Anti-clogging lubricating grease supply system based on self-cleaning structure

Abstract

The present invention discloses an anti-clogging lubricating grease supply system based on a self-cleaning structure, which includes six units: multimodal sensing and data acquisition, graph neural network feature extraction and processing. The multimodal sensing unit collects multidimensional data such as pressure and flow in real time, and converts them into node features through the graph neural network unit to construct a graph data model and deeply extract features. The improved Gaussian mixture model unit clusters and analyzes the system operation status based on this. The intelligent decision-making unit matches the control instructions according to the analysis results, and drives the self-cleaning execution unit to intercept impurities and clean the pipeline through a rotatable interception net and a high-pressure pulse cleaning device. At the same time, the grease supply adjustment unit accurately controls the grease pump speed, throttle valve opening and pressure compensation. The present invention combines multi-unit collaboration with innovative models to achieve intelligent monitoring of the operation status of the lubricating grease supply system, identification of blockage risks and active anti-blockage processing, thereby ensuring stable operation of the system.

Description

Technical Field

[0001] The present invention relates to the field of lubricating grease management in oil production, and in particular to an anti-clogging lubricating grease supply system based on a self-cleaning structure. Background Art

[0002] In the oil and gas industry, drilling equipment and oil production machinery must operate continuously under complex conditions of high temperature, high pressure, and high dust levels. Lubricant supply systems are crucial for ensuring their proper operation, and their performance directly impacts production efficiency and equipment lifespan. As oil and gas operations expand into complex areas like deep sea and shale formations, traditional lubricant supply systems are no longer able to meet the stringent requirements of modern oil and gas production, necessitating innovation.

[0003] Existing lubricant and grease supply systems have significant deficiencies in data processing. Most systems are equipped only with basic sensors, unable to fully capture multidimensional data such as pressure distribution and grease flow within the pipeline. Furthermore, traditional systems lack deep analytical architectures such as graph neural network feature extraction and processing units. They cannot convert data into node features to construct graph data models, making it difficult to mine spatial correlations between data based on pipeline topology. Furthermore, they cannot use improved Gaussian mixture model analysis units to perform cluster analysis on system operating status, making it difficult to promptly identify potential blockage risks.

[0004] In terms of anti-clogging and self-cleaning functions, traditional systems typically rely solely on simple filters to filter out impurities. This makes it difficult to dynamically intercept impurities such as sand and metal debris that can enter under complex operating conditions, easily causing pipe blockages. Furthermore, they are unable to automatically initiate cleaning procedures based on the system's operating status. Manual maintenance is not only inefficient and costly, but also results in incomplete cleaning, making it difficult to maintain the performance of the lubricant and the smooth flow of the pipes. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of the prior art, the present invention provides an anti-clogging lubricating grease supply system based on a self-cleaning structure.

[0006] The technical solution adopted by the present invention is an anti-clogging lubricating grease supply system based on a self-cleaning structure, comprising:

[0007] The multimodal sensing and data acquisition unit consists of a high-precision pressure sensor array, a flow monitoring module, a temperature sensing component, and an impurity concentration detection probe. The high-precision pressure sensor array is used to obtain real-time pressure distribution data within the pipeline; the flow monitoring module uses ultrasonic flow measurement technology to measure the grease flow rate; the temperature sensing component is used to obtain temperature information during the grease transportation process; and the impurity concentration detection probe uses the principle of optical scattering to detect the impurity concentration in the grease.

[0008] A graph neural network feature extraction and processing unit, which includes a multi-layer graph convolutional neural network architecture. It receives pressure, flow, temperature, and impurity concentration data transmitted by the multimodal sensing and data acquisition unit, converts them into node features, and constructs a graph data model containing pipeline topology information.

[0009] An improved Gaussian mixture model analysis unit, which introduces an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy, receives the feature vectors output by the graph neural network feature extraction and processing unit, and uses the improved Gaussian mixture model to perform cluster analysis on the operating status of the grease supply system, identifying normal operating modes and potential blockage risk modes.

[0010] An intelligent decision-making and control unit receives the analysis results of the improved Gaussian mixture model analysis unit, has a built-in preset control strategy library, and matches corresponding control instructions from the control strategy library based on the analysis results;

[0011] The self-cleaning execution unit consists of a rotatable impurity interception net, a high-pressure pulse cleaning device, and an impurity collection chamber. The rotatable impurity interception net is driven by a servo motor to dynamically intercept impurities in the grease. After receiving control instructions from the intelligent decision-making and control unit, the high-pressure pulse cleaning device sprays high-pressure pulse water into the pipeline to clean the interception net and the inner wall of the pipeline. Impurities generated by cleaning are drained into the impurity collection chamber through the pipeline.

[0012] The grease supply adjustment unit consists of a variable frequency driven grease pump, an adjustable throttle valve and a pressure compensation device. The variable frequency driven grease pump adjusts the speed according to the control instructions of the intelligent decision-making and control unit to change the grease supply amount; the adjustable throttle valve is installed in the delivery pipeline and the opening is adjusted by the electric actuator to control the grease flow; the pressure compensation device monitors the pipeline pressure in real time and adjusts the output pressure according to the control instructions to maintain stable pressure in the pipeline.

[0013] The forward propagation process of the graph convolutional neural network architecture used in the graph neural network feature extraction and processing unit satisfies the following formula: in, 、 Indicates the Layer, The node feature matrix of the layer graph convolutional neural network, each row represents the feature vector of a node; is the adjacency matrix with self-loops added, It is the original adjacency matrix constructed based on the topological structure of the lubricating grease delivery pipeline. is the identity matrix; yes The diagonal node degree matrix of It is The weight matrix of the layer is used to transform the node features; is the activation function.

[0014] Furthermore, the improved Gaussian mixture model analysis unit, the probability density function of the improved Gaussian mixture model is:

[0015]

[0016] in, is the probability density function, It is the feature vector output by the graph neural network feature extraction and processing unit, with a dimension of is the number of components of the Gaussian mixture model, which is pre-set according to the complexity of the operating state of the lubricating grease supply system; For the The weights of Gaussian components satisfy , its value is dynamically updated according to the system operation data through the adaptive weight adjustment mechanism; For the The mean vector of the Gaussian components; For the The covariance matrix of the Gaussian components adopts a dynamic covariance matrix update strategy and adjusts in real time according to the changes in data distribution. is the inverse of the covariance matrix, It is the index variable in the sum operation.

[0017] Furthermore, the graph neural network feature extraction and processing unit further includes an attention mechanism module, the calculation process of which satisfies the formula:

[0018]

[0019] in, and Represents the nodes in the graph data respectively and nodes The eigenvector of is the total number of nodes in the graph; Indicates the concatenation of two feature vectors; is the activation function; is the weight matrix of the attention mechanism module; For nodes For Node The attention coefficient reflects the node Feature pair nodes the importance of Nodes processed by the attention mechanism The new eigenvector of , T represents the transpose operation of the matrix, 、 jIt is the index variable in the sum operation.

[0020] Furthermore, the update formula of the adaptive weight adjustment mechanism of the improved Gaussian mixture model analysis unit is:

[0021]

[0022]

[0023] in, Indicates the number of iterations; is the number of samples in the current batch; 、 For the sequence The first iteration The weights of the Gaussian components; For the The sample at the iteration Belong to The posterior probability of the Gaussian components; For the The iteration is the mean vector, The covariance matrix of the Gaussian distribution is The probability density value at is the mean vector of the kth Gaussian component at the tth iteration, For the i Sample data, j 、 i It is the index variable in the sum operation.

[0024] Furthermore, the graph neural network feature extraction and processing unit takes into account the pipe diameter variation factor when constructing the graph data model. The node feature construction formula is: ,in, For nodes The eigenvector of For nodes The pressure value at is obtained by the pressure sensor in the multimodal perception and data acquisition unit; For nodes The grease flow rate value at is measured by the flow monitoring module; For nodes The temperature value at is collected by the temperature sensing component; For nodes The impurity concentration in the oil and fat in the corresponding area is detected by the impurity concentration detection probe; For nodes The pipe diameter value of the pipeline is pre-set according to the pipeline design parameters.

[0025] Furthermore, the improved Gaussian mixture model analysis unit, the dynamic covariance matrix update strategy satisfies the formula:

[0026]

[0027] in, Indicates the number of iterations; is the number of samples in the current batch; For the The sample at the iteration Belong to The posterior probability of the Gaussian components; is the sample feature vector; For the The first iteration The mean vector of the Gaussian components; is the regularization parameter used to prevent the covariance matrix from being singular; is the identity matrix; is the probability density value, and T is the matrix transpose operation.

[0028] Furthermore, the matching rules of the control strategy library in the intelligent decision-making and control unit are constructed using a graph neural network, and the matching process satisfies the formula:

[0029]

[0030] in, To improve the analysis result vector output by the Gaussian mixture model analysis unit; Represents the analysis result vector Perform graph neural network processing to explore the correlation between various factors in the analysis results; is the weight matrix of the matching rule; is the matching probability vector of each control strategy in the control strategy library. According to this vector, the control strategy with the largest probability is selected as the final control instruction. is the activation function.

[0031] Furthermore, the speed adjustment formula of the grease supply adjustment unit and the variable frequency driven grease pump is: ,in, is the adjusted grease pump speed; is the initial speed of the grease pump; Adjust the step size for the speed; Control instruction vector output by the intelligent decision-making and control unit; is the speed adjustment coefficient calculated according to the control command vector.

[0032] The anti-clogging lubricating grease supply system based on the self-cleaning structure includes the following steps:

[0033] Step 1: The multimodal sensing and data acquisition unit collects real-time data on pressure, grease flow, and temperature in the grease delivery pipeline, as well as grease impurity concentration at the oil tank outlet, at a preset sampling frequency. The collected data is then transmitted to the graph neural network feature extraction and processing unit via a high-speed data transmission bus.

[0034] Step 2: The graph neural network feature extraction and processing unit receives the data transmitted by the multimodal perception and data acquisition unit, converts it into node features, builds a graph data model based on the topological structure of the lubricating grease delivery pipeline, extracts and fuses features from the graph data through a multi-layer graph convolutional neural network architecture, mines the spatial correlation between the data, outputs the processed feature vector, and transmits it to the improved Gaussian mixture model analysis unit;

[0035] Step 3: The improved Gaussian mixture model analysis unit receives the feature vector output by the graph neural network feature extraction and processing unit, and uses the improved Gaussian mixture model to perform cluster analysis on the operating status of the oil supply system through an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy. This identifies the normal operating mode and potential blockage risk mode of the system, outputs the analysis results, and transmits them to the intelligent decision-making and control unit.

[0036] Step 4: The intelligent decision and control unit receives the analysis results of the improved Gaussian mixture model analysis unit, selects corresponding control instructions from the control strategy library based on a built-in control strategy library through specific matching rules, and transmits the control instructions to the self-cleaning execution unit and the grease supply adjustment unit respectively;

[0037] Step 5: After the self-cleaning execution unit receives the control instruction from the intelligent decision-making and control unit, the servo motor drives the rotatable impurity interception net to rotate, intercepting impurities in the grease. At the same time, the high-pressure pulse cleaning device sprays high-pressure pulse water flow into the conveying pipeline according to the control instruction to clean the interception net and the inner wall of the pipeline. The impurities generated by cleaning are drained into the impurity collection chamber;

[0038] Step 6: After the grease supply adjustment unit receives the control instruction from the intelligent decision-making and control unit, the variable frequency driven grease pump adjusts the speed according to the control instruction, the adjustable throttle valve adjusts the opening through the electric actuator, and the pressure compensation device monitors the pipeline pressure in real time and adjusts the output pressure according to the control instruction, completing the adjustment of the lubricating grease supply and pipeline pressure.

[0039] Beneficial Effects: The present invention proposes an anti-clogging lubricating grease supply system based on a self-cleaning structure. In terms of data processing and analysis, the multimodal sensing and data acquisition unit utilizes components such as a high-precision pressure sensor array and a flow monitoring module. Compared with the single and inefficient data acquisition method of traditional systems, it can comprehensively obtain multi-dimensional data such as pressure distribution, flow, temperature, and impurity concentration during the lubricating grease delivery process. The graph neural network feature extraction and processing unit combines the pipeline topology structure to construct a graph data model. Using the graph convolutional neural network architecture and attention mechanism, it deeply explores the spatial correlation between data, changing the situation where traditional systems cannot effectively process complex data. The improved Gaussian mixture model analysis unit accurately identifies the system operation mode through adaptive weight adjustment and dynamic covariance matrix update. Compared with the traditional fixed threshold judgment method, it greatly improves the ability to identify potential blockage risks. In terms of anti-clogging and self-cleaning functions, the rotatable impurity interception net and high-pressure pulse cleaning device of the self-cleaning execution unit replace the simple filter filtration of the traditional system, realizing dynamic interception and active cleaning of impurities, and avoiding impurity accumulation and clogging of the pipeline; the intelligent decision-making and control unit accurately matches instructions from the control strategy library based on the analysis results of the improved Gaussian mixture model, and links the self-cleaning execution unit and the grease supply adjustment unit to accurately adjust the grease pump speed, throttle valve opening and pressure compensation to ensure smooth pipeline and stable grease supply, solving the problem of the traditional system's lack of active cleaning mechanism and precise control capabilities, and significantly improving the reliability and stability of the lubricating grease supply system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a diagram of the system unit composition of the present invention;

[0041] Figure 2 It is a flow chart of the system operation steps of the present invention. DETAILED DESCRIPTION

[0042] like Figure 1 As shown, the anti-clogging lubricating grease supply system based on the self-cleaning structure includes:

[0043] A multimodal sensing and data acquisition unit, consisting of a high-precision pressure sensor array, a flow monitoring module, a temperature sensing component, and an impurity concentration detection probe. The high-precision pressure sensor array is installed in a distributed manner at key nodes of the lubricating grease delivery pipeline to obtain real-time pressure distribution data within the pipeline. The flow monitoring module is located on the main grease supply channel and uses ultrasonic flow measurement technology to accurately measure the grease flow rate. The temperature sensing component is tightly attached to the outer wall of the delivery pipeline to obtain temperature information during the grease delivery process. The impurity concentration detection probe is located at the oil outlet of the oil storage tank and uses the principle of optical scattering to detect the impurity concentration in the grease. The multimodal sensing and data acquisition unit is connected to subsequent units via a high-speed data transmission bus.

[0044] Specifically, the multimodal sensing and data acquisition unit is the basic module for the entire system to obtain operating information. It consists of a high-precision pressure sensor array, a flow monitoring module, a temperature sensing component, and an impurity concentration detection probe. In the lubricating grease delivery pipeline, pressure distribution is crucial to the operation of the system. The high-precision pressure sensor array is installed at each key node in a distributed layout. These sensors can sense the pressure values ​​at different locations in the pipeline in real time with extremely high accuracy. For example, at pipe bends, near valves, and other locations prone to pressure changes, they can accurately obtain pressure data and provide a basis for the system to judge the smoothness of grease delivery. The flow monitoring module is set on the main grease supply channel and uses ultrasonic flow measurement technology to accurately measure the grease flow rate per unit time. This data directly reflects the grease supply efficiency. Once the flow rate fluctuates abnormally, it may indicate a pipeline blockage or other fault.

[0045] The temperature sensing component fits tightly to the outer wall of the delivery pipeline and continuously monitors temperature changes during the grease delivery process. Excessively high temperatures may cause the viscosity of the grease to decrease, affecting the lubrication effect; excessively low temperatures may cause the grease to thicken, increase the delivery resistance, and even cause blockage. The impurity concentration detection probe is arranged at the oil outlet of the oil storage tank and uses the principle of optical scattering to detect the impurity content in the grease that is about to enter the delivery pipeline. Impurities are an important factor that causes pipeline blockage. By monitoring the impurity concentration in real time, blockage problems caused by impurity accumulation can be prevented in advance. The pressure, flow, temperature, impurity concentration and other data collected by this unit are quickly and stably transmitted to subsequent units through a high-speed data transmission bus, providing comprehensive and accurate raw data support for system analysis and decision-making.

[0046] A graph neural network feature extraction and processing unit, which includes a multi-layer graph convolutional neural network architecture. It receives pressure, flow, temperature, and impurity concentration data transmitted by the multimodal sensing and data acquisition unit, converts them into node features, constructs a graph data model containing pipeline topology information, extracts and fuses features from the graph data through graph convolution operations, mines spatial correlations between data, outputs processed feature vectors, and transmits them to the improved Gaussian mixture model analysis unit.

[0047] Specifically, the graph neural network feature extraction and processing unit receives data transmitted by the multimodal sensing and data acquisition unit and then deeply processes this raw data. It contains a multi-layer graph convolutional neural network architecture, which first converts data such as pressure, flow, temperature, and impurity concentration into node features. At the same time, it combines the topological structure information of the lubricating oil delivery pipeline to construct a graph data model. In this model, each node represents a monitoring point or key part in the pipeline, and the connection relationship between nodes reflects the actual layout of the pipeline.

[0048] Through graph convolution operations, the unit can deeply explore the spatial correlation between data based on the pipeline topology. For example, when the pressure at a certain node changes, the graph neural network can analyze the possible impact on the pressure, flow and other data of other nodes connected to the node, thereby identifying the patterns and potential trends of data changes. Compared with traditional data processing methods, this mining of spatial correlation relationships of data can more comprehensively and accurately understand the operating status of the lubricating grease supply system and extract more valuable feature information. After processing, the unit outputs the processed feature vector and transmits it to the improved Gaussian mixture model analysis unit to provide key data for subsequent analysis of the system operating status.

[0049] An improved Gaussian mixture model analysis unit, based on the traditional Gaussian mixture model, introduces an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy. This unit receives the feature vectors output by the graph neural network feature extraction and processing unit, performs cluster analysis on the operating status of the grease supply system using the improved Gaussian mixture model, identifies normal operating modes and potential blockage risk modes, and outputs the analysis results for transmission to the decision control unit.

[0050] Specifically, the improved Gaussian mixture model analysis unit innovates on the traditional Gaussian mixture model by introducing an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy. It receives the feature vectors output by the graph neural network feature extraction and processing unit, which contain key information about the operating status of the lubricating grease supply system.

[0051] The improved Gaussian mixture model performs cluster analysis on these eigenvectors, classifying the system's operating state into distinct modes. The adaptive weight adjustment mechanism dynamically adjusts the weights of each Gaussian component based on changes in the system's actual operating data, enabling the model to better adapt to the diversity of the system's operating states. The dynamic covariance matrix update strategy adjusts the covariance matrix in real time based on changes in the data distribution, thereby more accurately describing the characteristic distribution of the system's operating state. In this way, the unit can accurately identify the system's normal operating mode and potential congestion risk mode. Once a potential congestion risk mode is detected, the analysis results are promptly transmitted to the decision-making control unit, providing a basis for the system to take appropriate anti-congestion measures.

[0052] An intelligent decision-making and control unit receives the analysis results of the improved Gaussian mixture model analysis unit, has a built-in preset control strategy library, matches corresponding control instructions from the control strategy library based on the analysis results, and transmits the control instructions to the self-cleaning execution unit and the grease supply adjustment unit respectively;

[0053] Specifically, the intelligent decision-making and control unit is the core decision-making module of the entire system. It receives the system operating status analysis results output by the improved Gaussian mixture model analysis unit. This unit has a built-in preset control strategy library, which stores a variety of control strategies for different operating states and congestion risk situations.

[0054] After receiving the analysis results, the intelligent decision-making and control unit selects the control instructions that best suit the current system state from the control strategy library based on preset matching rules. For example, if a potential blockage risk is detected in the system, it will select corresponding control instructions, such as initiating a self-cleaning program and adjusting the grease supply volume or pressure, based on the degree and specific circumstances of the blockage risk. These control instructions are transmitted to the self-cleaning execution unit and the grease supply adjustment unit, respectively, to achieve precise control of the system. By accurately judging the system's operating status and making reasonable decisions, the intelligent decision-making and control unit ensures that the system can respond to various situations in a timely and effective manner, maintain the stable operation of the lubricating grease supply system, and avoid blockages.

[0055] The self-cleaning execution unit consists of a rotatable impurity interception net, a high-pressure pulse cleaning device, and an impurity collection chamber. The rotatable impurity interception net is installed in the conveying pipeline and is driven to rotate by a servo motor to achieve dynamic interception of impurities in the grease. The high-pressure pulse cleaning device is connected to the conveying pipeline and, upon receiving control instructions from the intelligent decision-making and control unit, sprays high-pressure pulse water into the pipeline to clean the interception net and the inner wall of the pipeline. Impurities generated by cleaning are drained into the impurity collection chamber through the pipeline. The self-cleaning execution unit is connected to the intelligent decision-making and control unit via a control signal line.

[0056] Specifically, the self-cleaning actuator is a key module for the system's anti-clogging function. It consists of a rotatable impurity interception net, a high-pressure pulse cleaning device, and an impurity collection chamber. The rotatable impurity interception net is installed in the delivery pipeline and is driven by a servo motor to continuously intercept impurities in the grease during delivery. The rotating design allows the interception net to more fully contact the grease, improving impurity interception efficiency and effectively preventing impurity accumulation in the pipeline.

[0057] When the intelligent decision-making and control unit issues a control command, the high-pressure pulse cleaning device begins operating. Connected to the delivery pipeline, it sprays a high-pressure pulse stream of water into the line. This high-pressure pulse stream has a powerful impact, flushing impurities from the surface of the interception net and the inner wall of the pipeline. Impurities generated by cleaning are then drained into the impurity collection chamber through a specially designed pipeline, completing the removal process. The self-cleaning actuator automatically intercepts impurities and performs regular cleaning, keeping the pipeline clean and unobstructed, fundamentally reducing the possibility of blockage and ensuring the normal delivery of lubricating grease.

[0058] The grease supply adjustment unit consists of a frequency-controlled grease pump, an adjustable throttle valve and a pressure compensation device. The frequency-controlled grease pump adjusts the speed according to the control instructions of the intelligent decision-making and control unit to change the grease supply amount; the adjustable throttle valve is installed in the delivery pipeline and the opening is adjusted by the electric actuator to control the grease flow; the pressure compensation device monitors the pipeline pressure in real time and adjusts the output pressure according to the control instructions to maintain stable pressure in the pipeline. The grease supply adjustment unit is connected to the intelligent decision-making and control unit through a control signal line.

[0059] Specifically, the grease supply regulation unit is responsible for precisely controlling the lubricating grease supply process. It consists of a variable-frequency-driven grease pump, an adjustable throttle valve, and a pressure compensation device. The variable-frequency-driven grease pump adjusts its speed based on the control instructions from the intelligent decision-making and control unit, precisely controlling the grease supply. When the system detects that a certain part requires more grease, the grease pump speed increases, increasing grease supply. Conversely, the speed decreases, reducing supply, ensuring that grease supply matches the actual needs of the equipment.

[0060] The adjustable throttle valve is installed in the delivery pipeline, and the opening is adjusted by the electric actuator to further accurately control the flow of grease. Under different working conditions, the throttle valve can adjust the opening size as needed to stabilize the grease flow within an appropriate range. The pressure compensation device monitors the pipeline pressure in real time. When the pipeline pressure changes due to various factors, the pressure compensation device will adjust the output pressure according to the control instructions to maintain stable pressure in the pipeline. Stable pressure is crucial for the smooth delivery of grease, and can avoid problems such as poor grease delivery or even blockage due to pressure fluctuations. The grease supply adjustment unit ensures that the lubricating grease supply system can operate stably and efficiently under various working conditions by coordinated adjustment of the grease supply amount, flow rate and pressure.

[0061] Preferably, in the graph neural network feature extraction and processing unit, the forward propagation process of the graph convolutional neural network architecture adopted satisfies the following formula:

[0062]

[0063] in, Indicates the The node feature matrix of the layer graph convolutional neural network, each row represents the feature vector of a node; is the adjacency matrix with self-loops added, It is the original adjacency matrix constructed based on the topological structure of the lubricating grease delivery pipeline. is the identity matrix; yes The diagonal node degree matrix of It is The weight matrix of the layer is used to transform the node features; The ReLU function is used as the activation function. This formula performs convolution operations on graph data to extract and fuse the features of pressure, flow, temperature and impurity concentration data in the lubricating grease supply system, and mines the spatial correlation between data based on the pipeline topology structure.

[0064] Specifically, in the anti-clogging lubricating grease supply system, in order to deeply explore the potential connections between data such as pressure, flow, temperature and impurity concentration, a graph convolutional neural network architecture is used to process the data. This architecture constructs a graph data model based on the topological structure of the lubricating grease delivery pipeline, and regards each monitoring point of the pipeline as a node. The connection relationship between the nodes reflects the pipeline layout. Through the convolution operation of the graph data, the spatial correlation between the data based on the pipeline topology can be fully considered. For example, when the pressure of a certain node is abnormal, the architecture can quickly analyze the changing trend of the surrounding node data and accurately extract the data features, providing a key basis for the subsequent analysis of the system operation status, and effectively improving the system's ability to predict the risk of blockage.

[0065] Preferably, in the improved Gaussian mixture model analysis unit, the probability density function of the improved Gaussian mixture model is:

[0066]

[0067] in, It is the feature vector output by the graph neural network feature extraction and processing unit, with a dimension of is the number of components of the Gaussian mixture model, which is pre-set according to the complexity of the operating state of the lubricating grease supply system; For the The weights of Gaussian components satisfy , its value is dynamically updated according to the system operation data through the adaptive weight adjustment mechanism; For the The mean vector of the Gaussian components; For the The covariance matrix of the Gaussian components is calculated, and a dynamic covariance matrix update strategy is adopted to adjust in real time according to the changes in data distribution. This formula is used to perform cluster analysis on the operating status of the lubricating grease supply system. By calculating the probability that the eigenvectors belong to different Gaussian components, the normal operating mode and potential blockage risk mode of the system are identified.

[0068] Specifically, an improved Gaussian mixture model plays a key role in accurately analyzing the operating status of the lubricating grease supply system. This model receives processed eigenvectors and, through cluster analysis of system operating data, distinguishes between normal operating modes and those with potential blockage risks. Its adaptive weight adjustment and dynamic covariance matrix update mechanisms optimize model parameters in real time based on changes in system operating conditions. In actual operation, when changes in equipment load cause fluctuations in grease pressure and flow, the model can rapidly adjust the weights of each Gaussian component and the covariance matrix, accurately identifying the system's operating status and providing reliable support for intelligent decision-making, thereby preventing equipment failures caused by blockages.

[0069] Preferably, the graph neural network feature extraction and processing unit further includes an attention mechanism module, the calculation process of which satisfies the formula:

[0070]

[0071] in, and Represents the nodes in the graph data respectively and nodes The eigenvector of is the total number of nodes in the graph; Indicates the concatenation of two feature vectors; is the weight matrix of the attention mechanism module; For nodes For Node The attention coefficient reflects the node Feature pair nodes the importance of Nodes processed by the attention mechanism The module calculates the attention coefficient between nodes to enhance the ability to extract key data features of the lubricating grease supply system and further optimize the feature extraction effect of the graph neural network.

[0072] Specifically, when processing data from the lubricating grease supply system, an attention mechanism module was introduced to enhance the extraction of key data features. Among the massive amounts of data generated by system operation, some data is more valuable for determining blockage risks. This module determines the importance of each node's features by calculating the attention coefficient between nodes. When conditions such as a sudden increase in impurity concentration or an abnormal decrease in flow rate, which could indicate a blockage, occur, the attention mechanism module can highlight these key data features, optimizing the graph neural network's feature extraction performance. This allows the system to more sensitively capture operational status changes and promptly identify potential blockages.

[0073] Preferably, in the improved Gaussian mixture model analysis unit, the updating formula of the adaptive weight adjustment mechanism is:

[0074]

[0075]

[0076] in, Indicates the number of iterations; is the number of samples in the current batch; For the The first iteration The weights of the Gaussian components; For the The sample at the iteration Belong to The posterior probability of the Gaussian components; For the The iteration is the mean vector, The covariance matrix of the Gaussian distribution is The formula dynamically adjusts the weights of each Gaussian component by calculating the posterior probability based on sample data, so that the improved Gaussian mixture model can better adapt to the changes in the operating state of the lubricating grease supply system and improve the ability to identify potential blockage risk patterns.

[0077] Specifically, the adaptive weight adjustment mechanism within the improved Gaussian mixture model plays a crucial role in addressing the ever-changing operating conditions of lubricating grease supply systems. This mechanism, based on the posterior probability calculation of sample data, recalculates the weights of each Gaussian component based on the new sample data, as changes in system operating data due to factors such as equipment status and environmental factors, such as changes in ambient temperature affecting grease flow and causing fluctuations in pressure and flow rates. This dynamic adjustment allows the model to better fit the current operating state, improve the accuracy of identifying potential blockage risk patterns, and ensure stable system operation.

[0078] Preferably, in the graph neural network feature extraction and processing unit, when constructing the graph data model, the pipe diameter change factor is taken into account, and the node feature construction formula is: ,in, For nodes The eigenvector of For nodes The pressure value at is obtained by the pressure sensor in the multimodal perception and data acquisition unit; For nodes The grease flow rate value at is measured by the flow monitoring module; For nodes The temperature value at is collected by the temperature sensing component; For nodes The impurity concentration in the oil and fat in the corresponding area is detected by the impurity concentration detection probe; For nodes The pipe diameter value of the pipeline is pre-set according to the pipeline design parameters; this formula incorporates the pipe diameter change factor into the node characteristics, allowing the graph data model to more accurately reflect the actual physical structure of the lubricating grease supply system and enhance the graph neural network's ability to extract system data features.

[0079] Specifically, when constructing the graph data model, the significant impact of changes in pipeline diameter on grease transportation was taken into account, and the pipe diameter factor was incorporated into the node features. During the actual transportation process, changes in pipe diameter will directly affect parameters such as grease pressure and flow rate. For example, where the pipe diameter narrows, the grease flow rate increases and the pressure changes. Pipe diameter values ​​are combined with data such as pressure, flow rate, temperature, and impurity concentration to form node features, making the graph data model more consistent with the actual physical structure of the system. When the graph neural network analyzes data based on this model, it can more accurately grasp the patterns of data changes and effectively extract features related to blockage risk, providing a comprehensive basis for blockage risk assessment.

[0080] Preferably, in the improved Gaussian mixture model analysis unit, the dynamic covariance matrix update strategy satisfies the formula:

[0081]

[0082] in, Indicates the number of iterations; is the number of samples in the current batch; For the The sample at the iteration Belong to The posterior probability of the Gaussian components; is the sample feature vector; For the The first iteration The mean vector of the Gaussian components; is the regularization parameter used to prevent the covariance matrix from being singular; is the unit matrix; this formula updates the covariance matrix based on sample data and posterior probability and introduces a regularization term, so that the improved Gaussian mixture model can adjust the covariance matrix in real time according to the distribution changes of the lubricating grease supply system operating data, more accurately describe the characteristic distribution of the system operating status, and improve the recognition accuracy of the blockage risk pattern.

[0083] Specifically, because the data distribution of the lubricating grease supply system changes with operating conditions, improving the dynamic covariance matrix update strategy of the Gaussian mixture model is crucial. The traditional fixed covariance matrix cannot accurately describe the complex changes in the system. This update strategy adjusts the covariance matrix in real time based on sample data and posterior probabilities, and introduces a regularization term to prevent matrix singularity. When the system's grease parameters fluctuate due to equipment wear, ambient temperature changes, and other factors, the update strategy can promptly optimize the covariance matrix, more accurately describing the distribution of system operating state characteristics, improving the model's recognition accuracy of blockage risk patterns, and ensuring reliable system operation.

[0084] Preferably, in the intelligent decision-making and control unit, the matching rules of the control strategy library are constructed using a graph neural network, and the matching process satisfies the formula:

[0085]

[0086] in, To improve the analysis result vector output by the Gaussian mixture model analysis unit; Represents the analysis result vector Perform graph neural network processing to explore the correlation between various factors in the analysis results; is the weight matrix of matching rules; is the matching probability vector of each control strategy in the control strategy library, and the control strategy with the highest probability is selected as the final control instruction based on the vector; the formula further processes the analysis results of the improved Gaussian mixture model through the graph neural network to construct a more accurate control strategy matching rule, so that the intelligent decision-making and control unit can more accurately select the appropriate control instruction according to the system operation status, and realize effective control of the self-cleaning execution unit and the grease supply adjustment unit.

[0087] Specifically, the intelligent decision-making and control unit uses a graph neural network to construct matching rules for the control strategy library to accurately select control strategies. After the improved Gaussian mixture model analysis unit outputs the system's operating status analysis results, the graph neural network deeply processes the resulting vectors to explore the complex correlations between various factors. When the system faces multiple potential blockage risk factors, such as the simultaneous occurrence of impurity accumulation and pressure anomalies, this matching rule can accurately analyze the mutual influence of these factors, enabling the intelligent decision-making unit to more accurately select control instructions, effectively control the self-cleaning execution unit and the grease supply adjustment unit, and promptly respond to blockage risks.

[0088] Preferably, in the grease supply adjustment unit, the speed adjustment formula of the variable frequency driven grease pump is: ,in, is the adjusted grease pump speed; is the initial speed of the grease pump; Adjust the step size for the speed; Control instruction vector output by the intelligent decision-making and control unit; The speed adjustment coefficient is calculated based on the control instruction vector. Its calculation process involves the analysis and operation of various parameters in the control instruction. This formula combines the control instructions of the intelligent decision-making and control unit to achieve precise adjustment of the speed of the variable frequency driven grease pump, thereby controlling the supply of lubricating grease to meet the needs of different system operating conditions and prevent pipeline blockage.

[0089] Specifically, in the grease supply adjustment unit, a specific speed adjustment method is used to achieve precise control of the speed of the variable frequency drive grease pump. This method combines the control instructions output by the intelligent decision-making and control unit, and calculates the speed adjustment coefficient to achieve precise adjustment of the grease pump speed. During system operation, when the lubrication demand in a certain area increases or there is a risk of blockage and the grease flow needs to be adjusted, the intelligent decision-making unit issues a command, and the grease pump adjusts the speed according to the command. At the same time, it works in conjunction with the adjustable throttle valve and pressure compensation device to ensure that the grease supply, flow, and pressure are in the optimal state, effectively preventing pipeline blockage and meeting the different working conditions of the system.

[0090] like Figure 2 As shown, the anti-clogging lubricating grease supply system based on the self-cleaning structure includes the following steps:

[0091] Step 1: The multimodal sensing and data acquisition unit collects real-time data on pressure, grease flow, and temperature in the grease delivery pipeline, as well as grease impurity concentration at the oil tank outlet, at a preset sampling frequency. The collected data is then transmitted to the graph neural network feature extraction and processing unit via a high-speed data transmission bus.

[0092] Step 2: The graph neural network feature extraction and processing unit receives the data transmitted by the multimodal perception and data acquisition unit, converts it into node features, builds a graph data model based on the topological structure of the lubricating grease delivery pipeline, extracts and fuses features from the graph data through a multi-layer graph convolutional neural network architecture, mines the spatial correlation between the data, outputs the processed feature vector, and transmits it to the improved Gaussian mixture model analysis unit;

[0093] Step 3: The improved Gaussian mixture model analysis unit receives the feature vector output by the graph neural network feature extraction and processing unit, and uses the improved Gaussian mixture model to perform cluster analysis on the operating status of the oil supply system through an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy. This identifies the normal operating mode and potential blockage risk mode of the system, outputs the analysis results, and transmits them to the intelligent decision-making and control unit.

[0094] Step 4: The intelligent decision and control unit receives the analysis results of the improved Gaussian mixture model analysis unit, selects corresponding control instructions from the control strategy library based on a built-in control strategy library through specific matching rules, and transmits the control instructions to the self-cleaning execution unit and the grease supply adjustment unit respectively;

[0095] Step 5: After the self-cleaning execution unit receives the control instruction from the intelligent decision-making and control unit, the servo motor drives the rotatable impurity interception net to rotate, intercepting impurities in the grease. At the same time, the high-pressure pulse cleaning device sprays high-pressure pulse water flow into the conveying pipeline according to the control instruction to clean the interception net and the inner wall of the pipeline. The impurities generated by cleaning are drained into the impurity collection chamber;

[0096] Step 6: After the grease supply adjustment unit receives the control instruction from the intelligent decision-making and control unit, the variable frequency driven grease pump adjusts the speed according to the control instruction, the adjustable throttle valve adjusts the opening through the electric actuator, and the pressure compensation device monitors the pipeline pressure in real time and adjusts the output pressure according to the control instruction to achieve the regulation of the lubricating grease supply and pipeline pressure.

[0097] The present invention addresses the shortcomings of weak data processing capabilities of traditional systems, and the system has built a powerful data perception and analysis system. The multimodal perception and data acquisition unit deploys high-precision pressure sensor arrays, flow monitoring modules and other equipment. Compared with traditional single-point monitoring, it can collect multi-dimensional data such as pressure distribution of key nodes in the pipeline and grease flow with higher precision and more comprehensiveness, laying a solid data foundation for system operation analysis. The graph neural network feature extraction and processing unit uses a graph convolutional neural network architecture to convert data into node features to construct a graph data model, and deeply mine the spatial correlation relationship between data based on pipeline topology; the improved Gaussian mixture model analysis unit uses adaptive weight adjustment and dynamic covariance matrix update strategy to accurately cluster and analyze the system operation status, realize early identification of potential blockage risks, and completely change the situation where traditional systems rely on experience judgment and lag in risk identification.

[0098] The system has achieved a major breakthrough in anti-clogging and self-cleaning functions. The self-cleaning execution unit's rotatable impurity interception net and high-pressure pulse cleaning device work together. Compared to the passive filtration of traditional filter screens, the rotatable impurity interception net can dynamically intercept impurities such as sand and metal debris mixed in during oil extraction. Under the instructions of the intelligent decision-making and control unit, the high-pressure pulse cleaning device regularly performs strong cleaning on the interception net and the inner wall of the pipeline to ensure that the pipeline is clean and unobstructed. The grease supply adjustment unit's variable frequency drive grease pump, adjustable throttle valve, and pressure compensation device accurately control the grease supply amount, flow rate, and pressure according to the instructions of the intelligent decision-making unit, maintaining stable system operation and avoiding blockages caused by abnormal grease delivery. This fundamentally solves the problems of frequent blockages and difficult maintenance in traditional systems, and greatly improves the operating efficiency and reliability of oil extraction equipment.

[0099] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0100] While embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Anti-clogging lubricating grease supply system based on self-cleaning structure, characterized in that: include: Multimodal perception and data acquisition unit, graph neural network feature extraction and processing unit, improved Gaussian mixture model analysis unit, intelligent decision-making and control unit, self-cleaning execution unit, grease supply adjustment unit; The multimodal sensing and data acquisition unit consists of a high-precision pressure sensor array, a flow monitoring module, a temperature sensing component, and an impurity concentration detection probe. The high-precision pressure sensor array is used to obtain real-time pressure distribution data in the pipeline; the flow monitoring module uses ultrasonic flow measurement technology to measure the grease flow rate; the temperature sensing component is used to obtain temperature information during the grease transportation process; and the impurity concentration detection probe uses the principle of optical scattering to detect the impurity concentration in the grease. The graph neural network feature extraction and processing unit includes a multi-layer graph convolutional neural network architecture, which receives pressure, flow, temperature and impurity concentration data transmitted by the multimodal sensing and data acquisition unit, converts them into node features, and constructs a graph data model containing pipeline topology information; The improved Gaussian mixture model analysis unit introduces an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy, receives the feature vector output by the graph neural network feature extraction and processing unit, and performs cluster analysis on the operating status of the oil supply system through the improved Gaussian mixture model to identify the normal operating mode and the potential blockage risk mode; The intelligent decision and control unit receives the analysis result of the improved Gaussian mixture model analysis unit, has a built-in preset control strategy library, and matches the corresponding control instruction from the control strategy library according to the analysis result; The self-cleaning execution unit consists of a rotatable impurity interception net, a high-pressure pulse cleaning device, and an impurity collection chamber. The rotatable impurity interception net is driven by a servo motor to rotate and dynamically intercept impurities in the grease. After receiving control instructions from the intelligent decision-making and control unit, the high-pressure pulse cleaning device sprays high-pressure pulse water into the pipeline to clean the interception net and the inner wall of the pipeline. Impurities generated by cleaning are drained into the impurity collection chamber through the pipeline. The grease supply adjustment unit consists of a variable-frequency driven grease pump, an adjustable throttle valve, and a pressure compensation device. The variable-frequency driven grease pump adjusts its speed according to the control instructions of the intelligent decision-making and control unit to change the grease supply amount; the adjustable throttle valve is installed in the delivery pipeline and its opening is adjusted by an electric actuator to control the grease flow rate; the pressure compensation device monitors the pipeline pressure in real time and adjusts the output pressure according to the control instructions to maintain stable pressure in the pipeline; The forward propagation process of the graph convolutional neural network architecture used in the graph neural network feature extraction and processing unit satisfies the following formula: Among them, H (l) 、H (l+1) Represents the node feature matrix of the l-th layer and the l+1-th layer graph convolutional neural network, where each row represents the feature vector of a node; To add the adjacency matrix of self-loops, A is the original adjacency matrix constructed according to the topological structure of the lubricating grease delivery pipeline, and I is the identity matrix; yes The diagonal node degree matrix of W (l) is the weight matrix of the lth layer, which is used to transform the node features; σ is the activation function; The probability density function of the improved Gaussian mixture model in the improved Gaussian mixture model analysis unit is: Where p(x) is the probability density function, x is the feature vector output by the graph neural network feature extraction and processing unit, with a dimension of d; K is the number of components of the Gaussian mixture model, which is pre-set according to the complexity of the operating state of the lubricating grease supply system; π k is the weight of the kth Gaussian component, and satisfies π k ≥0, its value is dynamically updated according to the system operation data through the adaptive weight adjustment mechanism; μ k is the mean vector of the kth Gaussian component; Σ k is the covariance matrix of the kth Gaussian component, and a dynamic covariance matrix update strategy is adopted to adjust in real time according to the changes in data distribution. is the inverse of the covariance matrix, and k is the index variable in the summation operation; The graph neural network feature extraction and processing unit further includes an attention mechanism module, the calculation process of which satisfies the formula: Among them, h i and h j Respectively represent the feature vectors of node i and node j in the graph data, N is the total number of nodes in the graph; [h i ‖h j ] indicates the concatenation of two feature vectors; LeakyReLU is the activation function; W a is the weight matrix of the attention mechanism module; α ij is the attention coefficient of node i to node j, reflecting the importance of the feature of node j to node i; h′ i is the new feature vector of node i after the attention mechanism is processed, T represents the transpose operation of the matrix, and k and j are the index variables in the summation operation.

2. The anti-clogging lubricating grease supply system based on the self-cleaning structure according to claim 1 is characterized in that: The updating formula of the adaptive weight adjustment mechanism in the improved Gaussian mixture model analysis unit is: Where t represents the number of iterations; n is the number of samples in the current batch; is the weight of the kth Gaussian component at the tth and t+1th iterations; is the sample x at the tth iteration i The posterior probability of belonging to the kth Gaussian component; For the tth iteration, is the mean vector, is the covariance matrix of the Gaussian distribution in x i The probability density value at is the mean vector of the kth Gaussian component at the tth iteration, x i is the i-th sample data, j and i are index variables in the summation operation.

3. The anti-clogging lubricating grease supply system based on the self-cleaning structure according to claim 1 is characterized in that: When constructing the graph data model in the graph neural network feature extraction and processing unit, the pipe diameter change factor is taken into account. The node feature construction formula is: i =[p i ,q i , T i , c i , d i ], where h i is the feature vector of node i; p i is the pressure value at node i, obtained by the pressure sensor in the multimodal perception and data acquisition unit; q i is the grease flow rate at node i, measured by the flow monitoring module; T i is the temperature value at node i, collected by the temperature sensing component; c i is the impurity concentration in the grease in the area corresponding to node i, detected by the impurity concentration detection probe; d i is the pipe diameter value of the pipeline where node i is located, which is pre-set according to the pipeline design parameters.

4. The anti-clogging lubricating grease supply system based on the self-cleaning structure according to claim 1 is characterized in that: The dynamic covariance matrix updating strategy in the improved Gaussian mixture model analysis unit satisfies the formula: Where t represents the number of iterations; n is the number of samples in the current batch; is the sample x at the tth iteration i The posterior probability of belonging to the kth Gaussian component; x i is the sample feature vector; is the mean vector of the kth Gaussian component at the tth iteration; λ is the regularization parameter used to prevent the covariance matrix from being singular; I is the identity matrix; is the probability density value, and T is the matrix transpose operation.

5. The anti-clogging lubricating grease supply system based on the self-cleaning structure according to claim 1 is characterized in that: The matching rules of the control strategy library in the intelligent decision-making and control unit are constructed using a graph neural network, and the matching process satisfies the formula: Where r is the analysis result vector output by the improved Gaussian mixture model analysis unit; GNN(r) represents the graph neural network processing of the analysis result vector r to mine the correlation between the various factors in the analysis results; W m is the weight matrix of the matching rule; s is the matching probability vector of each control strategy in the control strategy library. According to the vector, the control strategy with the largest probability is selected as the final control instruction. Softmax is the activation function.

6. The anti-clogging lubricating grease supply system based on the self-cleaning structure according to claim 1 is characterized in that: The speed adjustment formula of the variable frequency driven grease pump in the grease supply adjustment unit is: ω=ω0+Δω·f(s), where ω is the adjusted grease pump speed; ω0 is the initial speed of the grease pump; Δω is the speed adjustment step; s is the control instruction vector output by the intelligent decision and control unit; and f(s) is the speed adjustment coefficient calculated based on the control instruction vector.

7. The anti-clogging lubricating grease supply system based on a self-cleaning structure according to any one of claims 1 to 6, characterized in that: The system operation includes the following steps: Step 1: The multimodal sensing and data acquisition unit collects real-time data on pressure, grease flow, and temperature in the grease delivery pipeline, as well as grease impurity concentration at the oil tank outlet, at a preset sampling frequency. The collected data is then transmitted to the graph neural network feature extraction and processing unit via a high-speed data transmission bus. Step 2: The graph neural network feature extraction and processing unit receives the data transmitted by the multimodal perception and data acquisition unit, converts it into node features, builds a graph data model based on the topological structure of the lubricating grease delivery pipeline, extracts and fuses features from the graph data through a multi-layer graph convolutional neural network architecture, mines the spatial correlation between the data, outputs the processed feature vector, and transmits it to the improved Gaussian mixture model analysis unit; Step 3: The improved Gaussian mixture model analysis unit receives the feature vector output by the graph neural network feature extraction and processing unit, and uses the improved Gaussian mixture model to perform cluster analysis on the operating status of the oil supply system through an adaptive weight adjustment mechanism and a dynamic covariance matrix update strategy. This identifies the normal operating mode and potential blockage risk mode of the system, outputs the analysis results, and transmits them to the intelligent decision-making and control unit. Step 4: The intelligent decision and control unit receives the analysis results of the improved Gaussian mixture model analysis unit, selects corresponding control instructions from the control strategy library based on a built-in control strategy library through specific matching rules, and transmits the control instructions to the self-cleaning execution unit and the grease supply adjustment unit respectively; Step 5: After the self-cleaning execution unit receives the control instruction from the intelligent decision-making and control unit, the servo motor drives the rotatable impurity interception net to rotate, intercepting impurities in the grease. At the same time, the high-pressure pulse cleaning device sprays high-pressure pulse water flow into the conveying pipeline according to the control instruction to clean the interception net and the inner wall of the pipeline. The impurities generated by cleaning are drained into the impurity collection chamber; Step 6: After the grease supply adjustment unit receives the control instruction from the intelligent decision-making and control unit, the variable frequency driven grease pump adjusts the speed according to the control instruction, the adjustable throttle valve adjusts the opening through the electric actuator, and the pressure compensation device monitors the pipeline pressure in real time and adjusts the output pressure according to the control instruction, completing the adjustment of the lubricating grease supply and pipeline pressure.

Images