wind turbine gearbox vibration signal is converted into a time-frequency domain image using a Markov transfer field, and the time-frequency domain image is resampled to obtain a vibration two-dimensional time-frequency image having the same pixels as the two-dimensional image of the wear particle ring; Downsampling the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fusing the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsampling the fusion result; The wear status evaluation result is obtained based on the upsampling result.

[0060] Those skilled in the art will appreciate that the units and algorithmic steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0061] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. A computer program product comprises one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data processing device such as a server or data center that integrates one or more available media. Available media can include magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives).

[0062] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A wind turbine gearbox wear condition assessment method integrating vibration and oil, characterized in that: The following processes are included: Acquire synchronously collected wind turbine gearbox vibration signals and abrasive ring two-dimensional images used to characterize abrasive concentration changes; The wind turbine gearbox vibration signal is converted into a time-frequency domain image using a Markov transfer field, and the time-frequency domain image is resampled to obtain a vibration two-dimensional time-frequency image having the same pixels as the two-dimensional image of the wear particle ring; Downsampling the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fusing the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsampling the fusion result; The wear status evaluation result is obtained based on the upsampling result.

2. The wind turbine gearbox wear condition assessment method integrating vibration and oil as claimed in claim 1, characterized in that: The Markov transfer field is used to convert the wind turbine gearbox vibration signal into a time-frequency domain image, including: Discretizing the wind turbine gearbox vibration signal into a finite state sequence, and generating discrete state values ​​through a quantization function; According to the discrete state values, a Markov chain model is used to calculate the transition probability between each state value; Performing logarithmic transformation on the transition probability to generate a Markov transition field image containing time domain and frequency domain information.

3. The wind turbine gearbox wear condition assessment method integrating vibration and oil as claimed in claim 2, characterized in that: Use the Markov chain model to calculate the transition probability between each state value, including: ； in, The first in the sequence of finite states A signal, The first in the sequence of finite states A signal, Representative discrete state values, Representative discrete state values, is the counting function.

4. The wind turbine gearbox wear condition assessment method integrating vibration and oil as claimed in claim 3 is characterized in that: Perform a logarithmic transformation on the transition probability matrix, including: ； in, is the smoothing factor, Represents the logarithmic transformation result of the i-th discrete state value and the j-th discrete state value.

5. The wind turbine gearbox wear condition assessment method integrating vibration and oil as described in any one of claims 1 to 4, characterized in that: The wear status assessment results are obtained based on the upsampling results, including: The spatial aggregation layer aggregates the upsampling results globally to generate channel feature vectors; The fully connected layer maps the channel feature vector to the hidden representation space through linear transformation; The Dropout layer processes the hidden representation space; The Softmax layer performs probability calculation on the processing results of the Dropout layer to obtain the probability of each type of wear state, and the wear state with the highest probability is taken as the final wear state evaluation result.

6. The wind turbine gearbox wear condition assessment method integrating vibration and oil as claimed in claim 5, characterized in that: The spatial aggregation layer aggregates the upsampling results globally to generate channel feature vectors, including: ； in, Represents the height and width of the vibration two-dimensional time-frequency diagram, is the number of channels, Indicates location Place The characteristic value of each channel; The fully connected layer maps the channel feature vector to the hidden representation space through linear transformation ,include: ； in, and represents the fully connected layer weights and biases, represents the activation function; The Dropout layer processes the hidden representation space, including: ； in, is the drop probability.

7. The wind turbine gearbox wear condition assessment method integrating vibration and oil as claimed in claim 6, characterized in that: The Softmax layer performs probability calculations on the processing results of the Dropout layer, including: ； in, The probability of belonging to the kth wear state, The number of categories representing the wear status, Representative The result of the Dropout layer processing of the wear-like state, , Representative The result of the Dropout layer processing of the wear-like state.

8. A wind turbine gearbox wear condition assessment system integrating vibration and oil, characterized in that: include: An image acquisition unit is configured to: acquire synchronously collected wind turbine gearbox vibration signals and a two-dimensional image of an abrasive ring for characterizing a change in abrasive concentration; The two-dimensional time-frequency diagram generating unit is configured to: convert the wind turbine gearbox vibration signal into a time-frequency domain image using a Markov transfer field, and resample the time-frequency domain image to obtain a vibration two-dimensional time-frequency diagram having the same pixels as the two-dimensional image of the wear particle ring; The feature fusion unit is configured to: downsample the two-dimensional wear particle ring image and the two-dimensional vibration time-frequency map respectively, fuse the downsampled two-dimensional wear particle ring image and the downsampled two-dimensional vibration time-frequency map, and upsample the fusion result; The wear status generating unit is configured to obtain a wear status evaluation result according to the upsampling result.

9. A computer device, characterized in that: include: a processor and a computer-readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium having a computer program stored therein, wherein when the computer program is executed by the processor, the method for evaluating the wear state of a wind turbine gearbox integrating vibration and oil as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the wind turbine gearbox wear condition assessment method integrating vibration and oil according to any one of claims 1 to 7.

Images