Fault self-diagnosis and remote maintenance system of communication equipment

Abstract

The invention discloses a fault self-diagnosis and remote maintenance system and method for communication equipment, and relates to the technical field of communication networks. The system adopts a cloud edge collaborative architecture and comprises an edge intelligent agent module deployed on equipment and an intelligent operation and maintenance platform deployed on a cloud end. The edge agent is responsible for collecting multi-dimensional data of the equipment and executing a remote instruction; the cloud platform gathers data of the whole network, performs anomaly detection, root cause positioning driven by graph calculation and predictive health assessment through an intelligent fault diagnosis engine, and performs operation pre-verification by using digital twinning. The platform generates a security maintenance script through an automatic scheduling module, and issues the security maintenance script to an edge agent for execution to form a closed loop of perception, analysis, decision, execution and verification. According to the method, accurate positioning, prediction and early warning and safe remote repair of faults are realized, the operation and maintenance cost and dependence on field personnel are reduced, expert experience is continuously precipitated through the knowledge base, and the intelligence and automation level of communication network operation and maintenance is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of communication network technology, specifically to a fault self-diagnosis and remote maintenance system for communication equipment. Background Technology

[0002] With the rapid development of information technology, communication networks have become the infrastructure of modern society. Massive numbers of communication devices are deployed in various complex environments, from data centers to the city's edge, from indoors to the wilderness. The stable operation of these devices directly affects the quality and reliability of network services. However, the fault diagnosis and maintenance of communication equipment face unprecedented challenges: The sheer number and variety of devices: Modern networks consist of devices from different manufacturers, models, and technology generations, with varying log formats, management interfaces, and alarm standards, creating "data silos" that make unified monitoring and diagnosis difficult.

[0003] The symptoms of network failures are complex, and root cause identification is difficult: a single network performance problem may be caused by a combination of factors, including hardware failure, software defects, configuration errors, link congestion, or external attacks. Traditional alarm systems based on fixed rules have a high false alarm rate and struggle to reveal deep-seated fault correlations.

[0004] High reliance on on-site maintenance: Currently, many faults still require technicians to be physically present on-site to handle them by connecting to the control panel, checking indicator lights, and replacing circuit boards. This not only results in slow response times and high maintenance costs, but also presents safety risks due to difficulties in remote or high-risk environments.

[0005] Weak predictive maintenance capabilities: Existing maintenance models are mostly "reactive repairs" or periodic inspections, which cannot provide early warnings and interventions before equipment performance deteriorates or potential failures occur, leading to unplanned downtime and affecting business continuity.

[0006] Expert experience is difficult to accumulate and replicate: Fault diagnosis relies heavily on the personal experience of operation and maintenance experts. This experience lacks systematic and standardized refinement and inheritance, resulting in slow team capability growth and high risk of personnel turnover.

[0007] Existing technologies also include some remote monitoring and fault management solutions. For example, some network management systems (NMS) can centrally collect device status information and logs, and provide basic alarm functions. However, these systems typically lack in-depth intelligent analysis capabilities, resulting in prominent issues such as alarm storms and false alarms.

[0008] Existing technologies, such as CN117333151A (A Cloud-Edge-Based Predictive Maintenance Method and System for Industrial Equipment), propose using a cloud-edge-device architecture for data acquisition and predictive maintenance. This involves building a data lake in the cloud and using unsupervised learning for anomaly detection and health assessment, thereby achieving automated generation and scheduling of maintenance strategies. Other solutions, such as CN117201289A (A Fault Root Cause Localization Method, Device, Equipment, and Storage Medium), focus on improving fault root cause localization technology. By constructing a fault propagation graph and integrating causal inference and graph neural networks, they achieve accurate calculation of the fault source.

[0009] However, these existing technical solutions still have their limitations. The solution represented by CN117333151A, while achieving automation of predictive maintenance, has relatively weak root cause localization capabilities when faults occur, mainly relying on historical data pattern matching. For unknown or rare fault combinations in complex networks, its diagnostic accuracy needs improvement. The solution represented by CN117201289A, while providing advanced root cause localization algorithms, mainly focuses on the diagnostic phase and fails to seamlessly integrate accurate localization results into a complete remote maintenance closed loop that includes safety pre-verification, automated execution, and effect verification. Furthermore, neither solution fully explores the potential of digital twin technology in pre-verification of major operational risks, nor does it construct an evolutionary mechanism that continuously feeds back the diagnostic model with experience from each fault handling. Summary of the Invention

[0010] The purpose of this invention is to provide a fault self-diagnosis and remote maintenance system and method for communication equipment, in order to solve the problems of inaccurate fault diagnosis, difficulty in root cause location, reliance on on-site maintenance, lack of predictive ability, and difficulty in reusing expert experience in the prior art; to realize unified monitoring and intelligent analysis of cross-manufacturer and cross-model equipment, reduce the complexity of operation and maintenance; and to build a safe, controllable, and efficient remote maintenance channel and operation framework to minimize on-site intervention.

[0011] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a fault self-diagnosis and remote maintenance system for communication equipment. The system is characterized by adopting a layered and modular cloud-edge collaborative architecture, including: edge intelligent agent modules deployed on each communication device or near the device, and a cloud intelligent operation and maintenance platform deployed in a remote operation and maintenance center.

[0012] 1. Edge Intelligent Agent Module An edge intelligent agent module is a lightweight, highly reliable software or firmware component embedded in or residing in a communication device. Its main functions include: Multi-source heterogeneous data acquisition unit: Responsible for acquiring multi-dimensional operational status data from the communication devices it resides in. Acquisition methods include: Passive monitoring: Monitor the device's system logs (Syslog), debugging information, kernel messages, etc.

[0013] Active polling: Periodically obtain device performance metrics (CPU / memory utilization, interface traffic, packet error rate, temperature, power status, etc.), configuration information, routing table, session table, etc. through standard protocols (such as SNMP, NETCONF / YANG, gNMI) or device private CLI / API.

[0014] Event-driven capture: When a specific event occurs on the device (such as an interface state change or a process restart), a data snapshot is triggered for collection.

[0015] Data preprocessing: The collected raw data is cleaned (deduplicated, completed, formatted), normalized (mapped to a standard set of indicators from different vendors), and compressed to reduce network transmission overhead.

[0016] Local lightweight diagnostic unit: Integrates a simplified diagnostic model and rule base distributed from the cloud, which can perform preliminary diagnosis and emergency handling of common and urgent faults in network outage or low bandwidth environments (such as automatically shutting down the port after interface loop detection).

[0017] Secure Communication Unit: Establishes a two-way, encrypted, and authenticated secure channel with the cloud platform (such as the MQTT protocol based on TLS / DTLS). Supports reconnection after disconnection, data caching, and resume interrupted downloads. Simultaneously, it provides a secure sandbox environment for executing remote maintenance commands.

[0018] Remote Command Execution Unit: Serves as a secure execution terminal for cloud-based maintenance commands on the device's local machine. It receives and parses atomic maintenance commands from the cloud (such as "execute command show interface", "upload file flash: / config.cfg", "reboot module 1", "load firmware image.bin"), executes them under strict access control and operation log recording, and sends the execution results back to the cloud.

[0019] 2. Cloud-based intelligent operation and maintenance platform The cloud-based intelligent operations and maintenance platform is the "brain" of the entire system, providing centralized analysis, decision-making, and scheduling capabilities, including: The data lake and data governance module receives and stores massive amounts of time-series data, logs, and events uploaded from all edge agents across the network. It establishes a unified data model to further correlate, label, and index the data, providing a high-quality data foundation for upper-level analysis.

[0020] Intelligent Fault Diagnosis Engine Module (Core): Multimodal data fusion and analysis submodule: It integrates heterogeneous data such as time series indicators, event logs, configuration snapshots, and topology relationships in a spatiotemporal manner to build a panoramic operational view of devices and network services.

[0021] Fault Detection and Alarm Submodule: This module applies unsupervised learning algorithms (such as Isolation Forest and Auto Encoder) to perform baseline learning and anomaly detection on device behavior, enabling adaptive threshold alarms and reducing false alarms.

[0022] The root cause localization and analysis submodule employs a combination of graph-based computation and machine learning. First, a dynamic "fault propagation graph" is constructed based on network topology and service dependencies. When multiple abnormal events occur, causal inference algorithms (such as PC algorithms and Bayesian networks) or graph neural networks (GNNs) are used to calculate the probability of root cause nodes in the fault propagation graph, quickly locating the source of the fault (e.g., identifying a faulty line card as the root cause of the entire chassis's service interruption). Simultaneously, similarity matching is performed using a historical fault case database to aid in analysis.

[0023] The predictive health assessment submodule uses time series forecasting models (such as LSTM and Prophet) to predict future trends of key indicators based on historical performance data and operating load of the equipment. Combined with reliability models (such as Weibull distribution), it calculates the remaining useful life (RUL) and health score of the equipment or component, providing early warnings of potential failures.

[0024] Digital Twin and Simulation Verification Module: Creates high-fidelity digital twin models for critical networks or devices. Before implementing major remote maintenance operations (such as configuration changes or version upgrades), simulations and verifications are performed in the digital twin environment to assess the impact of the operations on services, generate optimal operation sequences and rollback plans, and greatly reduce operational risks.

[0025] Automated remote maintenance scheduling module: Maintenance Strategy Knowledge Base: Stores standardized and process-oriented maintenance operation playbooks for different failure modes and equipment models. These playbooks consist of atomic instructions, such as "collect logs -> analyze -> confirm root cause -> perform repair (such as restarting service, modifying configuration, switching spare parts) -> verify results".

[0026] Intelligent Scheduler: Based on the root cause conclusions and recommended remediation solutions output by the fault diagnosis engine, the scheduler automatically matches or dynamically generates maintenance scripts from the knowledge base. The scheduler is responsible for breaking down the scripts into specific, ordered atomic instructions, sending them to the corresponding edge agents for execution through a secure channel, and monitoring the entire execution process.

[0027] Safety and Compliance Controller: Implements functions such as "four-person eye" approval, operation time window restrictions, instruction signature verification, and full-process video recording (recording all inputs and outputs) for all remote maintenance operations to ensure that operations are safe, auditable, and traceable.

[0028] The Knowledge Management and Case Library module continuously and automatically extracts features (fault symptoms, root causes, resolution steps, and verification results) from successful fault handling cases, forming structured knowledge and storing it in the case library. Through Natural Language Processing (NLP) technology, it supports operations and maintenance personnel in retrieving cases and answering questions using natural language. This module enables the continuous accumulation and intelligent reuse of expert experience.

[0029] This invention provides a method for self-diagnosis and remote maintenance of communication equipment faults applied to the above-mentioned system, characterized in that the method includes the following steps: S1: Panoramic Data Perception and Convergence By deploying edge intelligent agents on various communication devices, the system continuously collects device operation status data in a multi-modal (active, passive, event-driven) manner. After preprocessing and normalization, the data is uploaded to the cloud data lake in real time or near real time through a secure channel.

[0030] S2: Intelligent Fault Analysis and Decision-Making The cloud platform performs integrated analysis on the aggregated data from across the entire network: S2.1: Anomaly Detection and Alarm: Utilizes machine learning models to dynamically learn the baseline of normal device behavior, detects anomalies that deviate from the baseline in real time, and generates accurate alarms.

[0031] S2.2: Root Cause Location: When a correlated anomaly occurs, the most likely root cause node is located in the fault propagation graph by combining network topology, service dependencies and historical data, using causal reasoning and graph algorithms, and the confidence level is given.

[0032] S2.3: Predictive early warning: Perform trend prediction and health assessment on key equipment indicators, and provide early warning of possible performance degradation or hardware failure.

[0033] S2.4: Recommended Repair Solution: Based on the root cause localization results, query the maintenance strategy knowledge base or call the digital twin for simulation to generate a safe and effective repair operation plan (or script).

[0034] S3: Secure Automated Remote Maintenance Execution S3.1: Script Scheduling and Command Issuance: The automated remote maintenance scheduling module transforms the repair plan into a specific maintenance script with an approval process, and securely issues the atomic commands in the script to the edge agent of the target device in sequence.

[0035] S3.2: Edge Security Execution and Feedback: The edge agent executes the received instructions in the local security sandbox and feeds back the execution status, progress and results to the cloud in real time.

[0036] S3.3: Execution process monitoring and closed-loop verification: The cloud platform monitors the execution of maintenance operations throughout the process, and automatically collects relevant indicators and logs after the operation is completed to verify whether the fault has been successfully repaired and whether the business has returned to normal, forming a complete closed loop of "perception-analysis-decision-execution-verification".

[0037] S4: Continuous Evolution of Operations and Maintenance Knowledge The entire process data of this fault handling (from the initial phenomenon to the final solution) is treated as a new case. The knowledge management module performs automated analysis and structured storage to enrich the case library and knowledge base. This is used to optimize future fault diagnosis models and maintenance strategies, enabling the system to learn and continuously improve itself.

[0038] Beneficial effects Compared with the prior art, the present invention has the following significant advantages: 1. By using multimodal data fusion and graph-based root cause localization, it is possible to penetrate complex fault symptoms, quickly and accurately locate the source of the problem, significantly shorten the mean time to fault location (MTTI), and reduce the immediate reliance on the experience of senior experts.

[0039] 2. By combining anomaly detection and predictive health models, early warnings can be issued before a failure occurs or its impact escalates, and contingency plans can be generated automatically or assisted, transforming "firefighting" into "fire prevention" and improving network availability.

[0040] 3. The powerful remote maintenance capability enables most software failures, configuration errors, and some hardware problems (such as remote reset and switching of redundant components) to be resolved without on-site intervention, significantly reducing travel costs, personnel costs, and business downtime.

[0041] 4. Through multiple security mechanisms such as digital twin pre-verification, operation approval workflow, instruction signing, and full-process video recording, the risks of remote maintenance operations are controllable, the process is traceable, and the results are reliable, thus avoiding secondary failures caused by misoperation.

[0042] 5. Transform implicit expert experience into explicit, searchable, and reusable structured knowledge bases and case libraries, empowering the entire operations and maintenance team, reducing the skill threshold for personnel and the risks brought about by personnel changes, and driving the continuous evolution of operations and maintenance capabilities.

[0043] 6. The cloud-edge collaborative architecture and modular design enable the system to flexibly adapt to devices from different manufacturers and models. By expanding new data acquisition adapters and diagnostic rules, new device types can be supported, protecting existing investments. Attached Figure Description

[0044] Figure 1 A system architecture flowchart of a fault self-diagnosis and remote maintenance system for communication equipment provided in an embodiment of the present invention; Figure 2 The present invention provides a flowchart of a method for a fault self-diagnosis and remote maintenance system for communication equipment. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Example 1: The fault self-diagnosis and remote maintenance system for communication equipment provided by this invention adopts a cloud-edge collaborative architecture. Edge intelligent agent modules 200 are deployed inside or near widely distributed communication equipment 101 (such as the operator's core router 102, access switch 103, base station 104, etc.). These agent modules 200 maintain a secure connection with a centrally deployed cloud-based intelligent operation and maintenance platform 300 via the Internet or a dedicated network. The platform 300 can be deployed in a public cloud, private cloud, or hybrid cloud environment.

[0047] The core of the Edge Intelligent Agent Module 200 is a miniature runtime environment.

[0048] The data collector 201 integrates multiple protocol adapters (SNMP adapter 2011, CLI adapter 2012, Syslog listener 2013, gNMI adapter 2014, etc.) and is responsible for collecting data from the host device. The data preprocessing engine 202 cleans and formats the collected messy data, for example, mapping the OIDs of CPU utilization from different vendors to a unified metric device.cpu.utilization, and serializing the data into an efficient binary format (such as Protocol Buffers).

[0049] The local diagnostic engine 203 loads model files (such as decision tree models) and rule scripts distributed from the cloud, and can independently determine simple faults such as "interface continuous CRC error exceeds the threshold" and trigger predefined local actions.

[0050] The secure communication client 204 uses MQTT over TLS to maintain a persistent connection with the cloud, responsible for data reporting and command reception. The remote executor 205 is the core security component. It parses JSON-formatted commands sent from the cloud, executes specific operations (such as calling the device's SSH client to execute commands) within a restricted container, and captures all standard input and output to form an "operation recording".

[0051] The data access layer 301 is responsible for receiving massive amounts of edge data and writing it into the data lake 302. The data lake 302 is built based on object storage and a time-series database, and stores historical data of all network devices throughout their entire lifecycle.

[0052] The intelligent fault diagnosis engine 303 is the core of the analysis. Its data fusion unit 3031 correlates metrics, logs, and configuration change events related to the same device within the same time window. The anomaly detector 3032 uses unsupervised learning to establish a dynamic baseline for each key metric of each type of device, generating anomaly events when real-time data deviates significantly from the baseline. The root cause analyzer 3033 maintains a real-time updated network topology and dependency graph. When multiple anomaly events occur simultaneously, it uses causal discovery algorithms (e.g., calculating Granger causality or causal relationships based on transition entropy) to reason on this graph, identify the "source" node most likely to trigger all other anomalies, and provide a diagnostic report. The predictor 3034 uses an LSTM network to predict the trend of key metrics for the next 7 days based on the device's performance data over the past 30 days, and calculates its health score by combining this with the device's runtime.

[0053] The Digital Twin Engine 304 models critical network areas at a 1:1 scale, including device logical models, configuration status, and simulated traffic. Before implementing major remote operations, operations personnel can conduct "sandbox simulations" using this engine.

[0054] The automated operations and maintenance scheduling center 305 includes a script library 3051, which stores standardized processes such as "fixing BGP neighbor jitter" and "cleaning up disk space". The workflow engine 3052 is responsible for orchestrating script execution. The security gateway 3053 digitally signs all outgoing maintenance commands and mandates that high-risk operations require two levels of approval.

[0055] Knowledge Center 306 uses NLP technology to analyze historical work orders and operation records, extract fault patterns, and build a semantically searchable case library 3061.

[0056] The workflow of this system is a continuous "OODA" (Observe, Judge, Decide, Act) cycle.

[0057] S1: Panoramic Data Perception and Aggregation. The edge agent operates continuously, for example, collecting performance metrics every 5 minutes and forwarding logs in real time. After preprocessing and compression, the data is published to the cloud via MQTT topics (such as / device / {device_id} / metric).

[0058] S2: Intelligent fault analysis and decision-making. This step is performed automatically on the cloud platform 300.

[0059] S201: Data fusion unit 3031 has received a new batch of data.

[0060] S202: The anomaly detector 3032 calculates key indicators (such as interface packet loss rate) and marks them as anomalies if they exceed the dynamic baseline (e.g., the baseline is 0.01% and the current value is 0.5%).

[0061] S203: Root cause analyzer 3033 was triggered. It discovered that another metric (CPU utilization) of the same device was also abnormally high at the same time, and that the traffic of several interfaces of its downstream devices suddenly dropped. Combining the topology map, it used graph algorithms to calculate and determined that the root cause was an overheating of a forwarding chip (ASIC) of the device, causing a sudden drop in performance, with a confidence level of 85%. At the same time, predictor 3034 reported that the fan speed of the device was showing a slow downward trend, which corroborated the possibility of a heat dissipation problem.

[0062] S204: Based on the root cause "chip overheating", the system matches a predefined "device overheating handling script" in the script library 3051.

[0063] S3: Secure automated remote maintenance execution.

[0064] S301: The system automatically generates a script instance containing steps such as "checking fan status", "reading temperature sensor", "switching to standby forwarding plane if main chip temperature exceeds X degrees", and "collecting diagnostic information".

[0065] S302: Since "switching forwarding planes" is a high-risk operation, security gateway 3053 generates an approval work order and sends it to the operations supervisor. The supervisor approves it on the mobile device.

[0066] S303: Workflow engine 3052 breaks down the approved script. The first instruction, "show environment temperature," is signed and sent to the edge agent 200 of the target device.

[0067] S304: After the remote executor 205 of the edge agent 200 verifies the instruction signature, it executes the CLI command in the sandbox and returns the result "Temperature: 85C (CRITICAL)".

[0068] S305: Based on the returned result (temperature criticality), the workflow engine decides to execute the next instruction "redundancy switchover to slot 1". This instruction is issued and executed, and the device service switches to the backup board within seconds.

[0069] S306: After execution, the system automatically triggers the verification step, collects the device interface status and service traffic, confirms that the service has been restored, and closes the fault loop.

[0070] S4: Continuous Evolution of Operations and Maintenance Knowledge. The entire process data was automatically packaged, and Knowledge Center 306 extracted key features from it: "Phenomenon: Increased interface packet loss rate, high CPU utilization; Root cause: Overheating of the main forwarding chip; Operation: Switching redundant boards; Result: Success." This new case was stored in Case Library 3061 and can be used for intelligent recommendations of similar faults in the future.

[0071] Example 2: Taking a specific metropolitan area network aggregation switch failure as an example, this illustrates the application edge agent reporting in this system: multiple user ports of aggregation switch SW1 experienced intermittent packet loss alarms.

[0072] Cloud-based analysis: The intelligent diagnostic engine 303's correlation analysis revealed that when packet loss occurred, the CRC error count on SW1's uplink optical port spiked synchronously, while the core router port at its peer showed no abnormalities. Simultaneously, the digital twin engine 304 simulated the link and found no configuration issues. The root cause analyzer 3033 deduced that the root cause was either a degradation of the SW1 uplink optical module or the fiber optic link, with a 90% confidence level.

[0073] Remote Maintenance: The system matches the "Optical Link Degradation Diagnostic Script". The script first issues a command to remotely collect DDMI (Digital Diagnostic Monitoring) information from the SW1 optical module, reporting that the received optical power is below the critical value. Subsequently, the system attempts to issue a command to remotely restart the SW1 optical port (soft reboot), but the problem persists. At this point, the script determines it is a hardware issue, requiring on-site replacement of the optical module or fiber. The system automatically generates an on-site maintenance work order containing precise location information (equipment name, slot port number, and recommended measures), dispatches it to frontline personnel, and provides detailed data before and after the fault for reference.

[0074] Knowledge Accumulation: The case and diagnostic path of "low optical power causing packet loss" have been recorded in the knowledge base, enriching the diagnostic experience for physical layer faults.

[0075] Example 3: In software version upgrade scenarios, the digital twin and simulation verification module of this system plays a crucial role. The plan is to upgrade the versions of 100 routers of the same model across the entire network.

[0076] The operations and maintenance personnel select the upgrade script on the platform interface, specify the target device group and upgrade image.

[0077] The system first selects three devices in a digital twin environment, injects real historical traffic data, and simulates the upgrade process. The simulation report shows that the session reconstruction time for one specific routing protocol may exceed the tolerance range after the upgrade.

[0078] Based on the simulation results, the system automatically optimized the original upgrade script, added a pre-configuration instruction to "modify the routing protocol Hello timer" before the upgrade instruction, and added a special check step after the upgrade.

[0079] The optimized script, after approval, was executed automatically in a "batch, gray-scale" manner. The system monitored the upgrade status and post-upgrade performance of each batch of devices, and only proceeded to the next batch after confirming there were no abnormalities, ensuring a smooth and controllable large-scale upgrade.

[0080] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A fault self-diagnosis and remote maintenance system for communication equipment, characterized in that, The cloud-edge collaborative architecture is adopted, including: Edge intelligent agent modules deployed on or near various communication devices; A cloud-based intelligent operations and maintenance platform deployed in a remote operations and maintenance center; The edge intelligent agent module is used to collect multi-dimensional operating status data of the communication device it resides in, preprocess it and upload it, and securely execute remote maintenance commands from the cloud. The cloud-based intelligent operation and maintenance platform is used to aggregate and analyze data from edge intelligent agent modules across the entire network, perform fault detection, root cause localization and prediction and early warning through an intelligent fault diagnosis engine, and generate and issue secure maintenance command sequences through an automated remote maintenance scheduling module to complete remote fault repair.

2. The fault self-diagnosis and remote maintenance system for communication equipment according to claim 1, characterized in that, The edge intelligent agent module specifically includes: The multi-source heterogeneous data acquisition unit is used to collect performance indicators, logs, configuration and event data from communication devices through active polling, passive listening and event-driven methods; The data preprocessing unit is used to clean, normalize, and compress the collected raw data; A local lightweight diagnostic unit is used to load and run diagnostic models and rules distributed from the cloud, and to perform preliminary fault diagnosis and emergency handling in environments with network outages or low bandwidth. The secure communication unit is used to establish and maintain an encrypted and authenticated two-way secure communication channel with the cloud-based intelligent operation and maintenance platform. The remote command execution unit is used to parse and execute atomic maintenance commands issued from the cloud in a local secure sandbox environment, and record the entire execution process logs and results.

3. The fault self-diagnosis and remote maintenance system for communication equipment according to claim 1, characterized in that, The cloud-based intelligent operation and maintenance platform specifically includes: The data lake and data governance module is used to store and govern time-series data, logs and events from devices across the network, and to build a unified data model. The intelligent fault diagnosis engine module, the core analysis module of the system, further includes: The multimodal data fusion and analysis submodule is used to perform spatiotemporal correlation fusion of metrics, logs, configuration, and topology data; The fault detection and alarm submodule uses an unsupervised learning algorithm to establish a dynamic baseline of device behavior, enabling adaptive threshold anomaly detection and alarm. The root cause localization and analysis submodule constructs a fault propagation graph based on network topology and dependencies, and uses causal inference algorithms or graph neural networks to locate the root cause nodes of the fault in the graph. The predictive health assessment submodule, based on time series prediction models and reliability models, predicts equipment indicator trends and assesses its health and remaining service life. The digital twin and simulation verification module is used to create high-fidelity digital models for critical networks or equipment and to simulate and pre-verify major remote maintenance operations. An automated remote maintenance scheduling module further includes: A maintenance strategy knowledge base is used to store standardized maintenance operation scripts for different failure modes; The intelligent scheduler is used to match or generate maintenance scripts based on fault diagnosis results, and decompose and sort them into atomic instruction sequences for scheduling and distribution. The security and compliance controller is used to approve, sign, record, and perform compliance checks on remote maintenance operations. The knowledge management and case library module is used to automatically extract and structure maintenance knowledge from historical fault handling cases, and supports intelligent retrieval and reuse.

4. The fault self-diagnosis and remote maintenance system for communication equipment according to claim 3, characterized in that, The root cause localization and analysis submodule is specifically used to: when multiple related abnormal events are detected, abstract network devices, links and services into nodes, abstract their physical connections and logical dependencies into edges, and construct a dynamic fault propagation graph; use a causal discovery algorithm based on a probabilistic graphical model or information theory to calculate the posterior probability of each node in the graph being the root cause, and output one or more nodes with the highest probability as the root cause localization result.

5. A fault self-diagnosis and remote maintenance system for communication equipment according to claim 3, characterized in that, The digital twin and simulation verification module is specifically used to: fully reproduce the operation steps in the digital twin environment before implementing configuration changes, software upgrades, or fault recovery operations; evaluate the impact of the operation on network services using historical traffic or simulated traffic; and automatically generate the optimal operation sequence, risk warnings, and rollback plans.

6. A method for self-diagnosis and remote maintenance of communication equipment faults applied to the system described in any one of claims 1-5, characterized in that, Includes the following steps: S1: Collect multi-dimensional operational status data of communication devices through the edge intelligent agent module, preprocess the data and upload it to the cloud; S2: The cloud-based intelligent operation and maintenance platform performs intelligent fault analysis and decision-making on the aggregated network data, including anomaly detection, root cause location, prediction and early warning, and repair solution recommendation; S3: Based on the recommended repair plan, the cloud platform generates a secure remote maintenance script through automated scheduling and sends it to the edge agent module of the target device for execution, while monitoring the execution process and verifying the repair results. S4: The entire process data of this fault handling will be used as a case study for structured analysis and stored in the knowledge base to optimize subsequent fault diagnosis and maintenance strategies.

7. The method for self-diagnosis and remote maintenance of communication equipment faults according to claim 6, characterized in that, Step S2 specifically includes: S2.1: Utilize unsupervised learning models to monitor key performance indicators of equipment in real time, identify abnormal points that deviate from the dynamic baseline, and generate accurate alarms; S2.2: When multiple spatiotemporally correlated anomalies occur, combine real-time network topology and service dependencies, and use graph computing and causal reasoning algorithms to locate the most likely root cause device or component in the fault propagation graph. S2.3: Based on the historical operating data of the equipment, use a time series forecasting model to predict the future trend of its key indicators, and combine the equipment load and age to calculate its comprehensive health score, and provide early warning of potential failures; S2.4: Based on the root cause localization results, query the maintenance strategy knowledge base or generate a remediation plan that includes specific operational steps, risk control, and verification methods through digital twin simulation.

8. The method for self-diagnosis and remote maintenance of communication equipment faults according to claim 6, characterized in that, Step S3 specifically includes: S3.1: The automated remote maintenance scheduling module instantiates the repair plan into an executable maintenance script and triggers the corresponding approval process according to the security policy; S3.2: After approval, the intelligent scheduler decomposes the script into ordered atomic instructions and sends them sequentially to the edge intelligent agent module of the target device through a secure channel; S3.3: The edge intelligent agent module executes each instruction in a local secure environment and feeds back the execution status and results to the cloud in real time; S3.4: The cloud platform monitors the entire script execution pipeline and automatically collects verification data after the script ends to confirm whether the fault has been eliminated and whether the business has returned to normal, thus realizing a closed loop of operation and maintenance.

9. The method for self-diagnosis and remote maintenance of communication equipment faults according to claim 6, characterized in that, Step S4 specifically includes: using natural language processing and machine learning technologies to automatically extract fault modes, root cause features, effective operations, and contextual environments from closed-loop fault work orders, operation records, and performance data to form structured case records; establishing a vectorized index of cases to support case retrieval and recommendation based on semantic similarity, which is used to assist in the analysis and decision-making of new faults.

10. A computing device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 6-9.

Images