Intelligent inspection method for communication transmission network equipment and related equipment

Abstract

The invention discloses an intelligent routing inspection method for communication transmission network equipment and related equipment, routing inspection operation is decoupled into a quintuple structure atomic power unit, a library is constructed, the routing inspection operation is endowed with high reusability and flexibility, different requirements can be quickly met, and customization and expansibility are improved. And an atomic energy combination mode is retrieved by means of an equipment knowledge base, so that accurate matching of inspection targets and capacities is realized, and support is provided for task planning. By means of a three-dimensional protocol capability matrix, historical data and equipment protocol knowledge are linked, an adaptive protocol is accurately screened, a task time sequence is planned, and inspection task flow execution efficiency is optimized. The inspection dimension is enriched by collecting multi-modal data, the corresponding analysis model is input after temporal space alignment, a structured result containing abnormal information, confidence and evidence is generated, and the analysis accuracy is improved. And finally, through multi-source evidence fusion, logic association and cross verification, the reliability of anomaly judgment is enhanced, and high efficiency, precision and intelligence of routing inspection diagnosis are realized.

Description

Technical Field

[0001] This application relates to the field of intelligent inspection technology, and more specifically, to an intelligent inspection method and related equipment for communication transmission network devices. Background Technology

[0002] With the rapid development of emerging technologies such as 5G, IoT, and cloud computing, the scale and complexity of communication transmission networks are growing explosively. The types and numbers of network devices are becoming increasingly diverse, leading to a sharp increase in operation and maintenance management pressure. This makes the efficiency and accuracy of inspection work increasingly crucial. Currently, existing inspection technologies are mainly divided into two categories. Traditional manual inspection solutions remain the mainstream choice for some scenarios. This method requires tedious steps such as developing inspection plans, preparing checklists, logging into each device on-site to verify its operating status and configuration interfaces, manually recording data and comparing it with standards, compiling problem reports, and subsequent rectification and verification. Its shortcomings are extremely significant and obvious, and it is completely unsuitable for the operation and maintenance needs of complex networks.

[0003] Another type of intelligent inspection solution that has emerged in recent years attempts to replace manual labor with technology. However, these solutions are mostly based on simple automated logic, with task generation relying on fixed templates. This results in rigid strategies, making it difficult to cope with the dynamic expansion of the network and the needs of equipment iteration, and their flexibility is extremely poor. Their analysis modes are singular and fixed, only able to complete basic data collection and simple comparison. The original results are not verified a second time, the depth of anomaly feature mining is insufficient, and the accuracy and reliability of professional analysis are limited. Protocol adaptation only meets basic connectivity requirements and lacks specificity. The data collection scope is narrow, making it difficult to achieve comprehensive and accurate inspection.

[0004] Therefore, there is an urgent need for a new intelligent inspection method for communication transmission network equipment to overcome the shortcomings of existing technologies and achieve a more efficient, accurate, and intelligent upgrade of communication transmission network inspection. Summary of the Invention

[0005] This application provides an intelligent inspection method and related equipment for communication transmission network devices, which realizes efficient, accurate and intelligent inspection and diagnosis, and significantly improves the technical level of network inspection and maintenance support capabilities.

[0006] A method for intelligent inspection of communication transmission network equipment, comprising:

[0007] The inspection operations of each network device are decoupled into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0008] In response to a user-inputted inspection request, the inspection request is parsed to determine the inspection target, and the associated atomic capability combination pattern is retrieved from a pre-built device knowledge base based on the inspection target.

[0009] In the three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, an adaptive protocol is selected from multiple candidate protocols for each atomic capability in the atomic capability combination mode based on the atomic capability combination mode and the target network device type, and task planning is performed to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0010] The inspection task flow is executed to collect multimodal data of the target network device;

[0011] After aligning the multimodal data in time and space according to the data type, the data is input into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0012] The structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation and cross-verification to generate inspection and diagnosis results.

[0013] Optionally, the process of generating the inspection task flow includes:

[0014] Based on the three-dimensional protocol capability matrix, obtain the multi-dimensional capability vector corresponding to the target network device type and each candidate protocol. The multi-dimensional capability vector includes quantitative indicators of data collection granularity, indicator coverage depth, historical reliability, and security level.

[0015] The weights of each capability dimension are determined based on the characteristics of atomic capabilities and the requirements of the inspection task.

[0016] Based on the multidimensional capability vector and the weights of each capability dimension, the weighted comprehensive score of each candidate protocol is calculated.

[0017] The primary execution protocol is selected based on the weighted composite score, and the atomic capabilities in the atomic capability combination mode are converted into operation commands supported by the primary execution protocol.

[0018] Based on equipment resource constraints and task priorities, all operation commands after atomic capability conversion are concurrently scheduled and timed to form the inspection task flow.

[0019] Optionally, the weighted composite score of each candidate protocol is calculated using the following formula:

[0020]

[0021] in, For the first One candidate protocol; For the first One target device type; For representing devices in multidimensional capability vectors Agreement In the Quantitative indicator values ​​for each capability dimension; For the corresponding to the first The weights of each capability dimension.

[0022] Optional, also includes:

[0023] The candidate protocol with the highest weighted composite score is selected as the main execution protocol.

[0024] If the difference between the highest and second-highest scores in the weighted composite scores of the candidate protocols is less than a preset redundancy threshold, then the main execution protocol and the protocol with the second-highest score are activated simultaneously to perform parallel redundant collection of key inspection indicators.

[0025] If the historical reliability index of the main execution protocol is lower than the preset reliability threshold, task degradation is triggered, non-critical atomic capability units are removed from the atomic capability combination mode, and a lightweight inspection task flow is generated.

[0026] Optional, also includes:

[0027] After each execution of the inspection task flow, the actual execution performance data of each called protocol is recorded, including response success rate and data collection integrity rate.

[0028] Based on the actual execution performance data, the capability vector values ​​of the corresponding protocols in the three-dimensional protocol capability matrix are dynamically adjusted using an iterative update algorithm.

[0029] Optionally, the inspection request is parsed to determine the inspection target, and based on the inspection target, associated atomic capability combination patterns are retrieved from a pre-built device knowledge base, including:

[0030] The inspection request is parsed to extract the equipment identifier, fault description, and maintenance operation intent contained therein, forming a structured query;

[0031] The structured query is used to retrieve the equipment knowledge base, which stores historical inspection cases and experience rules indexed by equipment type, fault scenario, and solution.

[0032] The retrieved historical training cases and empirical rules that match the inspection target are combined and generalized to infer a set of matching atomic capability units.

[0033] Based on the set of atomic capability units, the constraints of device resource limitations and task execution time windows are solved to determine the atomic capability combination mode.

[0034] Optionally, after aligning the multimodal data in time and space according to data type, the data is input into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, including:

[0035] The multimodal data is spatiotemporally aligned to ensure that data from different sources remain consistent in timestamps and spatiotemporal labels.

[0036] Input the device operation log into the log analysis model to identify abnormal patterns including continuous error codes and frequency mutations, and output the first structured analysis result containing the abnormal pattern code, confidence level and corresponding original log fragment index;

[0037] The device images captured by the camera are input into the visual diagnostic model to identify the color status of the indicator lights and the alarm text on the screen, and output a second structured analysis result containing visual anomaly labels, confidence scores and key frame image indexes.

[0038] The equipment performance index sequence and environmental sensor data are input into the time series analysis model to detect abnormal drift of the data relative to the prediction baseline, and output a third structured analysis result containing the data prediction value, anomaly score and data source identifier.

[0039] Optionally, the structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation, and cross-verification processing to generate inspection and diagnostic results, including:

[0040] The structured analysis results are formatted into standardized evidence objects that include anomaly type, confidence level, spatiotemporal label, and original data index.

[0041] Based on preset fusion rules and a fault knowledge base, the standardized evidence objects are weighted and fused to calculate the weighted comprehensive confidence level of each candidate fault type.

[0042] Based on the weighted comprehensive confidence level, a graded judgment is performed and a cross-verification process is initiated. Arbitration is conducted through logical consistency verification, supplementary data requests, or historical case backtracking to determine the inspection and diagnosis results.

[0043] Optionally, based on the weighted comprehensive confidence level, a graded judgment is performed and a cross-verification process is initiated. Arbitration is conducted through logical consistency verification, supplementary data requests, or historical case backtracking to determine the inspection and diagnosis results, including:

[0044] If the overall confidence level of a certain fault type is higher than the first predetermined threshold, and the difference between its confidence level and the confidence level of the second highest fault type is greater than the second predetermined threshold, then it is directly determined as the current fault type.

[0045] If the highest overall confidence level is lower than the first predetermined threshold, or the difference between the confidence level of the second highest fault type and the second predetermined threshold is less than or equal to the second predetermined threshold, then a secondary verification process is initiated. The secondary verification process includes requesting the data acquisition terminal to re-acquire the specified type of evidence data, and submitting the current contradictory evidence and the preliminary judgment result to the manual review interface.

[0046] If the overall confidence level of all candidate fault types is lower than the third predetermined threshold, the multimodal evidence is determined to be noise or invalid, and the analysis results are discarded.

[0047] Optional, also includes:

[0048] A pre-built device cause-effect graph is invoked, wherein the nodes of the device cause-effect graph represent device hardware components, software configuration items or environmental state factors, and the edges represent the causal relationships between the nodes;

[0049] Taking the node corresponding to the fault phenomenon determined in the inspection and diagnosis results as the starting node, a graph search algorithm is executed in the causal graph of the equipment to trace back along the causal edges.

[0050] The fault propagation path is determined based on the causal strength weights of each edge in the search path and the support of multimodal evidence for the path nodes, and the endpoint of the fault propagation path is determined as the root cause of the fault phenomenon.

[0051] A smart inspection device for communication transmission network equipment includes:

[0052] The atomic capability parsing module is used to decouple the inspection operations of each network device into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0053] The capability combination retrieval module is used to respond to the inspection request input by the user, parse the inspection request to determine the inspection target, and retrieve the associated atomic capability combination pattern from the pre-built device knowledge base based on the inspection target.

[0054] The inspection task planning module is used to select an appropriate protocol from multiple candidate protocols for each atomic capability in a three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, based on the atomic capability combination mode and the target network device type, and to perform task planning to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0055] The modal data acquisition module is used to execute the inspection task flow to acquire multimodal data of the target network device;

[0056] The multi-source anomaly analysis module is used to align the multimodal data in time and space according to the data type, and then input them into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0057] The fusion and correlation diagnosis module is used to sequentially perform multi-source evidence fusion, logical correlation and cross-verification processing on the structured analysis results to generate inspection diagnosis results.

[0058] A smart inspection device for communication transmission network equipment includes a memory and a processor;

[0059] The memory is used to store programs;

[0060] The processor is used to execute the program to implement the various steps of the intelligent inspection method for communication transmission network equipment as described in any of the above claims.

[0061] A readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of the intelligent inspection method for communication transmission network equipment as described in any of the preceding claims.

[0062] A computer program product includes a computer program that, when executed by a processor, performs the steps of the intelligent inspection method for communication transmission network equipment as described in any of the preceding claims.

[0063] As can be seen from the above technical solutions, the intelligent inspection method and related equipment for communication transmission network devices provided in this application decouple inspection operations into atomic capability units to construct an atomic capability library. This is combined with a device knowledge base and a three-dimensional protocol capability matrix to achieve task planning. Furthermore, a systematic intelligent inspection architecture is constructed using multi-modal data acquisition, multi-model collaborative analysis, and result fusion mechanisms, achieving comprehensive improvement in technical efficiency. Specifically, by decomposing inspection operations into five-tuple-structured atomic capability units and constructing a library, the inspection operations are given minimum-granularity reusability and flexibility. This allows for rapid matching based on different inspection needs, significantly improving the customization and compatibility of inspection tasks. By parsing inspection requests and retrieving atomic capability combination patterns from the device knowledge base, precise matching between inspection targets and atomic capabilities is achieved, providing knowledge support for subsequent task planning and improving the rationality and targeting of task generation. By leveraging a three-dimensional protocol capability matrix and linking historical inspection data, device protocol knowledge, and multi-dimensional capability dimensions, the system accurately selects compatible protocols for atomic capabilities and plans task timing and command sequences, effectively optimizing the execution efficiency of inspection task flows and ensuring a high degree of alignment between protocol adaptation and device type and inspection requirements. By collecting multimodal data from target network devices, the system overcomes the limitations of single data types, enriches the dimensions of inspection data, and provides a data foundation for comprehensively perceiving device operating status and surrounding environmental conditions. After spatiotemporal alignment of the multimodal data and inputting it into the corresponding analysis model, the system can accurately extract feature information from different types of data, generating structured results containing anomaly types, confidence levels, and related evidence, improving the professionalism and accuracy of data analysis. Through multi-source evidence fusion, logical correlation, and cross-verification of the structured analysis results, the system strengthens the reliability and comprehensiveness of anomaly judgment, reduces the limitations of single analysis results, and ultimately achieves efficient, accurate, and intelligent inspection and diagnosis, significantly improving the overall technical level and operation and maintenance capabilities of communication transmission network inspection. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0065] Figure 1 This is a flowchart of an intelligent inspection method for communication transmission network equipment disclosed in an embodiment of this application;

[0066] Figure 2 This is a schematic diagram of an intelligent inspection device for communication transmission network equipment disclosed in an embodiment of this application;

[0067] Figure 3This is a hardware structure block diagram of an intelligent inspection device for communication transmission network equipment disclosed in an embodiment of this application. Detailed Implementation

[0068] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0069] This application can be used in a wide variety of general-purpose or special-purpose computing device environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor devices, distributed computing environments including any of the above devices, etc.

[0070] The following section introduces the solution proposed in this application. The technical solution is as follows, and details are provided below.

[0071] Figure 1 This is a flowchart of an intelligent inspection method for communication transmission network equipment disclosed in an embodiment of this application.

[0072] like Figure 1 As shown, the method may include:

[0073] Step S1: Decouple the inspection operation of each network device into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0074] Specifically, the inspection scenarios, operation procedures, and functional requirements of various network devices are identified, and indivisible standardized inspection operations are defined as atomic capability units. This ensures that each unit carries only a single, independent, minimum-granularity inspection function, avoiding functional overlap and ensuring cross-scenario reusability, thus breaking down the rigid barriers of traditional inspection operations. Subsequently, a unified five-tuple structure is constructed for each atomic capability unit to achieve standardized and normalized capability management: the capability identifier is an encoding used to uniquely distinguish different atomic capabilities; the semantic tag is used to describe the inspection function category implemented by the atomic capability; the protocol binding information includes at least one network management protocol supported by the atomic capability, as well as the specific command template, object identifier, or application programming interface required to execute the atomic capability under the protocol; the preconditions include the network connectivity status, device authentication information, or other atomic capability identifiers that need to be completed before executing the atomic capability; and the output mode defines the data structure, field names, and types of the results returned after executing the atomic capability.

[0075] Step S2: In response to the inspection request input by the user, parse the inspection request to determine the inspection target, and retrieve the associated atomic capability combination pattern from the pre-built device knowledge base based on the inspection target.

[0076] Specifically, the system achieves technical integration between inspection requirements and atomic capabilities. Through semantic parsing algorithms and knowledge-driven retrieval mechanisms, it intelligently maps requirements to capability combinations, supporting precise scheduling of inspection tasks. After a user initiates an inspection request, the system launches a multi-dimensional parsing process based on semantic understanding and requirement feature extraction. Relying on natural language processing technology, it deconstructs the request text and accurately captures explicit requirements such as the scope of inspection equipment and core detection indicators through keyword extraction, semantic segmentation, and requirement classification algorithms. Simultaneously, based on a historical inspection requirement database and user profiles, it identifies potential requirements such as inspection priority, fault diagnosis focus, and result output format through association rule mining algorithms. After feature fusion, a structured inspection target dataset is generated, avoiding inspection redundancy or omissions caused by misunderstandings of requirements. Subsequently, the system calls a pre-built device knowledge base through an interface. This knowledge base adopts an architecture combining structured storage and unstructured indexing, integrating data such as network device attribute parameters, operation and maintenance characteristics, inspection scenario templates, fault association rules, and capability requirement mapping relationships. Through knowledge graph construction and iterative optimization, a standardized knowledge system is formed. Based on structured inspection targets, the system employs a dual algorithm of semantic similarity calculation and scene feature matching. It performs index retrieval in the knowledge base to select atomic capability combination patterns that meet the target matching threshold. These patterns are solidified into standardized capability matching templates through training with historical inspection cases and calibration with technical specifications. They can be directly used as input sources for subsequent task planning without manual intervention in capability combination. The algorithm-driven approach achieves precise matching of requirements and capabilities, providing standardized technical support for subsequent protocol selection and task flow generation.

[0077] Step S3: In the three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, based on the atomic capability combination mode and the target network device type, an adaptive protocol is selected from multiple candidate protocols for each atomic capability in the atomic capability combination mode, and task planning is performed to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0078] Specifically, the three-dimensional protocol capability matrix adopts a tensor storage structure, based on historical inspection datasets, device protocol specifications, and operational practice data. It is constructed using data normalization and adaptation quantification algorithms. The three dimensions correspond to device type codes, communication protocol identifiers, and multi-dimensional capability evaluation indicators, respectively. Matrix element values ​​are calculated using a weighted summation algorithm, representing the adaptation level of different devices and protocol combinations in terms of inspection coverage, execution latency, and transmission stability, providing a quantitative technical basis for protocol selection. The system first parses the protocol requirements and target device type codes of each capability combination mode, obtains the candidate protocol set through matrix index traversal, and then selects the optimal adaptable protocol for each atomic capability based on the adaptation degree ranking algorithm and inspection requirement constraints, avoiding problems such as link interruption and data packet loss caused by protocol incompatibility. After protocol adaptation is completed, the system initiates a task planning process based on a directed acyclic graph. Combining the preconditions and dependencies of each atomic capability, protocol execution timing constraints, device resource occupancy thresholds, and inspection priority weights, the system plans the execution order of atomic capabilities through a topology sorting algorithm. Simultaneously, the system encapsulates the protocol command template, object identifier, and operation logic into an executable instruction sequence, generating a standardized inspection task flow that includes instruction sets, timing nodes, resource allocation strategies, and exception rollback mechanisms. This achieves the technical coupling of atomic capabilities and protocols, supporting the automated execution of the inspection process.

[0079] Step S4: Execute the inspection task flow to collect multimodal data of the target network device.

[0080] Specifically, based on the timing nodes and instruction sets in the inspection task flow, the system sequentially triggers each atomic capability through the capability call interface, synchronously loads the adapted protocol stack, and establishes a communication link with the target device. It employs a combined polling and interrupt acquisition mechanism to conduct full-dimensional data collection. During the collection process, internal operational data, including configuration parameters, operation logs, performance indicators, and port status, is acquired through the device management interface and converted into a standardized format by the protocol parsing module. External status data, including physical layer indicator light status, device appearance images, ambient temperature and humidity, and abnormal noise characteristics, is collected through visual sensors and environmental sensing modules. This non-digital data is transformed into storable data through image preprocessing, audio feature extraction, and sensor signal calibration. A timestamp synchronization mechanism and spatial identification encoding are embedded during the collection process to bind traceability information such as collection time, device location, and collection module identifier to each type of data, storing it in a distributed data buffer. This ensures the temporal consistency and traceability of multimodal data. Through automated collection and standardized conversion, errors from manual collection are avoided, while simultaneously enriching the data dimensions.

[0081] Step S5: After aligning the multimodal data in time and space according to the data type, input them into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0082] Specifically, the process begins with spatiotemporal alignment preprocessing. Based on the timestamp information of the collected data, a time synchronization protocol is used to calibrate the temporal reference of different modalities, eliminating temporal deviations caused by differences in collection frequency. A spatial identifier mapping algorithm is then used to associate and bind various types of data with the physical location of the devices and the collection modules, achieving consistent alignment in the spatial dimension and laying the foundation for cross-modal correlation analysis. Subsequently, a data type-model adaptation mechanism is employed. Based on data characteristics (such as text logs, image data, and sensor signals), corresponding dedicated analysis models are invoked. Digital log data is integrated into a text classification and feature extraction model, visual data into an image recognition and anomaly detection model, and environmental sensor data into a signal analysis and threshold judgment model. Parallel analysis is achieved through a lightweight model deployment architecture. Each model, through feature extraction, anomaly identification, and confidence calculation algorithms, outputs structured results containing anomaly type codes, confidence scores, original data fragments, and feature indicators. Confidence scores are generated using a multi-feature fusion evaluation algorithm, and correlation evidence is stored using data hashing and index association technology, ensuring the traceability and standardization of the analysis results and achieving a technical transformation from raw data to structured information.

[0083] Step S6: Perform multi-source evidence fusion, logical correlation and cross-verification processing on each of the structured analysis results in sequence to generate inspection and diagnosis results.

[0084] Specifically, through multi-dimensional result processing, the system achieves more precise anomaly judgment and more intelligent decision-making, ultimately generating comprehensive and reliable inspection and diagnostic results. This provides strong support for operation and maintenance decisions and significantly improves network operation and maintenance capabilities. Specifically, the system first performs multi-source evidence fusion processing, integrating structured analysis results corresponding to different modalities of data. This breaks the information limitations of a single data source, complements the advantages of various types of evidence, and forms a comprehensive and complete evidence set, avoiding misjudgments caused by incomplete data. Next, logical association processing is performed. Based on fault association rules in the device knowledge base, the system mines the inherent logical relationships between different anomaly results, clarifying the correspondence between abnormal phenomena and potential fault causes and the scope of fault impact, achieving in-depth analysis from anomaly discovery to cause location. On this basis, cross-verification processing is conducted, using evidence from different sources and modalities to mutually verify each other, eliminating possible errors, misjudgments, or interference factors in single-model analysis, further improving the accuracy and reliability of anomaly judgment. After the above series of processing steps, the system generates complete inspection and diagnostic results, clearly presenting the overall operating status of the target network equipment, existing anomalies, root causes of faults, scope of impact, and rectification suggestions, along with a complete chain of evidence and the judgment process. This result can directly provide accurate decision-making basis for operation and maintenance personnel, significantly shorten the fault diagnosis and rectification cycle, improve operation and maintenance efficiency, realize intelligent decision-making for inspection and diagnosis, and significantly improve the technical level of network inspection and the overall operation and maintenance support capability.

[0085] As can be seen from the above technical solutions, the intelligent inspection method and related equipment for communication transmission network devices provided in this application decouple inspection operations into atomic capability units to construct an atomic capability library. This is combined with a device knowledge base and a three-dimensional protocol capability matrix to achieve task planning. Furthermore, a systematic intelligent inspection architecture is constructed using multi-modal data acquisition, multi-model collaborative analysis, and result fusion mechanisms, achieving comprehensive improvement in technical efficiency. Specifically, by decomposing inspection operations into five-tuple-structured atomic capability units and constructing a library, the inspection operations are given minimum-granularity reusability and flexibility. This allows for rapid matching based on different inspection needs, significantly improving the customization and compatibility of inspection tasks. By parsing inspection requests and retrieving atomic capability combination patterns from the device knowledge base, precise matching between inspection targets and atomic capabilities is achieved, providing knowledge support for subsequent task planning and improving the rationality and targeting of task generation. By leveraging a three-dimensional protocol capability matrix and linking historical inspection data, device protocol knowledge, and multi-dimensional capability dimensions, the system accurately selects compatible protocols for atomic capabilities and plans task timing and command sequences, effectively optimizing the execution efficiency of inspection task flows and ensuring a high degree of alignment between protocol adaptation and device type and inspection requirements. By collecting multimodal data from target network devices, the system overcomes the limitations of single data types, enriches the dimensions of inspection data, and provides a data foundation for comprehensively perceiving device operating status and surrounding environmental conditions. After spatiotemporal alignment of the multimodal data and inputting it into the corresponding analysis model, the system can accurately extract feature information from different types of data, generating structured results containing anomaly types, confidence levels, and related evidence, improving the professionalism and accuracy of data analysis. Through multi-source evidence fusion, logical correlation, and cross-verification of the structured analysis results, the system strengthens the reliability and comprehensiveness of anomaly judgment, reduces the limitations of single analysis results, and ultimately achieves efficient, accurate, and intelligent inspection and diagnosis, significantly improving the overall technical level and operation and maintenance capabilities of communication transmission network inspection.

[0086] Furthermore, since network device types continuously iterate and communication protocol versions are constantly upgraded in practical applications, and protocol execution performance fluctuates under different scenarios, this application also includes a matrix dynamic iterative update step, as follows:

[0087] ① Performance Data Acquisition and Recording: After each execution of the aforementioned inspection task flow, the system automatically records the actual execution performance data of each called protocol through the protocol monitoring interface. Key indicators include response success rate and data collection completeness rate. Simultaneously, it associates and stores contextual information such as corresponding device type, atomic capability identifier, inspection scenario, and execution time to form a standardized performance dataset, which is then stored in the matrix update data source library. The response success rate is calculated as the ratio of the number of protocol requests sent to the number of valid responses, and the data collection completeness rate is evaluated by the degree of matching between the actual number of collected data fields and the preset number of fields, ensuring the quantitative traceability of performance data.

[0088] ② Matrix Iterative Update: Based on the actual execution performance data, an iterative update algorithm is used to dynamically adjust the capability vector values ​​of the corresponding protocols in the three-dimensional protocol capability matrix. Specifically, firstly, an outlier removal and normalization process is performed on the performance data using a data preprocessing algorithm to eliminate the interference of extreme data on the evaluation results; then, based on a preset weight allocation rule, differentiated weights are assigned to the response success rate and data collection integrity rate, and a weighted evaluation model is constructed by combining historical performance data to calculate the updated protocol adaptation value; finally, through an incremental update mechanism, only the capability vector values ​​of the corresponding device-protocol combination are dynamically corrected without reconstructing the entire matrix, balancing update accuracy and computational efficiency, ensuring that the three-dimensional protocol capability matrix always matches the actual operation and maintenance scenario, and continuously optimizing the reliability of protocol adaptation decisions.

[0089] To further enhance the depth and accuracy of fault diagnosis and achieve precise tracing from fault symptoms to root causes, this application also includes a root cause localization step, as follows:

[0090] First, a pre-built device causal graph is invoked. This device causal graph is constructed based on the network device hardware architecture, software operation logic, and operation and maintenance failure cases. It is stored in a graph structure, where nodes represent device hardware components, software configuration items, or environmental state factors, and edges are directed edges used to represent the direct causal relationship between nodes. Each edge is associated with a preset causal strength weight. The weight value is calibrated based on the frequency of historical failure associations and expert knowledge to quantify the degree of causal influence between nodes.

[0091] Secondly, taking the node corresponding to the fault phenomenon identified in the inspection and diagnosis results as the starting node, a graph search algorithm is executed in the causal graph of the equipment to trace back along the causal edges. The search process adopts a depth-first strategy, and simultaneously records the causal edge weights of each path and the multimodal evidence support of the corresponding nodes. The evidence support is determined by retrieving the abnormal features, confidence levels, and associated evidence corresponding to the nodes in the structured analysis results, forming a multi-path tracing result set.

[0092] Finally, based on the causal strength weights of each edge in the search path and the support of multimodal evidence for the path nodes, a comprehensive evaluation model is constructed to determine the fault propagation path. Specifically, a weighted summation algorithm is used to calculate the comprehensive score of each tracing path, with weighting factors including causal strength weights and evidence confidence. The path with the highest comprehensive score is selected as the optimal fault propagation path, and the endpoint of this path is determined as the root cause of the fault phenomenon.

[0093] By employing the aforementioned root cause localization steps and combining them with existing multi-source evidence fusion and cross-verification mechanisms, the system achieves precise anomaly detection and intelligent decision-making, ultimately generating comprehensive and reliable inspection and diagnostic results to provide strong support for operational and maintenance decisions. The system integrates fault phenomena, propagation paths, root causes, and related evidence to form standardized diagnostic reports. These reports clearly present the overall operational status of the target network equipment, anomalies, root causes, scope of impact, and rectification suggestions, significantly shortening the fault investigation and rectification cycle, improving operational and maintenance efficiency, and substantially enhancing network operation and maintenance capabilities.

[0094] In some embodiments of this application, the process of generating the inspection task flow is described, which may specifically include:

[0095] ① Based on the three-dimensional protocol capability matrix, obtain the multi-dimensional capability vector corresponding to the target network device type and each candidate protocol. The multi-dimensional capability vector includes quantitative indicators of data collection granularity, indicator coverage depth, historical reliability, and security level.

[0096] ② Determine the weights of each capability dimension based on the characteristics of atomic capabilities and the requirements of the inspection task;

[0097] ③ Based on the multidimensional capability vector and the weights of each capability dimension, calculate the weighted comprehensive score of each candidate protocol;

[0098] ④ Select the primary execution protocol based on the weighted composite score, and convert the atomic capabilities in the atomic capability combination mode into operation commands supported by the primary execution protocol;

[0099] ⑤ Based on equipment resource constraints and task priorities, all operation commands after atomic capability conversion are concurrently scheduled and timed to form the inspection task flow.

[0100] Specifically, based on the constructed three-dimensional protocol capability matrix, multi-dimensional capability vectors corresponding to the target network device type and each candidate protocol are extracted. These multi-dimensional capability vectors encompass four core quantitative indicators: data collection granularity, indicator coverage depth, historical reliability, and security level. Data collection granularity characterizes the protocol's ability to collect refined device data; indicator coverage depth reflects the breadth of inspection indicators that the protocol can cover; historical reliability is based on statistical analysis of past inspection execution data; and the security level corresponds to the protocol's protective capabilities during data transmission and device operation. All indicators have been standardized to ensure comparability between dimensions, providing a unified benchmark for subsequent evaluation.

[0101] Subsequently, the weights of each capability dimension were determined by combining the characteristics of atomic capabilities with the requirements of inspection tasks. Core priorities were assigned based on the functional attributes of atomic capabilities, while also matching the core needs of inspection tasks. Differentiated weights were assigned to each capability dimension for different types of atomic capabilities and inspection scenarios, ensuring that the weight allocation aligns with actual application needs and makes subsequent protocol evaluation more targeted. For example, configuring verification-type atomic capabilities emphasizes data collection granularity and indicator coverage depth, while security inspection tasks emphasize security level and historical reliability. The evaluation logic was optimized through demand-driven weight settings.

[0102] Based on the acquired multidimensional capability vector and the determined dimension weights, a weighted comprehensive score is calculated for each candidate protocol. By fusing the quantitative indicators of each dimension with their corresponding weights, a comprehensive evaluation result for each type of candidate protocol is obtained. This result can intuitively reflect the overall suitability of the candidate protocol with the current atomic capabilities, target equipment, and inspection requirements, providing quantitative support for the selection of the subsequent main execution protocol.

[0103] The optimal main execution protocol is selected based on a weighted comprehensive score, prioritizing the protocol with the best overall performance as the core execution basis. After determining the main execution protocol, a preset command template library and protocol adaptation module are invoked to convert various atomic capabilities in the atomic capability combination mode into standardized operation commands that the main execution protocol can recognize and support. At the same time, key information such as association identifiers and parameter configurations required for protocol execution are supplemented, completing the technical coupling between atomic capabilities and execution commands, ensuring that commands can directly drive the equipment to perform inspection operations.

[0104] Finally, considering equipment resource constraints and task priorities, all converted operation commands are concurrently scheduled and timed. Resource constraints during equipment operation are fully considered to avoid excessive resource consumption during inspection operations affecting normal equipment operation. Furthermore, commands that can be executed in parallel are assigned independent execution chains according to task priority, and the execution order of dependent commands is planned based on preconditions. The execution timing and intervals of each command are reasonably set, ultimately forming a standardized inspection task flow that includes operation command sequences, execution logic, resource allocation schemes, and exception fallback mechanisms.

[0105] The formula for calculating the weighted composite score of each candidate protocol is as follows:

[0106]

[0107] in, For the first One candidate protocol; For the first One target device type; For representing devices in multidimensional capability vectors Agreement In the Quantitative indicator values ​​for each capability dimension; For the corresponding to the first The weights of each capability dimension.

[0108] Based on this, considering that in actual communication network inspection scenarios, the overall performance of candidate protocols may converge, and the reliability of a single main execution protocol is easily affected by factors such as network link fluctuations and equipment load changes, which may lead to risks such as missing inspection data and task execution interruption, this application further adds a dynamic protocol adaptation and optimization mechanism to improve the robustness of protocol adaptation, the reliability of data collection, and the fault tolerance of task execution. This mechanism enables intelligent control of protocol selection and task execution. The specific technical implementation is as follows:

[0109] ① Select the candidate protocol with the highest weighted composite score as the main execution protocol;

[0110] ②If the difference between the highest and second-highest scores in the weighted composite scores of each candidate protocol is less than a preset redundancy threshold, then the main execution protocol and the protocol with the second-highest score are activated simultaneously to perform parallel redundant collection of key inspection indicators.

[0111] ③ If the historical reliability index of the main execution protocol is lower than the preset reliability threshold, task degradation is triggered, non-critical atomic capability units are removed from the atomic capability combination mode, and a lightweight inspection task flow is generated.

[0112] Specifically, the first step is to implement a primary execution protocol optimization strategy. Based on the weighted comprehensive score ranking results, the candidate protocol with the highest score is selected as the primary execution protocol. This ensures optimal adaptation of the primary execution protocol to atomic capability characteristics, target network device types, and inspection task requirements, providing basic support for the efficient implementation of inspection tasks. This strategy serves as the benchmark scenario for protocol selection and can maximize the coverage depth and execution efficiency of inspection indicators.

[0113] Secondly, a protocol redundancy collection mechanism is triggered. A preset redundancy threshold is used to determine the difference between the highest and second-highest scores. This threshold is based on historical protocol adaptation data and inspection reliability requirements, representing a critical value for the overall performance differences of the protocols. If the difference is below this threshold, it indicates that the two protocols perform similarly in core dimensions such as data collection granularity and security level, both meeting key inspection requirements. In this case, the primary execution protocol and the second-highest-scoring alternative protocol are simultaneously activated to conduct parallel redundant collection of core inspection indicators. Through data source comparison and consistency verification logic between the two protocols, a dual data protection link is constructed to avoid the risk of key data distortion or loss due to link congestion or abnormal device response caused by a single protocol, thus enhancing the reliability and integrity of the collected data.

[0114] Finally, the dynamic task degradation mechanism is activated. After the main execution protocol is determined, historical reliability indicators in the three-dimensional protocol capability matrix are synchronously called for verification. If the indicator is lower than the preset reliability threshold (based on past protocol execution failure rate and data transmission stability statistical calibration), it indicates that the main execution protocol has potential execution risks. At this time, the degradation process is automatically triggered: according to the preset atomic capability priority classification system (based on the equipment operation and maintenance core indicator system, distinguishing between critical and non-critical capabilities), non-critical atomic capability units are accurately screened and removed, retaining only core atomic capabilities such as equipment core operating status monitoring and key configuration verification; then, based on the remaining core capabilities, the protocol command conversion and resource scheduling orchestration are re-completed to generate a lightweight inspection task flow, achieving a dynamic balance between reliability risk avoidance and core inspection requirement assurance, significantly improving the inspection system's adaptability under complex working conditions.

[0115] In some embodiments of this application, the process of parsing the inspection request to determine the inspection target and retrieving the associated atomic capability combination pattern from a pre-built device knowledge base based on the inspection target may specifically include:

[0116] ① Parse the inspection request, extract the equipment identifier, fault description, and maintenance operation intention contained therein, and form a structured query;

[0117] ②The structured query is used to retrieve the equipment knowledge base, which stores historical inspection cases and experience rules indexed by equipment type, fault scenario and solution;

[0118] ③ Combine and generalize the retrieved historical training cases and empirical rules that match the inspection target to output a set of matching atomic capability units;

[0119] ④ Based on the set of atomic capability units, perform constraint solving on device resource limitations and task execution time windows to determine the atomic capability combination mode.

[0120] Specifically, the received inspection requests are first processed through multi-dimensional analysis. Natural language processing (NLP) techniques are used to perform semantic segmentation, keyword extraction, and intent recognition on the inspection request text, accurately extracting the core information contained therein, specifically covering equipment identification, descriptions of fault phenomena, and operational intentions. Figure 3 Key elements by category. Equipment identifiers are used to locate specific inspection objects and their equipment types; fault symptom descriptions are textual representations of abnormal equipment behavior; and operation and maintenance intentions differentiate core needs such as fault diagnosis, routine checks, and performance monitoring. Extracted elements are encapsulated into structured queries according to a preset format, unifying field definitions and data types, eliminating ambiguity in natural language expressions, and providing standardized input for subsequent knowledge base retrieval.

[0121] Subsequently, a device knowledge base retrieval is performed based on structured queries. This device knowledge base employs a three-dimensional index architecture, classifying and indexing historical inspection cases and experience rules according to three dimensions: device type, fault scenario, and solution. Historical inspection cases include inspection processes, invoked atomic capabilities, and execution effects for similar past scenarios, while experience rules solidify the capability adaptation logic of domain experts for different device faults and maintenance needs. The retrieval process uses an algorithm combining semantic similarity matching and scenario feature mapping. The device identifier in the structured query identifies the corresponding device type category. Combined with the fault phenomenon description and maintenance operation intent, highly relevant historical inspection cases and experience rules are accurately selected within the corresponding category, forming a preliminary retrieval result set.

[0122] Based on the preliminary search results, combined generalization reasoning is performed to generate a matching set of atomic capability units. Through a fusion mechanism of case-based reasoning and rule-based reasoning, the retrieved historical inspection cases and empirical rules are decomposed and analyzed to extract core atomic capability units suitable for the corresponding inspection scenarios. At the same time, generalization expansion is performed for similar scenarios, and combined with the personalized needs of the current inspection target, suitable atomic capability units are supplemented, while redundant or mismatched capability units are eliminated. Finally, a set of atomic capability units covering the entire inspection process and adapting to the target needs is output, ensuring the completeness and relevance of capability coverage.

[0123] Finally, the final atomic capability combination pattern is determined through constraint solving. Based on the set of atomic capability units, the device resource status monitoring module is invoked to obtain the real-time resource limitations of the target device (including CPU usage threshold, port concurrency capacity, communication link bandwidth, etc.). Simultaneously, a constraint programming model is constructed and constraint solving is performed, combined with a preset task execution time window (i.e., the inspection period allowed by operations and maintenance). The combination logic of the atomic capability units is optimized through algorithms, eliminating capability combination schemes that exceed resource limitations or cannot be completed within the time window. Ultimately, an atomic capability combination pattern that balances feasibility, efficiency, and demand adaptability is determined, laying the foundation for subsequent protocol adaptation and task flow generation based on a three-dimensional protocol capability matrix.

[0124] In some embodiments of this application, the process of aligning the multimodal data in time and space according to data type, and then inputting it into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model may specifically include:

[0125] ① Perform spatiotemporal alignment processing on the multimodal data to ensure that data from different sources are consistent in timestamps and spatiotemporal labels;

[0126] ② Input the equipment operation log into the log analysis model, identify abnormal patterns including continuous error codes and frequency mutations, and output the first structured analysis result containing the abnormal pattern code, confidence level and corresponding original log fragment index;

[0127] ③ Input the device images captured by the camera into the visual diagnostic model, identify the color status of the indicator lights and the alarm text on the screen, and output the second structured analysis results containing visual anomaly labels, confidence levels and key frame image indexes;

[0128] ④ Input the equipment performance index sequence and environmental sensor data into the time series analysis model, detect the abnormal drift of the data relative to the prediction baseline, and output a third structured analysis result containing the data prediction value, anomaly score and data source identifier.

[0129] Specifically, the first step is to perform spatiotemporal alignment processing on the collected multimodal data. This process is a core prerequisite for ensuring the effectiveness of cross-modal data correlation analysis. A timestamp synchronization calibration algorithm is used to unify the time base of different data sources, eliminating timing deviations caused by asynchronous clocks and differences in acquisition frequencies of the acquisition devices. Simultaneously, based on information such as the physical location encoding of the devices and the deployment coordinates of the acquisition modules, spatial label mapping and association are completed. A unified spatial dimension identifier is bound to each type of data, ensuring that data from different acquisition sources maintain a high degree of consistency in timestamps and spatial labels. This constructs a standardized spatiotemporal benchmark for multimodal data, laying the foundation for subsequent sub-model analysis and cross-modal verification.

[0130] Secondly, different types of data are input into dedicated analysis models for processing. Device operation logs are input into a log analysis model, which, based on a log feature library and anomaly pattern recognition algorithm in the operations and maintenance field, deeply mines key information in the log text, accurately identifies typical anomaly patterns such as continuous error codes and sudden changes in indicator frequency, and quantitatively evaluates the identification results. Finally, it outputs a first structured analysis result containing anomaly pattern codes, confidence levels, and corresponding original log fragment indexes, achieving structured parsing and anomaly localization of log data. Device images captured by cameras are input into a visual diagnostic model, which integrates image preprocessing, target detection, and text recognition algorithms. It first performs preprocessing operations such as denoising and enhancement on the images, and then accurately... The system identifies visual features such as the color status of equipment indicator lights and screen alarm text, and outputs a second structured analysis result containing visual anomaly labels, confidence levels, and keyframe image indexes, thus completing a visual diagnosis of the equipment's physical status. It then inputs equipment performance index sequences and environmental sensor data into a time-series analysis model. This model, based on a prediction baseline generated from historical time-series data, uses a deviation detection algorithm to monitor the abnormal drift of current data relative to the prediction baseline in real time, quantifies the degree of data deviation, and outputs a third structured analysis result containing predicted data values, anomaly scores, and data source identifiers, thereby achieving a time-series analysis of equipment operating trends and environmental status.

[0131] In some embodiments of this application, the process of sequentially performing multi-source evidence fusion, logical correlation, and cross-verification processing on the structured analysis results to generate inspection and diagnostic results may specifically include:

[0132] ① The structured analysis results are formatted into standardized evidence objects that include anomaly type, confidence level, spatiotemporal label, and original data index;

[0133] ② Based on the preset fusion rules and fault knowledge base, the standardized evidence objects are weighted and fused to calculate the weighted comprehensive confidence level of each candidate fault type;

[0134] ③ Based on the weighted comprehensive confidence level, perform hierarchical judgment and initiate cross-verification process, and arbitrate through logical consistency verification, supplementary data request or historical case backtracking to determine the inspection diagnosis result.

[0135] Specifically, the structured analysis results are first standardized and formatted to eliminate format differences between different model outputs, thus constructing a unified evidence object system. Based on preset data specifications, each structured analysis result is encapsulated into a standardized evidence object containing anomaly type, confidence level, spatiotemporal label, and original data index. The anomaly type uses a unified encoding from a fault knowledge base to ensure consistency in cross-modal anomaly descriptions; the confidence level retains the quantitative results output by the original analysis model, serving as the basis for evidence reliability assessment; the spatiotemporal label uses the baseline identifier after spatiotemporal alignment of the multimodal data mentioned earlier, realizing the temporal and spatial correlation between evidence and original data; the original data index is associated with the original collected data (log fragments, keyframe images, time series) of the corresponding modality, providing support for subsequent traceability and verification. Simultaneously, invalid or incomplete evidence is eliminated through format validation algorithms, ensuring that all evidence objects participating in the fusion conform to data specifications, laying a standardized foundation for subsequent fusion processing.

[0136] Subsequently, based on preset fusion rules and a fault knowledge base, standardized evidence objects are weighted and fused to calculate the weighted comprehensive confidence level of each candidate fault type. The fusion rules are formulated by combining the reliability of the evidence source and the weight of the data type. The reliability of the evidence source is calibrated based on the historical accuracy of the corresponding analysis model, and the weight of the data type is set according to the contribution of different modalities of data to fault diagnosis (e.g., visual data has a higher weight in supporting hardware faults than log data). The fault knowledge base provides a list of candidate fault types and association rules between each fault and anomaly feature, providing knowledge support for evidence matching and fault mapping. During the fusion process, each evidence object is first associated with a candidate fault type using a feature matching algorithm. Then, weights are assigned to each association based on the fusion rules. The comprehensive confidence level of each candidate fault type is calculated by weighted summation of multiple features, achieving preliminary aggregation of cross-modal evidence and breaking the limitations of single-modal evidence.

[0137] Finally, a tiered judgment is executed based on the weighted overall confidence level, and a cross-verification process is initiated. The final inspection and diagnosis result is determined through a multi-level arbitration mechanism. The tiered judgment is based on confidence thresholds, classifying candidate fault types into three confidence levels: high, medium, and low. A high confidence level indicates sufficient evidence and a clear fault indication; a medium confidence level indicates some contradictions or insufficient evidence; and a low confidence level indicates fragmented evidence and an ambiguous fault indication. Corresponding cross-verification strategies are initiated for different levels: for high confidence levels, logical consistency checks are performed to verify whether the descriptions of the fault type by various modal evidences are consistent; if there are no contradictions, it is directly locked as a candidate conclusion. For medium confidence levels, a supplementary data request process is initiated, calling the corresponding capabilities in the atomic capability library based on missing evidence to supplement and collect key data to strengthen evidence support, or using historical case backtracking to match fault handling experience in similar scenarios to assist in the judgment. For low confidence levels, the association rules of the fault knowledge base and historical equipment operation data are combined to reconstruct the evidence association logic, eliminate interfering evidence, and recalculate the confidence level. After the above arbitration process, the fault conclusion with the highest confidence and the most sufficient evidence is selected. The anomaly details, confidence level, evidence chain and traceability index are integrated to generate a complete inspection and diagnosis result, realizing a closed-loop diagnosis from anomaly identification to fault location.

[0138] The process of performing a graded judgment based on the weighted comprehensive confidence level and initiating a cross-verification process, and arbitrating through logical consistency verification, supplementary data requests, or historical case backtracking to determine the inspection and diagnosis results, further includes:

[0139] If the overall confidence level of a certain fault type is higher than the first predetermined threshold, and the difference between its confidence level and the confidence level of the second highest fault type is greater than the second predetermined threshold, then it is directly determined as the current fault type.

[0140] If the highest overall confidence level is lower than the first predetermined threshold, or the difference between the confidence level of the second highest fault type and the second predetermined threshold is less than or equal to the second predetermined threshold, then a secondary verification process is initiated. The secondary verification process includes requesting the data acquisition terminal to re-acquire the specified type of evidence data, and submitting the current contradictory evidence and the preliminary judgment result to the manual review interface.

[0141] If the overall confidence level of all candidate fault types is lower than the third predetermined threshold, the multimodal evidence is determined to be noise or invalid, and the analysis results are discarded.

[0142] Specifically, a tiered judgment is performed based on a weighted comprehensive confidence level, and a cross-verification process is initiated. The final inspection and diagnosis result is determined through threshold quantification and a multi-level arbitration mechanism. The first, second, and third predetermined thresholds are all generated based on historical diagnostic data from the fault knowledge base, reliability requirements of the operation and maintenance scenario, and accuracy calibration of each analysis model. They can be dynamically adjusted according to equipment type and inspection priority to ensure the adaptability of the judgment logic.

[0143] The specific tiered arbitration process is as follows: If the overall confidence level of a certain fault type is higher than the first predetermined threshold, and the difference between its confidence level and the confidence level of the second highest fault type is greater than the second predetermined threshold, it indicates that the evidence support for this fault type is sufficient, the fault direction is unique and unambiguous, and there is no cross-modal evidence contradiction. At this time, this type is directly determined as the current fault type, and the corresponding standardized evidence object and original data index are synchronously associated to form a preliminary diagnostic conclusion. If the highest overall confidence level is lower than the first predetermined threshold, or the difference between its confidence level and the second highest fault type is less than or equal to the second predetermined threshold, it indicates that the evidence support is insufficient or there is a cross-modal evidence contradiction, and the fault direction is ambiguous. At this time, a secondary verification process is initiated: on the one hand, an instruction is sent to the data acquisition end through the atomic capability scheduling interface to request the re-collection of evidence data of a specified type (such as supplementing visual images or time-series performance data for contradictory evidence), and the evidence object is updated and the overall confidence level is recalculated based on the supplemented data; on the other hand, the current contradictory evidence, the preliminary judgment result, and the complete evidence chain are synchronized to the manual review interface, allowing maintenance personnel to view the original data, mark contradictory points, and provide review opinions, realizing collaborative arbitration of machine diagnosis and manual intervention. If the overall confidence level of all candidate fault types is lower than the third predetermined threshold, it indicates that the multimodal evidence has serious noise interference or data acquisition failure, and cannot form a valid fault indication. In this case, the multimodal evidence is judged as noise or invalid data, and the analysis results are directly discarded without generating an inspection diagnosis conclusion. At the same time, an anomaly alarm is triggered, prompting maintenance personnel to check the data acquisition link or re-initiate the inspection task to avoid invalid data misleading maintenance decisions. After the above-mentioned hierarchical arbitration process, the final fault type, overall confidence level, complete evidence chain, and traceability index are integrated to generate standardized inspection diagnosis results, realizing a closed-loop diagnosis from anomaly identification to fault location.

[0144] The following describes an intelligent inspection device for communication transmission network equipment provided in the embodiments of this application. The intelligent inspection device for communication transmission network equipment described below can be referred to in correspondence with the intelligent inspection method for communication transmission network equipment described above.

[0145] See Figure 2 , Figure 2 This is a schematic diagram of an intelligent inspection device for communication transmission network equipment disclosed in an embodiment of this application.

[0146] like Figure 2 As shown, the intelligent inspection device for communication transmission network equipment may include:

[0147] The atomic capability parsing module 110 is used to decouple the inspection operations of each network device into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0148] The capability combination retrieval module 120 is used to respond to the inspection request input by the user, parse the inspection request to determine the inspection target, and retrieve the associated atomic capability combination pattern from the pre-built device knowledge base based on the inspection target.

[0149] The inspection task planning module 130 is used to select an appropriate protocol from multiple candidate protocols for each atomic capability in a three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, based on the atomic capability combination mode and the target network device type, and to perform task planning to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0150] The modal data acquisition module 140 is used to execute the inspection task flow to acquire multimodal data of the target network device;

[0151] The multi-source anomaly analysis module 150 is used to align the multimodal data in time and space according to the data type, and then input them into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0152] The fusion and correlation diagnosis module 160 is used to sequentially perform multi-source evidence fusion, logical correlation and cross-verification processing on the structured analysis results to generate inspection diagnosis results.

[0153] As can be seen from the above technical solutions, the intelligent inspection method and related equipment for communication transmission network devices provided in this application decouple inspection operations into atomic capability units to construct an atomic capability library. This is combined with a device knowledge base and a three-dimensional protocol capability matrix to achieve task planning. Furthermore, a systematic intelligent inspection architecture is constructed using multi-modal data acquisition, multi-model collaborative analysis, and result fusion mechanisms, achieving comprehensive improvement in technical efficiency. Specifically, by decomposing inspection operations into five-tuple-structured atomic capability units and constructing a library, the inspection operations are given minimum-granularity reusability and flexibility. This allows for rapid matching based on different inspection needs, significantly improving the customization and compatibility of inspection tasks. By parsing inspection requests and retrieving atomic capability combination patterns from the device knowledge base, precise matching between inspection targets and atomic capabilities is achieved, providing knowledge support for subsequent task planning and improving the rationality and targeting of task generation. By leveraging a three-dimensional protocol capability matrix and linking historical inspection data, device protocol knowledge, and multi-dimensional capability dimensions, the system accurately selects compatible protocols for atomic capabilities and plans task timing and command sequences, effectively optimizing the execution efficiency of inspection task flows and ensuring a high degree of alignment between protocol adaptation and device type and inspection requirements. By collecting multimodal data from target network devices, the system overcomes the limitations of single data types, enriches the dimensions of inspection data, and provides a data foundation for comprehensively perceiving device operating status and surrounding environmental conditions. After spatiotemporal alignment of the multimodal data and inputting it into the corresponding analysis model, the system can accurately extract feature information from different types of data, generating structured results containing anomaly types, confidence levels, and related evidence, improving the professionalism and accuracy of data analysis. Through multi-source evidence fusion, logical correlation, and cross-verification of the structured analysis results, the system strengthens the reliability and comprehensiveness of anomaly judgment, reduces the limitations of single analysis results, and ultimately achieves efficient, accurate, and intelligent inspection and diagnosis, significantly improving the overall technical level and operation and maintenance capabilities of communication transmission network inspection.

[0154] The intelligent inspection device for communication transmission network equipment provided in this application embodiment can be applied to intelligent inspection equipment for communication transmission network equipment. Figure 3 The hardware structure block diagram of the intelligent inspection device for communication transmission network equipment is shown. (Refer to...) Figure 3 The hardware structure of the intelligent inspection equipment for communication transmission network devices may include: at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4.

[0155] In this embodiment of the application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 communicate with each other through communication bus 4;

[0156] Processor 1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0157] Memory 3 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device;

[0158] The memory stores a program, which the processor can call. The program is used for:

[0159] The inspection operations of each network device are decoupled into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0160] In response to a user-inputted inspection request, the inspection request is parsed to determine the inspection target, and the associated atomic capability combination pattern is retrieved from a pre-built device knowledge base based on the inspection target.

[0161] In the three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, an adaptive protocol is selected from multiple candidate protocols for each atomic capability in the atomic capability combination mode based on the atomic capability combination mode and the target network device type, and task planning is performed to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0162] The inspection task flow is executed to collect multimodal data of the target network device;

[0163] After aligning the multimodal data in time and space according to the data type, the data is input into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0164] The structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation and cross-verification to generate inspection and diagnosis results.

[0165] Optionally, the refined and extended functions of the program can be referred to the above description.

[0166] This application embodiment also provides a readable storage medium that can store a program suitable for execution by a processor, the program being used for:

[0167] The inspection operations of each network device are decoupled into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0168] In response to a user-inputted inspection request, the inspection request is parsed to determine the inspection target, and the associated atomic capability combination pattern is retrieved from a pre-built device knowledge base based on the inspection target.

[0169] In the three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, an adaptive protocol is selected from multiple candidate protocols for each atomic capability in the atomic capability combination mode based on the atomic capability combination mode and the target network device type, and task planning is performed to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0170] The inspection task flow is executed to collect multimodal data of the target network device;

[0171] After aligning the multimodal data in time and space according to the data type, the data is input into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0172] The structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation and cross-verification to generate inspection and diagnosis results.

[0173] Optionally, the refined and extended functions of the program can be referred to the above description.

[0174] This application also provides a computer program product, including a computer program, wherein the computer program is executed by a processor using the following method:

[0175] The inspection operations of each network device are decoupled into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode.

[0176] In response to a user-inputted inspection request, the inspection request is parsed to determine the inspection target, and the associated atomic capability combination pattern is retrieved from a pre-built device knowledge base based on the inspection target.

[0177] In the three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, an adaptive protocol is selected from multiple candidate protocols for each atomic capability in the atomic capability combination mode based on the atomic capability combination mode and the target network device type, and task planning is performed to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension.

[0178] The inspection task flow is executed to collect multimodal data of the target network device;

[0179] After aligning the multimodal data in time and space according to the data type, the data is input into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence.

[0180] The structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation and cross-verification to generate inspection and diagnosis results.

[0181] Optionally, the refined and extended functions of the program can be referred to the above description.

[0182] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0183] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0184] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent inspection of communication transmission network equipment, characterized in that, include: The inspection operations of each network device are decoupled into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode. In response to a user-inputted inspection request, the inspection request is parsed to determine the inspection target, and the associated atomic capability combination pattern is retrieved from a pre-built device knowledge base based on the inspection target. In the three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, an adaptive protocol is selected from multiple candidate protocols for each atomic capability in the atomic capability combination mode based on the atomic capability combination mode and the target network device type, and task planning is performed to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension. The inspection task flow is executed to collect multimodal data of the target network device; After aligning the multimodal data in time and space according to the data type, the data is input into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence. The structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation and cross-verification to generate inspection and diagnosis results.

2. The method according to claim 1, characterized in that, The process of generating the inspection task flow includes: Based on the three-dimensional protocol capability matrix, obtain the multi-dimensional capability vector corresponding to the target network device type and each candidate protocol. The multi-dimensional capability vector includes quantitative indicators of data collection granularity, indicator coverage depth, historical reliability, and security level. The weights of each capability dimension are determined based on the characteristics of atomic capabilities and the requirements of the inspection task. Based on the multidimensional capability vector and the weights of each capability dimension, the weighted comprehensive score of each candidate protocol is calculated. The primary execution protocol is selected based on the weighted composite score, and the atomic capabilities in the atomic capability combination mode are converted into operation commands supported by the primary execution protocol. Based on equipment resource constraints and task priorities, all operation commands after atomic capability conversion are concurrently scheduled and timed to form the inspection task flow.

3. The method according to claim 2, characterized in that, The formula for calculating the weighted composite score of each candidate protocol is as follows: in, For the first One candidate protocol; For the first One target device type; For representing devices in multidimensional capability vectors Agreement In the Quantitative indicator values ​​for each capability dimension; For the corresponding to the first The weights of each capability dimension.

4. The method according to claim 2, characterized in that, Also includes: The candidate protocol with the highest weighted composite score is selected as the main execution protocol. If the difference between the highest and second-highest scores in the weighted composite scores of the candidate protocols is less than a preset redundancy threshold, then the main execution protocol and the protocol with the second-highest score are activated simultaneously to perform parallel redundant collection of key inspection indicators. If the historical reliability index of the main execution protocol is lower than the preset reliability threshold, task degradation is triggered, non-critical atomic capability units are removed from the atomic capability combination mode, and a lightweight inspection task flow is generated.

5. The method according to claim 1, characterized in that, Also includes: After each execution of the inspection task flow, the actual execution performance data of each called protocol is recorded, including response success rate and data collection integrity rate. Based on the actual execution performance data, the capability vector values ​​of the corresponding protocols in the three-dimensional protocol capability matrix are dynamically adjusted using an iterative update algorithm.

6. The method according to claim 1, characterized in that, The inspection request is parsed to determine the inspection target, and based on the inspection target, the associated atomic capability combination pattern is retrieved from a pre-built device knowledge base, including: The inspection request is parsed to extract the equipment identifier, fault description, and maintenance operation intent contained therein, forming a structured query; The structured query is used to retrieve the equipment knowledge base, which stores historical inspection cases and experience rules indexed by equipment type, fault scenario, and solution. The retrieved historical training cases and empirical rules that match the inspection target are combined and generalized to infer a set of matching atomic capability units. Based on the set of atomic capability units, the constraints of device resource limitations and task execution time windows are solved to determine the atomic capability combination mode.

7. The method according to claim 1, characterized in that, After aligning the multimodal data in time and space according to data type, the data is input into the analysis model corresponding to the data type, respectively, to obtain the structured analysis results output by each analysis model, including: The multimodal data is spatiotemporally aligned to ensure that data from different sources remain consistent in timestamps and spatiotemporal labels. Input the device operation log into the log analysis model to identify abnormal patterns including continuous error codes and frequency mutations, and output the first structured analysis result containing the abnormal pattern code, confidence level and corresponding original log fragment index; The device images captured by the camera are input into the visual diagnostic model to identify the color status of the indicator lights and the alarm text on the screen, and output a second structured analysis result containing visual anomaly labels, confidence scores and key frame image indexes. The equipment performance index sequence and environmental sensor data are input into the time series analysis model to detect abnormal drift of the data relative to the prediction baseline, and output a third structured analysis result containing the data prediction value, anomaly score and data source identifier.

8. The method according to claim 1, characterized in that, The structured analysis results are sequentially subjected to multi-source evidence fusion, logical correlation, and cross-verification to generate inspection and diagnostic results, including: The structured analysis results are formatted into standardized evidence objects that include anomaly type, confidence level, spatiotemporal label, and original data index. Based on preset fusion rules and a fault knowledge base, the standardized evidence objects are weighted and fused to calculate the weighted comprehensive confidence level of each candidate fault type. Based on the weighted comprehensive confidence level, a graded judgment is performed and a cross-verification process is initiated. Arbitration is conducted through logical consistency verification, supplementary data requests, or historical case backtracking to determine the inspection and diagnosis results.

9. The method according to claim 8, characterized in that, Based on the weighted comprehensive confidence level, a tiered judgment is performed and a cross-verification process is initiated. Arbitration is conducted through logical consistency checks, supplementary data requests, or historical case backtracking to determine the inspection and diagnosis results, including: If the overall confidence level of a certain fault type is higher than the first predetermined threshold, and the difference between its confidence level and the confidence level of the second highest fault type is greater than the second predetermined threshold, then it is directly determined as the current fault type. If the highest overall confidence level is lower than the first predetermined threshold, or the difference between the confidence level of the second highest fault type and the second predetermined threshold is less than or equal to the second predetermined threshold, then a secondary verification process is initiated. The secondary verification process includes requesting the data acquisition terminal to re-acquire the specified type of evidence data, and submitting the current contradictory evidence and the preliminary judgment result to the manual review interface. If the overall confidence level of all candidate fault types is lower than the third predetermined threshold, the multimodal evidence is determined to be noise or invalid, and the analysis results are discarded.

10. The method according to claim 1, characterized in that, Also includes: A pre-built device cause-effect graph is invoked, wherein the nodes of the device cause-effect graph represent device hardware components, software configuration items or environmental state factors, and the edges represent the causal relationships between the nodes; Taking the node corresponding to the fault phenomenon determined in the inspection and diagnosis results as the starting node, a graph search algorithm is executed in the causal graph of the equipment to trace back along the causal edges. The fault propagation path is determined based on the causal strength weights of each edge in the search path and the support of multimodal evidence for the path nodes, and the endpoint of the fault propagation path is determined as the root cause of the fault phenomenon.

11. An intelligent inspection device for communication transmission network equipment, characterized in that, include: The atomic capability parsing module is used to decouple the inspection operations of each network device into multiple atomic capability units to build an atomic capability library. Each atomic capability unit is a five-tuple structure that performs the smallest granularity inspection operation. The five-tuple includes at least capability identifier, semantic tag, protocol binding information, preconditions and output mode. The capability combination retrieval module is used to respond to the inspection request input by the user, parse the inspection request to determine the inspection target, and retrieve the associated atomic capability combination pattern from the pre-built device knowledge base based on the inspection target. The inspection task planning module is used to select an appropriate protocol from multiple candidate protocols for each atomic capability in a three-dimensional protocol capability matrix constructed based on historical inspection data and device protocol knowledge, based on the atomic capability combination mode and the target network device type, and to perform task planning to generate an inspection task flow for the target network device that includes a specific operation command sequence and execution timing. The three-dimensional protocol capability matrix is ​​constructed based on historical inspection data and device protocol knowledge and is used to characterize the adaptability of different device types and different communication protocol combinations in the multi-dimensional capability dimension. The modal data acquisition module is used to execute the inspection task flow to acquire multimodal data of the target network device; The multi-source anomaly analysis module is used to align the multimodal data in time and space according to the data type, and then input them into the analysis model corresponding to the data type to obtain the structured analysis results output by each analysis model, which include anomaly type, confidence level and related evidence. The fusion and correlation diagnosis module is used to sequentially perform multi-source evidence fusion, logical correlation and cross-verification processing on the structured analysis results to generate inspection diagnosis results.

12. An intelligent inspection device for communication transmission network equipment, characterized in that, Including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the intelligent inspection method for communication transmission network equipment as described in any one of claims 1-10.

13. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the intelligent inspection method for communication transmission network equipment as described in any one of claims 1-10.

14. A computer program product, comprising a computer program, characterized in that, The computer program, when run by a processor, executes each step of the intelligent inspection method for communication transmission network equipment as described in any one of claims 1-10.

Images