Oil abrasive particle dynamic monitoring method and system based on deep learning

Abstract

The invention discloses an oil abrasive particle dynamic monitoring method and system based on deep learning, and belongs to the technical field of oil abrasive particle dynamic monitoring. The method comprises the following steps: S1, collecting an oil sample set in real time through a plurality of sensors arranged in an equipment lubrication system; s2, extracting an abrasive particle image in the oil sample set; s3, constructing a double-layer generative adversarial network model; s4, deploying the trained double-layer generative adversarial network model in a monitoring system, collecting and inputting an abrasive particle image in real time, and generating wear feature data; s5, calculating a wear state index of the equipment based on the identified wear feature data, and judging whether the equipment is in an abnormal state or not; s6, if it is detected that the abrasion state index exceeds a set threshold value, the system automatically triggers an early warning mechanism and generates early warning information to inform related personnel to take maintenance measures; and S7, storing a monitoring result in a database. According to the invention, accurate analysis and classification of oil abrasive particle data are realized, and the real-time monitoring and early warning capabilities of the system are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of dynamic monitoring of oil wear particles, and in particular relates to a method and system for dynamic monitoring of oil wear particles based on deep learning. Background Art

[0002] In existing technologies, oil wear particle monitoring mainly relies on wear particle detection and analysis based on traditional equipment such as optical microscopes and laser particle size analyzers. Traditional methods require manual intervention and collect oil samples for offline analysis to determine the lubrication status and wear of the equipment. This method can monitor equipment wear to a certain extent, but it has limitations in many aspects.

[0003] Traditional oil wear particle monitoring methods mostly require manual operation, have low data processing efficiency, and have long monitoring cycles. They make it difficult to achieve real-time equipment status monitoring. Equipment wear problems cannot be discovered in the early stages, which may result in maintenance measures being taken only when serious equipment failures occur, resulting in unnecessary downtime and economic losses.

[0004] In existing technologies, the detection of oil wear particles is mostly based on simple particle counting or size measurement methods. These technical means cannot fully capture the key information of the shape, composition, and type of the wear particles, which are the key indicators of the equipment wear characteristics. Traditional methods are difficult to fully and accurately characterize the wear status of the equipment. In addition, due to the complex morphology and chemical composition of the wear particles, traditional detection methods lack sufficient intelligent means to automatically analyze and classify complex data, which limits their application scope and detection accuracy.

[0005] Traditional oil wear particle monitoring systems cannot be automated and intelligent. Existing monitoring systems often rely on a single data collection and analysis method, lack processing capabilities based on big data and artificial intelligence, and are unable to cope with complex and changing industrial environments. In actual applications, as the equipment operating time increases and working conditions change, the wear pattern of the equipment often changes dynamically. Existing monitoring methods lack adaptability and responsiveness to dynamic changes.

[0006] In summary, the existing technologies have the following main shortcomings: First, traditional oil wear particle monitoring methods cannot achieve real-time monitoring, making it difficult to provide early warning of equipment wear problems; second, existing technologies cannot fully capture the multi-dimensional characteristics of wear particles, and the detection accuracy is limited; third, existing systems lack intelligent and automated processing capabilities, making them difficult to adapt to complex industrial application environments. Summary of the Invention

[0007] One purpose of the present invention is to propose a method and system for dynamic monitoring of oil wear particles based on deep learning. The present invention realizes accurate analysis and classification of oil wear particle data, greatly improving the real-time monitoring and early warning capabilities of the system.

[0008] A method for dynamic monitoring of oil wear particles based on deep learning according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect oil samples in real time through multiple sensors installed in the equipment lubrication system and transmit the oil samples to the data processing module;

[0010] S2. Preprocessing the oil sample set transmitted to the data processing module and extracting wear particle images from the oil sample set;

[0011] S3. Build a two-layer generative adversarial network model, including a first-layer generator, a second-layer generator, and a discriminator.

[0012] S4. Deploy the trained two-layer GAN model in the monitoring system, collect and input wear particle images in real time, and use the two-layer GAN model to automatically identify the type, size, and quantity of wear particles and generate wear characteristic data.

[0013] S5. Based on the identified wear characteristic data, the wear status index of the device is calculated and compared with the set threshold to determine whether the device is in an abnormal state. If the wear status index exceeds the set threshold, the process proceeds to S6. If the wear status index does not exceed the set threshold, the process proceeds to S7.

[0014] S6. Automatically trigger the early warning mechanism and generate early warning information to notify relevant personnel to take maintenance measures;

[0015] S7. The monitoring results are stored in a database, and the wear particle characteristic change trend and equipment operation status are displayed in a graphical form through a data visualization module.

[0016] Optionally, the S1 specifically includes:

[0017] S11. A sensor array is arranged at key nodes of the equipment lubrication system. The sensor array includes n sensors T1, T2, ..., T n Each sensor collects oil wear particle data in the oil sample in real time, including the volume flow rate Q of the oil i (t), number of abrasive particles Q i (t), abrasive particle size D i (t) and the chemical composition of the abrasive particles C i (t), where i is the sensor number and t is the time;

[0018] Sl2 the oil wear data collected by the n sensors is transmitted to the data processing module through the data transmission module;

[0019] S13. Perform preliminary formatting on the collected oil wear particle data to form an oil sample set:

[0020] D={(Q i (t), N i (t), D i (t), C i (t))|i=1,2,...,n;t=t1,t2,...,t m};

[0021] Where D represents all the oil wear particle data collected at different time points t and multiple sensor positions i, and m is the number of time points.

[0022] Optionally, the S2 specifically includes:

[0023] S21. De-noise the wear particle data in the oil sample set D transmitted to the data processing module, and use a bandpass filter to eliminate random noise and system noise in the sampling process to obtain the de-noised oil sample set D. denoise ;

[0024] S22. For the denoised oil sample set D denoise Standardization is performed by j And each sensor T i Abrasive grain size D i (t j ), number of abrasive particles N i (t j ) and abrasive chemical composition C i (t j ) to obtain the standardized oil sample set D norm ;

[0025] S23. Oil sample set D based on standardization norm Extract the wear particle image, and the wear particle size collected by each sensor Ti and number of abrasive particles Through the two-dimensional Laplacian operator Perform image gradient enhancement:

[0026]

[0027] in, is the Laplace edge detection result, Indicates that at time t j Sensor T i The collected standardized wear particle size, x and y are the spatial coordinates of the image;

[0028] S24. For each time point t j, wear particle image based on the detected edge information Energy function Optimize and then perform wear area segmentation:

[0029]

[0030] Among them, Ω i is the area of wear particles in the image, α and β are weight parameters, is the image gradient, indicating the strength of the edge, Represents chemical composition data, is the weight function related to chemical composition;

[0031] S25. Based on energy function Based on the minimization result, a fast matching algorithm is used to mark and classify each segmented area to generate the final wear particle image matrix:

[0032]

[0033] Among them, M i (t j ) is the matrix representation of the wear particle image, Label is the region labeling function, which means classifying the optimal segmentation region and finally outputs the labeled and classified wear particle image.

[0034] Optionally, the S3 specifically includes:

[0035] S31. Constructing the visual transformer model Vi(t based on the preprocessed wear particle image matrix j ), the wear particle image matrix is divided into several image blocks Pk, where k is the index of the image block, the size of each image block is p×p, and the embedded feature vector is obtained by embedding each image block The global dependencies between image patches are extracted through the self-attention mechanism, and the transformer model representation containing the global features of the wear particle image is output:

[0036]

[0037] S32. Construct a two-layer generative adversarial network model, including the first-layer generator G1 and the second-layer generator G2 and a discriminator D:

[0038] The first layer generator G1 generates wear particle images M of different types and sizes by inputting noise vector z1~N(0,1) gen1 (t j )=G1(z1), preliminary feature generation is performed on the generated wear particle images of different types and sizes through convolutional neural networks;

[0039] The second layer generator G2 generates the wear particle image M output by the first layer generator G1. gen1 (t j ) is used to generate the wear particle image moment M with more detailed features. gen2 (t j )=G2(M gen1 (t j )), so that the generated wear particle image matrix is close to the real wear particle image matrix;

[0040] The wear particle image matrix M generated by the discriminator D is input gen2 (t j ) and the real collected wear particle image matrix M i (t j ) The authenticity of the two is judged through adversarial training:

[0041]

[0042] in, is the loss function of the discriminator, is the expected operation, D(·) represents the classification probability output of the discriminator for the input image;

[0043] S33. Using the pre-labeled wear particle image dataset D label For the visual transformer model Vi(t j ) and the discriminator D are jointly trained to optimize the feature extraction capability and wear particle feature classification accuracy of the two-layer generative adversarial network model:

[0044]

[0045] in, is the classification loss function, and λ is the loss weight factor.

[0046] Optionally, the S4 specifically includes:

[0047] S41. Deploy the trained two-layer GAN model in the central data processing module of the monitoring system. The monitoring system uses a sensor array to collect a wear particle image matrix from the oil sample in real time and inputs the wear particle image matrix into the two-layer GAN model.

[0048] S42. The first layer generator is based on the collected wear particle image matrix M i (t j ) Generate preliminary wear particle image matrix features F gen1 (t j )=G1(M i (t j )), the generated wear particle image matrix M i (t j) includes initial grit size, grit shape and grit quantity;

[0049] S43. The second layer generator generates detail enhanced wear particle image matrix features F based on the preliminary wear particle image matrix features output by the first layer generator gen2 (t j )=G2(F gen1 (t j )), so that the enhanced wear particle image matrix is close to the real wear particle image matrix characteristics;

[0050] S44. Input the generated wear particle image matrix features and the real collected wear particle image matrix into the discriminator, and the discriminator judges the authenticity and classification of the wear particle image matrix and generates a classification result C i (t j );

[0051] S45. Based on classification results C i (t j ), generate wear feature data W i (t j ), including abrasive type, abrasive size, abrasive quantity and abrasive chemical composition.

[0052] Optionally, the S5 specifically includes:

[0053] S51. Based on the generated wear characteristic data W i (t j ) Calculate the wear status index of the equipment The wear state index is defined by combining the abrasive type, abrasive size, abrasive number and abrasive chemical composition:

[0054]

[0055] Among them, α1, α2, and α3 are weight parameters. is the number of normalized abrasive particles, is the standardized abrasive grain size;

[0056] S52, the calculated wear state index With the set threshold T w The comparison is performed, and the threshold is set based on the historical wear data and working conditions of the equipment to determine whether the equipment is in an abnormal state. The judgment conditions are:

[0057]

[0058] Among them, Status(t j ) indicates that the device is at time t j The state at time T w is the set threshold.

[0059] A deep learning-based dynamic monitoring system for oil wear particles includes the following modules:

[0060] The sensor module is used to collect real-time data on oil wear particles in the equipment lubrication system, including the number, size, shape and chemical composition of wear particles;

[0061] a data processing module configured to receive the oil wear particle data collected by the sensor module, and perform pre-processing, standardization, and denoising on the oil wear particle data;

[0062] A two-layer generative adversarial network model, including a first-layer generator, a second-layer generator, and a discriminator, generates and discriminates wear particle image features and classifies them to generate wear particle feature data;

[0063] The wear state assessment module calculates the wear state index of the equipment based on the wear particle characteristic data and determines whether the equipment is in an abnormal state;

[0064] The early warning module automatically generates abnormal warning information and notifies maintenance personnel when abnormal equipment wear is detected;

[0065] The data storage and visualization module stores monitoring data and displays the equipment operating status and wear particle characteristic change trends through visualization tools.

[0066] The beneficial effects of the present invention are:

[0067] (1) The present invention introduces a visual transformer model in wear particle image processing and uses a self-attention mechanism to realize the global feature extraction of wear particle images. Different from the traditional local image analysis method, the visual transformer divides the wear particle image into several image blocks and embeds each image block, and then extracts the global dependency of the wear particle image through the self-attention mechanism. It can capture the multi-dimensional features of the wear particles more comprehensively and accurately, including the shape, size and surface details of the wear particles.

[0068] (2) The present invention adopts a two-layer generative adversarial network model, which includes two generators and a discriminator. The first-layer generator generates a preliminary wear particle image, and the second-layer generator enhances the details of the wear particle image based on the first-layer generation, making it closer to the real wear particle image. Through adversarial training, the discriminator can effectively distinguish between real and generated wear particle images, thereby improving the system's ability to recognize wear particles in complex industrial environments. Compared with traditional rule-based recognition methods, the introduction of a two-layer generative adversarial network not only improves the detail fidelity of the wear particle image, but also optimizes the accuracy and reliability of wear particle feature classification.

[0069] (3) The present invention realizes real-time monitoring of the lubrication status and wear condition of the equipment by deploying the trained two-layer generative adversarial network model in the monitoring system. The system can dynamically calculate the wear status index of the equipment based on the type, size, quantity and chemical composition characteristics of the abrasive particles, and compare it with the set threshold in real time. Once the threshold is exceeded, the system will automatically trigger the early warning mechanism to notify the relevant maintenance personnel to take timely maintenance measures, thereby improving the real-time and automation level of the monitoring system and avoiding economic losses caused by equipment downtime due to failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0071] Figure 1 This is a flow chart of a method and system for dynamic monitoring of oil wear particles based on deep learning proposed by the present invention. DETAILED DESCRIPTION

[0072] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0073] Example 1: Reference Figure 1 A method for dynamic monitoring of oil wear particles based on deep learning includes the following steps:

[0074] S1. Collect oil samples in real time through multiple sensors installed in the equipment lubrication system and transmit the oil samples to the data processing module;

[0075] S2. Preprocessing the oil sample set transmitted to the data processing module and extracting wear particle images from the oil sample set;

[0076] S3. Build a two-layer generative adversarial network model, including a first-layer generator, a second-layer generator, and a discriminator.

[0077] S4. Deploy the trained two-layer GAN model in the monitoring system, collect and input wear particle images in real time, and use the two-layer GAN model to automatically identify the type, size, and quantity of wear particles and generate wear characteristic data.

[0078] S5. Based on the identified wear characteristic data, calculate the wear status index of the device and compare it with the set threshold to determine whether the device is in an abnormal state. If the wear status index is detected to exceed the set threshold, proceed to S6. If the wear status index is detected to not exceed the set threshold, proceed to S7.

[0079] S6. The system automatically triggers the early warning mechanism and generates early warning information to notify relevant personnel to take maintenance measures;

[0080] S7. The monitoring results are stored in a database, and the wear particle characteristic change trend and equipment operation status are displayed in a graphical form through a data visualization module.

[0081] In this embodiment, S1 specifically includes:

[0082] S11. A sensor array is arranged at key nodes of the equipment lubrication system. The sensor array includes n sensors T1, T2, ..., T n Each sensor collects oil wear particle data in the oil sample in real time, including the volume flow rate Q of the oil i (t), number of abrasive particles Q i (t), abrasive particle size D i (t) and the chemical composition of the abrasive particles C i (t), where i is the sensor number and t is the time;

[0083] S12. The oil wear particle data collected by n sensors is transmitted to the data processing module through the data transmission module;

[0084] S13. Perform preliminary formatting on the collected oil wear particle data to form an oil sample set:

[0085] D={(Q i (t), N i (t), D i (t), C i (t))|i=1,2,...,n;t=t1,t2,...,t m};

[0086] Where D represents all the oil wear particle data collected at different time points t and multiple sensor positions i, and m is the number of time points.

[0087] In this embodiment, S2 specifically includes:

[0088] S21. De-noise the wear particle data in the oil sample set D transmitted to the data processing module, and use a bandpass filter to eliminate random noise and system noise in the sampling process to obtain the de-noised oil sample set D. denoise ;

[0089] S22. For the denoised oil sample set D denoise Standardization is performed by j And each sensor T i Abrasive grain size D i (tj ), number of abrasive particles N i (t j ) and abrasive chemical composition C i (t j ) to obtain the standardized oil sample set D norm ;

[0090] S23. Oil sample set D based on standardization norm Extract the wear particle image, and the wear particle size collected by each sensor Ti and number of abrasive particles Through the two-dimensional Laplacian operator Perform image gradient enhancement:

[0091]

[0092] in, is the Laplace edge detection result, Indicates that at time t j Sensor T i The collected standardized wear particle size, x and y are the spatial coordinates of the image;

[0093] S24. For each time point t j , wear particle image based on the detected edge information Energy function Optimize and then perform wear area segmentation:

[0094]

[0095] Among them, Ω i is the area of wear particles in the image, α and β are weight parameters, is the image gradient, indicating the strength of the edge, Represents chemical composition data, is the weight function related to chemical composition;

[0096] S25. Based on energy function Based on the minimization result, a fast matching algorithm is used to mark and classify each segmented area to generate the final wear particle image matrix:

[0097]

[0098] Among them, M i (t j ) is the matrix representation of the wear particle image, Label is the region labeling function, which means classifying the optimal segmentation region and finally outputs the labeled and classified wear particle image.

[0099] In this embodiment, S3 specifically includes:

[0100] S31. Constructing the visual transformer model Vi(t based on the preprocessed wear particle image matrix j ), the wear particle image matrix is divided into several image blocks Pk, where k is the index of the image block and the size of each image block is p×p. Each image block is embedded to obtain the embedded feature vector εk. The global dependency between image blocks is extracted through the self-attention mechanism, and the transformer model representation containing the global features of the wear particle image is output:

[0101] Vi(t j ) = Attention(εk);

[0102] S32. Construct a two-layer generative adversarial network model, including the first-layer generator G1 and the second-layer generator G2 and a discriminator D:

[0103] The first layer generator G1 generates wear particle images M of different types and sizes by inputting noise vector z1~N(0,1) gen1 (t j )=G1(z1), preliminary feature generation is performed on the generated wear particle images of different types and sizes through convolutional neural networks;

[0104] The second layer generator G2 generates the wear particle image M output by the first layer generator G1. gen1 (t j ) is used to generate the wear particle image moment M with more detailed features. gen2 (t j )=G2(M gen1 (t j )), so that the generated wear particle image matrix is close to the real wear particle image matrix;

[0105] The wear particle image matrix M generated by the discriminator D is input gen2 (t j ) and the real collected wear particle image matrix M i (t j ) The authenticity of the two is judged through adversarial training:

[0106]

[0107] in, is the loss function of the discriminator, is the expected operation, D(·) represents the classification probability output of the discriminator for the input image;

[0108] S33. Using the pre-labeled wear particle image dataset D label For the visual transformer model Vi(t j) and the discriminator D are jointly trained to optimize the feature extraction capability and wear particle feature classification accuracy of the two-layer generative adversarial network model:

[0109]

[0110] in, is the classification loss function, and λ is the loss weight factor.

[0111] In this embodiment, S4 specifically includes:

[0112] S41. Deploy the trained two-layer GAN model in the central data processing module of the monitoring system. The monitoring system uses a sensor array to collect a wear particle image matrix from the oil sample in real time and inputs the wear particle image matrix into the two-layer GAN model.

[0113] S42. The first layer generator is based on the collected wear particle image matrix M i (t j ) Generate preliminary wear particle image matrix features F gen1 (t j )=G1(M i (t j )), the generated wear particle image matrix M i (t j ) includes initial grit size, grit shape and grit quantity;

[0114] S43. The second layer generator generates detail enhanced wear particle image matrix features F based on the preliminary wear particle image matrix features output by the first layer generator gen2 (t j )=G2(F gen1 (t j )), so that the enhanced wear particle image matrix is close to the real wear particle image matrix characteristics;

[0115] S44. Input the generated wear particle image matrix features and the real collected wear particle image matrix into the discriminator, and the discriminator judges the authenticity and classification of the wear particle image matrix and generates a classification result C i (t j );

[0116] S45. Based on classification results C i (t j ), generate wear feature data W i (t j ), including abrasive type, abrasive size, abrasive quantity and abrasive chemical composition.

[0117] In this embodiment, S5 specifically includes:

[0118] S51. Based on the generated wear characteristic data W i (t j ) Calculate the wear status index of the equipment The wear state index is defined by combining the abrasive type, abrasive size, abrasive number and abrasive chemical composition:

[0119]

[0120] Among them, α1, α2, and α3 are weight parameters. is the number of normalized abrasive particles, is the standardized abrasive grain size;

[0121] S52, the calculated wear state index With the set threshold T w The comparison is performed, and the threshold is set based on the historical wear data and working conditions of the equipment to determine whether the equipment is in an abnormal state. The judgment conditions are:

[0122]

[0123] Among them, Status(t j ) indicates that the device is at time t j The state at time T w is the set threshold.

[0124] A deep learning-based dynamic monitoring system for oil wear particles includes the following modules:

[0125] The sensor module is used to collect real-time data on oil wear particles in the equipment lubrication system, including the number, size, shape and chemical composition of wear particles;

[0126] a data processing module configured to receive the oil wear particle data collected by the sensor module, and perform pre-processing, standardization, and denoising on the oil wear particle data;

[0127] A two-layer generative adversarial network model, including a first-layer generator, a second-layer generator, and a discriminator, generates and discriminates wear particle image features and classifies them to generate wear particle feature data;

[0128] The wear state assessment module calculates the equipment's wear state index based on the wear particle characteristic data and determines whether the equipment is in an abnormal state;

[0129] The early warning module automatically generates abnormal warning information and notifies maintenance personnel when abnormal equipment wear is detected;

[0130] The data storage and visualization module stores monitoring data and displays the equipment operating status and wear particle characteristic change trends through visualization tools.

[0131] Example 2: Example 2 is the application of the present invention in wind power generation equipment. In a coastal wind farm, hundreds of wind turbines operate under harsh climatic conditions every day. The moisture, salt spray and high wind speed in the environment have a continuous impact on the lubrication system of the equipment. Monitoring abrasive particles in the lubricating oil is the key to ensuring the normal operation of wind turbines. However, traditional abrasive particle monitoring methods often rely on manual sampling, with a long detection cycle, and it is difficult to detect early wear problems of equipment in a timely manner.

[0132] To solve this problem, the wind farm decided to introduce the present invention to achieve real-time monitoring of the wear status of the lubrication system. The system is deployed on each wind turbine in the wind farm. Sensors are used to collect lubricating oil samples in real time and monitor the characteristics of abrasive particles in the oil. During operation, the system continuously analyzes multi-dimensional data on the number, size, and chemical composition of abrasive particles, thereby dynamically evaluating the wear status of the equipment.

[0133] During a monitoring process in May 2023, the monitoring system discovered an abnormality in the equipment of the wind turbine generator set numbered "W001". The system collected real-time data that the number of abrasive particles in the lubricating oil began to increase rapidly, from 50 per milliliter to 75. At the same time, the average size of the abrasive particles also increased from 5.2 microns to 6.2 microns. The system automatically determined that the abrasive data exceeded the preset threshold and generated the first abnormal wear warning report at 14:32 on the same day.

[0134] Monitoring data revealed a significant increase in the copper content of the abrasive particles. Analysis indicated that the bearings within the equipment might be experiencing abnormal wear. Equipment monitoring personnel confirmed this anomaly through system-generated reports and received an early warning at 2:35 PM. Based on the real-time data provided by the system, the maintenance team decided to shut down the equipment for inspection at 5:00 PM that same day.

[0135] Inspection revealed that the bearings of equipment W001 were severely worn, and the amount of metal particles in the lubricant exceeded the permitted limit. Further analysis of the particle shape, size, and composition determined that the primary cause of the wear was bearing lubrication failure and increased metal friction due to long-term operation. The maintenance team immediately repaired the equipment's lubrication system and replaced the damaged bearing assembly, successfully averting a further equipment failure.

[0136] During this period, another device numbered "W002" also had a similar wear warning. After monitoring that the number of abrasive particles in the W002 device increased from 65 to 85 per milliliter, and the abrasive particle size increased from 5.8 microns to 7.0 microns, the system generated another abnormality report at 10:12 on May 18. Through the report automatically generated by the system, the maintenance personnel quickly located the wear problem and decided to conduct further inspections on the equipment.

[0137] Test results for equipment W002 showed significantly elevated levels of iron in the abrasive particles. System analysis revealed that friction may have generated a large number of wear particles from the bearing's external material. The monitoring system recorded the growth trend of iron particles in the equipment's lubricating oil in real time and issued an emergency alert at 10:15. The maintenance team implemented timely downtime maintenance based on the data and successfully repaired damaged parts, preventing a more serious equipment failure.

[0138] To better verify the effectiveness of the system of the present invention, we selected 10 devices in a wind farm and conducted dynamic monitoring of oil wear particles for three months. The results were compared with traditional methods. Table 1 below shows the monitoring data of some devices:

[0139] Table 1 Comparison of data from oil wear dynamic monitoring system

[0140]

[0141] As can be seen from the data in Table 1 above, the system of the present invention can respond quickly when abnormal wear is detected. In the monitoring of device W001, the system generated an abnormal wear warning report at 14:32 when it found a rapid increase in the number and size of abrasive particles. The equipment inspection at 17:00 confirmed the wear of the bearing. In comparison, traditional manual inspection methods require monthly sampling and analysis, which takes up to 30 days. This means that with traditional methods, the wear problem of device W001 may not be discovered until 30 days later, which will put the equipment at greater risk of damage.

[0142] During the monitoring of equipment W002, the system found that the number of abrasive particles increased from 65 to 85 per milliliter and generated an abnormal warning report at 10:12. The maintenance team quickly performed shutdown maintenance based on this data and repaired the equipment wear problem at 13:00 that afternoon. Traditional methods, due to the long detection cycle, were unable to detect early signs of equipment wear in a timely manner.

[0143] The system of the present invention can not only monitor changes in the number and size of abrasive particles in real time, but also further identify the root cause of equipment wear by analyzing the chemical composition of the abrasive particles. During the monitoring of the W004 equipment, the system analyzed the copper content in the abrasive particles and found that the metal particles in the lubricating oil exceeded the standard. It generated an abnormal warning report at 11:00, reminding maintenance personnel to perform emergency maintenance operations. Traditional methods cannot provide real-time data due to the long sampling cycle, which may cause equipment wear problems to be delayed.

[0144] Through practical application in wind turbines, the proposed oil wear particle dynamic monitoring system, based on a two-layer generative adversarial network model, has demonstrated its exceptional real-time monitoring capabilities. By analyzing multidimensional data on the number, size, shape, and chemical composition of wear particles, the system generates timely warning reports for early signs of lubricant wear. This helps guide maintenance personnel in performing precise maintenance, thus preventing major equipment failures. Compared to traditional detection methods, this system boasts higher response speed and detection accuracy, effectively reducing equipment maintenance costs and improving operational reliability.

[0145] The present invention introduces a visual transformer model in wear particle image processing and uses a self-attention mechanism to realize the global feature extraction of wear particle images. Unlike traditional local image analysis methods, the visual transformer divides the wear particle image into several image blocks and embeds each image block, and then extracts the global dependency of the wear particle image through the self-attention mechanism. It can capture the multi-dimensional features of the wear particles more comprehensively and accurately, including the shape, size and surface details of the wear particles.

[0146] The present invention adopts a two-layer generative adversarial network model, which includes two generators and a discriminator. The first-layer generator generates a preliminary wear particle image, and the second-layer generator enhances the details of the wear particle image based on the first-layer generation, making it closer to the real wear particle image. Through adversarial training, the discriminator can effectively distinguish between real and generated wear particle images, thereby improving the system's ability to recognize wear particles in complex industrial environments. Compared with traditional rule-based recognition methods, the introduction of a two-layer generative adversarial network not only improves the detail fidelity of the wear particle image, but also optimizes the accuracy and reliability of wear particle feature classification.

[0147] The present invention realizes real-time monitoring of the equipment's lubrication status and wear conditions by deploying a trained two-layer generative adversarial network model in the monitoring system. The system can dynamically calculate the equipment's wear status index based on the type, size, quantity and chemical composition characteristics of the abrasive particles, and compare it with the set threshold in real time. Once the threshold is exceeded, the system will automatically trigger an early warning mechanism to notify relevant maintenance personnel to take timely maintenance measures, thereby improving the real-time and automation level of the monitoring system and avoiding economic losses caused by equipment downtime due to failures.

[0148] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for dynamic monitoring of oil wear particles based on deep learning, characterized in that: The steps include: S1. Collect oil sample data in real time through multiple sensors deployed in the equipment lubrication system and transmit the oil sample data to a data processing module to form an oil sample set; S2. Preprocessing the oil sample set transmitted to the data processing module and extracting wear particle images from the oil sample set; S3. Build a two-layer generative adversarial network model, including a first-layer generator, a second-layer generator, and a discriminator. S4. Deploy the trained two-layer GAN model in the monitoring system, collect and input wear particle images in real time, and use the two-layer GAN model to automatically identify the type, size, and number of wear particles in the wear particle images and generate wear feature data. S5. Based on the identified wear characteristic data, the wear status index of the device is calculated and compared with the set threshold to determine whether the device is in an abnormal state. If the wear status index exceeds the set threshold, the process proceeds to S6. If the wear status index does not exceed the set threshold, the process proceeds to S7. S6. Automatically trigger the early warning mechanism and generate early warning information to notify relevant personnel to take maintenance measures; S7. The monitoring results are stored in a database, and the wear particle characteristic change trend and equipment operation status are displayed in a graphical form through a data visualization module.

2. The method for dynamic monitoring of oil wear particles based on deep learning according to claim 1, characterized in that: Said S1 specifically includes: S11. A sensor array is arranged at key nodes of the equipment lubrication system. The sensor array includes n sensors T1, T2, ..., T n Each sensor collects oil wear particle data in the oil sample in real time, including the volume flow rate Q of the oil i (t), number of abrasive particles N i (t), abrasive particle size D i (t) and the chemical composition of the abrasive particles C i (t), where i is the sensor number and t is the time; S12. The oil wear particle data collected by the n sensors is transmitted to the data processing module through the data transmission module; S13. Perform preliminary formatting on the collected oil wear particle data to form an oil sample set: D y ={(Q i (t),N i (t),D i (t),C i (t))|i=1,2,...,n;t=t1,t2,...,t m }; Among them, D y represents all the oil wear particle data collected at different time points t and multiple sensor positions i, and m is the number of time points.

3. The method for dynamic monitoring of oil wear particles based on deep learning according to claim 1, characterized in that: The S2 specifically includes: S21. Transmit oil sample set D to the data processing module y The wear particle data in the denoising process is processed, and the random noise and system noise in the sampling process are eliminated using a bandpass filter to obtain the denoised oil sample set D denoise ; S22. For the denoised oil sample set D denoise Standardization is performed by j And each sensor T i Abrasive grain size D i (t j ), number of abrasive particles N i (t j ) and abrasive chemical composition C i (t j ) to obtain the standardized oil sample set D norm ; S23. Oil sample set D based on standardization norm Extract wear particle image, for each sensor Ti at time t j The collected standardized wear particle size and number of abrasive particles Through the two-dimensional Laplacian operator Perform image gradient enhancement: in, is the Laplace edge detection result, where x and y are the spatial coordinates of the image; S24. For each time point t j , wear particle image based on the detected edge information Energy function Optimize and then perform wear area segmentation: Among them, Ω i is the area of wear particles in the image, α and β are weight parameters, is the image gradient, indicating the strength of the edge, Represents chemical composition data, is the weight function related to chemical composition; S25. Based on energy function Based on the minimization result, a fast matching algorithm is used to mark and classify each segmented area to generate the final wear particle image matrix: Among them, M i (t j ) is the matrix representation of the wear particle image, Label is the region labeling function, which means classifying the optimal segmentation region and finally outputs the labeled and classified wear particle image.

4. The method for dynamic monitoring of oil wear particles based on deep learning according to claim 1, characterized in that: The S3 specifically includes: S31. Constructing the visual transformer model Vi(t based on the preprocessed wear particle image matrix j ), the wear particle image matrix is divided into several image blocks Pk, where k is the index of the image block and the size of each image block is p×p. Each image block is embedded to obtain the embedded feature vector εk. The global dependency between image blocks is extracted through the self-attention mechanism, and the transformer model representation containing the global features of the wear particle image is output: Vi(t j )=Attention(εk); S32. Construct a two-layer generative adversarial network model, including the first-layer generator G1 and the second-layer generator G2 and a discriminator D: The first layer generator G1 generates wear particle images M of different types and sizes by inputting noise vector z1~N(0,1) gen1 (t j )=G1(z1), preliminary feature generation is performed on the generated wear particle images of different types and sizes through convolutional neural networks; The second layer generator G2 generates the wear particle image M output by the first layer generator G1. gen1 (t j ) is used to generate the wear particle image moment M with more detailed features. gen2 (t j )=G2(M gen1 (t j )), so that the generated wear particle image matrix is close to the real wear particle image matrix; The wear particle image matrix M generated by the discriminator D is input gen2 (t j ) and the real collected wear particle image matrix M i (t j ) The authenticity of the two is judged through adversarial training: in, is the loss function of the discriminator, is the expected operation, D(·) represents the classification probability output of the discriminator for the input image; S33. Using the pre-labeled wear particle image dataset D label For the visual transformer model Vi(t j ) and the discriminator D are jointly trained to optimize the feature extraction capability and wear particle feature classification accuracy of the two-layer generative adversarial network model: in, is the classification loss function, and λ is the loss weight factor.

5. The method for dynamic monitoring of oil wear particles based on deep learning according to claim 1, characterized in that: The S4 specifically includes: S41. Deploy the trained two-layer GAN model in the central data processing module of the monitoring system. The monitoring system uses a sensor array to collect a wear particle image matrix from the oil sample in real time and inputs the wear particle image matrix into the two-layer GAN model. S42. The first layer generator is based on the collected wear particle image matrix M i (t j ) Generate preliminary wear particle image matrix features F gen1 (t j )=G1(M i (t j )), the generated wear particle image matrix M i (t j ) includes initial grit size, grit shape and grit quantity; S43. The second layer generator generates detail enhanced wear particle image matrix features F based on the preliminary wear particle image matrix features output by the first layer generator gen2 (t j )=G2(F gen1 (t j )), so that the enhanced wear particle image matrix is close to the real wear particle image matrix characteristics; S44. Input the generated wear particle image matrix features and the real collected wear particle image matrix into the discriminator, and the discriminator judges the authenticity and classification of the wear particle image matrix and generates a classification result C i (t j ); S45. Based on classification results C i (t j ), generate wear feature data W i (t j ), including abrasive type, abrasive size, abrasive quantity and abrasive chemical composition.

6. The method for dynamic monitoring of oil wear particles based on deep learning according to claim 1, characterized in that: The S5 specifically includes: S51. Based on the generated wear characteristic data W i (t j ) Calculate the wear status index of the equipment The wear state index is defined by combining the abrasive type, abrasive size, abrasive number and abrasive chemical composition: Among them, α1, α2, and α3 are weight parameters. is the number of normalized abrasive particles, is the standardized abrasive grain size; S52, the calculated wear state index With the set threshold T w The comparison is performed, and the threshold is set based on the historical wear data and working conditions of the equipment to determine whether the equipment is in an abnormal state. The judgment conditions are: Among them, Status(t j ) indicates that the device is at time t j The state at time T w is the set threshold.

7. A deep learning-based oil wear particle dynamic monitoring system, characterized in that: Includes the following modules: The sensor module is used to collect real-time data on oil wear particles in the equipment lubrication system, including the number, size, shape and chemical composition of wear particles; a data processing module configured to receive the oil wear particle data collected by the sensor module, and perform pre-processing, standardization, and denoising on the oil wear particle data; A two-layer generative adversarial network model, including a first-layer generator, a second-layer generator, and a discriminator, generates and discriminates wear particle image features and classifies them to generate wear particle feature data; The wear state assessment module calculates the wear state index of the equipment based on the wear particle characteristic data and determines whether the equipment is in an abnormal state; The early warning module automatically generates abnormal warning information and notifies maintenance personnel when abnormal equipment wear is detected; The data storage and visualization module stores monitoring data and displays the equipment operating status and wear particle characteristic change trends through visualization tools.

Images