System for automatic identification of device family defects based on spectral clustering and gnn

Abstract

This invention relates to an automatic identification system for equipment family defects based on spectral clustering and GNN, comprising a real-time data monitoring module, a graph model construction module, an adaptive learning module, a defect source tracing module, a predictive maintenance and early warning module, and a structured defect library and tag management module. After classifying equipment using spectral clustering, a graph model for predicting equipment family defects is constructed based on a graph neural network (GNN). For equipment exhibiting equipment family defects, a defect association graph is constructed to trace the origin and propagation path of the defects. Connecting to the hydropower industrial internet, leveraging the high interconnectivity, scalability, and visualization capabilities of the industrial internet platform, real-time online detection of equipment family defects is achieved, effectively identifying the root causes and propagation paths of these defects.

Description

Technical Field

[0001] This invention belongs to the field of hydropower industrial internet data processing, specifically involving an automatic identification system for equipment family defects based on spectral clustering and GNN. Background Technology

[0002] In the hydropower industry, the normal operation of equipment is crucial for ensuring safe production and improving economic efficiency. However, due to various reasons, these equipment may suffer from family-related defects—that is, defects or malfunctions of similar nature that are prevalent in a certain type or batch of products, equipment, or systems. These defects are usually related to common causes such as manufacturing, design, or material selection. The presence of such defects can potentially lead to serious accidents such as large-scale equipment shutdowns.

[0003] However, identifying familial defects presents significant challenges. Firstly, due to the unique nature of familial defects, many manufacturers and suppliers may not proactively disclose relevant information, making defect identification particularly difficult. Secondly, the hydropower industry involves a wide variety of equipment with diverse operating characteristics and data, resulting in significant differences between familial defects and a large overall number, further complicating defect identification.

[0004] Existing methods for identifying family defects in equipment often rely too heavily on feedback from operators and maintenance personnel, requiring long periods of accumulation to discover such defects. Furthermore, tracing the true cause of family defects is difficult, making them challenging to prevent and resolve. In addition, while regular inspections and maintenance are effective ways to detect family defects, these methods often require significant human and material resources and may lead to prolonged equipment downtime, greatly impacting production safety and efficiency. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned problems by providing an automatic identification system for equipment family defects based on spectral clustering and GNN. After classifying equipment using spectral clustering, a graph model for predicting equipment family defects is constructed based on a graph neural network (GNN). For equipment exhibiting equipment family defects, a defect association graph is built to track the origin and propagation path of the defects. Connecting to the hydropower industrial internet, and leveraging the high interconnectivity, scalability, and visualization capabilities of the industrial internet platform, real-time online detection of equipment family defects is achieved, effectively identifying the root causes and propagation paths of these defects.

[0006] The technical solution of this invention is an automatic identification system for family-based defects of equipment based on spectral clustering and GNN, including a real-time data monitoring module, a graph model construction module, an adaptive learning module, a defect source tracing module, a predictive maintenance and early warning module, a structured defect library and tag management module, and a user feedback module.

[0007] The real-time data monitoring module acquires equipment data through the hydropower industrial internet platform, transmits the collected equipment data to the database in real time, and stores the pre-processed data for use by the adaptive learning module and the defect source tracing module.

[0008] The graph model construction module uses spectral clustering and combines it with graph neural networks (GNNs) to construct a graph model, which is used to predict familial defects.

[0009] The adaptive learning module automatically updates and adjusts the graph model in the graph model construction module based on the new data output by the real-time data monitoring module.

[0010] The defect source tracing module is used to track the origin and propagation path of family defects in equipment.

[0011] The predictive maintenance and early warning module uses graph models to predict potential equipment family defects based on real-time data from the real-time data monitoring module, and issues timely early warnings for the predicted equipment family defects.

[0012] The structured defect library and tag management module are used to store, manage, and retrieve all information related to equipment family defects.

[0013] The user feedback module transmits user-reported equipment defects and related equipment feature data to the database, where they are preprocessed and stored in a structured defect database.

[0014] Preferably, the graph model construction module specifically includes:

[0015] The device classification submodule collects the data output by the real-time data detection module, generates a similarity matrix between devices based on the similarity of device data, and uses spectral clustering to classify similar devices into the same category.

[0016] The family defect prediction submodule uses a graph neural network (GNN) to predict equipment family defects. Based on the results of spectral clustering, it predicts family defects for each equipment category. i Constructing a subgraph ,in Let be the set of nodes in the subgraph, where each node represents a corresponding device. For the first i A set of edges in a subgraph of devices, where the edges represent the relationships between devices.

[0017] The graph model construction module, for nodes eigenvectors The iterative calculation is performed, and in each iteration, the feature of each node is updated as a function of its own and its neighbors' features. The iterative calculation formula is as follows:

[0018]

[0019] in, Represents a node Neighborhood set, u express The nodes in It is the first l The weight matrix of the order, For activation function, , They represent the first l sequence l The node obtained after +1 iterations eigenvectors.

[0020] After multiple iterations, each device will obtain the latest feature vector, and the output layer of the graph neural network (GNN) will predict the family defects of the devices based on the latest feature vector of each device.

[0021] Furthermore, the output layer of the graph neural network (GNN) predicts device family defects based on the latest feature vector of each device, specifically including:

[0022] Using the weight matrix of the output layer The latest feature vector of the device is transformed, and the transformation is calculated as follows:

[0023]

[0024] In the formula, Indicates the converted device v eigenvectors, For equipment v eigenvectors, This indicates the bias term.

[0025] The activation function of the output layer is based on The probability of a familial defect appearing in computing device v is calculated using the following formula:

[0026]

[0027] In the formula, Indicates equipment v The probability of having a family history of genetic defects.

[0028] like Then determine the device v There is a family history of the defect; otherwise, the equipment... vNo family history of defects.

[0029] Furthermore, during the training process of the Graph Neural Network (GNN), the output of the GNN is... Devices with training dataset v The true labels for familial defects are compared, and the loss is calculated using the binary cross-entropy loss function:

[0030]

[0031] In the formula, Indicates equipment v The true label for whether there is a familial defect is log(), which represents the logarithmic function.

[0032] Furthermore, the graph model in the graph model construction module is automatically updated and adjusted based on the new data output by the real-time data monitoring module. For the current graph model, where For the parameters of the graph model, The input data for the graph model is the device feature data matrix.

[0033] The goal of updating and adjusting the graph model in the graph model building module is to update Minimize the loss of the graphical model on the new data, and update The formula for calculation is:

[0034] ;

[0035] in, This represents a new device feature data matrix, where each row represents a device and each column represents a feature. express The corresponding family defect label vector, The loss function; This represents the parameters of the updated graph model; argmin is the index function for finding the minimum value.

[0036] After calculating the new parameters of the graphical model, the graphical model building module uses the new device feature data. Update the similarity matrix between devices based on existing data. S Use the updated version S Update the structure and weights of the graph neural network (GNN).

[0037] Furthermore, the defect source tracing module determines the equipment, components, and timestamps related to the familial defect based on the data obtained from the real-time data monitoring module and the user feedback module.

[0038] Construct a defect association graph for each known familial defect ,in This represents a set of defective nodes, where each defective node represents a defective device. For the set of defective edges, , Indicates equipment and equipment Defect correlation between them;

[0039] Based on the similarity between devices, supply chain information, and timestamps, defect edges are defined in the defect association graph. Assigning defect weights , Indicates equipment and equipment To determine the probability of defect propagation, Dijkstra's shortest path algorithm is used to find the defect edge with the largest defect weight, and the source of the defect is determined based on the defect edge.

[0040] Preferably, when the automatic identification system for equipment family defects detects a new equipment family defect or receives feedback on an equipment defect from a user, the structured defect library and tag management module stores the equipment defect information and equipment feature data in the structured defect library. The structured defect library includes not only the description information and feature data of the equipment defect, but also information about the related equipment, the time of occurrence of the defect, and the scope of the defect's impact.

[0041] Preferably, the structured defect library and tag management module further includes a tag management submodule, which is used to add, modify or delete tags for equipment defects. The tags are set based on various factors such as the type, severity and source of the equipment defect, and the tags are used to quickly retrieve specific types of equipment defects.

[0042] Preferably, the structured defect library and tag management module further includes a defect feedback submodule, which is used by the maintenance team or experts to provide feedback information on the data in the structured defect library; when existing equipment defects are further classified, the defect feedback submodule provides the function of editing, adding or modifying tags.

[0043] The above-mentioned method for automatic identification of device family defects based on spectral clustering and GNN includes the following steps:

[0044] S1. The collected equipment data is transmitted to the database in real time, the equipment data is preprocessed using a data preprocessing component, and the preprocessed equipment data is stored.

[0045] S2. After classifying the equipment using spectral clustering, a graph model of the equipment is constructed using a graph neural network (GNN). The graph model is used to predict family-related defects of the equipment.

[0046] S3. Based on the new data of the equipment obtained by the real-time data monitoring module, update and adjust the graph model in the graph model building module to adapt the graph model to the new operating conditions and changes of the equipment.

[0047] S4. Based on real-time equipment data, a graph model is used to predict equipment family defects, and timely warnings are issued for the predicted equipment family defects.

[0048] S5. For equipment exhibiting family-related defects, construct a defect association diagram to trace the origin and propagation path of the family-related defects.

[0049] Compared with the prior art, the beneficial effects of the present invention include:

[0050] 1) This invention constructs a graphical model for predicting equipment family defects and a defect association graph for tracking the origin and propagation path of equipment family defects. By integrating with an industrial internet platform, it realizes real-time online detection of equipment family defects, improves the accuracy of equipment family defect detection, reduces false alarms and false negatives, and can effectively identify the root causes and propagation paths of equipment family defects. The dual-channel analysis of real-time data stream (time dimension) and equipment association graph (spatial dimension) significantly improves the prediction accuracy compared with single-dimensional methods.

[0051] 2) This invention adopts a joint modeling strategy of spectral clustering (explicit topology) and GNN (implicit topology). By using the topological complementarity mechanism, it breaks through the cognitive limitations of traditional single-model representation. This design significantly improves the defect identification rate of the model in cross-supplier and cross-model equipment clusters, and shows a significantly better generalization ability than traditional methods, especially in the scenario of accessing new heterogeneous equipment.

[0052] 3) This invention introduces a parameter update formula to the graph model, realizing online incremental learning of model parameters, breaking through the bottleneck of traditional GNN static training, and enabling the device family defect detection model of this invention to have dynamic learning and adaptability. It can continuously learn and optimize based on new data and user feedback, ensuring a continuous improvement in defect detection accuracy.

[0053] 4) This invention constructs a defect association graph and uses the Dijkstra algorithm to search for the maximum defect weight path, combined with timestamps and supply chain information, to achieve efficient and accurate location of defect sources and effectively identify the root causes and propagation paths of equipment family defects.

[0054] 5) The user feedback module provided by this invention, combined with the structured defect library, forms a "data-model-knowledge" closed loop with the hydropower industrial internet platform. The autonomous evolution of the equipment defect knowledge base of the hydropower industrial internet platform is realized through dynamic labeling.

[0055] 6) This invention achieves technological breakthroughs and synergistic efficiency in the entire defect detection process through deep integration with the industrial internet platform. By connecting with the industrial internet platform, it enables real-time data synchronization between equipment and SCADA and MIS systems within seconds, improving the coverage of detection data. Combined with the industrial internet platform, the defect detection results are visualized, mapping equipment family defects, sources, and propagation paths to digital twin objects on the industrial internet platform, making it easier for maintenance personnel to quickly locate the source of the fault and take targeted measures. It also realizes the structured and tagged management of the equipment defect database and synchronizes it with the equipment defect knowledge base of the hydropower industrial internet platform, forming a quality traceability chain for the entire equipment lifecycle. Attached Figure Description

[0056] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0057] Figure 1 This is a schematic diagram of an automatic identification system for device family defects based on spectral clustering and GNN, according to an embodiment of the present invention.

[0058] Figure 2 This is a flowchart illustrating the device family defect identification method according to an embodiment of the present invention. Detailed Implementation

[0059] Example 1

[0060] like Figure 1 As shown, the automatic identification system for family-based defects of equipment based on spectral clustering and GNN includes a real-time data monitoring module, a graph model construction module, an adaptive learning module, a defect source tracing module, a predictive maintenance and early warning module, a structured defect library and tag management module, and a user feedback module.

[0061] The real-time data monitoring module transmits the collected equipment data to the database in real time and stores the pre-processed data for use by the adaptive learning module and the defect source tracing module.

[0062] In this embodiment, the real-time data monitoring module is connected to the hydropower industrial internet platform, and through the industrial internet platform, it connects to the sensor group, SCADA system, and MIS system to obtain data for the detection of equipment family defects.

[0063] The graph model construction module uses spectral clustering and combines it with graph neural networks (GNNs) to construct a graph model, which is used to predict familial defects.

[0064] The graph model construction module specifically includes:

[0065] The device classification submodule collects the data output by the real-time data detection module, generates a similarity matrix between devices based on the similarity of device data, and uses spectral clustering to classify similar devices into the same category.

[0066] The family defect prediction submodule uses a graph neural network (GNN) to predict equipment family defects. Based on the results of spectral clustering, it predicts family defects for each equipment category. i Constructing a subgraph ,in Let be the set of nodes in the subgraph, where each node represents a corresponding device. For the first i A set of edges in a subgraph of devices, where the edges represent the relationships between devices.

[0067] For nodes eigenvectors The iterative calculation is performed, and in each iteration, the feature of each node is updated as a function of its own and its neighbors' features. The iterative calculation formula is as follows:

[0068]

[0069] in, Represents a node Neighborhood set, u express The nodes in It is the first l The weight matrix of the order, For activation function, , They represent the first l sequence l The node obtained after +1 iterations eigenvectors.

[0070] After multiple iterations, each device will obtain the latest feature vector, and the output layer of the graph neural network (GNN) will predict the family defects of the devices based on the latest feature vector of each device.

[0071] The output layer of the Graph Neural Network (GNN) predicts device family defects based on the latest feature vector of each device, specifically including:

[0072] Using the weight matrix of the GNN output layer The latest feature vector of the device is transformed, and the transformation is calculated as follows:

[0073]

[0074] In the formula, Indicates the converted device veigenvectors, For equipment v eigenvectors, This indicates the bias term.

[0075] The activation function of the output layer is based on The probability of a familial defect appearing in computing device v is calculated using the following formula:

[0076]

[0077] In the formula, Indicates equipment v The probability of having a family history of genetic defects.

[0078] like Then determine the device v There is a family history of the defect; otherwise, the equipment... v No family history of defects.

[0079] During the training of a Graph Neural Network (GNN), the output of the GNN is... Devices with training dataset v The true labels for familial defects are compared, and the loss is calculated using the binary cross-entropy loss function:

[0080]

[0081] In the formula, Indicates equipment v The true label for whether there is a familial defect is log(), which represents the logarithmic function.

[0082] The adaptive learning module automatically updates and adjusts the graph model in the graph model building module based on the new data output by the real-time data monitoring module.

[0083] set up For the current graph model, where For the parameters of the graph model, The input data for the graph model is the device feature data matrix.

[0084] The goal of updating and adjusting the graph model in the graph model building module is to update Minimize the loss of the graphical model on the new data, and update The formula for calculation is:

[0085] ;

[0086] in, This represents a new device feature data matrix, where each row represents a device and each column represents a feature. express The corresponding family defect label vector, The loss function; This represents the parameters of the updated graph model; argmin is the index function for finding the minimum value.

[0087] After calculating the new parameters of the graphical model, the graphical model building module uses the new device feature data. Update the similarity matrix between devices based on existing data. S Use the updated version S Update the structure and weights of the graph neural network (GNN).

[0088] The defect source tracing module is used to track the origin and propagation path of family defects in equipment.

[0089] Based on data obtained from the real-time data monitoring module and the user feedback module, the devices, components, and timestamps associated with familial defects are identified.

[0090] Construct a defect association graph for each known familial defect ,in This represents a set of defective nodes, where each defective node represents a defective device. For the set of defective edges, , Indicates equipment and equipment Defect correlation between them;

[0091] Based on the similarity between devices, supply chain information, and timestamps, defect edges are defined in the defect association graph. Assigning defect weights , Indicates equipment and equipment To determine the probability of defect propagation between nodes, Dijkstra's shortest path algorithm is used to find the defect edge with the largest defect weight, and the source of the defect is determined based on the defect edge.

[0092] The predictive maintenance and early warning module uses graph models to predict potential equipment family defects based on real-time data from the real-time data monitoring module, and issues timely warnings for the predicted equipment family defects.

[0093] The structured defect library and tag management module are used to store, manage, and retrieve all information related to equipment family defects.

[0094] The structured defect library and tag management module includes a tag management submodule and a defect feedback submodule.

[0095] The tag management submodule is used to add, modify, or delete tags for equipment defects. Tags are set based on various factors such as the type, severity, and source of the equipment defect. These tags are used to quickly retrieve specific types of equipment defects.

[0096] The defect feedback submodule is used to provide feedback information to the maintenance team or experts on the data in the structured defect library; when existing equipment defects are further classified, the defect feedback submodule provides the function of editing, adding or modifying tags.

[0097] The user feedback module transmits user-reported equipment defects and related equipment feature data to the database, where they are preprocessed and stored in a structured defect database.

[0098] When the automatic identification system for equipment family defects detects a new equipment family defect or receives feedback on an equipment defect from a user, the structured defect library and tag management module stores the equipment defect information and equipment feature data in the structured defect library. The structured defect library includes not only the description information and feature data of the equipment defect, but also information on the related equipment, the time of occurrence of the defect, and the scope of the defect's impact.

[0099] In this embodiment, the user feedback module and the structured defect library are combined with the hydropower industrial internet platform to form a "data-model-knowledge" closed loop. Through online real-time detection of equipment family defects, defect root cause identification, structured management of the equipment defect library and dynamic labeling of tags, as well as synchronization with the equipment defect knowledge base of the hydropower industrial internet platform, the autonomous evolution of the equipment defect knowledge base of the hydropower industrial internet platform is realized.

[0100] The results show that the combination of this invention with the hydropower industrial internet platform forms a quality traceability chain for the entire equipment lifecycle, which reduces the defect reproduction rate by 82% and increases the accuracy of spare parts replacement to 57.3%, generating synergistic benefits that significantly exceed the expectations of the technology combination.

[0101] Example 2

[0102] like Figure 2 As shown, the method for automatic identification of device family defects based on spectral clustering and GNN includes the following steps:

[0103] S1. The collected device data is transmitted to the database in real time, and the device data is preprocessed using a data preprocessing component and then stored.

[0104] S2. After classifying the equipment using spectral clustering, a graph model of the equipment is constructed using a graph neural network (GNN). The graph model is used to predict family-related defects of the equipment.

[0105] Based on the results of spectral clustering, for each device categoryi Constructing a subgraph ,in Let be the set of nodes in the subgraph, where each node represents a corresponding device. For the first i A set of edges in a subgraph of devices, where the edges represent the relationships between devices.

[0106] For nodes eigenvectors The iterative calculation is performed, and in each iteration, the feature of each node is updated as a function of its own and its neighbors' features. The iterative calculation formula is as follows:

[0107]

[0108] in, Represents a node Neighborhood set, u express The nodes in It is the first l The weight matrix of the order, For activation function,

[0109] , They represent the first l sequence l The node obtained after +1 iterations eigenvectors;

[0110] After multiple iterations, each device will obtain the latest feature vector, and the output layer of the graph neural network (GNN) will predict the family defects of the devices based on the latest feature vector of each device.

[0111] The output layer of the Graph Neural Network (GNN) predicts device family defects based on the latest feature vector of each device, specifically including:

[0112] Using the weight matrix of the output layer The latest feature vector of the device is transformed, and the transformation is calculated as follows:

[0113]

[0114] In the formula, Indicates the converted device v eigenvectors, For equipment v eigenvectors, Indicates the bias term;

[0115] The activation function of the output layer is based on The probability of a familial defect appearing in computing device v is calculated using the following formula:

[0116]

[0117] In the formula, Indicates equipment v The probability of having a family history of genetic defects.

[0118] like Then determine the device v There is a family history of the defect; otherwise, the equipment... v No family history of defects.

[0119] During the training of a Graph Neural Network (GNN), the output of the GNN is... Devices with training dataset v The true labels for familial defects are compared, and the loss is calculated using the binary cross-entropy loss function:

[0120]

[0121] In the formula, Indicates equipment v The true label for whether there is a familial defect is log(), which represents the logarithmic function.

[0122] S3. Based on the new data of the equipment obtained by the real-time data monitoring module, update and adjust the graph model in the graph model building module to adapt the graph model to the new operating conditions and changes of the equipment.

[0123] set up For the current graph model, where For the parameters of the graph model, The input data for the graph model is the device feature data matrix;

[0124] The goal of updating and adjusting the graph model in the graph model building module is to update Minimize the loss of the graphical model on the new data, and update The formula for calculation is:

[0125] ;

[0126] in, This represents a new device feature data matrix, where each row represents a device and each column represents a feature. express The corresponding family defect label vector, The loss function; This represents the parameters of the updated graph model; argmin is the index function for finding the minimum value.

[0127] S4. Based on the equipment data in the database, a graph model is used to predict equipment family defects, and timely warnings are issued for the predicted equipment family defects.

[0128] S5. For equipment exhibiting family-related defects, construct a defect association diagram to trace the origin and propagation path of the family-related defects.

[0129] Based on real-time device-related data and user feedback data, identify the devices, components, and timestamps associated with familial defects.

[0130] Construct a defect association graph for each known familial defect ,in This represents a set of defective nodes, where each defective node represents a defective device. For the set of defective edges, , Indicates equipment and equipment Defect correlation between them;

[0131] Based on the similarity between devices, supply chain information, and timestamps, defect edges are defined in the defect association graph. Assigning defect weights , Indicates equipment and equipment To determine the probability of defect propagation, Dijkstra's shortest path algorithm is used to find the defect edge with the largest defect weight, and the source of the defect is determined based on the defect edge.

[0132] S6. Store, manage, and retrieve information and data related to device family defects.

Claims

1. An automatic identification system for equipment family defects based on spectral clustering and GNN, characterized in that, Includes the following modules: The real-time data monitoring module transmits the collected equipment data to the database in real time and stores the pre-processed data for use by the adaptive learning module and the defect source tracing module. The real-time data monitoring module communicates with the hydropower industrial internet platform and connects with the sensor group and SCADA system through the industrial internet platform to obtain data for the detection of equipment family defects. The graph model construction module uses spectral clustering and combines it with graph neural networks (GNNs) to construct a graph model, which is used to predict familial defects. The adaptive learning module automatically updates and adjusts the graph model in the graph model building module based on the new data output by the real-time data monitoring module. The defect source tracing module is used to track the origin and propagation path of family defects in equipment; The predictive maintenance and early warning module uses graph models to predict equipment family defects based on real-time data obtained from the real-time data monitoring module, and issues timely warnings for the predicted equipment family defects. The structured defect library and tag management module are used to store, manage, and retrieve all information related to equipment family defects; The user feedback module transmits user-reported equipment defects and related equipment feature data to the database, where they are preprocessed and then stored in the structured defect database. The graph model construction module specifically includes: The equipment classification submodule collects the data output by the real-time data monitoring module, generates a similarity matrix between equipment based on the similarity of equipment data, and uses spectral clustering to classify similar equipment into the same category. The family defect prediction submodule uses a graph neural network (GNN) to predict equipment family defects. Based on the results of spectral clustering, it predicts the family defects for each equipment category. i Constructing a subgraph ,in Let be the set of nodes in the subgraph, where each node represents a corresponding device. For the first i The set of edges in a subgraph of devices, where the edges represent the relationships between devices; The defect source tracing module determines the equipment, components, and timestamps related to the family defect based on data obtained from the real-time data monitoring module and the user feedback module. Construct a defect association diagram for each device family defect ,in This represents a set of defective nodes, where each defective node represents a defective device. For the set of defective edges, , Indicates device and equipment Defect correlation between them; Based on the similarity between devices, supply chain information, and timestamps, defect edges are defined in the defect association graph. Assigning defect weights , Indicates device and equipment To determine the probability of defect propagation between nodes, Dijkstra's shortest path algorithm is used to find the defect edge with the largest defect weight, and the source of the defect is determined based on the defect edge. The user feedback module, combined with the structured defect database, forms a "data-model-knowledge" closed loop on the hydropower industrial internet platform. Through online real-time detection of equipment family defects, defect root cause identification, structured management of the equipment defect database, dynamic labeling of tags, and synchronization with the equipment defect knowledge base of the hydropower industrial internet platform, the autonomous evolution of the equipment defect knowledge base of the hydropower industrial internet platform is achieved.

2. The automatic identification system for device family defects based on spectral clustering and GNN according to claim 1, characterized in that, The graph model construction module, for nodes eigenvectors The iterative calculation is performed, and in each iteration, the feature of each node is updated as a function of its own and its neighbors' features. The iterative calculation formula is as follows: ； in, Represents a node Neighborhood set, u express The nodes in It is the first l The weight matrix of the order, For activation function, , They represent the first l sequence l The node obtained after +1 iterations eigenvectors; After multiple iterations, the latest feature vector for each device is obtained. The output layer of the graph neural network (GNN) then predicts device family defects based on the latest feature vector for each device.

3. The automatic identification system for device family defects based on spectral clustering and GNN according to claim 2, characterized in that, In the graph model construction module, the output layer of the graph neural network (GNN) predicts device family defects based on the latest feature vector of each device, specifically including: Using the weight matrix of the GNN output layer The latest feature vector of the device is transformed, and the transformation is calculated as follows: ； In the formula, Indicates the converted device v eigenvectors, For equipment v eigenvectors, Indicates the bias term; The activation function of the output layer is based on The probability of a familial defect appearing in computing device v is calculated using the following formula: ； In the formula, Indicates device v The probability of having a family history of genetic defects.

4. The automatic identification system for device family defects based on spectral clustering and GNN according to claim 3, characterized in that, In the adaptive learning module, the graph model in the graph model construction module is automatically updated and adjusted based on the new data output by the real-time data monitoring module. For the current graph model, where For the parameters of the graph model, The input data for the graph model is the device feature data matrix; The goal of updating and adjusting the graph model in the graph model building module is to update Minimize the loss of the graphical model on the new data, and update The formula for calculation is: ； in, This indicates new device characteristic data. express The corresponding family defect label vector, The loss function; This represents the parameters of the updated graph model; argmin is the index function for finding the minimum value.

5. The automatic identification system for device family defects based on spectral clustering and GNN according to claim 4, characterized in that, After calculating the new parameters of the graphical model, the graphical model building module uses the new device feature data. Update the similarity matrix between devices based on existing data. S Use the updated version S Update the structure and weights of the graph neural network (GNN).

6. The automatic identification system for device family defects based on spectral clustering and GNN according to claim 1, characterized in that, When the automatic identification system for equipment family defects detects a new equipment family defect or receives feedback on an equipment defect from a user, the structured defect library and tag management module stores the equipment defect information and equipment feature data in the structured defect library. The structured defect library includes not only the description information and feature data of the equipment defect, but also information on the related equipment, the time of occurrence of the defect, and the scope of the defect's impact.

7. The automatic identification system for device family defects based on spectral clustering and GNN according to any one of claims 1-6, characterized in that, The structured defect library and tag management module also includes a tag management submodule, which is used to add, modify or delete tags for equipment defects. The tags are set based on various factors such as the type, severity and source of the equipment defect, and are used to quickly retrieve specific types of equipment defects.

8. The method for automatic identification of device family defects based on spectral clustering and GNN as described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Transmit the collected device data to the database in real time, preprocess the device data, and store the preprocessed device data; S2. After classifying the equipment using spectral clustering, a graph model of the equipment is constructed using a graph neural network (GNN). The graph model is used to predict family-related defects of the equipment. S3. Based on the new data of the equipment obtained by the real-time data monitoring module, update and adjust the graph model in the graph model building module to adapt the graph model to the new operating conditions and changes of the equipment. S4. Based on real-time equipment data, a graph model is used to predict equipment family defects, and timely warnings are issued for the predicted equipment family defects. S5. For equipment exhibiting family-related defects, construct a defect association diagram to trace the origin and propagation path of the family-related defects; S6. Store, manage, and retrieve information and data related to device family defects.

Images