Big model-based multi-dimensional multi-index adaptive inspection report generation method and system and medium

Abstract

This application provides a method, system, and medium for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model. The method includes: acquiring multi-dimensional, multi-indicator raw inspection data; cleaning and normalizing the raw inspection data; generating a service topology diagram; establishing component relationships; constructing a dynamic system state diagram; and generating a full-link related data chain. Based on multi-dimensional evaluation indicators, the full-link related data chain is analyzed to select Top-K key indicator data with diagnostic value in the current scenario. A causal discovery algorithm is used to mine potential data relationships to obtain related data, and a problem propagation path diagram is generated based on the related data. The inspection health status is determined, and the report structure and content depth are adjusted according to the status level to generate an inspection report. Through multi-source data association modeling and problem propagation path diagram, the end-to-end operating status of the system is comprehensively reflected, achieving a unified state assessment of the entire system and improving the efficiency of inspection report generation.

Description

Technical Field

[0001] This application relates to the field of inspection report generation technology, specifically to a method, system, and medium for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model. Background Technology

[0002] Traditional operation and maintenance inspection reports rely on manual data collection, analysis, and report writing. Limited by labor costs, professional capabilities, and analytical dimensions, the core pain points are concentrated in the following areas: Inefficient and with low fault tolerance: Inspection data is scattered across various components and systems. Manual collection, organization, and analysis require a lot of time and effort, and are prone to data omissions and analysis biases due to human negligence, making it impossible to meet the needs of rapid inspection of large-scale systems.

[0003] The content is simplistic and lacks synergy: the reports focus on isolated analysis of single components and single metrics, failing to achieve synergistic analysis across services, middleware, and links, thus failing to reflect the end-to-end operational status of the system and making it difficult to locate cross-component correlation issues.

[0004] The analysis is superficial and lacks depth: it can only achieve basic data statistics and anomaly listing, but it lacks intelligent in-depth analysis capabilities, cannot uncover the root causes, transmission paths and potential risks behind the anomalies, and is difficult to complete a unified status assessment of the entire system.

[0005] The conclusions are unreliable and pose security risks: lacking specific constraints for the operation and maintenance domain, large-scale model applications are prone to problems such as illusions, uncontrollable reasoning logic, and inconsistencies between analysis conclusions and actual data, which may mislead operation and maintenance decisions and lead to the expansion of system failures. Summary of the Invention

[0006] The purpose of this application is to provide a method, system and medium for generating multi-dimensional and multi-indicator adaptive inspection reports based on a large model. By using multi-source data association modeling and problem propagation path diagrams, it comprehensively reflects the end-to-end operating status of the system, realizes unified status assessment of the entire system, improves the efficiency of inspection report generation, reduces manual intervention and lowers the probability of human error.

[0007] This application also provides a method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model, including: Obtain raw inspection data with multiple dimensions and indicators, and clean and normalize the raw inspection data. Based on the inspection data after cleaning and normalization, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. Based on multi-dimensional evaluation index analysis of the entire data chain, the Top-K key index data with diagnostic value in the current scenario are selected. The processed related data is converted into a preset data format, and a causal discovery algorithm is used to mine potential data correlations to obtain related data. Based on the related data, a problem propagation path diagram is generated. Based on the correlation data and problem propagation path diagram, the health status of the inspection is determined, and the report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making.

[0008] Optionally, in the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model described in this application embodiment, the original inspection data with multiple dimensions and indicators is obtained, and the original inspection data is cleaned and normalized, specifically including: Traverse the original inspection data, analyze the data missing status information, and analyze the data missing type based on the data missing status information; If the data missing type is completely missing, the data is determined to be abnormal, and the original inspection data is collected again. If the data is missing in part, interpolation and mean imputation methods are used to fill in the missing data. The supplemented data is converted to the same standard range based on the standardization and normalization method; The normalized data is validated, and the converted data is corrected based on the validation results.

[0009] Optionally, in the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model described in this application embodiment, a service topology graph is generated based on the cleaned and normalized inspection data, component associations are established, a dynamic system state graph is constructed, key data is marked, and a full-link association data chain is generated, specifically including: Based on the inspection data after cleaning and normalization, a visualization topology construction algorithm is used to generate a visualization service topology map. Based on a visualized service topology diagram, a multi-dimensional component association system is constructed to form a component association network. Based on the service topology diagram, component association system and real-time normalized data, a dynamically updatable system state diagram is constructed. Based on dynamic system state diagrams and normalized data, key data with core diagnostic value are marked; Based on the component association system, dynamic system state diagram and key data marking results, the flow of the entire link is analyzed to form a structured full-link association data chain. Optionally, in the multi-dimensional, multi-index adaptive inspection report generation method based on a large model described in this application embodiment, the multi-dimensional evaluation indicators include: Topological prior weights: Basic weights are set based on the centrality and rank of components in the service dependency graph; Time-series dynamic sensitivity: Calculate the coefficient of variation or score at the point of change for each indicator over a preset period, and temporarily increase the weight of highly volatile indicators; Business semantic activation: Based on the loading of predefined scenario templates, relevant indicator sets are automatically activated; User feedback reinforcement learning: Allows operations and maintenance personnel to score the relevance of indicators in historical reports, and continuously optimizes the indicator selection strategy based on the strategy gradient.

[0010] Optionally, in the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model described in this application embodiment, the processed associated data is converted into a preset data format, a causal discovery algorithm is used to mine potential data correlations to obtain associated data, and a problem propagation path diagram is generated based on the associated data, specifically including: The processed associated data is converted to a unified data format according to a preset data format; The algorithm for causal discovery is used to process related data in a preset format, to uncover potential causal relationships between the data, and to organize the causal relationship results to generate standardized related data. Each piece of related data contains core attributes such as causal node, correlation strength, confidence level, and time delay; The core attributes are sorted by confidence level, and the core associations with a confidence level greater than or equal to the set confidence threshold are marked to generate complete association data. A problem propagation path diagram is generated based on complete related data.

[0011] Optionally, in the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model described in this application embodiment, the inspection health status is determined based on the associated data and the problem propagation path diagram, and the report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making, specifically including: Analysis of inspection health status information based on associated data and problem transmission path diagram; The health status information of the inspection is compared with the set status information to obtain the status deviation rate; The health status level is analyzed based on the state deviation rate and the set deviation rate threshold. Based on the determined health status level, the structure and module proportions of the inspection report are dynamically adjusted to generate the inspection report. Secondly, embodiments of this application provide a multi-dimensional, multi-indicator adaptive inspection report generation system based on a large model. The system includes a memory and a processor. The memory includes a program for a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model. When the program for the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model is executed by the processor, it performs the following steps: Obtain raw inspection data with multiple dimensions and indicators, and clean and normalize the raw inspection data. Based on the inspection data after cleaning and normalization, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. Based on multi-dimensional evaluation index analysis of the entire data chain, the Top-K key index data with diagnostic value in the current scenario are selected. The processed related data is converted into a preset data format, and a causal discovery algorithm is used to mine potential data correlations to obtain related data. Based on the related data, a problem propagation path diagram is generated. Based on the correlation data and problem propagation path diagram, the health status of the inspection is determined, and the report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making.

[0012] Optionally, in the multi-dimensional, multi-indicator adaptive inspection report generation system based on a large model described in this application embodiment, the original inspection data with multiple dimensions and indicators is obtained, and the original inspection data is cleaned and normalized, specifically including: Traverse the original inspection data, analyze the data missing status information, and analyze the data missing type based on the data missing status information; If the data missing type is completely missing, the data is determined to be abnormal, and the original inspection data is collected again. If the data is missing in part, interpolation and mean imputation methods are used to fill in the missing data. The supplemented data is converted to the same standard range based on the standardization and normalization method; The normalized data is validated, and the converted data is corrected based on the validation results.

[0013] Optionally, in the multi-dimensional, multi-indicator adaptive inspection report generation system based on a large model described in this application embodiment, a service topology diagram is generated based on the cleaned and normalized inspection data, component associations are established, a dynamic system state diagram is constructed, key data is marked, and a full-link associated data chain is generated, specifically including: Based on the inspection data after cleaning and normalization, a visualization topology construction algorithm is used to generate a visualization service topology map. Based on a visualized service topology diagram, a multi-dimensional component association system is constructed to form a component association network. Based on the service topology diagram, component association system and real-time normalized data, a dynamically updatable system state diagram is constructed. Based on dynamic system state diagrams and normalized data, key data with core diagnostic value are marked; Based on the component association system, dynamic system state diagram and key data marking results, the flow of the entire link is analyzed to form a structured full-link association data chain. Thirdly, embodiments of this application also provide a computer-readable storage medium, which includes a method program for generating multi-dimensional and multi-indicator adaptive inspection reports based on a large model. When the method program is executed by a processor, it implements the steps of the method program for generating multi-dimensional and multi-indicator adaptive inspection reports based on a large model as described in any of the above claims.

[0014] As can be seen from the above, the multi-dimensional, multi-indicator adaptive inspection report generation method, system, and medium provided in this application embodiment are as follows: First, the original inspection data with multiple dimensions and indicators is acquired and cleaned and normalized. Then, a service topology diagram is generated based on the cleaned and normalized inspection data, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. Next, the full-link related data chain is analyzed based on multi-dimensional evaluation indicators to select Top-K key indicator data with diagnostic value in the current scenario. The processed related data is converted into a preset data format, and a causal discovery algorithm is used to mine potential data relationships to obtain related data. A problem propagation path diagram is generated based on the related data. The inspection health status is determined based on the related data and the problem propagation path diagram. The report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decisions. Through multi-source data association modeling and the problem propagation path diagram, the end-to-end operating status of the system is comprehensively reflected, achieving a unified system status assessment across the entire link, improving the efficiency of inspection report generation, reducing manual intervention, and lowering the probability of human error. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart of a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model provided in this application embodiment; Figure 2 A flowchart illustrating the original inspection data cleaning and normalization process of the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model provided in this application embodiment; Figure 3The flowchart illustrates the end-to-end data chain generation method for the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model, as provided in the embodiments of this application. Detailed Implementation

[0017] 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 a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0018] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0019] Please refer to Figure 1 , Figure 1 This is a flowchart of a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model, as described in some embodiments of this application. This multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model is used in a terminal device and includes the following steps: S101: Obtain raw inspection data with multiple dimensions and indicators, and clean and normalize the raw inspection data. S102, Based on the inspection data after cleaning and normalization, generate a service topology diagram, establish component associations, construct a dynamic system state diagram, mark key data, and generate a full-link association data chain; S103 analyzes the entire data chain based on multi-dimensional evaluation indicators to select Top-K key indicator data with diagnostic value in the current scenario; S104: Convert the processed associated data into a preset data format, use a causal discovery algorithm to mine potential data associations, obtain associated data, and generate a problem propagation path diagram based on the associated data; S105 determines the health status of the inspection based on the associated data and the problem propagation path diagram, and adaptively adjusts the report structure and content depth according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making.

[0020] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating the raw inspection data cleaning and normalization process of a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model, as described in some embodiments of this application. According to embodiments of the present invention, obtaining multi-dimensional, multi-indicator raw inspection data and performing cleaning and normalization processing on the raw inspection data specifically includes: S201, Traverse the original inspection data, analyze the data missing status information, and analyze the data missing type based on the data missing status information; S202. If the data missing type is completely missing, the data is determined to be abnormal, and the original inspection data is collected again. S203. If the data missing type is partial missing, interpolation and mean imputation methods are used to supplement the missing data. S204, based on the standardization and normalization method, converts the supplemented data into the same standard range; S205 verifies the normalized data and corrects the converted data based on the verification results.

[0021] It should be noted that the core objective of data normalization is to eliminate invalid and abnormal data, correct data deviations, and ensure the accuracy and completeness of the original inspection data, laying the foundation for subsequent normalization processing. The specific steps are as follows: Data integrity verification: Traverse the original inspection data in temporary storage, check and mark missing data, and adopt corresponding handling methods for different missing types: Completely missing data: If the data for a certain indicator is completely missing (e.g., no response time data for a certain service was collected), mark the data as missing and remove the related logic of that indicator during subsequent correlation analysis; For partially missing data: If some data collection time points of a certain indicator are missing, interpolation methods (such as linear interpolation) or mean filling methods (based on the mean of historical normal data in the same time period) are used to supplement the data to avoid missing data affecting the overall analysis; for partially missing data of core key indicators (such as error rate and timeout indicators), historical data from similar scenarios in the same period are used as the priority to fill the missing data.

[0022] Anomaly identification and removal: Based on experience in operations and maintenance, reasonable value ranges and normal fluctuation thresholds for each indicator are preset to identify and process anomaly data. Data outside the reasonable range: For example, if the normal response time of a service is 50ms-500ms, and abnormal data of 10000ms is collected, it is judged as invalid abnormal data and is removed. Abnormal fluctuation data: By calculating the rate of change of the indicator at adjacent collection time points, if the rate of change exceeds the preset threshold (e.g., a single fluctuation exceeds 50%), and there is no corresponding business scenario to support it (e.g., no major promotion, sudden increase in traffic, etc.), it is judged as abnormal fluctuation data and is removed. Abnormal log data: Filter invalid logs (such as duplicate logs, blank logs, and debug logs), and only retain log data with diagnostic value such as error logs and warning logs, and mark the corresponding components, metrics and abnormal types (such as timeouts and errors) of the logs.

[0023] Redundant data filtering: Removes duplicate and invalid redundant data to reduce data processing pressure. Duplicate data: Investigate and delete identical duplicate data (duplicate data at the same time point, the same indicator, or the same data source), and keep only one valid data. Redundant data: Delete redundant data that has no practical diagnostic value (such as redundant system logs that are irrelevant to inspection and analysis, and invalid empty field data), and focus on core indicator data and key log data.

[0024] Data correction: Corrective actions are taken for correctable deviations identified during verification (such as incorrect data format or incorrect units). Unit error correction: If some response time data is in seconds and some is in milliseconds, it will be uniformly converted to milliseconds to ensure data consistency.

[0025] Data verification after cleaning: The cleaned data is verified a second time to ensure that there is no missing, abnormal, or redundant data. If the verification is successful, the data will proceed to the normalization process. If the verification fails, the data will be returned to the corresponding cleaning step for reprocessing.

[0026] The data normalization process is as follows: Core objective: To eliminate format and metric differences between different indicators and data sources, making various inspection data comparable and correlateable, and providing standardized data for subsequent entity association, indicator weight evaluation, and large-scale model analysis. Specific steps are as follows: Establish normalization standards: Based on the characteristics of operation and maintenance inspection indicators, formulate unified normalization standards covering three core dimensions: Format standards: Standardize the format of all data, including timestamp format, numerical format (consistently retaining 2 decimal places), and log format (consistently organized according to the format of "time-component-metric-anomaly type-details"). Measurement Standards: For indicator data with different dimensions, formulate unified dimension conversion rules, such as uniformly converting all time-related indicators (response time, query latency) to milliseconds (ms), and uniformly converting all ratio-related indicators (error rate, success rate) to percentages (%). Semantic standards: unify the semantic descriptions of indicators and exception types, such as uniformly naming "interface call failure rate" and "service error rate" as "error rate", and uniformly marking "timeout" and "response timeout" as "timeout exception" to avoid semantic ambiguity.

[0027] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating the full-link associated data chain generation method of a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model, as described in some embodiments of this application. According to embodiments of the present invention, a service topology diagram is generated based on the cleaned and normalized inspection data, component associations are established, a dynamic system state diagram is constructed, key data is marked, and a full-link associated data chain is generated. Specifically, this includes: S301, based on the inspection data after cleaning and normalization, uses a visualization topology construction algorithm to generate a visualization service topology map; S302, based on a visualized service topology diagram, constructs a multi-dimensional component association system to form a component association network; S303, based on the service topology diagram, component association system and real-time normalized data, constructs a dynamically updatable system state diagram; S304, based on dynamic system state diagrams and normalized data, marks key data with core diagnostic value; S305 analyzes the entire link flow based on the component association system, dynamic system state diagram and key data marking results, and forms a structured full-link association data chain. According to embodiments of the present invention, the multi-dimensional evaluation indicators include: Topological prior weights: Basic weights are set based on the centrality and rank of components in the service dependency graph; Time-series dynamic sensitivity: Calculate the coefficient of variation or score at the point of change for each indicator over a preset period, and temporarily increase the weight of highly volatile indicators; Business semantic activation: Based on the loading of predefined scenario templates, relevant indicator sets are automatically activated; User feedback reinforcement learning: Allows operations and maintenance personnel to score the relevance of indicators in historical reports, and continuously optimizes the indicator selection strategy based on the strategy gradient.

[0028] According to an embodiment of the present invention, the processed associated data is converted into a preset data format, a causal discovery algorithm is used to mine potential data correlations to obtain associated data, and a problem propagation path diagram is generated based on the associated data, specifically including: The processed associated data is converted to a unified data format according to a preset data format; The algorithm for causal discovery is used to process related data in a preset format, to uncover potential causal relationships between the data, and to organize the causal relationship results to generate standardized related data. Each piece of related data contains core attributes such as causal node, correlation strength, confidence level, and time delay; The core attributes are sorted by confidence level, and the core associations with a confidence level greater than or equal to the set confidence threshold are marked to generate complete association data. A problem propagation path diagram is generated based on complete related data.

[0029] It should be noted that the associated data format conversion involves calling the full-link associated data chain stored in the target database and converting it according to the preset data format to ensure compatibility with the causal discovery algorithm and subsequent large-scale model analysis requirements. Preset format definition: Combining the characteristics of the operation and maintenance field and algorithm requirements, a unified data format is preset, covering six core fields: data identifier, component information, indicator name, indicator value, association type, and timestamp. The field format is consistent with the normalization standard to ensure semantic ambiguity. Data transformation execution: The structured data and unstructured data (such as tagged abnormal logs) in the related data chain are converted into preset formats respectively. Among them, the numerical data (indicator values, correlation strength) retains the normalized results, and the text data (log details, correlation description) is simplified and refined according to the preset semantic specifications to ensure that the data is concise and that the core information is not lost. Post-conversion verification: Perform a comprehensive verification on the converted data, focusing on format consistency, data integrity, and consistency of association logic to ensure no format deviations, missing data, or disordered associations. After passing the verification, output the pre-formatted association data for subsequent causal discovery algorithm calculations.

[0030] Data correlation mining based on causal discovery algorithms: Using causal discovery algorithms (such as PC algorithm and LiNGAM algorithm) adapted to operation and maintenance scenarios, the algorithm is used to perform calculations on the pre-formatted related data to mine the potential causal relationships between the data and generate complete related data. Algorithm initialization configuration: Combine historical operation and maintenance data and component association system to configure the core parameters of the algorithm (such as association significance threshold and causal confidence threshold), filter weak association and pseudo-association data, and improve the accuracy of causal mining; Causal correlation mining: Input pre-formatted correlated data into the algorithm model. The algorithm mines causal relationships between data based on the time series changes of indicators and component dependencies (such as "increased MySQL query latency → payment service response timeout → order service error rate exceeding the standard"), which is different from simple statistical correlation. Data processing: The causal relationships mined by the algorithm are processed to generate standardized data. Each data entry contains core attributes such as causal nodes (cause components + indicators, result components + indicators), correlation strength, confidence level, and time delay. Data is sorted from high to low confidence level and high-confidence core associations (confidence level ≥ 90%) are marked.

[0031] Problem propagation path diagram generated based on correlation data: Taking standardized correlation data obtained from causal discovery as the core, combined with service topology diagram and dynamic system state diagram, a visual problem propagation path diagram is generated to clearly present the anomaly propagation logic.

[0032] Path graph basic construction: Adopting the component node and edge association logic of dynamic system state graph, the causal nodes in the standardized association data are mapped to the corresponding component nodes, the causal relationship is mapped to the edge between the nodes, and the association strength, confidence and time delay of the edge are labeled. Anomaly Node Location: Based on the key data marking results, locate anomaly nodes in the system (such as component nodes with excessive error rate or timeout), use the anomaly nodes as the core starting point of the path graph, and mark the anomaly type, anomaly degree and occurrence time. Path analysis and pruning: Starting from the abnormal node, analyze all potential transmission paths along the causal relationship logic (from cause to effect), eliminate weak correlation, low confidence (confidence <60%) and transmission paths with no actual diagnostic value, and retain only the core transmission paths with the most information and high confidence for judging the system state; Path graph optimization and output: The pruned path graph is optimized by marking abnormal nodes, core transmission paths, and normal nodes with different colors to clearly distinguish the scope of anomalies and transmission logic; visual charts and structured data are generated, allowing users to click on nodes / paths to view related data and anomaly details, while also storing the path graph data for subsequent large-scale model anomaly diagnosis and report generation.

[0033] According to an embodiment of the present invention, the health status of an inspection is determined based on associated data and a problem propagation path diagram. The report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making. Specifically, this includes: Analysis of inspection health status information based on associated data and problem transmission path diagram; The health status information of the inspection is compared with the set status information to obtain the status deviation rate; The health status level is analyzed based on the state deviation rate and the set deviation rate threshold. Based on the determined health status level, the structure and module proportions of the inspection report are dynamically adjusted to generate the inspection report. Secondly, embodiments of this application provide a multi-dimensional, multi-indicator adaptive inspection report generation system based on a large model. The system includes a memory and a processor. The memory includes a program for a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model. When the program for the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model is executed by the processor, it performs the following steps: Obtain raw inspection data with multiple dimensions and indicators, and clean and normalize the raw inspection data. Based on the inspection data after cleaning and normalization, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. Based on multi-dimensional evaluation index analysis of the entire data chain, the Top-K key index data with diagnostic value in the current scenario are selected. The processed related data is converted into a preset data format, and a causal discovery algorithm is used to mine potential data correlations to obtain related data. Based on the related data, a problem propagation path diagram is generated. Based on the correlation data and problem propagation path diagram, the health status of the inspection is determined, and the report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making.

[0034] According to an embodiment of the present invention, multi-dimensional and multi-indicator raw inspection data is obtained, and the raw inspection data is cleaned and normalized, specifically including: Traverse the original inspection data, analyze the data missing status information, and analyze the data missing type based on the data missing status information; If the data missing type is completely missing, the data is determined to be abnormal, and the original inspection data is collected again. If the data is missing in part, interpolation and mean imputation methods are used to fill in the missing data. The supplemented data is converted to the same standard range based on the standardization and normalization method; The normalized data is validated, and the converted data is corrected based on the validation results.

[0035] According to an embodiment of the present invention, a service topology diagram is generated based on the inspection data after cleaning and normalization, component associations are established, a dynamic system state diagram is constructed, key data is marked, and a full-link associated data chain is generated, specifically including: Based on the inspection data after cleaning and normalization, a visualization topology construction algorithm is used to generate a visualization service topology map. Based on a visualized service topology diagram, a multi-dimensional component association system is constructed to form a component association network. Based on the service topology diagram, component association system and real-time normalized data, a dynamically updatable system state diagram is constructed. Based on dynamic system state diagrams and normalized data, key data with core diagnostic value are marked; Based on the component association system, dynamic system state diagram and key data marking results, the flow of the entire link is analyzed to form a structured full-link association data chain.

[0036] In summary, the present invention has the following beneficial effects: Unified Perspective Across the Entire Link: Breaking down barriers between components and data sources, and through multi-source data association modeling and dynamic system state diagrams, the system's end-to-end operational status is comprehensively reflected, enabling a unified state assessment of the entire system across the entire link.

[0037] Significantly improved problem tracing capabilities: Through cause-and-effect analysis and problem propagation path diagram generation, the root cause and scope of impact of cross-component anomalies can be quickly located, reducing the time cost for operation and maintenance personnel to troubleshoot problems and improving the efficiency of anomaly handling.

[0038] Risk warning and collaborative optimization: Based on the trend of indicator changes, it can identify potential operational risks of the system in advance, and generate cross-component collaborative optimization suggestions to help operation and maintenance personnel proactively prevent and control risks and optimize system performance; at the same time, the whole process is automated, which greatly improves the efficiency of inspection report generation, reduces manual intervention, and reduces the probability of human error.

[0039] The report is highly adaptive: the report structure and content depth are intelligently adjusted according to the system's health status, which avoids information overload in normal scenarios and ensures information integrity in abnormal scenarios, thus meeting the actual reading and decision-making needs of operation and maintenance personnel.

[0040] The analysis conclusions are credible and controllable: The hybrid large model architecture for the operation and maintenance field, combined with domain constraints and data verification mechanisms, significantly reduces the "illusion" rate of the large model, ensuring that every analysis conclusion is supported by data and is logically rigorous, and avoiding misleading operation and maintenance decisions.

[0041] The third aspect of the present invention provides a computer-readable storage medium, the storage medium including a method program for generating multi-dimensional and multi-indicator adaptive inspection reports based on a large model, wherein when the method program is executed by a processor, it implements the steps of the method program for generating multi-dimensional and multi-indicator adaptive inspection reports based on a large model as described above.

[0042] This invention discloses a method, system, and medium for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model. The method involves acquiring multi-dimensional, multi-indicator raw inspection data and cleaning and normalizing it. Based on the cleaned and normalized data, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. The full-link related data chain is analyzed based on multi-dimensional evaluation indicators to select Top-K key indicators with diagnostic value in the current scenario. The processed related data is converted to a preset data format, and a causal discovery algorithm is used to mine potential data relationships to obtain related data. A problem propagation path diagram is generated based on the related data. The inspection health status is determined based on the related data and the problem propagation path diagram. The report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decisions. Through multi-source data association modeling and the problem propagation path diagram, the system's end-to-end operating status is comprehensively reflected, achieving a unified system status assessment across the entire link, improving the efficiency of inspection report generation, reducing manual intervention, and lowering the probability of human error.

[0043] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0044] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0045] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0046] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0047] Alternatively, if the integrated units of the present invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model, characterized in that, include: Obtain raw inspection data with multiple dimensions and indicators, and clean and normalize the raw inspection data. Based on the inspection data after cleaning and normalization, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. Based on multi-dimensional evaluation index analysis of the entire data chain, the Top-K key index data with diagnostic value in the current scenario are selected. The processed related data is converted into a preset data format, and a causal discovery algorithm is used to mine potential data correlations to obtain related data. Based on the related data, a problem propagation path diagram is generated. Based on the correlation data and problem propagation path diagram, the health status of the inspection is determined, and the report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making.

2. The method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model according to claim 1, characterized in that, Obtain raw inspection data from multiple dimensions and indicators, and perform cleaning and normalization processing on the raw inspection data, specifically including: Traverse the original inspection data, analyze the data missing status information, and analyze the data missing type based on the data missing status information; If the data missing type is completely missing, the data is determined to be abnormal, and the original inspection data is collected again. If the data is missing in part, interpolation and mean imputation methods are used to fill in the missing data. The supplemented data is converted to the same standard range based on the standardization and normalization method; The normalized data is validated, and the converted data is corrected based on the validation results.

3. The method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model according to claim 2, characterized in that, Based on the cleaned and normalized inspection data, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link interconnected data chain is generated, specifically including: Based on the inspection data after cleaning and normalization, a visualization topology construction algorithm is used to generate a visualization service topology map. Based on a visualized service topology diagram, a multi-dimensional component association system is constructed to form a component association network. Based on the service topology diagram, component association system and real-time normalized data, a dynamically updatable system state diagram is constructed. Based on dynamic system state diagrams and normalized data, key data with core diagnostic value are marked; Based on the component association system, dynamic system state diagram and key data marking results, the flow of the entire link is analyzed to form a structured full-link association data chain.

4. The method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model according to claim 3, characterized in that, Multi-dimensional evaluation indicators include: Topological prior weights: Basic weights are set based on the centrality and rank of components in the service dependency graph; Time-series dynamic sensitivity: Calculate the coefficient of variation or score at the point of change for each indicator over a preset period, and temporarily increase the weight of highly volatile indicators; Business semantic activation: Based on the loading of predefined scenario templates, relevant indicator sets are automatically activated; User feedback reinforcement learning: Allows operations and maintenance personnel to score the relevance of indicators in historical reports, and continuously optimizes the indicator selection strategy based on the strategy gradient.

5. The method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model according to claim 4, characterized in that, The processed correlated data is converted into a preset data format, and a causal discovery algorithm is used to mine potential correlations in the data to obtain correlated data. Based on the correlated data, a problem propagation path diagram is generated, which specifically includes: The processed associated data is converted to a unified data format according to a preset data format; The algorithm for causal discovery is used to process related data in a preset format, to uncover potential causal relationships between the data, and to organize the causal relationship results to generate standardized related data. Each piece of related data contains core attributes such as causal node, correlation strength, confidence level, and time delay; The core attributes are sorted by confidence level, and the core associations with a confidence level greater than or equal to the set confidence threshold are marked to generate complete association data. A problem propagation path diagram is generated based on complete related data.

6. The method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model according to claim 5, characterized in that, Based on correlated data and problem propagation path diagrams, the health status of inspections is determined. The report structure and content depth are adaptively adjusted according to the status level to generate inspection reports that meet the needs of operational decision-making, specifically including: Analysis of inspection health status information based on associated data and problem transmission path diagram; The health status information of the inspection is compared with the set status information to obtain the status deviation rate; The health status level is analyzed based on the state deviation rate and the set deviation rate threshold. Based on the determined health status level, the structure and module proportions of the inspection report are dynamically adjusted to generate the inspection report.

7. A multi-dimensional, multi-indicator adaptive inspection report generation system based on a large model, characterized in that, The system includes a memory and a processor. The memory contains a program for a multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model. When the processor executes the program for the multi-dimensional, multi-indicator adaptive inspection report generation method based on a large model, it performs the following steps: Obtain raw inspection data with multiple dimensions and indicators, and clean and normalize the raw inspection data. Based on the inspection data after cleaning and normalization, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link related data chain is generated. Based on multi-dimensional evaluation index analysis of the entire data chain, the Top-K key index data with diagnostic value in the current scenario are selected. The processed related data is converted into a preset data format, and a causal discovery algorithm is used to mine potential data correlations to obtain related data. Based on the related data, a problem propagation path diagram is generated. Based on the correlation data and problem propagation path diagram, the health status of the inspection is determined, and the report structure and content depth are adaptively adjusted according to the status level to generate an inspection report that meets the needs of operation and maintenance decision-making.

8. The multi-dimensional, multi-index adaptive inspection report generation system based on a large model according to claim 7, characterized in that, Obtain raw inspection data from multiple dimensions and indicators, and perform cleaning and normalization processing on the raw inspection data, specifically including: Traverse the original inspection data, analyze the data missing status information, and analyze the data missing type based on the data missing status information; If the data missing type is completely missing, the data is determined to be abnormal, and the original inspection data is collected again. If the data is missing in part, interpolation and mean imputation methods are used to fill in the missing data. The supplemented data is converted to the same standard range based on the standardization and normalization method; The normalized data is validated, and the converted data is corrected based on the validation results.

9. The multi-dimensional, multi-index adaptive inspection report generation system based on a large model according to claim 8, characterized in that, Based on the cleaned and normalized inspection data, a service topology diagram is generated, component relationships are established, a dynamic system state diagram is constructed, key data is marked, and a full-link interconnected data chain is generated, specifically including: Based on the inspection data after cleaning and normalization, a visualization topology construction algorithm is used to generate a visualization service topology map. Based on a visualized service topology diagram, a multi-dimensional component association system is constructed to form a component association network. Based on the service topology diagram, component association system and real-time normalized data, a dynamically updatable system state diagram is constructed. Based on dynamic system state diagrams and normalized data, key data with core diagnostic value are marked; Based on the component association system, dynamic system state diagram and key data marking results, the flow of the entire link is analyzed to form a structured full-link association data chain.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a method program for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model. When the method program is executed by a processor, it implements the steps of the method for generating multi-dimensional, multi-indicator adaptive inspection reports based on a large model as described in any one of claims 1 to 6.

Images