Automatic management method and system for operation and maintenance fault detection

Abstract

The invention belongs to the field of fault detection, and particularly relates to an automatic management method and system for operation and maintenance fault detection, and the method comprises the steps of collection network monitoring, data storage, operation and maintenance self-inspection, user query and maintenance request. According to the scheme, the GTN model is adopted, the input log message is converted into the directed graph, the fault is detected through the node features and the spatial structure of the graph, the structure of the directed graph and the semantics of the log message are combined, and the accuracy and stability of fault detection are improved; a binary file system is constructed, disk space and redundant information are reduced, a search engine can conveniently and deeply dig specific content, and needed information can be found more quickly; the multi-stage drill-down search engine is adopted, the search range is gradually refined and focused according to the real-time search requirement of the user, the accuracy of the search result is improved, and the interactive experience of user query is optimized.

Description

Technical Field

[0001] This invention belongs to the field of fault detection, specifically referring to an automated management method and system for operation and maintenance fault detection. Background Technology

[0002] As gas station networks grow in scale and business demands, they require continuous maintenance, operation, and fault monitoring to ensure data security, help maintenance personnel quickly identify and handle faults, and reduce economic losses and repair costs. However, traditional log-based fault detection methods cannot fully utilize the spatial structure relationships between log events, increasing the possibility of false alarms and leading to performance instability. Traditional storage file systems have limitations in file format, making them unsuitable for storing large amounts of data specifically for search engines. Traditional search engines used for maintenance fault detection cannot reasonably infer user search intent based on historical search records. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an automated management method and system for operation and maintenance fault detection. Addressing the limitations of traditional log-based fault detection methods, which fail to fully utilize the spatial structural relationships between log events, increasing the possibility of false alarms and leading to performance instability, this solution employs a GTN model. It transforms input log messages into a directed graph, utilizing node features and the spatial structure of the graph to detect faults. By combining the structure of the directed graph with the semantics of log messages, it improves the accuracy and stability of fault detection, overcoming the limitations of under-reach and over-suppression in graph neural network stacking layers during message transmission. Furthermore, addressing the limitations of traditional storage file systems in terms of file format restrictions and inability to specifically support large-scale data storage for search engines, this solution constructs a binary file system. Storing data in binary form, files can be directly loaded into main memory with the help of basic array structures in any language, reducing disk space and redundant information. This facilitates search engines in deeply exploring specific content and finding the required information more quickly. Finally, addressing the limitation of traditional search engines used for operation and maintenance fault detection in their inability to reasonably infer user search intent based on historical search records, this solution employs a multi-level drill-down search engine. Based on the user's real-time search needs, it progressively refines and focuses the search scope, improving the accuracy of search results and optimizing the user's query experience.

[0004] The technical solution adopted by this invention is as follows: This invention provides an automated management method and system for operation and maintenance fault detection, wherein the automated management system for operation and maintenance fault detection includes a network monitoring module, a data storage module, an operation and maintenance self-inspection module, a user query module, and a maintenance reporting module; the user query module is equipped with a server;

[0005] The branch monitoring module monitors the log messages of all branch equipment and software and transmits them to the data storage module;

[0006] The data storage module receives log messages from the site monitoring module, establishes a binary file system, and sends the log messages to the operation and maintenance self-test module.

[0007] The operation and maintenance self-test module receives log messages from the data storage module, uses the GTN model to find the fault types of the network equipment and network software, and sends the fault types to the maintenance and repair module.

[0008] The user query module uses a multi-level drill-down search engine, allowing users to search and view log messages and fault types in real time.

[0009] The maintenance and repair module automatically assigns pre-defined solutions for the corresponding fault types to the faulty equipment and software at the service points.

[0010] The automated management method for operation and maintenance fault detection includes the following steps:

[0011] Step S1: Site monitoring, monitor the log messages of all site devices and site software, and transmit them to the data storage module;

[0012] Step S2: Data storage. The data storage module receives log messages from the site monitoring module, establishes a binary file system, and sends the log messages to the operation and maintenance self-test module.

[0013] Step S3: Operation and maintenance self-check, receive log messages from the data storage module, use the GTN model to find the fault type of the network equipment and network software, and send the fault type to the maintenance and repair module;

[0014] Step S4: User query, using a multi-level drill-down search engine, allows users to search and view log messages and fault types in real time;

[0015] Step S5: Maintenance and Repair Reporting. Receive the fault type from the operation and maintenance self-test module, automatically assign corresponding backup solutions to the faulty network equipment and software, and generate automated operation and maintenance reports.

[0016] As a further improvement to this solution, in step S1, the network monitoring specifically includes the following steps:

[0017] Step S11: Monitor the log messages of all network equipment, including oil dispensers, level gauges, and local databases. Transmit the log messages of the network equipment to the data storage module and update them once a day.

[0018] Step S12: Monitor the log messages of all branch software displaying the branch service status. The branch software includes the management system, POS server, POS client and invoice system. The service status includes the installation version, branch user status, running status and software environment. Transmit the log messages of the branch software to the data storage module and update them every three hours.

[0019] As a further improvement to this solution, in step S2, data storage specifically includes the following steps:

[0020] Step S21: The data storage module receives log messages from the branch monitoring module, compresses and backs up the log messages, and automatically cleans them up after seven days.

[0021] Step S22: Extract keywords from log messages and classify the keywords using the following classification formula:

[0022] ;

[0023] In the formula, It is a keyword. It's a classification of keywords. It's about keyword ranking. It refers to the number of times the keyword appears. It is the cumulative sum of keyword rankings; keywords with similar cumulative rankings are considered to belong to the same category.

[0024] Step S23: Create binary header and data files for each category, load the header files into main memory, and store the data files on disk.

[0025] As a further improvement to this solution, in step S3, the operation and maintenance self-check includes the following steps:

[0026] Step S31: Graph representation, converting log messages into a directed graph, specifically including the following steps:

[0027] Step S311: Semantic awareness. The log interpreter Drain interprets log messages as log events. The BERT model is used to extract node features from log events and generate fixed-dimensional semantic vectors. A semantic vector is defined as a node. All semantic vectors are arranged in chronological order to form a log sequence.

[0028] Step S312: Perceive the structure. Based on the log sequence, generate a degree matrix reflecting the local topology, a distance matrix reflecting the global spatial structure, and an edge weight matrix reflecting the number of edges between nodes.

[0029] Step S32: Graph construction. The GTN module is used to determine the fault type of the log message. The GTN module includes self-focused units and a location feedforward network, and specifically includes the following steps:

[0030] Step S321: The self-focused unit performs independent linear transformations on the node features, projecting the node features into a query matrix, a key matrix, and a numerical matrix, as shown in the following formula:

[0031] ;

[0032] ;

[0033] ;

[0034] in, , , These are the query matrix, key matrix, and value matrix, respectively. It is a node feature. , , These are the learnable parameters for query, key, and value, respectively.

[0035] Step S322: Merge the edge weight matrix into and middle:

[0036] ;

[0037] ;

[0038] ;

[0039] In the formula, , , These are the first-level query matrix, the first-level key matrix, and the first-level numerical matrix of the GTN module, respectively. This represents the sum of the edge weight matrix along the row direction. This represents the Hadamard product operation of matrices;

[0040] Step S323: Calculate the spatial bias term, using the following formula:

[0041] ;

[0042] In the formula, For spatial bias, It is the first query matrix Query elements, It is the first of the key matrix Each key element It's a dot product operation. It is the relative path distance between nodes;

[0043] Step S324: Calculate the attention coefficient matrix, using the following formula:

[0044] ;

[0045] In the formula, Pay attention to the coefficient matrix. It is the output dimension of the linear transformation;

[0046] Step S325: Calculate the attention weight matrix, using the following formula:

[0047] ;

[0048] In the formula, It is the attention weight matrix. It is an exponential operation with base e. It is the total number of nodes. Indicates the base e of the first... Pay attention to the cumulative sum of the elements of the coefficient matrix;

[0049] Step S326: Calculate hidden features:

[0050] ;

[0051] In the formula, It is the first One hidden feature, Indicates the first Column traversal from 1 to The cumulative sum over time, It is the first of the numerical matrices One value element;

[0052] Step S327: Aggregate nodes and node features into a directed graph representation, as shown in the following formula:

[0053] ;

[0054] In the formula, It is represented as a directed graph. It is the first node set 1 node yes The Layer node characteristics, yes The cumulative sum, Is seeking The maximum value, It is a parameterized abstract aggregate function with parameters;

[0055] Step S33: Graph classification, outputting fault types, specifically including the following steps:

[0056] Step S331: Construct and train the GTN model. The location feedforward network contains three fully connected layers with Gaussian error linear units as activation functions, using cross-entropy as the loss function. The directed graph represents the normalized input feedforward network, and the output is the predicted fault type, as shown in the following formula:

[0057] ;

[0058] In the formula, It predicts the type of failure. It is a function of probability transformation. It is a location feedforward network;

[0059] Step S332: Input the directed graph representation extracted from the log message into the trained GTN model, and the GTN model outputs the fault type.

[0060] As a further improvement to this solution, in step S4, the user queries, specifically including the following steps:

[0061] Step S41: Capture the keywords searched by the user and search for them in the header file to obtain the offset of the keywords;

[0062] Step S42: Based on the offset, locate the data file on the disk to obtain the main search page;

[0063] Step S43: Find the cached available pages from the server, intersect them with the main search page to generate a common page, and cache the common page as a new available page in parallel on the server in the form of a key-value dictionary object;

[0064] Step S44: If the user's search has not yet ended, repeat steps S41 to S43; otherwise, render and display the final searched public page on the user interface.

[0065] The beneficial effects achieved by the present invention using the above solution are as follows:

[0066] (1) In view of the problem that traditional log-based fault detection methods cannot make full use of the spatial structural relationship between log events, which increases the possibility of potential false alarms and leads to unstable performance, this solution adopts the GTN model to convert the input log message into a directed graph, and uses node features and the spatial structure of the graph to detect faults. It combines the structure of the directed graph with the semantics of the log message to improve the accuracy and stability of fault detection and overcome the limitations of underreach and oversuppression of graph neural network stacked layers during message transmission.

[0067] (2) In view of the problem that traditional storage file systems are limited by file format and cannot be used to store large data for search engines, this solution constructs a binary file system to store data in binary form. Files can be directly loaded into main memory with the help of basic array structures in any language, reducing disk space and redundant information, making it easier for search engines to dig deeper into specific content and find the information they need more quickly.

[0068] (3) Traditional search engines used for fault detection in operation and maintenance have the problem of not being able to reasonably guess the user's search intent based on the user's historical search records. This solution adopts a multi-level drill-down search engine, which gradually refines and focuses the search scope according to the user's real-time search needs, thereby improving the accuracy of search results and optimizing the user's interactive experience. Attached Figure Description

[0069] Figure 1 A flowchart illustrating an automated management method for operation and maintenance fault detection provided by the present invention;

[0070] Figure 2 This is a schematic diagram of an automated management system for operation and maintenance fault detection provided by the present invention.

[0071] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0072] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0073] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0074] Example 1, see Figure 1 and Figure 2This invention provides an automated management method and system for operation and maintenance fault detection. The automated management system for operation and maintenance fault detection includes a network monitoring module, a data storage module, an operation and maintenance self-inspection module, a user query module, and a maintenance reporting module; the user query module is equipped with a server.

[0075] The branch monitoring module monitors the log messages of all branch equipment and software and transmits them to the data storage module;

[0076] The data storage module receives log messages from the site monitoring module, establishes a binary file system, and sends the log messages to the operation and maintenance self-test module.

[0077] The operation and maintenance self-test module receives log messages from the data storage module, uses the GTN model to find the fault types of the network equipment and network software, and sends the fault types to the maintenance and repair module.

[0078] The user query module uses a multi-level drill-down search engine, allowing users to search and view log messages and fault types in real time.

[0079] The maintenance and repair module automatically assigns pre-defined solutions for the corresponding fault types to the faulty equipment and software at the service points.

[0080] The automated management method for operation and maintenance fault detection includes the following steps:

[0081] Step S1: Site monitoring, monitor the log messages of all site devices and site software, and transmit them to the data storage module;

[0082] Step S2: Data storage. The data storage module receives log messages from the site monitoring module, establishes a binary file system, and sends the log messages to the operation and maintenance self-test module.

[0083] Step S3: Operation and maintenance self-check, receive log messages from the data storage module, use the GTN model to find the fault type of the network equipment and network software, and send the fault type to the maintenance and repair module;

[0084] Step S4: User query, using a multi-level drill-down search engine, allows users to search and view log messages and fault types in real time;

[0085] Step S5: Maintenance and Repair Reporting. Receive the fault type from the operation and maintenance self-test module, automatically assign corresponding backup solutions to the faulty network equipment and software, and generate automated operation and maintenance reports.

[0086] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the network monitoring method specifically includes the following steps:

[0087] Step S11: Monitor the log messages of all network equipment, including oil dispensers, level gauges, and local databases. Transmit the log messages of the network equipment to the data storage module and update them once a day.

[0088] Step S12: Monitor the log messages of all branch software displaying the branch service status. The branch software includes the management system, POS server, POS client and invoice system. The service status includes the installation version, branch user status, running status and software environment. Transmit the log messages of the branch software to the data storage module and update them every three hours.

[0089] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, data storage is performed, and the specific method includes the following steps:

[0090] Step S21: The data storage module receives log messages from the branch monitoring module, compresses and backs up the log messages, and automatically cleans them up after seven days.

[0091] Step S22: Extract keywords from log messages and classify the keywords using the following classification formula:

[0092] ;

[0093] In the formula, It is a keyword. It's a classification of keywords. It's about keyword ranking. It refers to the number of times the keyword appears. It is the cumulative sum of keyword rankings; keywords with similar cumulative rankings are considered to belong to the same category.

[0094] Step S23: Create binary header and data files for each category, load the header files into main memory, and store the data files on disk.

[0095] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, the operation and maintenance self-check includes the following steps:

[0096] Step S31: Graph representation, converting log messages into a directed graph, specifically including the following steps:

[0097] Step S311: Semantic awareness. The log interpreter Drain interprets log messages as log events. The BERT model is used to extract node features from log events and generate fixed-dimensional semantic vectors. A semantic vector is defined as a node. All semantic vectors are arranged in chronological order to form a log sequence.

[0098] Step S312: Perceive the structure. Based on the log sequence, generate a degree matrix reflecting the local topology, a distance matrix reflecting the global spatial structure, and an edge weight matrix reflecting the number of edges between nodes.

[0099] Step S32: Graph construction. The GTN module is used to determine the fault type of the log message. The GTN module includes self-focused units and a location feedforward network, specifically including the following steps:

[0100] Step S321: The self-focused unit performs independent linear transformations on the node features, projecting the node features into a query matrix, a key matrix, and a numerical matrix, as shown in the following formula:

[0101] ;

[0102] ;

[0103] ;

[0104] in, , , These are the query matrix, key matrix, and value matrix, respectively. It is a node feature. , , These are the learnable parameters for query, key, and value, respectively.

[0105] Step S322: Merge the edge weight matrix into and middle:

[0106] ;

[0107] ;

[0108] ;

[0109] In the formula, , , These are the first-level query matrix, the first-level key matrix, and the first-level numerical matrix of the GTN module, respectively. This represents the sum of the edge weight matrix along the row direction. This represents the Hadamard product operation of matrices;

[0110] Step S323: Calculate the spatial bias term, using the following formula:

[0111] ;

[0112] In the formula, For spatial bias, It is the first query matrix Query elements, It is the first of the key matrix Each key element It's a dot product operation. It is the relative path distance between nodes;

[0113] Step S324: Calculate the attention coefficient matrix, using the following formula:

[0114] ;

[0115] In the formula, Pay attention to the coefficient matrix. It is the output dimension of the linear transformation;

[0116] Step S325: Calculate the attention weight matrix, using the following formula:

[0117] ;

[0118] In the formula, It is the attention weight matrix. It is an exponential operation with base e. It is the total number of nodes. Indicates the base e of the first... Pay attention to the cumulative sum of the elements of the coefficient matrix;

[0119] Step S326: Calculate hidden features:

[0120] ;

[0121] In the formula, It is the first One hidden feature, Indicates the first Column traversal from 1 to The cumulative sum over time, It is the first of the numerical matrices One value element;

[0122] Step S327: Aggregate nodes and node features into a directed graph representation, as shown in the following formula:

[0123] ;

[0124] In the formula, It is represented as a directed graph. It is the first node set 1 node yes The Layer node characteristics, yes The cumulative sum, Is seeking The maximum value, It is a parameterized abstract aggregate function with parameters;

[0125] Step S33: Graph classification, outputting fault types, specifically including the following steps:

[0126] Step S331: Construct and train the GTN model. The location feedforward network contains three fully connected layers with Gaussian error linear units as activation functions, using cross-entropy as the loss function. The directed graph represents the normalized input feedforward network, and the output is the predicted fault type, as shown in the following formula:

[0127] ;

[0128] In the formula, It predicts the type of failure. It is a function of probability transformation. It is a location feedforward network;

[0129] Step S332: Input the directed graph representation extracted from the log message into the trained GTN model, and the GTN model outputs the fault type.

[0130] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, the user queries, and the specific method includes the following steps:

[0131] Step S41: Capture the keywords searched by the user and search for them in the header file to obtain the offset of the keywords;

[0132] Step S42: Based on the offset, locate the data file on the disk to obtain the main search page;

[0133] Step S43: Find the cached available pages from the server, intersect them with the main search page to generate a common page, and cache the common page as a new available page in parallel on the server in the form of a key-value dictionary object;

[0134] Step S44: If the user's search has not yet ended, repeat steps S41 to S43; otherwise, render and display the final searched public page on the user interface.

[0135] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0136] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0137] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An automated management method for operation and maintenance fault detection, applied to an automated management system for operation and maintenance fault detection, characterized in that: The automated management system for operation and maintenance fault detection includes a site monitoring module, a data storage module, an operation and maintenance self-inspection module, a user query module, and a maintenance reporting module. The user query module is equipped with a server; The automated management method for operation and maintenance fault detection includes the following steps: Step S1: Branch monitoring. The branch monitoring module monitors the log messages of all branch equipment and branch software and transmits them to the data storage module. Step S2: Data storage. The data storage module receives log messages from the site monitoring module, establishes a binary file system, and sends the log messages to the operation and maintenance self-test module. Step S3: Operation and maintenance self-check. The operation and maintenance self-check module receives log messages from the data storage module, uses the GTN model to find the fault types of the network equipment and network software, and sends the fault types to the maintenance and repair module. Step S4: User query, using a multi-level drill-down search engine, allows users to search and view log messages and fault types in real time; Step S5: Maintenance and Repair Reporting. The maintenance and repair reporting module receives the fault type from the operation and maintenance self-test module, automatically assigns a preliminary solution to the faulty equipment and software at the site, and generates an automated operation and maintenance report.

2. The automated management method for operation and maintenance fault detection according to claim 1, characterized in that: In step S1, the site monitoring specifically includes the following steps: Step S11: Monitor the log messages of all network equipment, including oil dispensers, level gauges, and local databases. Transmit the log messages of the network equipment to the data storage module and update them once a day. Step S12: Monitor the log messages of all branch software displaying the branch service status. The branch software includes the management system, POS server, POS client and invoice system. The service status includes the installation version, branch user status, running status and software environment. Transmit the log messages of the branch software to the data storage module and update them every three hours.

3. The automated management method for operation and maintenance fault detection according to claim 2, characterized in that: In step S2, data storage specifically includes the following steps: Step S21: The data storage module receives log messages from the branch monitoring module, compresses and backs up the log messages, and automatically cleans them up after seven days. Step S22: Extract keywords from log messages and classify the keywords using the following classification formula: ； In the formula, It is a keyword. It's a classification of keywords. It's about keyword ranking. It refers to the number of times the keyword appears. It is the cumulative sum of keyword rankings; keywords with similar cumulative rankings are considered to belong to the same category. Step S23: Create binary header and data files for each category, load the header files into main memory, and store the data files on disk.

4. The automated management method for operation and maintenance fault detection according to claim 3, characterized in that: In step S3, the operation and maintenance self-check specifically includes the following steps: Step S31: Graph representation, converting log messages into a directed graph, specifically including the following steps: Step S311: Semantic awareness. The log interpreter Drain interprets log messages as log events. The BERT model is used to extract node features from log events and generate fixed-dimensional semantic vectors. A semantic vector is defined as a node. All semantic vectors are arranged in chronological order to form a log sequence. Step S312: Perceive the structure. Based on the log sequence, generate a degree matrix reflecting the local topology, a distance matrix reflecting the global spatial structure, and an edge weight matrix reflecting the number of edges between nodes. Step S32: Graph construction. The GTN module is used to determine the fault type of the log message. The GTN module includes self-focused units and a location feedforward network, specifically including the following steps: Step S321: The self-focused unit performs independent linear transformations on the node features, projecting the node features into a query matrix, a key matrix, and a numerical matrix, as shown in the following formula: ； ； ； in, , , These are the query matrix, key matrix, and value matrix, respectively. It is a node feature. , , These are the learnable parameters for query, key, and value, respectively. Step S322: Merge the edge weight matrix into and middle: ； ； ； In the formula, , , These are the first-level query matrix, the first-level key matrix, and the first-level numerical matrix of the GTN module, respectively. This represents the sum of the edge weight matrix along the row direction. This represents the Hadamard product operation of matrices; Step S323: Calculate the spatial bias term, using the following formula: ； In the formula, For spatial bias, It is the first query matrix Query elements, It is the first of the key matrix Each key element It's a dot product operation. It is the relative path distance between nodes; Step S324: Calculate the attention coefficient matrix, using the following formula: ； In the formula, Pay attention to the coefficient matrix. It is the output dimension of the linear transformation; Step S325: Calculate the attention weight matrix, using the following formula: ； In the formula, It is the attention weight matrix. It is an exponential operation with base e. It is the total number of nodes. Indicates the base e of the first... Pay attention to the cumulative sum of the elements of the coefficient matrix; Step S326: Calculate hidden features: ； In the formula, It is the first One hidden feature, Indicates the first Column traversal from 1 to The cumulative sum over time, It is the first of the numerical matrices One value element; Step S327: Aggregate nodes and node features into a directed graph representation, as shown in the following formula: ； In the formula, It is represented as a directed graph. It is the first node set 1 node yes The Layer node characteristics, yes The cumulative sum, Is seeking The maximum value, It is a parameterized abstract aggregate function with parameters; Step S33: Graph classification, outputting fault types, specifically including the following steps: Step S331: Construct and train the GTN model. The location feedforward network contains three fully connected layers with Gaussian error linear units as activation functions, using cross-entropy as the loss function. The directed graph represents the normalized input feedforward network, and the output is the predicted fault type, as shown in the following formula: ； In the formula, It predicts the type of failure. It is a function of probability transformation. It is a location feedforward network; Step S332: Input the directed graph representation extracted from the log message into the trained GTN model, and the GTN model outputs the fault type.

5. The automated management method for operation and maintenance fault detection according to claim 4, characterized in that: In step S4, the user query specifically includes the following steps: Step S41: Capture the keywords searched by the user and search for them in the header file to obtain the offset of the keywords; Step S42: Based on the offset, locate the data file on the disk to obtain the main search page; Step S43: Find the cached available pages from the server, intersect them with the main search page to generate a common page, and cache the common page as a new available page in parallel on the server in the form of a key-value dictionary object; Step S44: If the user's search has not yet ended, repeat steps S41 to S43; otherwise, render and display the final searched public page on the user interface.

6. An automated management system for operation and maintenance fault detection, used to implement the automated management method for operation and maintenance fault detection as described in any one of claims 1-5, characterized in that: The branch monitoring module monitors the log messages of all branch equipment and software and transmits them to the data storage module; The data storage module receives log messages from the site monitoring module, establishes a binary file system, and sends the log messages to the operation and maintenance self-test module. The operation and maintenance self-test module receives log messages from the data storage module, uses the GTN model to find the fault types of the network equipment and network software, and sends the fault types to the maintenance and repair module. The user query module uses a multi-level drill-down search engine, allowing users to search and view log messages and fault types in real time. The maintenance and repair module automatically assigns pre-defined solutions for the corresponding fault types to the faulty equipment and software at the service points.

Images