Intelligent diagnosis and self-healing method for abnormal transmission of communication system

Abstract

The invention discloses an intelligent diagnosis and self-healing method for transmission abnormity of a communication system, and the method comprises the steps: carrying out the real-time analysis of collected mass signal data according to a pre-established abnormity type judgment rule base, and recognizing a potential transmission abnormity type; a fault diagnosis decision tree model is constructed, a system state evaluation index is comprehensively considered, and a root cause causing transmission abnormity is reasoned; according to a self-healing strategy configuration file, an optimal recovery measure is matched, and self-healing of the system is achieved through a backup data recovery mechanism or a redundant module hot switching function; for different types of signals, key feature parameters of the signals are extracted, and a transmission interruption risk prediction model is constructed in combination with a communication link topology logic representation method; mining abnormal transmission modes from a historical transmission event data set to form a transmission abnormal mode library as a knowledge base for interruption early warning; and calculating an interruption risk assessment index of each signal, and according to a risk grade division standard, judging a risk grade to which the signal belongs.

Description

Technical Field

[0001] This invention relates to the field of substation signal transmission, specifically to an intelligent diagnosis and self-healing method for communication system transmission anomalies. Background Technology

[0002] The continuity and reliability of signal transmission in substations are crucial for ensuring the safe and stable operation of the power grid. Currently, various status variables, measurements, and alarm information within substations need to be transmitted to the dispatch center in real time via communication networks. However, due to the complexity of communication links and the variety of equipment, ensuring the continuity of signal transmission is difficult. Communication interruptions can lead to delays or errors in dispatch decisions, and even trigger serious power grid accidents.

[0003] Traditional signal transmission continuity checks rely primarily on manual periodic inspections and sampling tests, which are inefficient and lack comprehensive coverage. To address this issue, there is an urgent need to research and develop a system capable of comprehensively monitoring the real-time transmission status of all signals within a substation and automatically analyzing and diagnosing potential continuity issues based on various protocols and standards. Furthermore, to handle communication link failures and anomalies, the system must be able to automatically detect and manage redundant paths in the communication network and automatically switch to backup links upon detecting a fault, thereby maximizing the uninterrupted signal transmission. In addition, as an online monitoring and acceptance system, its reliability and security must meet high standards. It requires real-time monitoring of the operational status of each module and the establishment of robust fault diagnosis and recovery mechanisms to ensure that the verification system itself does not become a new threat to signal transmission. Summary of the Invention

[0004] The purpose of this invention is to solve the above-mentioned problems and provide an intelligent diagnosis and self-healing method for abnormal transmission in communication systems.

[0005] The technical solution adopted by this invention to solve its technical problem is: A method for intelligent diagnosis and self-healing of transmission anomalies in a communication system includes the following steps: S101 analyzes the collected signal data based on a pre-established rule base for judging anomaly types and identifies potential transmission anomaly types. S102 constructs a fault diagnosis decision tree model, comprehensively considers system status evaluation indicators, and infers the root cause of the transmission anomaly. Based on the self-healing strategy configuration file, S103 matches the optimal recovery measures and achieves system self-repair through backup data recovery mechanism or redundant module hot-swapping function. S104 uses a system health scoring algorithm to quantify and verify the reliable operation status of the system. When the health level is lower than the preset threshold, a warning message is sent. S105 extracts key feature parameters for different types of signals and constructs a transmission interruption risk prediction model by combining the communication link topology logic representation method. S106 extracts abnormal transmission patterns from historical transmission event datasets to form a transmission abnormal pattern library, which serves as a knowledge base for interruption early warning. S107 calculates the interruption risk assessment index for each signal and determines its risk level according to the risk level classification standard. If the risk level exceeds the preset threshold, S108 will generate a template based on the early warning information, automatically generate early warning information, and decompose the preventive maintenance task into several sub-tasks and issue them for execution.

[0006] Step S101 includes: Acquire a pre-established rule base for judging anomaly types and massive amounts of real-time signal data; For the massive amount of signal data, data preprocessing techniques are used to clean the data and obtain the first dataset; Based on the first dataset, data features are extracted to obtain the second dataset; Based on the rules in the rule base for determining the anomaly type, the second dataset is matched and analyzed; If the second dataset satisfies the anomaly type judgment rule, then it is determined that the massive signal data has potential transmission anomalies, and the anomaly type is marked to obtain the third dataset; Cluster analysis was performed on the third dataset to identify different types of transmission anomalies, resulting in the fourth dataset; Machine learning algorithms are used to classify and predict the anomaly types in the fourth dataset; Based on the classification and prediction results, the identified potential transmission anomaly types are output, resulting in the fifth dataset.

[0007] Further, step S102 includes: Obtain data on multiple state evaluation metrics of the system as input feature vectors for the decision tree model; Based on the predefined decision tree node partitioning rules, the feature vectors are recursively classified and judged. If the current node meets the leaf node condition, then the fault cause corresponding to that node will be output as the diagnostic result. If the current node does not meet the leaf node condition, then select the optimal splitting feature according to the node splitting rule; Based on the threshold of the selected splitting feature, the feature vector is divided into left and right child nodes, and step 2 is executed recursively. Repeat steps 2 to 5 until all feature vectors are classified to leaf nodes to obtain a complete diagnostic result. By comprehensively analyzing all diagnostic results, the root cause of the system transmission anomaly was determined.

[0008] Further, step S103 includes: According to the self-healing strategy configuration file, obtain the system fault type and fault level; For the fault type and fault level, the optimal recovery measure is matched from the preset recovery measure library; If the optimal recovery measure is to recover using backup data, then the required backup data is obtained from the backup data repository, and the data recovery module is used to perform data recovery operations to repair the faulty system. If the optimal recovery measure is to use redundant modules for hot switching, then the fault detection module monitors the operating status of each redundant module in real time to identify the faulty module and the backup module. According to the redundancy module switching rules, the system operation is switched to the backup module to ensure business continuity; The support vector machine algorithm is used to train and optimize the self-healing strategy configuration file and recovery measure matching rules based on historical system fault data; By using decision tree algorithms, the switching rules for redundant modules are dynamically adjusted to improve system fault tolerance and reliability, ultimately achieving intelligent system self-repair.

[0009] Further, step S104 includes: The system obtains real-time operating status data and uses a pre-trained system health scoring model to calculate the current system health score. The calculated system health score is compared with the preset health threshold to determine whether the current health status of the system is below the warning line. If the system health score is lower than the preset threshold, an alarm notification mechanism will be triggered. When sending alert messages, based on the system's historical operating data and current abnormal indicators, anomaly detection algorithms are used to locate key indicators and anomalies that are causing a decline in system health. The identified key indicators and anomalies, along with the current system health score, will be sent to the relevant operations and maintenance personnel.

[0010] Further, step S105 includes: Acquire signal data of different types and determine their key characteristic parameters; Based on the topology of the communication link, establish a link topology logical model; The extracted key signal feature parameters are input into the link topology logic model, and the transmission interruption risk value of each link node is calculated through the model. Set a transmission interruption risk threshold. If the risk value of a link node exceeds the threshold, the node is judged to have a transmission interruption risk. Based on the transmission interruption risk value of each link node, a transmission interruption risk prediction model is generated by training a machine learning algorithm. When new signal data is acquired, its key feature parameters are extracted and input into the prediction model to obtain the prediction result of the transmission interruption risk of the link.

[0011] Further, step S106 includes: Obtain the historical transmission event dataset, preprocess the data, and remove noisy data and outliers; An association rule mining algorithm is used to mine frequent itemsets and association rules from historical transmission event datasets to obtain a preliminary candidate set of abnormal transmission patterns. Based on the preset abnormal transmission mode evaluation index, the preliminary abnormal transmission mode candidate set is evaluated and scored, and abnormal transmission modes that meet the score threshold are selected. The selected abnormal transmission patterns are organized and stored according to a predefined knowledge base architecture to build a transmission abnormal pattern library.

[0012] Further, step S107 includes: Obtain the various attribute parameters of the signal as input to the risk assessment model; Based on the pre-established interruption risk assessment model and combined with the signal attribute parameters, the interruption risk assessment index of the signal is obtained. Obtain the preset risk level classification threshold, compare the signal interruption risk assessment index with the risk level threshold, and determine which risk level range the index belongs to. Based on the risk level assessment results, output the risk level assessment conclusion of the signal, and attach the corresponding interruption risk assessment index value. The risk assessment process described above is executed sequentially for all signals to obtain the risk level classification results for each signal.

[0013] The beneficial effects of this invention are: 1. This invention analyzes massive amounts of signal data in real time to identify potential transmission anomaly types and constructs a fault diagnosis decision tree model to deduce the root cause of the transmission anomaly. Simultaneously, based on a self-healing strategy configuration file, this invention matches the optimal recovery measures and achieves self-repair through a backup data recovery mechanism or redundant module hot-swapping functionality. Furthermore, this invention employs a system health scoring algorithm to quantitatively verify the reliable operating status of the invention. When the health level falls below a preset threshold, an alarm notification is sent via a push notification channel.

[0014] 2. For different types of signals, this invention extracts their key feature parameters and, combined with the communication link topology logic representation method, constructs a transmission interruption risk prediction model. It mines abnormal transmission patterns from historical transmission event datasets to form a transmission anomaly pattern library, calculates the interruption risk assessment index for each signal, determines its risk level, and automatically generates early warning information if it exceeds a preset threshold, and assigns preventative maintenance tasks to relevant maintenance personnel. This invention enables intelligent diagnosis, self-healing, and early warning of communication system transmission anomalies, improving system reliability and availability, reducing manual intervention, and lowering maintenance costs. Attached Figure Description

[0015] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0016] like Figure 1 As shown, an intelligent diagnosis and self-healing method for transmission anomalies in a communication system includes the following steps: S101 analyzes the collected signal data based on a pre-established rule base for judging anomaly types and identifies potential transmission anomaly types. S102 constructs a fault diagnosis decision tree model, comprehensively considers system status evaluation indicators, and infers the root cause of the transmission anomaly. Based on the self-healing strategy configuration file, S103 matches the optimal recovery measures and achieves system self-repair through backup data recovery mechanism or redundant module hot-swapping function. S104 uses a system health scoring algorithm to quantify and verify the reliable operation status of the system. When the health level is lower than the preset threshold, a warning message is sent. S105 extracts key feature parameters for different types of signals and constructs a transmission interruption risk prediction model by combining the communication link topology logic representation method. S106 extracts abnormal transmission patterns from historical transmission event datasets to form a transmission abnormal pattern library, which serves as a knowledge base for interruption early warning. S107 calculates the interruption risk assessment index for each signal and determines its risk level according to the risk level classification standard. If the risk level exceeds the preset threshold, S108 will automatically generate a warning message based on the warning information template, and at the same time decompose the preventive maintenance task into several sub-tasks and distribute them to the operation and maintenance personnel for execution.

[0017] Step S101 includes: acquiring a pre-established rule base for judging anomaly types and massive amounts of real-time signal data; the system will acquire the pre-established rule base for judging anomaly types, which contains judgment conditions and thresholds for various anomaly types, such as signal strength below -90dBm, signal-to-noise ratio less than 10dB, etc.

[0018] For the massive signal data, data preprocessing techniques are used to clean the data, remove invalid data and noise interference, and extract key features such as signal strength, signal-to-noise ratio, and bit error rate to obtain the first dataset.

[0019] Based on the first dataset, data features are extracted to obtain the second dataset.

[0020] Based on the rules in the anomaly type judgment rule base, the second dataset is matched and analyzed.

[0021] If the second dataset satisfies the anomaly type judgment rule, then it is determined that the massive signal data has potential transmission anomalies, and the anomaly type is marked to obtain the third dataset.

[0022] Cluster analysis was performed on the third dataset using the K-means algorithm. The abnormal signal data were divided into different clusters based on the similarity of abnormal features. Each cluster represents a type of transmission anomaly. Different types of transmission anomalies were identified, resulting in the fourth dataset.

[0023] Machine learning algorithms are used to classify and predict the anomaly types of the fourth dataset. A classification model is built by training historical anomaly data, and the type of new anomaly signal data is predicted. The machine learning algorithms include support vector machines and random forests.

[0024] Based on the classification and prediction results, the identified potential transmission anomaly types are output to obtain the fifth dataset. Based on the fifth dataset, an anomaly report is generated, which describes in detail the time, location, anomaly type, and scope of impact of the anomaly, providing data support for subsequent anomaly localization and processing.

[0025] Through the above steps, the system can achieve real-time anomaly detection and analysis of massive signal data, promptly identify potential transmission anomalies, and ensure the stable operation of the communication network.

[0026] Step S102 includes: The system collects multiple status evaluation indicators such as network transmission rate, packet loss rate, and latency in real time, which are used as input feature vectors for the decision tree model.

[0027] Based on predefined decision tree node partitioning rules, such as information gain and Gini index, the feature vectors are recursively classified.

[0028] If the current node meets the leaf node condition, the fault cause corresponding to that node will be output as the diagnostic result. If the number of samples is less than the preset threshold or the node depth reaches the maximum value, the fault cause corresponding to that node, such as network congestion or equipment failure, will be output as the diagnostic result.

[0029] If the current node does not meet the leaf node condition, then according to the node splitting rules, such as maximizing information gain, the optimal splitting feature is selected. For example, with a packet loss rate of 5% as the threshold, the feature vector is divided into two child nodes with a packet loss rate greater than or equal to 5% and less than 5%, and the judgment process is executed recursively.

[0030] Based on the threshold of the selected splitting feature, the feature vector is divided into left and right child nodes, and step 2 is executed recursively.

[0031] Repeat steps 2 to 5 until all feature vectors are classified to leaf nodes to obtain a complete diagnostic result.

[0032] Finally, by comprehensively analyzing all diagnostic results, if the diagnostic results of most samples are network congestion, then network congestion is determined to be the root cause of the system transmission anomaly, and a diagnostic report containing the cause of the fault, diagnostic basis, and handling suggestions is generated to provide decision support for system optimization.

[0033] Step S103 includes: The system first classifies faults into different types such as network faults, hardware faults, and software faults according to the self-healing strategy configuration file and through fault classification algorithms such as decision trees or support vector machines. Based on the degree of impact of the fault on the system operation, the fault level is divided into four levels: minor, general, severe, and fatal.

[0034] For the fault type and fault level, a similarity matching algorithm, such as cosine similarity or Jaccard similarity, is used in a preset recovery measure library to match the optimal recovery measure based on the fault type and level.

[0035] If the optimal recovery measure is to restore using backup data, the system will automatically retrieve the most recent complete backup data from the backup data repository, perform data recovery operations through the data recovery module, repair the faulty system, and achieve a recovery speed of up to 1TB per second.

[0036] If the optimal recovery measure is to use redundant modules for hot switching, the fault detection module monitors the operating status of each redundant module in real time to identify the faulty module and the backup module. According to the redundant module switching rules, the system operation is switched to the backup module to ensure business continuity. The system monitors indicators such as CPU utilization, memory usage, and response time of each redundant module in real time. Using anomaly detection algorithms such as Isolation Forest, it identifies the faulty module and selects the module with the lowest resource utilization and lightest load from the backup modules as the switching target. According to the preset redundant module switching rules, business traffic is seamlessly transferred from the faulty module to the backup module, with the switching time controlled within 100 milliseconds.

[0037] The system continuously collects historical fault data and uses the support vector machine algorithm to train and optimize the self-healing strategy configuration file and recovery measure matching rules based on historical system fault data, thereby improving the accuracy of fault handling.

[0038] The system also improves fault tolerance and reliability by dynamically adjusting the configuration and switching rules of redundant modules through decision tree algorithms, ultimately achieving intelligent system self-repair. Through continuous self-learning and optimization, the system's self-healing ability is constantly enhanced, and the average fault recovery time can be shortened by more than 50%.

[0039] Step S104 includes: acquiring real-time operating status data of the system, including indicators such as CPU utilization, memory utilization, disk I / O, and network traffic; Based on the acquired real-time operational status data, a pre-trained system health scoring model is used to calculate the current system health score. The calculated system health score is then compared with a preset health threshold to determine whether the current system health status is below the warning line. For example, if the system health scoring model built using the Support Vector Machine (SVM) algorithm calculates a current system health score of 75 points by inputting these real-time indicator data, and the preset health threshold is 80 points, then by comparison, it can be determined that the current system health status is below the warning line.

[0040] If the system health score is lower than the preset threshold, an alarm notification mechanism will be triggered, and warning information will be sent to relevant maintenance personnel according to the pre-configured alarm notification channels.

[0041] When sending alert messages, based on the system's historical operating data and current abnormal indicators, the anomaly detection algorithm is used to locate the key indicators and anomalies that cause the system's health to decline. The anomaly detection algorithm built by the system using historical operating data compares the differences between the current indicator data and the historical normal data to locate the key indicators that cause the health to decline as CPU utilization and memory utilization, and the anomalies as CPU utilization of 85% and memory utilization of 70%.

[0042] The key indicators and anomalies identified, along with the current system health score of 75, are sent to the relevant operations and maintenance personnel to facilitate their rapid analysis and location of system faults.

[0043] The alarm notification mechanism also uses a role matching algorithm to automatically and accurately push alarm information to the on-duty maintenance personnel responsible for the system based on the job responsibilities and duty schedules of different roles such as system administrator, database administrator, and network administrator. This improves the timeliness and accuracy of alarm notifications and ensures the stable operation of the system.

[0044] Step S105 includes: acquiring signal data of different types, and determining the key characteristic parameters for each signal type.

[0045] Based on the topology of the communication link, a logical network topology model consisting of nodes and connections can be established. Nodes can be devices such as servers, switches, and routers, while connections represent the physical links between devices.

[0046] The extracted key signal feature parameters are input into the link topology logic model. Through the fault propagation mechanism in the model, the impact of faults on each node is analyzed, and the transmission interruption risk value of each node is calculated.

[0047] A transmission interruption risk threshold is set. If the risk value of a link node exceeds the threshold, the node is judged to have a transmission interruption risk.

[0048] Based on the transmission interruption risk value of each link node, a transmission interruption risk prediction model is trained and generated using machine learning algorithms such as support vector machines.

[0049] When new signal data is acquired, its key feature parameters are extracted and input into the prediction model to obtain the prediction result of the transmission interruption risk of the link.

[0050] Based on the prediction results, preventative measures are taken for nodes at risk of interruption, such as adding redundant links, limiting signal transmission power, and optimizing transmission routes, thereby reducing the probability of transmission interruption and improving the reliability of communication links. Through the above technical solutions, intelligent prediction and prevention of communication link transmission interruption risks can be achieved, ensuring the stable operation of the communication system.

[0051] Step S106 includes: first, obtaining a dataset of transmission events from the historical transmission event database for the past three months, and then using data cleaning techniques to remove approximately 5% of the noisy data and outliers by setting data threshold ranges and data format specifications.

[0052] Then, the Apriori association rule mining algorithm was used, with a minimum support of 0.5 and a minimum confidence of 8, to mine frequent itemsets and association rules from the historical transmission event dataset, thus obtaining a preliminary candidate set of abnormal transmission patterns.

[0053] Based on the pre-set evaluation index system for abnormal transmission patterns, including pattern support, confidence, and novelty, a weighted scoring method is used to evaluate and score the preliminary candidate set of abnormal transmission patterns. The weights are set by expert experience, and abnormal transmission patterns with scores greater than 6 are selected.

[0054] The selected abnormal transmission patterns are organized and stored according to a predefined knowledge base architecture to build a transmission abnormal pattern library.

[0055] For real-time transmitted event data, a rule-based pattern matching algorithm is used to extract event features and search for matches in the abnormal pattern library. If the similarity of the match is greater than 8, it is determined to be an abnormal event, triggering an interruption warning, generating warning information, outputting the interruption reason, and associating the abnormal pattern with typical cases in historical data.

[0056] To continuously optimize and update the transmission anomaly pattern library, new anomaly patterns are regularly mined from newly added historical transmission event data every week. A similarity comparison method between new and old patterns is used to remove outdated and invalid anomaly patterns with a similarity of less than 2, so as to ensure the effectiveness and real-time nature of the anomaly pattern library.

[0057] Step S107 includes: When conducting a signal interruption risk assessment, firstly, various attribute parameters of the signal are collected through sensors, such as signal strength, signal-to-noise ratio, and bit error rate. Specifically, the signal strength ranges from -90dBm to -10dBm, the signal-to-noise ratio ranges from 0dB to 30dB, and the bit error rate ranges from 10^-6 to 10^-2. These parameters are then input into a pre-established BP neural network risk assessment model, which has been trained to optimize the weights and thresholds of each parameter.

[0058] Based on the pre-established interruption risk assessment model and combined with the signal attribute parameters, the interruption risk assessment index of the signal is obtained, with a value range of 0 to 100.

[0059] Obtain preset risk level classification thresholds, including indicator thresholds corresponding to high risk, medium risk and low risk levels, where the high risk threshold is 80, the medium risk threshold is 60 and the low risk threshold is 40.

[0060] The signal interruption risk assessment index is compared with the risk level threshold to determine which risk level range the index belongs to. If the interruption risk assessment index is higher than the high-risk threshold, the signal is classified as high-risk. If the index is between the medium-risk and high-risk thresholds, it is classified as medium-risk. If the index is lower than the low-risk threshold, it is classified as low-risk. For example, if the index is 85, it is higher than the high-risk threshold of 80 and classified as high-risk. If the index is 70, it is between 60 and 80 and classified as medium-risk. If the index is 30, it is lower than the low-risk threshold of 40 and classified as low-risk.

[0061] Based on the risk level assessment results, output the risk level assessment conclusion of the signal, and attach the corresponding interruption risk assessment index value. The risk assessment process described above is executed sequentially for all signals to obtain the risk level classification results for each signal, forming a complete signal outage risk assessment report, which provides a basis for the decision-making of operation and maintenance personnel.

[0062] Step S106 includes: Acquiring risk level data: Real-time collection of equipment operating status data via sensor networks, monitoring systems, or other data sources, and prediction of risk levels using pre-trained machine learning models (e.g., regression models). Determining if the risk level exceeds a threshold: If the acquired risk level exceeds a preset threshold, an early warning process is triggered; if the risk level does not exceed the preset threshold, monitoring continues. Generating early warning information: Based on a preset early warning information template, key data such as risk level, equipment information, and time are automatically filled in to generate early warning information. Decomposing maintenance tasks: Based on the risk type and equipment information contained in the early warning information, preventative maintenance tasks are automatically decomposed into multiple sub-tasks, each containing specific execution steps and required resources. A decision tree model can be used here to decompose tasks based on risk type and equipment information. Identifying relevant maintenance personnel: Matching and identifying relevant maintenance personnel based on the type of sub-task and required skills. Using a pre-established skill matrix, sub-tasks are assigned to maintenance personnel with the corresponding skills. Assigning sub-tasks: The decomposed sub-tasks are assigned to the identified relevant maintenance personnel through a work order system or other communication tools. Subtask information includes task description, deadline, and required resources. Task execution status is monitored in real time, and progress and results are reported back to relevant management personnel. If problems arise during subtask execution, a new early warning process is triggered. Support vector machine models can be used to predict task execution status and identify potential problems in advance.

Claims

1. A method for intelligent diagnosis and self-healing of transmission anomalies in a communication system, characterized in that, Includes the following steps: S101 analyzes the collected signal data based on a pre-established rule base for judging anomaly types and identifies potential transmission anomaly types. S102 constructs a fault diagnosis decision tree model, comprehensively considers system status evaluation indicators, and infers the root cause of the transmission anomaly. Based on the self-healing strategy configuration file, S103 matches the optimal recovery measures and achieves system self-repair through backup data recovery mechanism or redundant module hot-swapping function. S104 uses a system health scoring algorithm to quantify and verify the reliable operation status of the system. When the health level is lower than the preset threshold, a warning message is sent. S105 extracts key feature parameters for different types of signals and constructs a transmission interruption risk prediction model by combining the communication link topology logic representation method. S106 extracts abnormal transmission patterns from historical transmission event datasets to form a transmission abnormal pattern library, which serves as a knowledge base for interruption early warning. S107 calculates the interruption risk assessment index for each signal and determines its risk level according to the risk level classification standard. If the risk level exceeds the preset threshold, S108 will generate a template based on the early warning information, automatically generate early warning information, and decompose the preventive maintenance task into several sub-tasks and issue them for execution.

2. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S101 includes: Acquire a pre-established rule base for judging anomaly types and massive amounts of real-time signal data; For the massive amount of signal data, data preprocessing techniques are used to clean the data and obtain the first dataset; Based on the first dataset, data features are extracted to obtain the second dataset; Based on the rules in the rule base for determining the anomaly type, the second dataset is matched and analyzed; If the second dataset satisfies the anomaly type judgment rule, then it is determined that the massive signal data has potential transmission anomalies, and the anomaly type is marked to obtain the third dataset; Cluster analysis was performed on the third dataset to identify different types of transmission anomalies, resulting in the fourth dataset; Machine learning algorithms are used to classify and predict the anomaly types in the fourth dataset; Based on the classification and prediction results, the identified potential transmission anomaly types are output, resulting in the fifth dataset.

3. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S102 includes: Obtain data on multiple state evaluation metrics of the system as input feature vectors for the decision tree model; Based on the predefined decision tree node partitioning rules, the feature vectors are recursively classified and judged. If the current node meets the leaf node condition, then the fault cause corresponding to that node will be output as the diagnostic result. If the current node does not meet the leaf node condition, then select the optimal splitting feature according to the node splitting rule; Based on the threshold of the selected splitting feature, the feature vector is divided into left and right child nodes, and step 2 is executed recursively. Repeat steps 2 to 5 until all feature vectors are classified to leaf nodes to obtain a complete diagnostic result. By comprehensively analyzing all diagnostic results, the root cause of the system transmission anomaly was determined.

4. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S103 includes: According to the self-healing strategy configuration file, obtain the system fault type and fault level; For the fault type and fault level, the optimal recovery measure is matched from the preset recovery measure library; If the optimal recovery measure is to recover using backup data, then the required backup data is obtained from the backup data repository, and the data recovery module is used to perform data recovery operations to repair the faulty system. If the optimal recovery measure is to use redundant modules for hot switching, then the fault detection module monitors the operating status of each redundant module in real time to identify the faulty module and the backup module. According to the redundancy module switching rules, the system operation is switched to the backup module to ensure business continuity; The support vector machine algorithm is used to train and optimize the self-healing strategy configuration file and recovery measure matching rules based on historical system fault data; By using decision tree algorithms, the switching rules for redundant modules are dynamically adjusted to improve system fault tolerance and reliability, ultimately achieving intelligent system self-repair.

5. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S104 includes: The system obtains real-time operating status data and uses a pre-trained system health scoring model to calculate the current system health score. The calculated system health score is compared with the preset health threshold to determine whether the current health status of the system is below the warning line. If the system health score is lower than the preset threshold, an alarm notification mechanism will be triggered. When sending alert messages, based on the system's historical operating data and current abnormal indicators, anomaly detection algorithms are used to locate key indicators and anomalies that are causing a decline in system health. The identified key indicators and anomalies, along with the current system health score, will be sent to the relevant operations and maintenance personnel.

6. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S105 includes: Acquire signal data of different types and determine their key characteristic parameters; Based on the topology of the communication link, establish a link topology logical model; The extracted key signal feature parameters are input into the link topology logic model, and the transmission interruption risk value of each link node is calculated through the model. Set a transmission interruption risk threshold. If the risk value of a link node exceeds the threshold, the node is judged to have a transmission interruption risk. Based on the transmission interruption risk value of each link node, a transmission interruption risk prediction model is generated by training a machine learning algorithm. When new signal data is acquired, its key feature parameters are extracted and input into the prediction model to obtain the prediction result of the transmission interruption risk of the link.

7. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S106 includes: Obtain the historical transmission event dataset, preprocess the data, and remove noisy data and outliers; An association rule mining algorithm is used to mine frequent itemsets and association rules from historical transmission event datasets to obtain a preliminary candidate set of abnormal transmission patterns. Based on the preset abnormal transmission mode evaluation index, the preliminary abnormal transmission mode candidate set is evaluated and scored, and abnormal transmission modes that meet the score threshold are selected. The selected abnormal transmission patterns are organized and stored according to a predefined knowledge base architecture to build a transmission abnormal pattern library.

8. The intelligent diagnosis and self-healing method for communication system transmission anomalies as described in claim 1, characterized in that, Step S107 includes: Obtain the various attribute parameters of the signal as input to the risk assessment model; Based on the pre-established interruption risk assessment model and combined with the signal attribute parameters, the interruption risk assessment index of the signal is obtained. Obtain the preset risk level classification threshold, compare the signal interruption risk assessment index with the risk level threshold, and determine which risk level range the index belongs to. Based on the risk level assessment results, output the risk level assessment conclusion of the signal, and attach the corresponding interruption risk assessment index value. The risk assessment process described above is executed sequentially for all signals to obtain the risk level classification results for each signal.

Images