Optical access equipment management method and system based on gateway and edge computing

Abstract

The invention provides an optical access equipment management method and system based on gateway and edge computing in the technical field of optical access equipment management. The method comprises the following steps: S1, creating an equipment fault diagnosis model; s2, acquiring a large amount of historical state data of the optical access equipment to construct a data set; s3, dividing the data set into a training set, a verification set and a test set, training the equipment fault diagnosis model through the training set, and verifying and testing the trained equipment fault diagnosis model through the verification set and the test set; s4, compressing the equipment fault diagnosis model passing the test, and deploying the equipment fault diagnosis model to a gateway; s5, the gateway collects real-time state data of the optical access equipment and inputs the real-time state data into the deployed equipment fault diagnosis model to obtain a fault diagnosis report; and step S6, the gateway sends a fault repair instruction to the optical access device based on the fault diagnosis report. The method has the advantages that the timeliness and the safety of management of the optical access equipment are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of optical access device management technology, and in particular to a method and system for managing optical access devices based on gateways and edge computing. Background Technology

[0002] In the field of Information and Communication Technology (ICT), Fiber to the Home (FTTH) technology has become a key infrastructure for the evolution of modern broadband networks. As a high-bandwidth, low-latency transmission solution, FTTH achieves gigabit-level data transmission rates by extending optical fiber to user terminals (such as homes or businesses). With its significant advantage of providing stable network connections, this technology effectively supports emerging applications such as high-definition video streaming, cloud gaming, remote collaboration, Internet of Things (IoT) device access, and 5G enhanced networks, leading to the widespread application of optical access equipment.

[0003] Optical access devices, typically referring to Optical Network Units (ONUs), are responsible for the core functions of converting fiber optic signals into Ethernet signals and managing local network resources on the user side. In home scenarios, ONUs often integrate Wi-Fi router functionality to connect various smart terminals (such as smartphones, laptops, and smart home devices); while in enterprise scenarios, ONUs need to support large-scale device interconnection (such as video conferencing systems and industrial control systems). It is worth noting that home networks often deploy multiple ONU devices to build complex access topologies; while enterprise networks face the challenge of managing dozens or even thousands of devices, placing higher demands on real-time monitoring and fine-grained control capabilities.

[0004] Traditionally, the management of optical access devices relies on centralized cloud platforms. Specifically, a gateway (or CPE) residing at the user's premises collects status data from the optical access devices (such as bandwidth utilization, fault alarm codes, optical signal quality indicators, security logs, etc.) and uploads it to the cloud platform for processing. The cloud platform then issues control commands to the gateway, ultimately achieving centralized management and control of the optical access devices. In this architecture, the gateway primarily acts as a data relay node: real-time status data collected by the optical access devices is aggregated to the gateway via the uplink (such as the WAN interface) and forwarded to the cloud platform; the cloud platform analyzes the received real-time status data, generates optimization strategies, and then transmits the commands to the gateway for execution via the downlink. While this centralized management model has certain advantages in the initial deployment phase when the number of devices is limited (such as ease of deployment and convenient global optimization), its inherent shortcomings become increasingly apparent as the number of devices grows exponentially, mainly manifested in the following ways:

[0005] 1. Significant bandwidth bottlenecks easily lead to network congestion: Massive optical access devices (e.g., a single ONU can generate tens of KB of telemetry data per second) need to report status data to the cloud in real time, consuming a large amount of valuable network bandwidth. In home scenarios, while FTTH can provide high downlink bandwidth (e.g., 1Gbps), uplink bandwidth is usually limited (e.g., 100Mbps). In enterprise networks, when multiple devices concurrently report event logs and performance indicators, the uplink saturation point is easily reached. Relevant statistics show that during peak hours, such uplink traffic can cause latency jitter of more than 20%, affecting the smoothness of real-time applications on the user end (such as video conferencing), and in severe cases, may even cause service degradation or interruption. In addition, control commands issued by the cloud platform (such as configuration updates) also consume downlink bandwidth, forming a bidirectional bandwidth bottleneck. This management model that relies on centralized data transmission inherently lacks good scalability, and the problems of soaring bandwidth costs and degraded user experience are particularly significant in bandwidth-constrained areas or during peak network hours.

[0006] 2. Significant response latency, failing to meet real-time requirements: The inherent processing latency introduced by centralized architecture is insufficient to meet the requirements of high real-time applications. Changes in device status (such as ONU signal anomalies) require multiple processing steps: data is first transmitted from the optical access device to the gateway, then uploaded to the cloud via the internet router (this physical transmission takes an average of 100-300ms); the cloud platform completes analysis and processing; finally, control commands are sent to the gateway for execution via the downlink. The end-to-end latency of the entire process often exceeds 500ms, far exceeding the threshold required for demanding real-time scenarios such as industrial automation control (usually below 100ms).

[0007] 3. High Data Privacy and Security Risks: Uploading data from optical access devices, which contains sensitive information such as user connection records, behavioral data, and even location trajectories, to a cloud platform for processing poses significant privacy and security risks. On the one hand, data is vulnerable to man-in-the-middle attacks (such as DNS hijacking and packet sniffing) during transmission via public internet and other pathways; insufficiently encrypted information is easily stolen. On the other hand, third-party service providers relied upon by cloud storage may have security vulnerabilities (such as improper configuration), leading to unauthorized access. More seriously, the cloud platform itself is susceptible to advanced persistent threats (APTs). Once attackers infiltrate the cloud platform, they can obtain large amounts of real-time status data from optical access devices, which can then be used to infer user privacy (such as family routines and sensitive corporate information) or to carry out large-scale malicious manipulation.

[0008] Therefore, how to provide a management method and system for optical access devices based on gateways and edge computing to improve the timeliness and security of optical access device management has become an urgent technical problem to be solved. Summary of the Invention

[0009] The technical problem to be solved by the present invention is to provide a method and system for managing optical access devices based on gateways and edge computing, so as to improve the timeliness and security of optical access device management.

[0010] In a first aspect, the present invention provides a method for managing optical access devices based on gateways and edge computing, comprising the following steps:

[0011] Step S1: Create a device fault diagnosis model for outputting fault diagnosis reports based on a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and set the loss function of the device fault diagnosis model;

[0012] Step S2: Obtain a large amount of historical status data of optical access devices, preprocess and label each historical status data to construct a dataset;

[0013] Step S3: Divide the dataset into a training set, a validation set, and a test set. Train the equipment fault diagnosis model using the training set and the loss function. Validate and test the trained equipment fault diagnosis model using the validation set and the test set.

[0014] Step S4: After compressing the tested device fault diagnosis model, deploy it to the gateway;

[0015] Step S5: The gateway collects real-time status data of the optical access device, inputs the real-time status data into the deployed device fault diagnosis model, and obtains a fault diagnosis report.

[0016] Step S6: The gateway sends a fault repair instruction to the optical access device based on the fault diagnosis report and obtains the fault repair result fed back by the optical access device.

[0017] Step S7: The gateway encrypts the fault diagnosis report and fault repair results into an encrypted data packet and uploads the encrypted data packet to the cloud platform;

[0018] Step S8: The cloud platform decrypts the received encrypted data packet to obtain the fault diagnosis report and fault repair results. The cloud platform aggregates and analyzes the fault diagnosis reports and fault repair results uploaded by each gateway to generate a device optimization strategy. The device optimization strategy is then distributed to the optical access device through the gateway.

[0019] Furthermore, in step S1, the multi-source feature extraction layer is constructed based on the device operation channel, optical module channel, optical fiber link channel, user port channel, network management channel, and timing processing channel;

[0020] The device operation channel is used to extract device operation status features from device operation status data through a dual-layer LSTM unit; the optical module channel is used to extract optical module status features from optical module status data through a CNN-Transformer hybrid unit; the fiber optic link channel is used to extract fiber optic link status features from fiber optic link status data through a graph convolutional network unit; the user port channel is used to extract user port status features from user port status data through a conditional random field unit; and the network management channel is used to extract network management status features from network management status data through a BERT fine-tuning unit.

[0021] The cross-modal fusion layer is used to fuse the device operating status features, optical module status features, optical fiber link status features, user port status features, and network management status features through the feature cross-module and dependency modeling module to obtain fused features;

[0022] The multi-task output layer is used to output a fault diagnosis report, including fault identifier, fault type, fault code, and fault handling suggestions, based on the fusion features.

[0023] Furthermore, step S2 specifically includes:

[0024] Acquire a large amount of historical status data for optical access devices. This historical status data includes device operating status data, optical module status data, fiber optic link status data, user port status data, and network management status data. The device operating status data includes device online status, device temperature, and device power status. The optical module status data includes optical module type, optical module transmit power, optical module receive power, and optical module temperature. The fiber optic link status data includes fiber connection status, fiber link loss, and fiber link length. The user port status data includes port rate, port connection status, port traffic, and port bit error rate. The network management status data includes device configuration information, device alarm information, and device log information.

[0025] The historical state data is preprocessed, including at least deduplication, missing value handling, error correction and data standardization. The preprocessed historical state data is then labeled with at least fault identifiers, fault types, fault codes and fault handling suggestions. A dataset is constructed based on the labeled historical state data.

[0026] Furthermore, step S3 specifically includes:

[0027] Based on stratified sampling, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The equipment fault diagnosis model is trained using the training set and a loss function. During the training process, the hyperparameters of the equipment fault diagnosis model are continuously optimized until the loss value of the loss function is less than a preset loss threshold.

[0028] The diagnostic accuracy is calculated using the validation set to validate the trained equipment fault diagnosis model. If the validation fails, the training set is expanded and training continues; if the validation passes, then:

[0029] The F1 score is calculated using the test set to test the validated equipment fault diagnosis model. If the test fails, the training set is expanded and training continues; if the test passes, training ends.

[0030] Furthermore, step S4 specifically includes:

[0031] The tested equipment fault diagnosis model is compressed using knowledge distillation technology, the performance of the compressed equipment fault diagnosis model is verified, and the equipment fault diagnosis model that has passed the performance verification is deployed to the gateway.

[0032] Step S5 specifically involves:

[0033] The gateway collects real-time status data of optical access devices through GPON or XGPON, preprocesses the real-time status data, and inputs it into the deployed device fault diagnosis model to obtain a fault diagnosis report including fault identifier, fault type, fault code, and fault handling suggestions.

[0034] Step S6 specifically involves:

[0035] Based on the fault handling suggestions carried in the fault diagnosis report, the gateway sends a fault repair command to the optical access device in real time and obtains the fault repair result from the optical access device, indicating whether the repair was successful or failed.

[0036] Step S7 specifically involves:

[0037] The gateway obtains the current timestamp T, concatenates the fault diagnosis report and fault repair results into initial data D, calculates the hash value H of the initial data D using the hash256 algorithm, and uses the first 128 bits of the hash value H as key K1 and the last 128 bits as key K2.

[0038] The initial data D is divided into data block D1 and data block D2. Data block D1 is encrypted using the AES algorithm with key K1 to obtain encrypted data block E1. Data block D2 is encrypted using the RSA algorithm with key K2 to obtain encrypted data block E2. Encrypted data block E1 and encrypted data block E2 are concatenated to obtain concatenated data P.

[0039] A random string is generated based on the timestamp T. The random string is added to the beginning and end of the concatenated data P to obtain obfuscated data C. The obfuscated data C is then encrypted using the Blowfish algorithm to obtain obfuscated ciphertext data E3.

[0040] Calculate the signature data Sig of the timestamp T, and encrypt the obfuscated ciphertext data E3, hash value H, timestamp T and signature data Sig into an encrypted data packet using the AES algorithm, and upload the encrypted data packet to the cloud platform.

[0041] Secondly, the present invention provides an optical access device management system based on gateways and edge computing, comprising the following modules:

[0042] The equipment fault diagnosis model creation module is used to create an equipment fault diagnosis model for outputting fault diagnosis reports based on a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and to set the loss function of the equipment fault diagnosis model.

[0043] The dataset construction module is used to acquire a large amount of historical status data of optical access devices, and to construct a dataset after preprocessing and labeling the historical status data.

[0044] The equipment fault diagnosis model training module is used to divide the dataset into a training set, a validation set, and a test set, train the equipment fault diagnosis model using the training set and a loss function, and validate and test the trained equipment fault diagnosis model using the validation set and the test set.

[0045] The equipment fault diagnosis model deployment module is used to compress the tested equipment fault diagnosis model and deploy it to the gateway.

[0046] The fault diagnosis module is used by the gateway to collect real-time status data of the optical access device, input the real-time status data into the deployed device fault diagnosis model, and obtain a fault diagnosis report.

[0047] The fault repair module is used by the gateway to send a fault repair instruction to the optical access device based on the fault diagnosis report and to obtain the fault repair result fed back by the optical access device.

[0048] The report encryption upload module is used by the gateway to encrypt the fault diagnosis report and fault repair results into an encrypted data packet, and then upload the encrypted data packet to the cloud platform;

[0049] The device optimization strategy distribution module is used by the cloud platform to decrypt the received encrypted data packets to obtain fault diagnosis reports and fault repair results. The cloud platform aggregates and analyzes the fault diagnosis reports and fault repair results uploaded by each gateway to generate device optimization strategies, and distributes the device optimization strategies to the optical access devices through the gateways.

[0050] Furthermore, in the equipment fault diagnosis model creation module, the multi-source feature extraction layer is constructed based on the equipment operation channel, optical module channel, optical fiber link channel, user port channel, network management channel, and timing processing channel;

[0051] The device operation channel is used to extract device operation status features from device operation status data through a dual-layer LSTM unit; the optical module channel is used to extract optical module status features from optical module status data through a CNN-Transformer hybrid unit; the fiber optic link channel is used to extract fiber optic link status features from fiber optic link status data through a graph convolutional network unit; the user port channel is used to extract user port status features from user port status data through a conditional random field unit; and the network management channel is used to extract network management status features from network management status data through a BERT fine-tuning unit.

[0052] The cross-modal fusion layer is used to fuse the device operating status features, optical module status features, optical fiber link status features, user port status features, and network management status features through the feature cross-module and dependency modeling module to obtain fused features;

[0053] The multi-task output layer is used to output a fault diagnosis report, including fault identifier, fault type, fault code, and fault handling suggestions, based on the fusion features.

[0054] Furthermore, the dataset construction module is specifically used for:

[0055] Acquire a large amount of historical status data for optical access devices. This historical status data includes device operating status data, optical module status data, fiber optic link status data, user port status data, and network management status data. The device operating status data includes device online status, device temperature, and device power status. The optical module status data includes optical module type, optical module transmit power, optical module receive power, and optical module temperature. The fiber optic link status data includes fiber connection status, fiber link loss, and fiber link length. The user port status data includes port rate, port connection status, port traffic, and port bit error rate. The network management status data includes device configuration information, device alarm information, and device log information.

[0056] The historical state data is preprocessed, including at least deduplication, missing value handling, error correction and data standardization. The preprocessed historical state data is then labeled with at least fault identifiers, fault types, fault codes and fault handling suggestions. A dataset is constructed based on the labeled historical state data.

[0057] Furthermore, the equipment fault diagnosis model training module is specifically used for:

[0058] Based on stratified sampling, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The equipment fault diagnosis model is trained using the training set and a loss function. During the training process, the hyperparameters of the equipment fault diagnosis model are continuously optimized until the loss value of the loss function is less than a preset loss threshold.

[0059] The diagnostic accuracy is calculated using the validation set to validate the trained equipment fault diagnosis model. If the validation fails, the training set is expanded and training continues; if the validation passes, then:

[0060] The F1 score is calculated using the test set to test the validated equipment fault diagnosis model. If the test fails, the training set is expanded and training continues; if the test passes, training ends.

[0061] Furthermore, the equipment fault diagnosis model deployment module is specifically used for:

[0062] The tested equipment fault diagnosis model is compressed using knowledge distillation technology, the performance of the compressed equipment fault diagnosis model is verified, and the equipment fault diagnosis model that has passed the performance verification is deployed to the gateway.

[0063] The fault diagnosis module is specifically used for:

[0064] The gateway collects real-time status data of optical access devices through GPON or XGPON, preprocesses the real-time status data, and inputs it into the deployed device fault diagnosis model to obtain a fault diagnosis report including fault identifier, fault type, fault code, and fault handling suggestions.

[0065] The fault repair module is specifically used for:

[0066] Based on the fault handling suggestions carried in the fault diagnosis report, the gateway sends a fault repair command to the optical access device in real time and obtains the fault repair result from the optical access device, indicating whether the repair was successful or failed.

[0067] The encrypted report upload module is specifically used for:

[0068] The gateway obtains the current timestamp T, concatenates the fault diagnosis report and fault repair results into initial data D, calculates the hash value H of the initial data D using the hash256 algorithm, and uses the first 128 bits of the hash value H as key K1 and the last 128 bits as key K2.

[0069] The initial data D is divided into data block D1 and data block D2. Data block D1 is encrypted using the AES algorithm with key K1 to obtain encrypted data block E1. Data block D2 is encrypted using the RSA algorithm with key K2 to obtain encrypted data block E2. Encrypted data block E1 and encrypted data block E2 are concatenated to obtain concatenated data P.

[0070] A random string is generated based on the timestamp T. The random string is added to the beginning and end of the concatenated data P to obtain obfuscated data C. The obfuscated data C is then encrypted using the Blowfish algorithm to obtain obfuscated ciphertext data E3.

[0071] Calculate the signature data Sig of the timestamp T, and encrypt the obfuscated ciphertext data E3, hash value H, timestamp T and signature data Sig into an encrypted data packet using the AES algorithm, and upload the encrypted data packet to the cloud platform.

[0072] The advantages of this invention are:

[0073] 1. A device fault diagnosis model is created through a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and a loss function is set for the model. Then, a dataset is constructed by acquiring a large amount of historical status data from optical access devices. This dataset is divided into training, validation, and test sets. The device fault diagnosis model is trained using the training set and the loss function, and validated and tested using the validation and test sets. The tested model is compressed and deployed to the gateway. The gateway collects real-time status data from the optical access devices and inputs it into the fault diagnosis model to obtain a fault diagnosis report. Based on the report, the gateway sends a fault repair command to the optical access devices, obtains the repair results, and encrypts the report and repair results into an encrypted data packet, which is then uploaded to the cloud platform. The cloud platform decrypts the encrypted data packet to retrieve the report and repair results. The cloud platform aggregates and analyzes fault diagnosis reports and fault repair results uploaded by each gateway, generates device optimization strategies, and distributes these strategies to optical access devices through the gateways. In other words, the device fault diagnosis model is compressed and deployed to the gateway, enabling the gateway to directly collect and process real-time status data of the optical access devices locally, generate fault diagnosis reports, and immediately issue fault repair instructions. This achieves localized closed-loop processing, avoiding the data transmission delays of traditional cloud processing, thus compressing response time to the millisecond level and significantly improving timeliness. Simultaneously, the gateway only encrypts lightweight fault diagnosis reports and repair results into encrypted data packets before uploading, rather than the original status data, reducing the exposure of sensitive information. End-to-end encryption mechanisms prevent man-in-the-middle attacks during transmission, and local processing avoids remote transmission of raw data, fundamentally reducing privacy leaks and security risks, ultimately greatly improving the timeliness and security of optical access device management.

[0074] 2. By deploying a compressed and optimized multimodal fault diagnosis model at the edge of the gateway node, millisecond-level real-time fault diagnosis and automated repair of optical access equipment are achieved, significantly improving response speed and equipment availability. Simultaneously, by utilizing multi-source feature extraction and cross-modal fusion technologies, combined with high-precision analysis of multi-dimensional data such as equipment operation, optical links, and user ports, diagnostic accuracy is greatly enhanced. In terms of data closed-loop, relying on the cloud to aggregate and analyze encrypted uploaded repair results and generate optimization strategies to feed back to the equipment, a secure collaborative mechanism of "edge real-time processing - cloud intelligent optimization" is formed, effectively reducing operation and maintenance costs and extending equipment lifespan. Its modular design is compatible with mainstream protocols such as GPON / XGPON, and combined with knowledge distillation model compression and hybrid encryption technologies, it ensures data security while also considering efficient deployment and sustainable iteration capabilities in resource-constrained scenarios, comprehensively improving the management intelligence level and operational economy of the optical access network.

[0075] 3. By constructing multiple channels (device operation, optical module, fiber optic link, user port, network management, and timing processing channels), multi-dimensional status data (such as timing data, text data, graph structure data, etc.) are integrated. This multi-source feature extraction avoids the limitations of relying on a single data source and can cover all aspects of device operation (such as physical layer, network layer, user layer, and management layer), thereby significantly improving the accuracy and robustness of fault diagnosis and reducing false alarm rate and false negative rate.

[0076] 4. The dual-layer LSTM unit in the equipment operation channel: Specifically designed for processing time-series data, it captures long-term dependencies (such as continuous changes in equipment operating status), improving the ability to identify periodic faults; the CNN-Transformer hybrid unit in the optical module channel: Combining the local feature extraction of convolutional neural networks (CNN) and the global context understanding of Transformers, it efficiently processes image or sequence data (such as optical module performance parameters), reducing computational redundancy; the graph convolutional network (GCN) unit in the fiber optic link channel: Adapted to the graph structure data of fiber optic links (such as network topology), it effectively... Modeling dependencies between nodes improves sensitivity to link failures (such as interruptions or interference); the Conditional Random Field (CRF) unit in the user port channel is suitable for sequence labeling tasks (such as port state changes), enhancing the detection accuracy of user-end anomalies (such as congestion or attacks) by modeling dependencies between states; the BERT fine-tuning unit in the network management channel uses a pre-trained language model to process text data (such as network logs), and quickly understands semantics after fine-tuning, accelerating the analysis of management-level faults; this customized extraction strategy not only improves data processing efficiency (such as GPU resource optimization) but also reduces model complexity.

[0077] 5. The cross-modal fusion layer's feature cross-module enables interaction between features (such as associating device status with fiber optic links), discovers hidden patterns (such as link problems caused by device anomalies), and improves the diagnosis of complex faults (such as fault chains caused by multiple factors); the dependency modeling module explicitly models the dependencies between different data sources (such as the impact of network management data on user ports), ensuring that the fused features have high-dimensional representation and reducing redundancy; this fusion mechanism is superior to simple feature splicing, providing stronger generalization ability, especially in variable network environments, and can adapt to new fault types.

[0078] 6. By constructing a multi-source feature extraction layer (covering six channels: equipment operation, optical modules, fiber optic links, user ports, network management, and time-series processing), and employing customized, high-efficiency feature extraction units (such as LSTM for time-series data processing, graph convolution for network topology adaptation, and BERT for parsing text logs) tailored to the data type characteristics of each channel, deep feature mining of multi-dimensional state data is achieved. Its innovative cross-modal fusion layer effectively integrates the correlation between heterogeneous features through feature cross-referencing and dependency modeling, significantly improving the identification accuracy and robustness of complex faults. Finally, the multi-task output layer generates an end-to-end diagnostic report containing fault identifiers, types, codes, and processing suggestions, ensuring system scalability while comprehensively optimizing the efficiency, accuracy, and practicality of fault diagnosis for optical access equipment.

[0079] 7. By cleverly integrating multi-layered encryption, data obfuscation, and integrity verification mechanisms, the security and reliability of data transmission are significantly improved. The core advantages are: using hash values ​​to derive dynamic keys to achieve efficient and self-consistent key management; combining AES and RSA hybrid encryption strategies to differentiate the processing of segmented data blocks to balance security and efficiency; generating random strings through timestamps for header and tail obfuscation; and employing Blowfish encryption to destroy data, greatly enhancing the ability to resist reverse analysis and eavesdropping; while integrating hash verification, timestamp signatures, and other multiple protections, it not only ensures the integrity and freshness of data but also provides non-repudiable audit trails, ultimately constructing a robust and secure transmission system that can effectively resist tampering, replay attacks, and single point of failure risks. Attached Figure Description

[0080] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0081] Figure 1 This is a flowchart of a method for managing optical access devices based on gateways and edge computing according to the present invention.

[0082] Figure 2 This is a schematic diagram of the structure of an optical access device management system based on gateway and edge computing according to the present invention. Detailed Implementation

[0083] The overall concept of the technical solution in this application embodiment is as follows: The device fault diagnosis model is compressed and deployed to the gateway, enabling the gateway to directly collect and process real-time status data of the optical access device locally, generate a fault diagnosis report, and immediately issue fault repair instructions. This achieves localized closed-loop processing, avoiding the data transmission delay of traditional cloud processing, thereby compressing the response time to the millisecond level and significantly improving timeliness. Simultaneously, the gateway only encrypts the lightweight fault diagnosis report and repair results into encrypted data packets before uploading, rather than the original status data. This reduces the exposure of sensitive information and prevents man-in-the-middle attacks during transmission through end-to-end encryption. Combined with local processing to avoid remote transmission of original data, this fundamentally reduces privacy leaks and security risks, thereby improving the timeliness and security of optical access device management.

[0084] Please refer to Figures 1 to 2 As shown, a preferred embodiment of the optical access device management method based on gateway and edge computing of the present invention includes the following steps:

[0085] Step S1: Create a device fault diagnosis model for outputting fault diagnosis reports based on a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and set the loss function of the device fault diagnosis model;

[0086] Step S2: Obtain a large amount of historical status data of optical access devices, preprocess and label each historical status data to construct a dataset;

[0087] Step S3: Divide the dataset into a training set, a validation set, and a test set. Train the equipment fault diagnosis model using the training set and the loss function. Validate and test the trained equipment fault diagnosis model using the validation set and the test set.

[0088] Step S4: After compressing the tested device fault diagnosis model, deploy it to the gateway;

[0089] Step S5: The gateway collects real-time status data of the optical access device, inputs the real-time status data into the deployed device fault diagnosis model, and obtains a fault diagnosis report.

[0090] Step S6: The gateway sends a fault repair instruction to the optical access device based on the fault diagnosis report and obtains the fault repair result fed back by the optical access device.

[0091] Step S7: The gateway encrypts the fault diagnosis report and fault repair results into an encrypted data packet and uploads the encrypted data packet to the cloud platform;

[0092] Step S8: The cloud platform decrypts the received encrypted data packet to obtain the fault diagnosis report and fault repair results. The cloud platform aggregates and analyzes the fault diagnosis reports and fault repair results uploaded by each gateway to generate a device optimization strategy. The device optimization strategy is then distributed to the optical access device through the gateway.

[0093] In step S1, the multi-source feature extraction layer is constructed based on the device operation channel, optical module channel, optical fiber link channel, user port channel, network management channel, and timing processing channel.

[0094] The device operation channel is used to extract device operation status features from device operation status data through a dual-layer LSTM unit; the optical module channel is used to extract optical module status features from optical module status data through a CNN-Transformer hybrid unit; the fiber optic link channel is used to extract fiber optic link status features from fiber optic link status data through a graph convolutional network unit; the user port channel is used to extract user port status features from user port status data through a conditional random field unit; and the network management channel is used to extract network management status features from network management status data through a BERT fine-tuning unit.

[0095] By constructing multiple channels (device operation, optical module, fiber optic link, user port, network management, and timing processing channels), multi-dimensional status data (such as timing data, text data, and graph structure data) are integrated. This multi-source feature extraction avoids the limitations of relying on a single data source and can cover all aspects of device operation (such as the physical layer, network layer, user layer, and management layer), thereby significantly improving the accuracy and robustness of fault diagnosis and reducing false alarm and false negative rates.

[0096] The dual-layer LSTM unit in the equipment operation channel is specifically designed for processing time-series data, capturing long-term dependencies (such as continuous changes in equipment operating status) and improving the ability to identify periodic faults. The CNN-Transformer hybrid unit in the optical module channel combines the local feature extraction of convolutional neural networks (CNNs) with the global context understanding of Transformers, efficiently processing image or sequence data (such as optical module performance parameters) and reducing computational redundancy. The graph convolutional network (GCN) unit in the fiber optic link channel adapts to the graph structure data of fiber optic links (such as network topology), effectively building... The dependency relationships between module nodes enhance the sensitivity to link failures (such as interruptions or interference); the Conditional Random Field (CRF) unit in the user port channel is suitable for sequence labeling tasks (such as port state changes), and enhances the detection accuracy of user-end anomalies (such as congestion or attacks) by modeling the dependencies between states; the BERT fine-tuning unit in the network management channel uses a pre-trained language model to process text data (such as network logs), and quickly understands semantics after fine-tuning, accelerating the analysis of management-level faults; this customized extraction strategy not only improves data processing efficiency (such as GPU resource optimization), but also reduces model complexity.

[0097] The cross-modal fusion layer is used to fuse the device operating status features, optical module status features, optical fiber link status features, user port status features, and network management status features through the feature cross-module and dependency modeling module to obtain fused features;

[0098] The feature cross-addressing module is built based on multi-head cross-attention (8 heads), taking 128-dimensional features from 5 channels as input, and is used to calculate cross-source feature weights such as optical module-fiber and port-network management. The dependency modeling module is built based on hierarchical graph neural network (HGNN), used to establish a three-order relationship graph of device-port-fiber. The message passing layer performs 3 rounds of node updates, and the graph readout layer aggregates global features and outputs a 256-dimensional fused feature vector.

[0099] The cross-modal fusion layer's feature cross-module enables interaction between features (such as associating device status with fiber optic links), discovers hidden patterns (such as link problems caused by device anomalies), and improves the diagnosis of complex faults (such as fault chains caused by multiple factors). The dependency modeling module explicitly models the dependencies between different data sources (such as the impact of network management data on user ports), ensuring that the fused features have high-dimensional representation and reducing redundancy. This fusion mechanism is superior to simple feature splicing, providing stronger generalization capabilities, especially in dynamic network environments, and can adapt to new fault types.

[0100] The multi-task output layer is used to output a fault diagnosis report, including fault identifier, fault type, fault code, and fault handling suggestions, based on the fusion features.

[0101] The multi-task output layer is constructed based on a fault identifier head, a joint fault decoding module, and a processing suggestion generator. The fault identifier head has the following structure: fully connected (256→64) + Dropout (0.3) → Sigmoid output, used for binary classification to determine the existence of faults. The joint fault decoding module has the following structure: a) Feature routing layer: dynamically reweighted based on the fault identifier output; b) Fault type branch: fully connected (64→32) → Softmax (K classes); c) Fault code branch: Graph Attention Decoder (GAT), generating M-class code distributions. The processing suggestion generator uses a Conditional Generative Adversarial Network (CGAN) generator, with the input being a concatenated vector of fault type + code, and the output being an N-dimensional suggestion vector (Sigmoid multi-label).

[0102] By constructing a multi-source feature extraction layer (covering six channels: equipment operation, optical modules, fiber optic links, user ports, network management, and time-series processing), and employing customized, high-efficiency feature extraction units (such as LSTM for time-series data processing, graph convolution for network topology adaptation, and BERT for parsing text logs) tailored to the data type characteristics of each channel, deep feature mining of multi-dimensional state data is achieved. Its innovative cross-modal fusion layer effectively integrates the correlation between heterogeneous features through feature cross-referencing and dependency modeling, significantly improving the identification accuracy and robustness of complex faults. Finally, the multi-task output layer generates an end-to-end diagnostic report containing fault identifiers, types, codes, and processing suggestions, ensuring system scalability while comprehensively optimizing the efficiency, accuracy, and practicality of fault diagnosis for optical access equipment.

[0103] The formula for the loss function is:

[0104] L=λ1*L_id+λ2*L_type+λ3*L_code+λ4*L_adv;

[0105] Where L represents the loss value of the loss function; L_id represents the fault identification loss, which uses Focal Loss (γ=2, α=0.8) to alleviate sample imbalance; L_type represents the type loss, which uses label smoothing cross-entropy (smoothing factor=0.1); L_code represents the code loss, which uses constrained KL divergence (constraining the fault type-code logical relationship); and L_adv represents the suggestion loss, which uses generative adversarial loss (Wasserstein distance) + reconstruction loss.

[0106] Step S2 specifically involves:

[0107] Acquire a large amount of historical status data for optical access devices. This historical status data includes device operating status data, optical module status data, fiber optic link status data, user port status data, and network management status data. The device operating status data includes device online status (indicating whether the device is properly connected to the network. For example, an optical access device is "online" if it is powered on and has a valid network connection; if the device loses power or the fiber optic link is interrupted, its status is "offline") and device temperature (this is an important indicator of the device's internal heat dissipation. Optical access devices typically contain various electronic components; excessively high temperatures can lead to performance degradation or even damage. For example, the normal operating temperature range of the device may be 0℃-50℃). If the temperature exceeds 50℃, the device may issue a high-temperature alarm. The device power status includes information such as input voltage and current. For example, the normal input voltage range is -48V ± 5%. If the voltage exceeds this range, it may cause abnormal device operation, and the power status data can reflect this situation in a timely manner. The optical module status data includes the optical module type (different optical access devices may support multiple types of optical modules, such as SFP (Small Form Factor), SFP+, XFP, etc. Status data can identify the type of the currently inserted optical module to confirm whether it meets the device requirements), and the optical module transmit power (this is an indicator of the strength of the optical signal transmitted by the optical module. For example, for a 1000BASE-S...). X-ray modules typically have several power ranges: transmit power (ranging from -10dBm to -3dBm; transmit power below -10dBm can cause excessive signal attenuation during transmission, affecting communication quality); receive power (representing the strength of the received optical signal; for example, receive power ranges from -10dBm to -20dBm; receive power below -20dBm may result in signal loss or increased bit error rate); and temperature (optical modules also contain temperature-sensitive components; excessively high or low temperatures can affect performance; for example, the normal operating temperature range for optical modules is -5℃ to 70℃; temperatures exceeding this range may lead to performance degradation or even damage). The fiber optic link status data includes fiber optic connection status (indicating whether the fiber optic cable is correctly connected. For example, if the fiber optic plug is not properly inserted or the fiber is damaged, the fiber optic connection status may show "not connected" or "link fault"), fiber optic link loss (this is an important indicator of fiber optic link transmission efficiency. Fiber optic link loss is mainly caused by factors such as the attenuation of the fiber itself, connector loss, and fusion splice loss. For example, for single-mode fiber, the attenuation is approximately 0.2dB-0.5dB per kilometer. If the fiber optic link loss is too high, it may cause the optical signal to fail to transmit normally), and fiber optic link length (the length of the fiber optic link can be measured using devices such as an optical time domain reflectometer (OTDR). This is very important for the maintenance and fault location of fiber optic networks).For example, in an optical access network, the length of the fiber optic link may vary from several kilometers to tens of kilometers. If the fiber optic link length does not match the designed length, it may be because the fiber has been accidentally cut or incorrectly connected. The user port status data includes port rate (representing the transmission rate of the user port. For example, common rates are 100Mbps, 1Gbps, 10Gbps, etc. If the port rate is set incorrectly, it may cause users to be unable to communicate normally), port connection status (indicating whether the user port is connected to a terminal device. For example, if the user port is connected to a computer, the port connection status may show "connected"; if no device is connected, the status is "not connected"), port traffic (including uplink and downlink traffic. For example, in a home broadband access scenario, the user's downlink traffic may reach tens to hundreds of Mbps, while the uplink traffic is relatively small. By monitoring port traffic, we can understand the user's network usage and also detect abnormal traffic in a timely manner, such as network attacks or malicious downloads), and port bit error rate (representing the probability of errors occurring during data transmission at the user port. For example, the bit error rate is generally required to be below 10^-9. If the bit error rate is low...). Excessive temperatures may lead to data transmission errors and affect the user's network experience. The network management status data includes device configuration information (including network parameters such as the device's IP address, subnet mask, and gateway. For example, the device's IP address might be 192.168.1.1, the subnet mask 255.255.255.0, and the gateway 192.168.1.254. If the configuration information is incorrect, the device may not be able to access the network normally), device alarm information (alarm information is generated when the device malfunctions. For example, the device may generate high temperature alarms, optical module failure alarms, and fiber optic link interruption alarms. By reviewing alarm information, maintenance personnel can quickly locate problems and take measures), and device log information (recording various events during the device's operation, including user logins, configuration modifications, and fault occurrences. For example, the device log may record that a user logged into the device at a certain time and made configuration modifications, or that the device detected a fiber optic link interruption at a certain time. By analyzing the device log, the device's operating history can be understood, facilitating fault diagnosis and security auditing).

[0108] The historical state data is preprocessed, including at least deduplication, missing value handling, error correction and data standardization. The preprocessed historical state data is then labeled with at least fault identifiers, fault types, fault codes and fault handling suggestions. A dataset is constructed based on the labeled historical state data.

[0109] Step S3 specifically involves:

[0110] Based on stratified sampling, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The equipment fault diagnosis model is trained using the training set and a loss function. During the training process, the hyperparameters of the equipment fault diagnosis model are continuously optimized until the loss value of the loss function is less than a preset loss threshold.

[0111] The diagnostic accuracy is calculated using the validation set to validate the trained equipment fault diagnosis model. If the validation fails, the training set is expanded and training continues; if the validation passes, then:

[0112] The F1 score is calculated using the test set to test the validated equipment fault diagnosis model. If the test fails, the training set is expanded and training continues; if the test passes, training ends.

[0113] Step S4 specifically involves:

[0114] The tested equipment fault diagnosis model is compressed using knowledge distillation technology, the performance of the compressed equipment fault diagnosis model is verified, and the equipment fault diagnosis model that has passed the performance verification is deployed to the gateway.

[0115] Step S5 specifically involves:

[0116] The gateway collects real-time status data of optical access devices through GPON or XGPON, preprocesses the real-time status data, and inputs it into the deployed device fault diagnosis model to obtain a fault diagnosis report including fault identifier, fault type, fault code, and fault handling suggestions.

[0117] Step S6 specifically involves:

[0118] Based on the fault handling suggestions carried in the fault diagnosis report, the gateway sends a fault repair command to the optical access device in real time and obtains the fault repair result from the optical access device, indicating whether the repair was successful or failed.

[0119] Step S7 specifically involves:

[0120] The gateway obtains the current timestamp T, concatenates the fault diagnosis report and fault repair results into initial data D, calculates the hash value H of the initial data D using the hash256 algorithm, and uses the first 128 bits of the hash value H as key K1 and the last 128 bits as key K2.

[0121] The initial data D is divided into data block D1 and data block D2. Data block D1 is encrypted using the AES algorithm with key K1 to obtain encrypted data block E1. Data block D2 is encrypted using the RSA algorithm with key K2 to obtain encrypted data block E2. Encrypted data block E1 and encrypted data block E2 are concatenated to obtain concatenated data P.

[0122] A random string is generated based on the timestamp T. The random string is added to the beginning and end of the concatenated data P to obtain obfuscated data C. The obfuscated data C is then encrypted using the Blowfish algorithm to obtain obfuscated ciphertext data E3.

[0123] Calculate the signature data Sig of the timestamp T, and encrypt the obfuscated ciphertext data E3, hash value H, timestamp T and signature data Sig into an encrypted data packet using the AES algorithm, and upload the encrypted data packet to the cloud platform.

[0124] By cleverly integrating multi-layered encryption, data obfuscation, and integrity verification mechanisms, the security and reliability of data transmission are significantly improved. The core advantages are: efficient and self-consistent key management through hash-derived dynamic keys; differentiated processing of segmented data blocks using a hybrid AES and RSA encryption strategy to balance security and efficiency; head and tail obfuscation through timestamp-generated random strings; and a Blowfish encryption corruption mode to significantly enhance resistance to reverse engineering and eavesdropping. Simultaneously, the integration of hash verification, timestamp signatures, and other multiple protections ensures data integrity and freshness while providing non-repudiable audit trails, ultimately constructing a robust and secure transmission system that effectively resists tampering, replay attacks, and single-point-of-failure risks.

[0125] By deploying a compressed and optimized multimodal fault diagnosis model at the edge of the gateway node, millisecond-level real-time fault diagnosis and automated repair of optical access equipment are achieved, significantly improving response speed and equipment availability. Simultaneously, multi-source feature extraction and cross-modal fusion technologies, combined with high-precision analysis of multi-dimensional data such as equipment operation, optical links, and user ports, greatly enhance diagnostic accuracy. In terms of data closed-loop, the cloud aggregates and analyzes encrypted repair results and generates optimization strategies that are fed back to the equipment, forming a secure collaborative mechanism of "edge real-time processing - cloud intelligent optimization," effectively reducing maintenance costs and extending equipment lifespan. Its modular design is compatible with mainstream protocols such as GPON / XGPON, and combined with knowledge distillation model compression and hybrid encryption technologies, it ensures data security while also considering efficient deployment and sustainable iteration capabilities in resource-constrained scenarios, comprehensively improving the management intelligence and operational economy of the optical access network.

[0126] A preferred embodiment of the optical access device management system based on gateway and edge computing of the present invention includes the following modules:

[0127] The equipment fault diagnosis model creation module is used to create an equipment fault diagnosis model for outputting fault diagnosis reports based on a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and to set the loss function of the equipment fault diagnosis model.

[0128] The dataset construction module is used to acquire a large amount of historical status data of optical access devices, and to construct a dataset after preprocessing and labeling the historical status data.

[0129] The equipment fault diagnosis model training module is used to divide the dataset into a training set, a validation set, and a test set, train the equipment fault diagnosis model using the training set and a loss function, and validate and test the trained equipment fault diagnosis model using the validation set and the test set.

[0130] The equipment fault diagnosis model deployment module is used to compress the tested equipment fault diagnosis model and deploy it to the gateway.

[0131] The fault diagnosis module is used by the gateway to collect real-time status data of the optical access device, input the real-time status data into the deployed device fault diagnosis model, and obtain a fault diagnosis report.

[0132] The fault repair module is used by the gateway to send a fault repair instruction to the optical access device based on the fault diagnosis report and to obtain the fault repair result fed back by the optical access device.

[0133] The report encryption upload module is used by the gateway to encrypt the fault diagnosis report and fault repair results into an encrypted data packet, and then upload the encrypted data packet to the cloud platform;

[0134] The device optimization strategy distribution module is used by the cloud platform to decrypt the received encrypted data packets to obtain fault diagnosis reports and fault repair results. The cloud platform aggregates and analyzes the fault diagnosis reports and fault repair results uploaded by each gateway to generate device optimization strategies, and distributes the device optimization strategies to the optical access devices through the gateways.

[0135] In the equipment fault diagnosis model creation module, the multi-source feature extraction layer is constructed based on the equipment operation channel, optical module channel, optical fiber link channel, user port channel, network management channel, and timing processing channel;

[0136] The device operation channel is used to extract device operation status features from device operation status data through a dual-layer LSTM unit; the optical module channel is used to extract optical module status features from optical module status data through a CNN-Transformer hybrid unit; the fiber optic link channel is used to extract fiber optic link status features from fiber optic link status data through a graph convolutional network unit; the user port channel is used to extract user port status features from user port status data through a conditional random field unit; and the network management channel is used to extract network management status features from network management status data through a BERT fine-tuning unit.

[0137] By constructing multiple channels (device operation, optical module, fiber optic link, user port, network management, and timing processing channels), multi-dimensional status data (such as timing data, text data, and graph structure data) are integrated. This multi-source feature extraction avoids the limitations of relying on a single data source and can cover all aspects of device operation (such as the physical layer, network layer, user layer, and management layer), thereby significantly improving the accuracy and robustness of fault diagnosis and reducing false alarm and false negative rates.

[0138] The dual-layer LSTM unit in the equipment operation channel is specifically designed for processing time-series data, capturing long-term dependencies (such as continuous changes in equipment operating status) and improving the ability to identify periodic faults. The CNN-Transformer hybrid unit in the optical module channel combines the local feature extraction of convolutional neural networks (CNNs) with the global context understanding of Transformers, efficiently processing image or sequence data (such as optical module performance parameters) and reducing computational redundancy. The graph convolutional network (GCN) unit in the fiber optic link channel adapts to the graph structure data of fiber optic links (such as network topology), effectively building... The dependency relationships between module nodes enhance the sensitivity to link failures (such as interruptions or interference); the Conditional Random Field (CRF) unit in the user port channel is suitable for sequence labeling tasks (such as port state changes), and enhances the detection accuracy of user-end anomalies (such as congestion or attacks) by modeling the dependencies between states; the BERT fine-tuning unit in the network management channel uses a pre-trained language model to process text data (such as network logs), and quickly understands semantics after fine-tuning, accelerating the analysis of management-level faults; this customized extraction strategy not only improves data processing efficiency (such as GPU resource optimization), but also reduces model complexity.

[0139] The cross-modal fusion layer is used to fuse the device operating status features, optical module status features, optical fiber link status features, user port status features, and network management status features through the feature cross-module and dependency modeling module to obtain fused features;

[0140] The feature cross-addressing module is built based on multi-head cross-attention (8 heads), taking 128-dimensional features from 5 channels as input, and is used to calculate cross-source feature weights such as optical module-fiber and port-network management. The dependency modeling module is built based on hierarchical graph neural network (HGNN), used to establish a three-order relationship graph of device-port-fiber. The message passing layer performs 3 rounds of node updates, and the graph readout layer aggregates global features and outputs a 256-dimensional fused feature vector.

[0141] The cross-modal fusion layer's feature cross-module enables interaction between features (such as associating device status with fiber optic links), discovers hidden patterns (such as link problems caused by device anomalies), and improves the diagnosis of complex faults (such as fault chains caused by multiple factors). The dependency modeling module explicitly models the dependencies between different data sources (such as the impact of network management data on user ports), ensuring that the fused features have high-dimensional representation and reducing redundancy. This fusion mechanism is superior to simple feature splicing, providing stronger generalization capabilities, especially in dynamic network environments, and can adapt to new fault types.

[0142] The multi-task output layer is used to output a fault diagnosis report, including fault identifier, fault type, fault code, and fault handling suggestions, based on the fusion features.

[0143] The multi-task output layer is constructed based on a fault identifier head, a joint fault decoding module, and a processing suggestion generator. The fault identifier head has the following structure: fully connected (256→64) + Dropout (0.3) → Sigmoid output, used for binary classification to determine the existence of faults. The joint fault decoding module has the following structure: a) Feature routing layer: dynamically reweighted based on the fault identifier output; b) Fault type branch: fully connected (64→32) → Softmax (K classes); c) Fault code branch: Graph Attention Decoder (GAT), generating M-class code distributions. The processing suggestion generator uses a Conditional Generative Adversarial Network (CGAN) generator, with the input being a concatenated vector of fault type + code, and the output being an N-dimensional suggestion vector (Sigmoid multi-label).

[0144] By constructing a multi-source feature extraction layer (covering six channels: equipment operation, optical modules, fiber optic links, user ports, network management, and time-series processing), and employing customized, high-efficiency feature extraction units (such as LSTM for time-series data processing, graph convolution for network topology adaptation, and BERT for parsing text logs) tailored to the data type characteristics of each channel, deep feature mining of multi-dimensional state data is achieved. Its innovative cross-modal fusion layer effectively integrates the correlation between heterogeneous features through feature cross-referencing and dependency modeling, significantly improving the identification accuracy and robustness of complex faults. Finally, the multi-task output layer generates an end-to-end diagnostic report containing fault identifiers, types, codes, and processing suggestions, ensuring system scalability while comprehensively optimizing the efficiency, accuracy, and practicality of fault diagnosis for optical access equipment.

[0145] The formula for the loss function is:

[0146] L=λ1*L_id+λ2*L_type+λ3*L_code+λ4*L_adv;

[0147] Where L represents the loss value of the loss function; L_id represents the fault identification loss, which uses Focal Loss (γ=2, α=0.8) to alleviate sample imbalance; L_type represents the type loss, which uses label smoothing cross-entropy (smoothing factor=0.1); L_code represents the code loss, which uses constrained KL divergence (constraining the fault type-code logical relationship); and L_adv represents the suggestion loss, which uses generative adversarial loss (Wasserstein distance) + reconstruction loss.

[0148] The dataset construction module is specifically used for:

[0149] Acquire a large amount of historical status data for optical access devices. This historical status data includes device operating status data, optical module status data, fiber optic link status data, user port status data, and network management status data. The device operating status data includes device online status (indicating whether the device is properly connected to the network. For example, an optical access device is "online" if it is powered on and has a valid network connection; if the device loses power or the fiber optic link is interrupted, its status is "offline") and device temperature (this is an important indicator of the device's internal heat dissipation. Optical access devices typically contain various electronic components; excessively high temperatures can lead to performance degradation or even damage. For example, the normal operating temperature range of the device may be 0℃-50℃). If the temperature exceeds 50℃, the device may issue a high-temperature alarm. The device power status includes information such as input voltage and current. For example, the normal input voltage range is -48V ± 5%. If the voltage exceeds this range, it may cause abnormal device operation, and the power status data can reflect this situation in a timely manner. The optical module status data includes the optical module type (different optical access devices may support multiple types of optical modules, such as SFP (Small Form Factor), SFP+, XFP, etc. Status data can identify the type of the currently inserted optical module to confirm whether it meets the device requirements), and the optical module transmit power (this is an indicator of the strength of the optical signal transmitted by the optical module. For example, for a 1000BASE-S...). X-ray modules typically have several power ranges: transmit power (ranging from -10dBm to -3dBm; transmit power below -10dBm can cause excessive signal attenuation during transmission, affecting communication quality); receive power (representing the strength of the received optical signal; for example, receive power ranges from -10dBm to -20dBm; receive power below -20dBm may result in signal loss or increased bit error rate); and temperature (optical modules also contain temperature-sensitive components; excessively high or low temperatures can affect performance; for example, the normal operating temperature range for optical modules is -5℃ to 70℃; temperatures exceeding this range may lead to performance degradation or even damage). The fiber optic link status data includes fiber optic connection status (indicating whether the fiber optic cable is correctly connected. For example, if the fiber optic plug is not properly inserted or the fiber is damaged, the fiber optic connection status may show "not connected" or "link fault"), fiber optic link loss (this is an important indicator of fiber optic link transmission efficiency. Fiber optic link loss is mainly caused by factors such as the attenuation of the fiber itself, connector loss, and fusion splice loss. For example, for single-mode fiber, the attenuation is approximately 0.2dB-0.5dB per kilometer. If the fiber optic link loss is too high, it may cause the optical signal to fail to transmit normally), and fiber optic link length (the length of the fiber optic link can be measured using devices such as an optical time domain reflectometer (OTDR). This is very important for the maintenance and fault location of fiber optic networks).For example, in an optical access network, the length of the fiber optic link may vary from several kilometers to tens of kilometers. If the fiber optic link length does not match the designed length, it may be because the fiber has been accidentally cut or incorrectly connected. The user port status data includes port rate (representing the transmission rate of the user port. For example, common rates are 100Mbps, 1Gbps, 10Gbps, etc. If the port rate is set incorrectly, it may cause users to be unable to communicate normally), port connection status (indicating whether the user port is connected to a terminal device. For example, if the user port is connected to a computer, the port connection status may show "connected"; if no device is connected, the status is "not connected"), port traffic (including uplink and downlink traffic. For example, in a home broadband access scenario, the user's downlink traffic may reach tens to hundreds of Mbps, while the uplink traffic is relatively small. By monitoring port traffic, we can understand the user's network usage and also detect abnormal traffic in a timely manner, such as network attacks or malicious downloads), and port bit error rate (representing the probability of errors occurring during data transmission at the user port. For example, the bit error rate is generally required to be below 10^-9. If the bit error rate is low...). Excessive temperatures may lead to data transmission errors and affect the user's network experience. The network management status data includes device configuration information (including network parameters such as the device's IP address, subnet mask, and gateway. For example, the device's IP address might be 192.168.1.1, the subnet mask 255.255.255.0, and the gateway 192.168.1.254. If the configuration information is incorrect, the device may not be able to access the network normally), device alarm information (alarm information is generated when the device malfunctions. For example, the device may generate high temperature alarms, optical module failure alarms, and fiber optic link interruption alarms. By reviewing alarm information, maintenance personnel can quickly locate problems and take measures), and device log information (recording various events during the device's operation, including user logins, configuration modifications, and fault occurrences. For example, the device log may record that a user logged into the device at a certain time and made configuration modifications, or that the device detected a fiber optic link interruption at a certain time. By analyzing the device log, the device's operating history can be understood, facilitating fault diagnosis and security auditing).

[0150] The historical state data is preprocessed, including at least deduplication, missing value handling, error correction and data standardization. The preprocessed historical state data is then labeled with at least fault identifiers, fault types, fault codes and fault handling suggestions. A dataset is constructed based on the labeled historical state data.

[0151] The equipment fault diagnosis model training module is specifically used for:

[0152] Based on stratified sampling, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The equipment fault diagnosis model is trained using the training set and a loss function. During the training process, the hyperparameters of the equipment fault diagnosis model are continuously optimized until the loss value of the loss function is less than a preset loss threshold.

[0153] The diagnostic accuracy is calculated using the validation set to validate the trained equipment fault diagnosis model. If the validation fails, the training set is expanded and training continues; if the validation passes, then:

[0154] The F1 score is calculated using the test set to test the validated equipment fault diagnosis model. If the test fails, the training set is expanded and training continues; if the test passes, training ends.

[0155] The equipment fault diagnosis model deployment module is specifically used for:

[0156] The tested equipment fault diagnosis model is compressed using knowledge distillation technology, the performance of the compressed equipment fault diagnosis model is verified, and the equipment fault diagnosis model that has passed the performance verification is deployed to the gateway.

[0157] The fault diagnosis module is specifically used for:

[0158] The gateway collects real-time status data of optical access devices through GPON or XGPON, preprocesses the real-time status data, and inputs it into the deployed device fault diagnosis model to obtain a fault diagnosis report including fault identifier, fault type, fault code, and fault handling suggestions.

[0159] The fault repair module is specifically used for:

[0160] Based on the fault handling suggestions carried in the fault diagnosis report, the gateway sends a fault repair command to the optical access device in real time and obtains the fault repair result from the optical access device, indicating whether the repair was successful or failed.

[0161] The encrypted report upload module is specifically used for:

[0162] The gateway obtains the current timestamp T, concatenates the fault diagnosis report and fault repair results into initial data D, calculates the hash value H of the initial data D using the hash256 algorithm, and uses the first 128 bits of the hash value H as key K1 and the last 128 bits as key K2.

[0163] The initial data D is divided into data block D1 and data block D2. Data block D1 is encrypted using the AES algorithm with key K1 to obtain encrypted data block E1. Data block D2 is encrypted using the RSA algorithm with key K2 to obtain encrypted data block E2. Encrypted data block E1 and encrypted data block E2 are concatenated to obtain concatenated data P.

[0164] A random string is generated based on the timestamp T. The random string is added to the beginning and end of the concatenated data P to obtain obfuscated data C. The obfuscated data C is then encrypted using the Blowfish algorithm to obtain obfuscated ciphertext data E3.

[0165] Calculate the signature data Sig of the timestamp T, and encrypt the obfuscated ciphertext data E3, hash value H, timestamp T and signature data Sig into an encrypted data packet using the AES algorithm, and upload the encrypted data packet to the cloud platform.

[0166] By cleverly integrating multi-layered encryption, data obfuscation, and integrity verification mechanisms, the security and reliability of data transmission are significantly improved. The core advantages are: efficient and self-consistent key management through hash-derived dynamic keys; differentiated processing of segmented data blocks using a hybrid AES and RSA encryption strategy to balance security and efficiency; head and tail obfuscation through timestamp-generated random strings; and a Blowfish encryption corruption mode to significantly enhance resistance to reverse engineering and eavesdropping. Simultaneously, the integration of hash verification, timestamp signatures, and other multiple protections ensures data integrity and freshness while providing non-repudiable audit trails, ultimately constructing a robust and secure transmission system that effectively resists tampering, replay attacks, and single-point-of-failure risks.

[0167] By deploying a compressed and optimized multimodal fault diagnosis model at the edge of the gateway node, millisecond-level real-time fault diagnosis and automated repair of optical access equipment are achieved, significantly improving response speed and equipment availability. Simultaneously, multi-source feature extraction and cross-modal fusion technologies, combined with high-precision analysis of multi-dimensional data such as equipment operation, optical links, and user ports, greatly enhance diagnostic accuracy. In terms of data closed-loop, the cloud aggregates and analyzes encrypted repair results and generates optimization strategies that are fed back to the equipment, forming a secure collaborative mechanism of "edge real-time processing - cloud intelligent optimization," effectively reducing maintenance costs and extending equipment lifespan. Its modular design is compatible with mainstream protocols such as GPON / XGPON, and combined with knowledge distillation model compression and hybrid encryption technologies, it ensures data security while also considering efficient deployment and sustainable iteration capabilities in resource-constrained scenarios, comprehensively improving the management intelligence and operational economy of the optical access network.

[0168] In summary, the advantages of this invention are as follows:

[0169] 1. A device fault diagnosis model is created through a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and a loss function is set for the model. Then, a dataset is constructed by acquiring a large amount of historical status data from optical access devices. This dataset is divided into training, validation, and test sets. The device fault diagnosis model is trained using the training set and the loss function, and validated and tested using the validation and test sets. The tested model is compressed and deployed to the gateway. The gateway collects real-time status data from the optical access devices and inputs it into the fault diagnosis model to obtain a fault diagnosis report. Based on the report, the gateway sends a fault repair command to the optical access devices, obtains the repair results, and encrypts the report and repair results into an encrypted data packet, which is then uploaded to the cloud platform. The cloud platform decrypts the encrypted data packet to retrieve the report and repair results. The cloud platform aggregates and analyzes fault diagnosis reports and fault repair results uploaded by each gateway, generates device optimization strategies, and distributes these strategies to optical access devices through the gateways. In other words, the device fault diagnosis model is compressed and deployed to the gateway, enabling the gateway to directly collect and process real-time status data of the optical access devices locally, generate fault diagnosis reports, and immediately issue fault repair instructions. This achieves localized closed-loop processing, avoiding the data transmission delays of traditional cloud processing, thus compressing response time to the millisecond level and significantly improving timeliness. Simultaneously, the gateway only encrypts lightweight fault diagnosis reports and repair results into encrypted data packets before uploading, rather than the original status data, reducing the exposure of sensitive information. End-to-end encryption mechanisms prevent man-in-the-middle attacks during transmission, and local processing avoids remote transmission of raw data, fundamentally reducing privacy leaks and security risks, ultimately greatly improving the timeliness and security of optical access device management.

[0170] 2. By deploying a compressed and optimized multimodal fault diagnosis model at the edge of the gateway node, millisecond-level real-time fault diagnosis and automated repair of optical access equipment are achieved, significantly improving response speed and equipment availability. Simultaneously, by utilizing multi-source feature extraction and cross-modal fusion technologies, combined with high-precision analysis of multi-dimensional data such as equipment operation, optical links, and user ports, diagnostic accuracy is greatly enhanced. In terms of data closed-loop, relying on the cloud to aggregate and analyze encrypted uploaded repair results and generate optimization strategies to feed back to the equipment, a secure collaborative mechanism of "edge real-time processing - cloud intelligent optimization" is formed, effectively reducing operation and maintenance costs and extending equipment lifespan. Its modular design is compatible with mainstream protocols such as GPON / XGPON, and combined with knowledge distillation model compression and hybrid encryption technologies, it ensures data security while also considering efficient deployment and sustainable iteration capabilities in resource-constrained scenarios, comprehensively improving the management intelligence level and operational economy of the optical access network.

[0171] 3. By constructing multiple channels (device operation, optical module, fiber optic link, user port, network management, and timing processing channels), multi-dimensional status data (such as timing data, text data, graph structure data, etc.) are integrated. This multi-source feature extraction avoids the limitations of relying on a single data source and can cover all aspects of device operation (such as physical layer, network layer, user layer, and management layer), thereby significantly improving the accuracy and robustness of fault diagnosis and reducing false alarm rate and false negative rate.

[0172] 4. The dual-layer LSTM unit in the equipment operation channel: Specifically designed for processing time-series data, it captures long-term dependencies (such as continuous changes in equipment operating status), improving the ability to identify periodic faults; the CNN-Transformer hybrid unit in the optical module channel: Combining the local feature extraction of convolutional neural networks (CNN) and the global context understanding of Transformers, it efficiently processes image or sequence data (such as optical module performance parameters), reducing computational redundancy; the graph convolutional network (GCN) unit in the fiber optic link channel: Adapted to the graph structure data of fiber optic links (such as network topology), it effectively... Modeling dependencies between nodes improves sensitivity to link failures (such as interruptions or interference); the Conditional Random Field (CRF) unit in the user port channel is suitable for sequence labeling tasks (such as port state changes), enhancing the detection accuracy of user-end anomalies (such as congestion or attacks) by modeling dependencies between states; the BERT fine-tuning unit in the network management channel uses a pre-trained language model to process text data (such as network logs), and quickly understands semantics after fine-tuning, accelerating the analysis of management-level faults; this customized extraction strategy not only improves data processing efficiency (such as GPU resource optimization) but also reduces model complexity.

[0173] 5. The cross-modal fusion layer's feature cross-module enables interaction between features (such as associating device status with fiber optic links), discovers hidden patterns (such as link problems caused by device anomalies), and improves the diagnosis of complex faults (such as fault chains caused by multiple factors); the dependency modeling module explicitly models the dependencies between different data sources (such as the impact of network management data on user ports), ensuring that the fused features have high-dimensional representation and reducing redundancy; this fusion mechanism is superior to simple feature splicing, providing stronger generalization ability, especially in variable network environments, and can adapt to new fault types.

[0174] 6. By constructing a multi-source feature extraction layer (covering six channels: equipment operation, optical modules, fiber optic links, user ports, network management, and time-series processing), and employing customized, high-efficiency feature extraction units (such as LSTM for time-series data processing, graph convolution for network topology adaptation, and BERT for parsing text logs) tailored to the data type characteristics of each channel, deep feature mining of multi-dimensional state data is achieved. Its innovative cross-modal fusion layer effectively integrates the correlation between heterogeneous features through feature cross-referencing and dependency modeling, significantly improving the identification accuracy and robustness of complex faults. Finally, the multi-task output layer generates an end-to-end diagnostic report containing fault identifiers, types, codes, and processing suggestions, ensuring system scalability while comprehensively optimizing the efficiency, accuracy, and practicality of fault diagnosis for optical access equipment.

[0175] 7. By cleverly integrating multi-layered encryption, data obfuscation, and integrity verification mechanisms, the security and reliability of data transmission are significantly improved. The core advantages are: using hash values ​​to derive dynamic keys to achieve efficient and self-consistent key management; combining AES and RSA hybrid encryption strategies to differentiate the processing of segmented data blocks to balance security and efficiency; generating random strings through timestamps for header and tail obfuscation; and employing Blowfish encryption to destroy data, greatly enhancing the ability to resist reverse analysis and eavesdropping; while integrating hash verification, timestamp signatures, and other multiple protections, it not only ensures the integrity and freshness of data but also provides non-repudiable audit trails, ultimately constructing a robust and secure transmission system that can effectively resist tampering, replay attacks, and single point of failure risks.

[0176] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for managing optical access devices based on gateways and edge computing, characterized in that: Includes the following steps: Step S1: Create a device fault diagnosis model for outputting fault diagnosis reports based on a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and set the loss function of the device fault diagnosis model; Step S2: Obtain a large amount of historical status data of optical access devices, preprocess and label each historical status data to construct a dataset; Step S3: Divide the dataset into a training set, a validation set, and a test set. Train the equipment fault diagnosis model using the training set and the loss function. Validate and test the trained equipment fault diagnosis model using the validation set and the test set. Step S4: After compressing the tested device fault diagnosis model, deploy it to the gateway; Step S5: The gateway collects real-time status data of the optical access device, inputs the real-time status data into the deployed device fault diagnosis model, and obtains a fault diagnosis report. Step S6: The gateway sends a fault repair instruction to the optical access device based on the fault diagnosis report and obtains the fault repair result fed back by the optical access device. Step S7: The gateway encrypts the fault diagnosis report and fault repair results into an encrypted data packet and uploads the encrypted data packet to the cloud platform; Step S8: The cloud platform decrypts the received encrypted data packet to obtain the fault diagnosis report and fault repair results. The cloud platform aggregates and analyzes the fault diagnosis reports and fault repair results uploaded by each gateway to generate a device optimization strategy. The device optimization strategy is then distributed to the optical access device through the gateway.

2. The optical access device management method based on gateway and edge computing as described in claim 1, characterized in that: In step S1, the multi-source feature extraction layer is constructed based on the device operation channel, optical module channel, optical fiber link channel, user port channel, network management channel, and timing processing channel. The device operation channel is used to extract device operation status features from device operation status data through a dual-layer LSTM unit; The optical module channel is used to extract optical module state features from optical module state data through a CNN-Transformer hybrid unit; the optical fiber link channel is used to extract optical fiber link state features from optical fiber link state data through a graph convolutional network unit; the user port channel is used to extract user port state features from user port state data through a conditional random field unit; and the network management channel is used to extract network management state features from network management state data through a BERT fine-tuning unit. The cross-modal fusion layer is used to fuse the device operating status features, optical module status features, optical fiber link status features, user port status features, and network management status features through the feature cross-module and dependency modeling module to obtain fused features; The multi-task output layer is used to output a fault diagnosis report, including fault identifier, fault type, fault code, and fault handling suggestions, based on the fusion features.

3. The optical access device management method based on gateway and edge computing as described in claim 1, characterized in that: Step S2 specifically involves: Acquire a large amount of historical status data for optical access devices. This historical status data includes device operating status data, optical module status data, fiber optic link status data, user port status data, and network management status data. The device operating status data includes device online status, device temperature, and device power status. The optical module status data includes optical module type, optical module transmit power, optical module receive power, and optical module temperature. The fiber optic link status data includes fiber connection status, fiber link loss, and fiber link length. The user port status data includes port rate, port connection status, port traffic, and port bit error rate. The network management status data includes device configuration information, device alarm information, and device log information. The historical state data is preprocessed, including at least deduplication, missing value handling, error correction and data standardization. The preprocessed historical state data is then labeled with at least fault identifiers, fault types, fault codes and fault handling suggestions. A dataset is constructed based on the labeled historical state data.

4. The optical access device management method based on gateway and edge computing as described in claim 1, characterized in that: Step S3 specifically involves: Based on stratified sampling, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:

1. The equipment fault diagnosis model is trained using the training set and a loss function. During the training process, the hyperparameters of the equipment fault diagnosis model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The diagnostic accuracy is calculated using the validation set to validate the trained equipment fault diagnosis model. If the validation fails, the training set is expanded and training continues; if the validation passes, then: The F1 score is calculated using the test set to test the validated equipment fault diagnosis model. If the test fails, the training set is expanded and training continues; if the test passes, training ends.

5. The optical access device management method based on gateway and edge computing as described in claim 1, characterized in that: Step S4 specifically involves: The tested equipment fault diagnosis model is compressed using knowledge distillation technology, the performance of the compressed equipment fault diagnosis model is verified, and the equipment fault diagnosis model that has passed the performance verification is deployed to the gateway. Step S5 specifically involves: The gateway collects real-time status data of optical access devices through GPON or XGPON, preprocesses the real-time status data, and inputs it into the deployed device fault diagnosis model to obtain a fault diagnosis report including fault identifier, fault type, fault code, and fault handling suggestions. Step S6 specifically involves: Based on the fault handling suggestions carried in the fault diagnosis report, the gateway sends a fault repair command to the optical access device in real time and obtains the fault repair result from the optical access device, indicating whether the repair was successful or failed. Step S7 specifically involves: The gateway obtains the current timestamp T, concatenates the fault diagnosis report and fault repair results into initial data D, calculates the hash value H of the initial data D using the hash256 algorithm, and uses the first 128 bits of the hash value H as key K1 and the last 128 bits as key K2. The initial data D is divided into data block D1 and data block D2. Data block D1 is encrypted using the AES algorithm with key K1 to obtain encrypted data block E1. Data block D2 is encrypted using the RSA algorithm with key K2 to obtain encrypted data block E2. Encrypted data block E1 and encrypted data block E2 are concatenated to obtain concatenated data P. A random string is generated based on the timestamp T. The random string is added to the beginning and end of the concatenated data P to obtain obfuscated data C. The obfuscated data C is then encrypted using the Blowfish algorithm to obtain obfuscated ciphertext data E3. Calculate the signature data Sig of the timestamp T, and encrypt the obfuscated ciphertext data E3, hash value H, timestamp T and signature data Sig into an encrypted data packet using the AES algorithm, and upload the encrypted data packet to the cloud platform.

6. A management system for optical access devices based on gateways and edge computing, characterized in that: Includes the following modules: The equipment fault diagnosis model creation module is used to create an equipment fault diagnosis model for outputting fault diagnosis reports based on a multi-source feature extraction layer, a cross-modal fusion layer, and a multi-task output layer, and to set the loss function of the equipment fault diagnosis model. The dataset construction module is used to acquire a large amount of historical status data of optical access devices, and to construct a dataset after preprocessing and labeling the historical status data. The equipment fault diagnosis model training module is used to divide the dataset into a training set, a validation set, and a test set, train the equipment fault diagnosis model using the training set and a loss function, and validate and test the trained equipment fault diagnosis model using the validation set and the test set. The equipment fault diagnosis model deployment module is used to compress the tested equipment fault diagnosis model and deploy it to the gateway. The fault diagnosis module is used by the gateway to collect real-time status data of the optical access device, input the real-time status data into the deployed device fault diagnosis model, and obtain a fault diagnosis report. The fault repair module is used by the gateway to send a fault repair instruction to the optical access device based on the fault diagnosis report and to obtain the fault repair result fed back by the optical access device. The report encryption upload module is used by the gateway to encrypt the fault diagnosis report and fault repair results into an encrypted data packet and upload the encrypted data packet to the cloud platform; The device optimization strategy distribution module is used by the cloud platform to decrypt the received encrypted data packets to obtain fault diagnosis reports and fault repair results. The cloud platform aggregates and analyzes the fault diagnosis reports and fault repair results uploaded by each gateway to generate device optimization strategies, and distributes the device optimization strategies to the optical access devices through the gateways.

7. The optical access device management system based on gateway and edge computing as described in claim 6, characterized in that: In the equipment fault diagnosis model creation module, the multi-source feature extraction layer is constructed based on the equipment operation channel, optical module channel, optical fiber link channel, user port channel, network management channel, and timing processing channel; The device operation channel is used to extract device operation status features from device operation status data through a dual-layer LSTM unit; The optical module channel is used to extract optical module state features from optical module state data through a CNN-Transformer hybrid unit; the optical fiber link channel is used to extract optical fiber link state features from optical fiber link state data through a graph convolutional network unit; the user port channel is used to extract user port state features from user port state data through a conditional random field unit; and the network management channel is used to extract network management state features from network management state data through a BERT fine-tuning unit. The cross-modal fusion layer is used to fuse the device operating status features, optical module status features, optical fiber link status features, user port status features, and network management status features through the feature cross-module and dependency modeling module to obtain fused features; The multi-task output layer is used to output a fault diagnosis report, including fault identifier, fault type, fault code, and fault handling suggestions, based on the fusion features.

8. The optical access device management system based on gateway and edge computing as described in claim 6, characterized in that: The dataset construction module is specifically used for: Acquire a large amount of historical status data for optical access devices. This historical status data includes device operating status data, optical module status data, fiber optic link status data, user port status data, and network management status data. The device operating status data includes device online status, device temperature, and device power status. The optical module status data includes optical module type, optical module transmit power, optical module receive power, and optical module temperature. The fiber optic link status data includes fiber connection status, fiber link loss, and fiber link length. The user port status data includes port rate, port connection status, port traffic, and port bit error rate. The network management status data includes device configuration information, device alarm information, and device log information. The historical state data is preprocessed, including at least deduplication, missing value handling, error correction and data standardization. The preprocessed historical state data is then labeled with at least fault identifiers, fault types, fault codes and fault handling suggestions. A dataset is constructed based on the labeled historical state data.

9. The optical access device management system based on gateway and edge computing as described in claim 6, characterized in that: The equipment fault diagnosis model training module is specifically used for: Based on stratified sampling, the dataset is divided into a training set, a validation set, and a test set in a ratio of 8:1:

1. The equipment fault diagnosis model is trained using the training set and a loss function. During the training process, the hyperparameters of the equipment fault diagnosis model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The diagnostic accuracy is calculated using the validation set to validate the trained equipment fault diagnosis model. If the validation fails, the training set is expanded and training continues; if the validation passes, then: The F1 score is calculated using the test set to test the validated equipment fault diagnosis model. If the test fails, the training set is expanded and training continues; if the test passes, training ends.

10. The optical access device management system based on gateway and edge computing as described in claim 6, characterized in that: The equipment fault diagnosis model deployment module is specifically used for: The tested equipment fault diagnosis model is compressed using knowledge distillation technology, the performance of the compressed equipment fault diagnosis model is verified, and the equipment fault diagnosis model that has passed the performance verification is deployed to the gateway. The fault diagnosis module is specifically used for: The gateway collects real-time status data of optical access devices through GPON or XGPON, preprocesses the real-time status data, and inputs it into the deployed device fault diagnosis model to obtain a fault diagnosis report including fault identifier, fault type, fault code, and fault handling suggestions. The fault repair module is specifically used for: Based on the fault handling suggestions carried in the fault diagnosis report, the gateway sends a fault repair command to the optical access device in real time and obtains the fault repair result from the optical access device, indicating whether the repair was successful or failed. The encrypted report upload module is specifically used for: The gateway obtains the current timestamp T, concatenates the fault diagnosis report and fault repair results into initial data D, calculates the hash value H of the initial data D using the hash256 algorithm, and uses the first 128 bits of the hash value H as key K1 and the last 128 bits as key K2. The initial data D is divided into data block D1 and data block D2. Data block D1 is encrypted using the AES algorithm with key K1 to obtain encrypted data block E1. Data block D2 is encrypted using the RSA algorithm with key K2 to obtain encrypted data block E2. Encrypted data block E1 and encrypted data block E2 are concatenated to obtain concatenated data P. A random string is generated based on the timestamp T. The random string is added to the beginning and end of the concatenated data P to obtain obfuscated data C. The obfuscated data C is then encrypted using the Blowfish algorithm to obtain obfuscated ciphertext data E3. Calculate the signature data Sig of the timestamp T, and encrypt the obfuscated ciphertext data E3, hash value H, timestamp T and signature data Sig into an encrypted data packet using the AES algorithm, and upload the encrypted data packet to the cloud platform.

Images