A Method and System for Intelligent Dialogue Q&A in Operation and Maintenance Technical Services Based on a Large-Scale Knowledge Base

Abstract

This application provides an intelligent dialogue Q&A method and system for operation and maintenance technical services based on a large model knowledge base. It relates to the field of large model operation and maintenance dialogue technology. The method involves receiving operation and maintenance consultation requests, capturing the dynamic evolution characteristics of the problem, and associating them with the real-time updated knowledge modules of the large model knowledge base to obtain corresponding relationships. Based on these relationships and phased descriptions, dynamic scenario adaptation processing is performed. An operation and maintenance Q&A generation module with knowledge iteration functionality is invoked, combining the adaptation results and incremental knowledge to generate multiple rounds of preliminary Q&A content. The preliminary Q&A content is dynamically optimized and adjusted based on feedback information to obtain dynamically optimized Q&A content. The content is fed back to the initiator, and relevant information is associated with the knowledge iteration module to drive incremental updates, completing the intelligent dialogue Q&A process and improving the accuracy and practicality of operation and maintenance Q&A.

Description

Technical Field

[0001] This application relates to the field of large-scale model operation and maintenance dialogue technology, and more specifically, to an intelligent dialogue Q&A method and system for operation and maintenance technical services based on a large-scale model knowledge base. Background Technology

[0002] In the field of operations and maintenance (O&M) technical services, with the increasing complexity of systems and the growing diversity of O&M scenarios, O&M personnel face increasingly complex and varied problems. Traditional O&M Q&A methods mainly rely on human experience or fixed knowledge base queries, which have many limitations.

[0003] On the one hand, traditional methods struggle to capture the dynamic evolution of operational issues. Operational problems are often not static but evolve over time, influenced by factors such as changes in system operating status. However, current technologies cannot track these changes in real time, resulting in answers that may not accurately match the needs of the problem at different stages, thus reducing the effectiveness and relevance of the responses.

[0004] On the other hand, existing technologies have shortcomings in terms of knowledge updating and iteration. Operation and maintenance knowledge is constantly evolving and being updated, with new solutions and troubleshooting methods constantly emerging. However, updating traditional knowledge bases usually requires manual intervention, which is time-consuming and inefficient, and cannot promptly incorporate the latest knowledge into the Q&A process, potentially making the Q&A content outdated. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide an intelligent dialogue Q&A method and system for operation and maintenance technical services based on a large model knowledge base.

[0006] According to a first aspect of this application, a method for intelligent dialogue and Q&A of operation and maintenance technical services based on a large model knowledge base is provided, the method comprising:

[0007] Receive consultation requests for operation and maintenance technical services, synchronously capture the dynamic evolution characteristics of operation and maintenance issues in the consultation requests, associate them with real-time updated knowledge modules in the large model knowledge base that match the dynamic evolution characteristics, and obtain the correspondence between consultation requests marked with dynamic evolution characteristics and real-time updated knowledge modules.

[0008] Based on the correspondence with dynamic evolution feature tags, and combined with the phased description information of operation and maintenance issues in the consultation request, the operation and maintenance knowledge in the real-time updated knowledge module is dynamically adapted to the scenario, and dynamic scenario adaptation results containing evolution stage matching information are generated.

[0009] Call the operation and maintenance Q&A generation module with knowledge iteration function in the large model, input the dynamic scenario adaptation result containing evolution stage matching information into the operation and maintenance Q&A generation module, and import the incremental operation and maintenance knowledge corresponding to the evolution stage from the real-time updated knowledge module to generate multiple rounds of preliminary Q&A content as the operation and maintenance problem evolves.

[0010] Based on the initial Q&A content from multiple rounds and the feedback from the parties initiating the consultation requests, the initial Q&A content from multiple rounds was dynamically optimized and adjusted. Combined with the incremental knowledge supplementation of the knowledge module in real time, the dynamically optimized Q&A content was obtained in sync with the evolution of operation and maintenance issues.

[0011] The system dynamically optimizes and feeds back the Q&A content to the party initiating the consultation request. At the same time, it links the dynamic evolution characteristics of the operation and maintenance issues, the content of multiple rounds of preliminary Q&A, feedback information, and dynamically optimized Q&A content to the knowledge iteration module of the large model knowledge base, driving the incremental updates of the knowledge module in real time, and completing the intelligent dialogue Q&A process for operation and maintenance technical services.

[0012] According to a second aspect of this application, an intelligent dialogue and Q&A system for operation and maintenance technical services based on a large model knowledge base is provided. The intelligent dialogue and Q&A system for operation and maintenance technical services based on a large model knowledge base includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the intelligent dialogue and Q&A system for operation and maintenance technical services based on a large model knowledge base implements the aforementioned intelligent dialogue and Q&A method for operation and maintenance technical services based on a large model knowledge base.

[0013] According to a third aspect of this application, a computer-readable storage medium is provided, wherein computer-executable instructions are stored therein, and when the computer-executable instructions are executed, the aforementioned intelligent dialogue Q&A method for operation and maintenance technical services based on a large model knowledge base is implemented.

[0014] Based on any of the above aspects, the technical effect of this application is as follows:

[0015] By receiving O&M technical service consultation requests and simultaneously capturing the dynamic evolution characteristics of O&M issues, the system can accurately associate them with real-time updated knowledge modules in a large model knowledge base, establishing a correspondence marked with dynamic evolution characteristics. Based on this correspondence and the stage-specific description information of the O&M issues, the system performs dynamic scenario adaptation processing on the real-time updated knowledge modules, generating dynamic scenario adaptation results containing evolution stage matching information. This ensures that O&M knowledge closely matches the specific scenarios of the issues at different stages, improving the relevance of Q&A responses. The system calls an O&M Q&A generation module with knowledge iteration capabilities, combining the dynamic scenario adaptation results and corresponding incremental O&M knowledge to generate multiple rounds of preliminary Q&A content. This provides targeted answers as the O&M issues evolve. Dynamic optimization and adjustments are made based on the multiple rounds of preliminary Q&A content and feedback information, and incremental knowledge is supplemented to obtain dynamically optimized Q&A content synchronized with the evolution of the O&M issues, ensuring that the Q&A content remains accurate and effective. Finally, relevant information is associated with the knowledge iteration module to drive the incremental updates of the real-time updated knowledge modules, continuously adapting to changes in O&M issues and significantly improving the accuracy and timeliness of O&M Q&A responses. Attached Figure Description

[0016] Figure 1 The flowchart of the intelligent dialogue Q&A method for operation and maintenance technical services based on a large model knowledge base provided in the embodiments of this application is shown.

[0017] Figure 2 This paper illustrates a schematic diagram of the component structure of an intelligent dialogue Q&A system for operation and maintenance technical services based on a large model knowledge base, as provided in an embodiment of this application. Detailed Implementation

[0018] Figure 1 This paper illustrates a flowchart of an intelligent dialogue Q&A method and system for operation and maintenance technical services based on a large model knowledge base, as provided in an embodiment of this application. The detailed steps include:

[0019] Step S110: Receive consultation requests for operation and maintenance technical services, synchronously capture the dynamic evolution characteristics of operation and maintenance issues in the consultation requests, associate them with real-time updated knowledge modules in the large model knowledge base that match the dynamic evolution characteristics, and obtain the correspondence between consultation requests marked with dynamic evolution characteristics and real-time updated knowledge modules.

[0020] In intelligent operations and maintenance (O&M) service scenarios, the system receives O&M technical service consultation requests initiated by on-site engineers via a mobile O&M app through a pre-defined API interface. These requests are transmitted in a structured data format, comprising a request header, a request body, and metadata. The request body is the text describing the O&M problem, while the metadata includes information such as the request initiation time and the terminal device identifier. During reception, the text is processed in real-time, and natural language processing (NLP) technology is used to simultaneously capture the dynamic evolution characteristics of the O&M problem. Specifically, a sequence labeling model based on a bidirectional long short-term memory (LSTM) network is employed to label entity words, attribute words, and relational words in the text sequence, identifying dynamic elements such as core fault phenomena, parameter indicators, and changes in requirements in the problem description. Simultaneously, a large model knowledge base is linked using knowledge graph retrieval technology. This large model knowledge base employs a distributed storage architecture and contains multiple real-time updated knowledge modules, each corresponding to a different O&M domain and problem type. The metadata of a knowledge module records the dynamic evolution feature tags it covers. By calculating the cosine similarity between the dynamic evolution features of a consultation request and the tags of the knowledge module, the knowledge module with the highest similarity is selected for real-time updates. The identifier of the knowledge module and the matching dynamic evolution feature tag are added to the consultation request, and finally a corresponding relationship data structure with dynamic evolution feature tags is formed and stored in a temporary cache area.

[0021] Step S111: Extract the text describing the operation and maintenance issues in the consultation request, split the core issue description section and the supplementary explanation description section in the text, and distinguish between the fixed description part of the core issue description section and the variable description part of the supplementary explanation description section.

[0022] After extracting the text describing the operational issues from the request body of the consultation request, a method combining rule-based and machine learning is used for text segmentation. First, regular expressions are used to match sentences guided by keywords such as "fault," "problem," and "anomaly," initially identifying them as candidate sentences for the core problem description section. Sentences containing keywords such as "environment," "parameters," and "attempts" are considered candidate sentences for supplementary description sections. Then, a pre-trained text classification model is used to perform secondary classification on the candidate sentences. This model is based on the BERT architecture, fine-tuned on the operational domain corpus; the input is a sentence vector, and the output is the probability of being either a core problem section or a supplementary description section. For the classified core problem description section, a keyword extraction algorithm (such as TF-IDF) is used to identify the fixed description parts, i.e., stable information such as fault type and core phenomena. The variable description parts in the supplementary description sections are linked to a dynamic attribute dictionary, such as variable parameter descriptions like temperature and pressure, using entity linking technology, thus distinguishing between the fixed and variable description parts.

[0023] Step S112: Perform time-series analysis on the variable representation part, extract the new operation and maintenance phenomenon descriptions, parameter change descriptions and requirement adjustment descriptions added as the representation progresses, and determine the order and correlation of the new operation and maintenance phenomenon descriptions, parameter change descriptions and requirement adjustment descriptions.

[0024] When performing time-series analysis on the variable description portion, the generation time of each description segment is first marked with a timestamp to construct a description time series. Then, a sliding window technique is used to segment the time series, with the window size dynamically adjusted according to the description speed. Within each window, a named entity recognition model is used to extract descriptions of operational phenomena (such as "abnormal noise" and "vibration"), parameter changes (such as "temperature increase" and "pressure decrease"), and demand adjustment (such as "priority recovery" and "delayed processing"). For newly extracted descriptions, a directed graph model is used to construct relationships, where nodes represent new descriptions and edges represent causal, temporal, or inclusion relationships between descriptions. For example, if "temperature increase" appears after "abnormal noise," and both describe the same equipment component, a temporal association edge is established between them, and the relationship type "accompanying occurrence" is labeled.

[0025] Step S113: Based on the order and correlation of the newly added descriptions of operation and maintenance phenomena, parameter changes, and demand adjustments, construct an evolution path model for operation and maintenance problems, and divide the evolution path model into an initial stage, a development stage, and a stable stage. Each stage in the initial stage, development stage, and stable stage corresponds to a set of aggregated features of newly added descriptions.

[0026] Based on the order of appearance and correlation of newly added descriptions, an evolutionary path model for operation and maintenance problems is constructed. First, the newly added descriptions are arranged chronologically to form an evolutionary sequence. Then, a hierarchical clustering algorithm is used to cluster the evolutionary sequence, with clustering features including description type, correlation strength, and time interval. Based on the clustering results, the evolutionary path is divided into an initial stage, a development stage, and a stable stage. The initial stage corresponds to clusters containing core fault phenomena, with aggregation features such as fault type and initial phenomena; the development stage corresponds to clusters containing derived phenomena and parameter changes, with aggregation features such as derived phenomenon type and parameter change trend; the stable stage corresponds to clusters containing demand adjustments and final demands, with aggregation features such as demand priority and target state. The boundary of each stage is determined by calculating the silhouette coefficient of adjacent clusters; when the silhouette coefficient is less than a preset threshold, it is determined to be a stage boundary.

[0027] Step S1131: The order in which the new descriptions appear is assigned a time sequence number. The new descriptions are arranged in the order of the numbers to form a time sequence of new descriptions. Each new description in the time sequence of new descriptions corresponds to a time sequence node.

[0028] When assigning time-series numbers to the order in which new descriptions appear, the starting time is the moment the inquiry request is initiated. Each new description received is assigned a unique time-series number in the order of receipt. These new descriptions are then arranged in ascending order of their time-series numbers to form a time-series new description sequence. Each new description corresponds to a time-series node in the sequence, and the node contains attributes such as description text, timestamp, and description type. For example, the node with time-series number 1 corresponds to the first new description, the node with time-series number 2 corresponds to the second new description, and so on.

[0029] Step S1132: Analyze the relationship between the new descriptions of adjacent time series nodes in the new description sequence, and determine whether the new description of the next time series node is based on the new description of the previous time series node, or whether the new description of the next time series node is a supplement to the new description of the previous time series node.

[0030] When analyzing the relationships between adjacent time-series nodes in the newly added time-series description sequence, a method combining semantic similarity calculation and rule matching is adopted. First, the semantic similarity of the description text of adjacent nodes is calculated using a cosine similarity algorithm, converting the text into word vectors before calculating the similarity value. If the similarity value is greater than a preset threshold, the two are considered to have a semantic relationship. Then, based on operational domain knowledge rules, the type of relationship is determined. If the description of the later node further explains a detail in the description of the earlier node, it is determined to be a supplementary explanation relationship; if the description of the later node introduces a new situation or phenomenon based on the description of the earlier node, it is determined to be an expansion relationship. For example, if the earlier node describes "abnormal equipment noise" and the later node describes "the abnormal noise is an intermittent clicking sound," it is determined to be a supplementary explanation relationship.

[0031] Step S1133: Group adjacent time-series nodes with direct relationships into a node group. Each node group contains at least two time-series nodes. New descriptions within the node group focus on the details of the same operation and maintenance problem.

[0032] Adjacent time-series nodes with direct relationships (such as supplementary descriptions or expansion relationships) are grouped into a node group. First, the newly added time-series description sequence is traversed, starting from the first node, checking if there is a direct relationship between it and the next node. If so, they are grouped into a temporary node group. The relationship between the last node in the temporary node group and the next node is then checked. If a direct relationship still exists, that node is added to the temporary node group, and so on, until no direct relationship exists. Then, starting from the first node after the temporary node group, the above process is repeated to form multiple node groups. New descriptions within each node group must revolve around the same operational issue details. Thematic keywords for the node group are extracted using a topic model (such as LDA) to ensure consistency of thematic keywords within the same node group.

[0033] Step S1134: Count the number of time-series nodes contained in each node group, and calculate the information density of the newly added description in each node group. The information density is determined based on the ratio of the number of valid operation and maintenance information entries in the newly added description to the length of the description text of the newly added description.

[0034] After counting the number of time-series nodes in each node group, the information density of newly added descriptions within the node group is calculated. First, valid O&M information entries are defined as information units directly related to O&M issues, such as fault phenomena, parameter indicators, and operational attempts, contained in the description. The number of valid O&M information entries in each newly added description is extracted using regular expressions and an entity recognition model. Then, the length of the description text for each newly added description, i.e., the number of characters, is counted. The information density is the ratio of the total number of valid O&M information entries in all newly added descriptions within the node group to the total length of the description text. For example, if a node group contains 2 time-series nodes, has a total of 5 valid O&M information entries, and a total description text length of 100, then the information density is 5 / 100.

[0035] Step S1135: Based on the temporal sequence of the node groups and the trend of information density change, determine the stage division boundary of the evolution path: the first node group with information density increasing from low to high is divided into the initial stage, and the new description of the initial stage is mainly based on the preliminary description of the core operation and maintenance phenomena.

[0036] Based on the temporal sequence of node groups, the information density change rate of adjacent node groups is calculated sequentially. The information density change rate is (information density of the next node group - information density of the previous node group) / information density of the previous node group. Node groups with an information density change rate greater than a preset positive threshold are identified as information density increasing node groups. The first node group in the temporal sequence to show an increase in information density is designated as the initial stage. The newly added description of the initial stage mainly includes a preliminary description of core operational phenomena, such as equipment model, failure time, and main abnormal manifestations.

[0037] Step S1136: The node group whose information density is continuously stable and higher than the initial stage information density, and whose newly added descriptions are supplemented with details around the core phenomenon, is divided into the development stage. The newly added descriptions in the development stage include detailed information such as changes in operation and maintenance parameters and extension of related phenomena.

[0038] After determining the initial stage, the information density change trend of subsequent node groups is analyzed. Node groups whose information density change rate is within a preset small fluctuation range (e.g., the absolute value is less than a certain threshold) and whose information density value is higher than that of the initial stage are classified as development stages. New descriptions for development stages need to be supplemented with details around the core phenomenon, verified by checking the correlation between the node group's keywords and the core phenomenon. New descriptions for development stages include details such as changes in operational parameters (e.g., changes in specific values ​​of temperature, pressure, current, etc.) and extensions of related phenomena (e.g., anomalies in other components caused by the core phenomenon).

[0039] Step S1137: The node group whose information density tends to be stable and whose new descriptions no longer introduce new details, and which only confirm or adjust existing information, is divided into the stable stage. The new descriptions in the stable stage are mainly based on clear requirements and expected effects.

[0040] For node groups that have progressed beyond the development stage, if the rate of change in information density approaches zero (i.e., it tends to stabilize), and the newly added descriptions no longer extract new operational phenomena, parameter changes, or other detailed information, but only contain descriptions confirming existing information or adjusting requirements, then the node group is classified as the stable stage. New descriptions in the stable stage primarily focus on clearly defined requirements (e.g., "Needs to be fixed by XX time") and expected results (e.g., "Hopefully restored to normal operating status"). The presence of new details is determined by checking whether the newly added descriptions contain new entities or relationships.

[0041] Step S1138: Extract the common features of the newly added descriptions in all node groups in the initial stage. The common features of the newly added descriptions in all node groups in the initial stage include core phenomenon keywords, preliminary parameter range descriptions and basic requirement directions. The common features of the newly added descriptions in all node groups in the initial stage are used as the aggregated features of the initial stage.

[0042] When extracting common features from newly added descriptions across all node groups in the initial stage, keywords are first extracted from the new descriptions of each node group. The TF-IDF algorithm is used to calculate the term frequency-inverse document frequency, and the top few most frequent words are selected as candidate keywords. Then, the intersection of the candidate keywords across all node groups is taken to obtain the core phenomenon keywords. Simultaneously, preliminary parameter range descriptions, such as vague parameter descriptions like "high temperature" and "low pressure," are extracted from the new descriptions. Basic demand indications are obtained by analyzing verb and noun collocations in the descriptions, such as "needs to be checked" and "hopes to be resolved." The core phenomenon keywords, preliminary parameter range descriptions, and basic demand indications are combined to form the aggregated features of the initial stage.

[0043] Step S1139: Extract the common features of newly added descriptions in all node groups in the development stage in the same way as the initial stage aggregated feature extraction. The common features of newly added descriptions in all node groups in the development stage include keywords of detailed parameter changes, descriptions of related phenomena and extended requirements. Use the common features of newly added descriptions in all node groups in the development stage as the aggregated features of the development stage.

[0044] Following the same method as the initial stage's aggregated feature extraction, common features of newly added descriptions within all node groups in the development stage are extracted. Keywords indicating detailed parameter changes are obtained by extracting specific parameter names and trend terms from the descriptions, such as "temperature rise" and "current increase." Related phenomena are described as other abnormal phenomena associated with the core phenomenon in the description, such as "abnormal noise accompanied by vibration." Extended requirements refer to specific demands further proposed on top of the basic requirements, such as "detailed inspection procedures are needed" or "priority treatment of a certain component is desired." These common features are combined as the aggregated features for the development stage.

[0045] Step S11310: Extract the common features of newly added descriptions in all node groups in the stable stage in the same way as the initial stage aggregate features and the development stage aggregate features. The common features of newly added descriptions in all node groups in the stable stage include demand confirmation keywords, expected effect descriptions and adjustment direction statements. Use the common features of newly added descriptions in all node groups in the stable stage as the aggregate features of the stable stage.

[0046] Similarly, extract the common features of newly added descriptions within all node groups during the stable phase. Requirement confirmation keywords are words that explicitly state the requirement in the description, such as "confirmed need for component replacement" or "agree to the above solution." Expected results descriptions express the desired operational outcome, such as "the equipment will operate normally after repair" or "hope to achieve XX performance indicators." Adjustment direction descriptions explain adjustments to existing solutions or requirements, such as "adjusting priority from high to medium" or "adding XX inspection items." Combine these common features as the aggregated features for the stable phase.

[0047] Step S11311: Integrate the node group sequence, information density change curve, and aggregation characteristics of the initial stage, development stage, and stable stage to construct an evolution path model for operation and maintenance problems.

[0048] The node groups in the initial, development, and stable stages are arranged chronologically, and information density variation curves are plotted, with the horizontal axis representing the node group's chronological order and the vertical axis representing the information density value. Boundary points for each stage are marked on the curves. Simultaneously, the aggregation features of each stage are added as labels to the corresponding stage's curve. The node group order, information density variation curves, and aggregation features of each stage are integrated into a structured data model to construct an operational problem evolution path model. This operational problem evolution path model can be visualized using a directed graph, where nodes represent stages, edges represent the evolutionary relationships between stages, and node attributes include aggregation features and information density.

[0049] Step S114: Take the aggregated features of each stage in the initial stage, development stage and stable stage as the dynamic evolution features of the operation and maintenance problem, add stage identifier and occurrence time sequence identifier to each dynamic evolution feature to form a set of identified dynamic evolution features.

[0050] The aggregated features of the initial, development, and stable stages are used as dynamic evolution features of the operation and maintenance problem at different stages. A stage identifier, such as "Initial Stage," "Development Stage," or "Stable Stage," and an occurrence time sequence identifier, i.e., the time sequence range of the node group corresponding to that stage, are added to each dynamic evolution feature. For example, if the node group time sequence corresponding to the initial stage is 1-3, then the occurrence time sequence identifier is "1-3." These identified dynamic evolution features are combined to form an identified dynamic evolution feature set, stored as a dictionary data structure. The key is the stage identifier, and the values ​​are sub-dictionaries containing the aggregated features and the occurrence time sequence identifier.

[0051] Step S115: Query the knowledge update logs of all real-time updated knowledge modules in the large model knowledge base, and extract the operation and maintenance problem stage adaptation information corresponding to the update content of each real-time updated knowledge module; the operation and maintenance problem stage adaptation information includes the evolution stage that the real-time updated knowledge module can match and the corresponding aggregation features.

[0052] The system queries the knowledge update logs of all real-time updated knowledge modules in the large model knowledge base. The knowledge update logs record the update time, update content summary, and applicable maintenance problem types for each knowledge module. By parsing the "Adaptation Stage" and "Feature Tag" fields in the knowledge update logs, the system extracts the matching evolutionary stages (e.g., "Initial Stage," "Development Stage") and corresponding aggregated features (e.g., core phenomenon keywords, parameter change trends, etc.) for each real-time updated knowledge module. This information is then organized into maintenance problem stage adaptation information, using the knowledge module ID as the key and a list containing matching evolutionary stages and corresponding aggregated features as the values.

[0053] Step S116: Compare the tagged set of dynamic evolution features with the operation and maintenance problem stage adaptation information of each real-time updated knowledge module, and select real-time updated knowledge modules whose operation and maintenance problem stage adaptation information contains at least one dynamic evolution feature. Select the selected real-time updated knowledge modules as candidate real-time updated knowledge modules.

[0054] Each dynamic evolution feature in the tagged dynamic evolution feature set is compared with the operational problem stage adaptation information of each real-time updated knowledge module. During the comparison, the similarity between the aggregated features of the dynamic evolution features and the aggregated adaptation features of the knowledge module is calculated using a cosine similarity algorithm. If the similarity is greater than a preset threshold, the knowledge module is determined to contain that dynamic evolution feature. Real-time updated knowledge modules containing at least one dynamic evolution feature are selected as candidate real-time updated knowledge modules.

[0055] Step S117: Add a matching dynamic evolution feature identifier to each candidate real-time updated knowledge module, and construct an association mapping table between consultation requests and candidate real-time updated knowledge modules. The association mapping table contains consultation request identifier, candidate real-time updated knowledge module identifier, matching dynamic evolution feature, and corresponding stage identifier.

[0056] For each candidate real-time updated knowledge module, a matching dynamic evolution feature identifier is added. This identifier includes the stage identifier and aggregate feature of the dynamic evolution feature matching the knowledge module within the set of identified dynamic evolution features. Then, a mapping table is constructed between consultation requests and candidate real-time updated knowledge modules. This mapping table is a two-dimensional table structure, containing columns for consultation request identifier (uniquely identifying this consultation request), candidate real-time updated knowledge module identifier (the unique ID of the knowledge module), matching dynamic evolution feature (aggregate feature content), and corresponding stage identifier (e.g., "development stage").

[0057] Step S118: Prioritize the candidate real-time updated knowledge modules in the association mapping table. Determine the priority order based on the number of matches between each candidate real-time updated knowledge module and the dynamic evolution feature, and the time difference between the update time of the candidate real-time updated knowledge module and the consultation request reception time.

[0058] When prioritizing candidate real-time updated knowledge modules in the association mapping table, two factors are considered: the number of matches and the time difference. The number of matches is the number of dynamically evolving features matched by the knowledge module; the time difference is the difference between the knowledge module's update time and the consultation request reception time. Different weights are assigned to these two factors, for example, the weight of the number of matches is 0.6, and the weight of the time difference is 0.4. The comprehensive score of each knowledge module is calculated as follows: Comprehensive Score = Number of Matches * Weight of Number of Matches + (1 / (1+Time Difference)) * Weight of Time Difference. The candidate real-time updated knowledge modules are then sorted from highest to lowest comprehensive score to obtain the priority order.

[0059] Step S119: Select the candidate real-time updated knowledge modules in the priority ranking, retain the association mapping relationship corresponding to the candidate real-time updated knowledge modules, and form a correspondence between consultation requests with dynamic evolution feature tags and real-time updated knowledge modules.

[0060] Based on the priority ranking, a predetermined number (e.g., 1-3) of candidate real-time updated knowledge modules are selected. The association mapping relationships of these candidate real-time updated knowledge modules in the association mapping table are retained, including the consultation request identifier, knowledge module identifier, matching dynamic evolution characteristics, and corresponding stage identifier. These association mapping relationships are combined to form a correspondence between consultation requests marked with dynamic evolution characteristics and real-time updated knowledge modules, and stored as JSON format data.

[0061] Step S120: Based on the correspondence with dynamic evolution feature tags, and combined with the phased description information of the operation and maintenance issues in the consultation request, perform dynamic scenario adaptation processing on the operation and maintenance knowledge in the real-time updated knowledge module to generate dynamic scenario adaptation results containing evolution stage matching information.

[0062] Based on the correspondence marked with dynamic evolutionary features, matching, real-time updated knowledge modules are obtained. Then, combined with the stage-specific descriptions of the maintenance issues in the consultation request at different evolutionary stages, dynamic scenario adaptation processing is performed on the maintenance knowledge in the knowledge modules. Dynamic scenario adaptation processing includes knowledge filtering, knowledge recombination, and scenario-based description transformation. Knowledge filtering selects relevant knowledge entries from the knowledge modules based on specific scenario elements (such as equipment model and fault phenomena) in the stage-specific description information; knowledge recombination organizes knowledge entries according to the evolutionary stages of the maintenance issues; scenario-based description transformation converts general knowledge into specific descriptions related to the current maintenance scenario, such as replacing "equipment" with a specific equipment model. Dynamic scenario adaptation results containing evolutionary stage matching information (such as the evolutionary stage matched by each knowledge entry and the matching degree score) are generated and stored in XML format.

[0063] Step S121: Extract the identifier of the real-time updated knowledge module and the matching dynamic evolution feature stage identifier from the correspondence with dynamic evolution feature tags, and retrieve the operation and maintenance knowledge structure in the real-time updated knowledge module based on the identifier; the operation and maintenance knowledge structure includes a knowledge classification directory, stage adaptation tags and knowledge association links.

[0064] Extract the identifiers (e.g., knowledge module IDs) and matching dynamic evolution feature stage identifiers (e.g., "development stage") of real-time updated knowledge modules from the correspondence with dynamically evolved feature tags. Based on the knowledge module identifiers, retrieve the operation and maintenance knowledge structure within the real-time updated knowledge module through the knowledge base retrieval interface. The operation and maintenance knowledge structure is hierarchically organized, including a knowledge classification directory (e.g., categorized by fault type or equipment component), stage adaptation tags (marking the evolution stage of each knowledge adaptation), and knowledge association links (representing the causal and dependency relationships between knowledge items). The knowledge association links are represented by a directed graph, where nodes are knowledge item IDs and edges are association types (e.g., "cause" or "required").

[0065] Step S122: Decompose the phased description information of the operation and maintenance problem in the consultation request, and divide it into initial phase description text, development phase description text and stable phase description text according to the phase identifier of dynamic evolution characteristics; each phase description text in the initial phase description text, development phase description text and stable phase description text corresponds to a set of operation and maintenance problem details description.

[0066] When breaking down the phased descriptions of maintenance issues in consultation requests, the text is segmented into initial phase, development phase, and stable phase descriptions based on the phase identifiers and their occurrence sequence identifiers according to dynamic evolution characteristics. For example, the description text for time sequence nodes 1-3 is classified as the initial phase description text, the description text for time sequence nodes 4-6 as the development phase description text, and the description text for time sequence nodes 7-9 as the stable phase description text. Each phase description text corresponds to a set of detailed descriptions of the maintenance issue, obtained by extracting entities, attributes, and relationships from the text. For example, the detailed description of the initial phase includes equipment model, initial fault symptoms, etc., while the detailed description of the development phase includes parameter changes, derived phenomena, etc.

[0067] Step S123: Compare the detailed description of the operation and maintenance problem in the initial stage description text with the operation and maintenance knowledge with the initial stage adaptation tag in the real-time updated knowledge module, extract the initial stage knowledge subdirectory that matches the detailed description of the operation and maintenance problem in the knowledge classification directory, and record the matching points between the knowledge entries in the initial stage knowledge subdirectory and the detailed description of the operation and maintenance problem.

[0068] The detailed descriptions of operational issues in the initial stage text are compared with operational knowledge tagged with the initial stage in the real-time updated knowledge module. During the comparison, keywords in the detailed descriptions are matched with the titles and summaries of the knowledge entries. The BM25 algorithm is used to calculate text similarity, and the knowledge category directory containing the knowledge entry with the highest similarity is selected as the initial stage knowledge subdirectory. Matching points between knowledge entries and detailed descriptions of operational issues in this subdirectory—i.e., common keywords, entities, or attributes—are recorded and stored as a list, with each matching point containing the matching text and the matching location.

[0069] Step S124: In the same way as the initial stage description text processing, compare the detailed description of operation and maintenance issues in the development stage description text with the operation and maintenance knowledge with development stage adaptation tags in the real-time updated knowledge module, extract the development stage knowledge subdirectory that matches the detailed description of operation and maintenance issues in the knowledge classification directory, and record the matching points between the knowledge entries in the development stage knowledge subdirectory and the detailed description of operation and maintenance issues.

[0070] The development stage description text is processed in the same way as the initial stage description text. The detailed descriptions of operational issues in the development stage description text are compared with the operational knowledge in the real-time updated knowledge module that includes development stage-appropriate tags. The similarity is calculated using the BM25 algorithm, and matching development stage knowledge subdirectories are extracted. Matching points between knowledge entries and detailed descriptions are recorded. Matching points in the development stage focus more on details such as parameter changes and related phenomena.

[0071] Step S125: Following the same method as the initial stage description text and the development stage description text, compare the detailed description of operation and maintenance issues in the stable stage description text with the operation and maintenance knowledge with stable stage adaptation tags in the real-time updated knowledge module, extract the stable stage knowledge subdirectories that match the detailed description of operation and maintenance issues in the knowledge classification directory, and record the matching points between the knowledge entries in the stable stage knowledge subdirectories and the detailed description of operation and maintenance issues.

[0072] Similarly, the detailed descriptions of operational issues in the stable phase description text are compared with the operational knowledge tagged with stable phase in the real-time updated knowledge module. Matching stable phase knowledge subdirectories are extracted and the matching points are recorded. The matching points for the stable phase mainly include information such as requirement descriptions and expected results.

[0073] Step S126: Based on the initial stage knowledge subdirectory, the development stage knowledge subdirectory, and the stable stage knowledge subdirectory and their corresponding matching points, construct the stage knowledge adaptation link; the stage knowledge adaptation link includes the relationship between each knowledge subdirectory in the initial stage knowledge subdirectory, the development stage knowledge subdirectory, and the stable stage knowledge subdirectory, the cross-reference information of the matching points of each knowledge subdirectory, and the connection logic of operation and maintenance knowledge between different knowledge subdirectories.

[0074] Based on the knowledge subdirectories and their matching points across three stages, a stage-based knowledge adaptation link is constructed. First, the relationships between the knowledge subdirectories are analyzed. For example, the initial stage knowledge subdirectories provide basic information for the development stage knowledge subdirectories, and the development stage knowledge subdirectories provide a basis for the stable stage knowledge subdirectories, thus establishing a sequential relationship of "initial—development—stable." Then, recurring keywords or entities are extracted from the matching points of each knowledge subdirectory as cross-reference information, establishing reference relationships between matching points. Finally, based on the relationships and cross-reference information of the knowledge subdirectories, the connection logic of operational knowledge between different knowledge subdirectories is outlined, such as the logical order from fault phenomenon description to cause analysis to solution. The above relationships, cross-reference information, and connection logic are combined to form the stage-based knowledge adaptation link, represented by a directed graph data structure.

[0075] Step S1261: Arrange the knowledge subdirectories of the initial stage, development stage, and stable stage in the order of the stages, and determine the core knowledge entries of each knowledge subdirectory of the initial stage, development stage, and stable stage knowledge subdirectories; the core knowledge entries of each knowledge subdirectory are determined based on the number of matching points between the knowledge entries in the knowledge subdirectory and the corresponding stage description text.

[0076] The knowledge subcategories for the three stages are arranged in the order of initial stage, development stage, and stable stage. For each knowledge subcategory, the number of matching points between each knowledge entry and the corresponding stage's description text is counted. The knowledge entry with the most matching points is determined as the core knowledge entry for that subcategory. If multiple knowledge entries have the same number of matching points and are the most numerous, the core knowledge entry is determined by calculating the knowledge entry's authority score (based on the knowledge entry's update time, citation count, etc.), and the one with the highest authority score is the core knowledge entry.

[0077] Step S1262: Establish the association between the core entries of the initial stage knowledge subdirectory and the core entries of the development stage knowledge subdirectory; analyze the dependency between the operation and maintenance solutions contained in the core entries of the initial stage knowledge subdirectory and the detailed supplementary solutions contained in the core entries of the development stage knowledge subdirectory; determine that the operation and maintenance solutions of the core entries of the initial stage knowledge subdirectory are the basic solutions for the detailed supplementary solutions of the core entries of the development stage knowledge subdirectory, and that the detailed supplementary solutions of the core entries of the development stage knowledge subdirectory are the extension solutions for the operation and maintenance solutions of the core entries of the initial stage knowledge subdirectory.

[0078] Establish the relationships between the core entries of the knowledge subdirectories in the initial and development phases. Analyze the dependencies between the operational solutions (such as preliminary fault location methods) included in the core entries of the initial phase and the detailed supplementary solutions (such as specific parameter measurement methods) included in the core entries of the development phase. Determine the dependencies between the two by checking whether the detailed supplementary solutions reference or are based on the operational solutions. If a detailed supplementary solution is a detailed description of a step in the operational solution, then the operational solution is determined to be the basic solution, and the detailed supplementary solution is an extension solution.

[0079] Step S1263: Establish the association between the core entries of the development stage knowledge subdirectory and the core entries of the stable stage knowledge subdirectory in the same way as the association between the core entries of the initial stage knowledge subdirectory and the core entries of the development stage knowledge subdirectory. Analyze the integration relationship between the detailed schemes contained in the core entries of the development stage knowledge subdirectory and the comprehensive schemes contained in the core entries of the stable stage knowledge subdirectory. Determine that the detailed schemes of the core entries of the development stage knowledge subdirectory are components of the comprehensive schemes of the core entries of the stable stage knowledge subdirectory, and that the comprehensive schemes of the core entries of the stable stage knowledge subdirectory are a summary scheme of the detailed schemes of the core entries of the development stage knowledge subdirectory.

[0080] Establish the relationship between the core items of the development stage and the stable stage in the same manner as in step S1262. Analyze the integration relationship between the detailed solutions (such as inspection methods for each component) of the core items in the development stage and the comprehensive solutions (such as the overall solution) of the core items in the stable stage. If the comprehensive solution includes multiple steps or methods from the detailed solutions, then the detailed solutions are determined to be components of the comprehensive solution, and the comprehensive solution is a summary of the detailed solutions.

[0081] Step S1264: Extract the overlapping descriptions of the matching points in the initial stage knowledge subdirectory with the matching points in the development stage knowledge subdirectory. Use the overlapping descriptions of the matching points in the initial stage knowledge subdirectory with the matching points in the development stage knowledge subdirectory as cross-reference identifiers to establish cross-reference relationships between the matching points in the initial stage knowledge subdirectory and the matching points in the development stage knowledge subdirectory. Record the reference direction and reference basis from the matching points in the initial stage knowledge subdirectory to the matching points in the development stage knowledge subdirectory.

[0082] Extract overlapping descriptions from the matching points in the initial stage knowledge subdirectory and the development stage knowledge subdirectory, such as common equipment component names and fault phenomenon terminology. Use these overlapping descriptions as cross-reference identifiers. Establish cross-reference relationships from initial stage matching points to development stage matching points, with the reference direction pointing from the initial stage to the development stage. The reference basis is the logical relationship between the two in the evolution of operational issues; for example, the development stage matching point is a further refinement of the initial stage matching point. Record the reference direction and reference basis in the attributes of the cross-reference relationship.

[0083] Step S1265: Extract the overlapping descriptions of the corresponding matching points in the development stage knowledge subdirectory with the corresponding matching points in the stable stage knowledge subdirectory. Use the overlapping descriptions of the corresponding matching points in the development stage knowledge subdirectory as cross-reference identifiers to establish cross-reference relationships between the corresponding matching points in the development stage knowledge subdirectory and the corresponding matching points in the stable stage knowledge subdirectory. Record the reference direction and reference basis from the corresponding matching points in the development stage knowledge subdirectory to the corresponding matching points in the stable stage knowledge subdirectory.

[0084] Similarly, the overlapping descriptions of the matching points corresponding to the knowledge subdirectories of the development stage and the stable stage are extracted as cross-reference identifiers to establish cross-reference relationships between the matching points of the development stage and the stable stage. The reference direction (development stage points to stable stage) and the reference basis (e.g., the matching point of the stable stage is a requirement proposed based on the matching point of the development stage) are recorded.

[0085] Step S1266: Construct the connection logic from the initial stage knowledge subdirectory to the development stage knowledge subdirectory. The construction process is based on the basic scheme of the core entries of the initial stage knowledge subdirectory, and supplements the detailed scheme of the core entries of the development stage knowledge subdirectory, so that the basic scheme is expanded into an intermediate scheme containing detailed parameters; the source and order of the supplemented detailed parameters are recorded in the connection logic.

[0086] Construct a logical connection between the knowledge subdirectories of the initial and development stages. Starting with the basic scheme of the core items in the initial stage, supplement the basic scheme with detailed parameters (such as temperature thresholds, pressure ranges, etc.) from the detailed schemes of the core items in the development stage, thus expanding the basic scheme into an intermediate scheme containing detailed parameters. The connection logic records the source of each supplemented detailed parameter, such as references and equipment manuals in the knowledge module, as well as the order of supplementation, i.e., supplementing according to the importance and order of appearance of the parameters in the scheme.

[0087] Step S1267: Construct the connection logic from the development stage knowledge subdirectory to the stable stage knowledge subdirectory. The construction process is based on the intermediate solution of the core entries of the development stage knowledge subdirectory, and integrates the comprehensive solution of the core entries of the stable stage knowledge subdirectory, so that the intermediate solution is summarized into a complete solution including effect verification and subsequent maintenance; the integrated solution modules and the summary order are recorded in the connection logic.

[0088] Construct a logical connection between the knowledge subdirectories of the development and stable phases. Based on the intermediate solutions of the core items in the development phase, integrate the comprehensive solutions of the core items in the stable phase. For example, integrate the cause analysis from the intermediate solutions with the solutions, effect verification steps, and follow-up maintenance suggestions from the comprehensive solutions to form a complete solution including effect verification and follow-up maintenance. Record the names of the integrated solution modules and their summarization order in the connection logic. The summarization order follows the natural order of the operation and maintenance process, such as first the solution, then the effect verification, and finally the follow-up maintenance.

[0089] Step S1268: Combine the relationships between each knowledge subdirectory in the initial stage knowledge subdirectory, the development stage knowledge subdirectory, and the stable stage knowledge subdirectory, the cross-reference relationships between the corresponding matching points in the initial stage knowledge subdirectory and the development stage knowledge subdirectory, the cross-reference relationships between the corresponding matching points in the development stage knowledge subdirectory and the stable stage knowledge subdirectory, and the connection logic from the initial stage knowledge subdirectory to the development stage knowledge subdirectory and from the development stage knowledge subdirectory to the stable stage knowledge subdirectory in sequence, and add stage knowledge adaptation link node identifiers; each node identifier corresponds to a core knowledge item or a cross-reference point, forming a stage knowledge adaptation link.

[0090] The relationships between knowledge subdirectories, the cross-references of matching points, and the connection logic between knowledge subdirectories are combined according to the stage sequence. A unique stage knowledge adaptation link node identifier is added to each core knowledge item and cross-reference point, such as "Initial Core Item" or "Cross-reference Point 1". The stage knowledge adaptation link is constructed using these node identifiers and their relationships (relationships, references, and connection logic), forming a complete directed graph structure.

[0091] Step S127: Supplement the missing stage knowledge connection nodes in the stage knowledge adaptation link, and improve the association reference of knowledge entries in each knowledge subdirectory of the initial stage knowledge subdirectory, development stage knowledge subdirectory, and stable stage knowledge subdirectory.

[0092] During the knowledge adaptation process, if logical gaps or missing necessary connecting nodes are found (e.g., jumping directly from a phenomenon description to a solution, or lacking a causal analysis node), the corresponding connecting nodes for that stage are added. These added nodes can be content extracted from other knowledge items within the knowledge module, or intermediate conclusions generated based on common-sense reasoning. Simultaneously, the interconnected references between knowledge items within each knowledge subdirectory are improved by adding internal links to link related knowledge items, enhancing the coherence of the knowledge.

[0093] Step S128: Add an evolutionary stage matching identifier to the stage knowledge adaptation link. The evolutionary stage matching identifier includes the matching degree of each stage in the initial stage, development stage, and stable stage, the coverage of knowledge items in each stage, and the completeness information of the connecting nodes of each stage.

[0094] Add an evolutionary stage matching identifier to the stage-based knowledge adaptation link. For each stage, calculate the matching degree, i.e., the proportion of matching points between the knowledge subdirectory and the description text in that stage relative to the total number of detailed descriptions; the knowledge item coverage, i.e., the operational issues covered by the knowledge subdirectory (such as phenomena, causes, solutions, etc.); and the integrity information of the connecting nodes, i.e., the number of missing and supplementary connecting nodes in that stage. Combine the above information into an evolutionary stage matching identifier and add it to the attributes of the stage-based knowledge adaptation link.

[0095] Step S129: Integrate the stage knowledge adaptation links with evolution stage matching identifiers with the initial stage knowledge subdirectory, the development stage knowledge subdirectory, the stable stage knowledge subdirectory and their corresponding matching points to form a dynamic scene adaptation result containing evolution stage matching information.

[0096] The stage knowledge adaptation links with evolution stage matching identifiers are integrated with the knowledge subdirectories of the three stages and their matching points. The content of the knowledge subdirectories, matching points, stage knowledge adaptation links, and evolution stage matching identifiers are combined to form a dynamic scene adaptation result containing evolution stage matching information. This dynamic scene adaptation result is presented in HTML format for easy parsing and display by subsequent modules.

[0097] Step S130: Call the operation and maintenance Q&A generation module with knowledge iteration function in the large model, input the dynamic scenario adaptation result containing evolution stage matching information into the operation and maintenance Q&A generation module, and import the incremental operation and maintenance knowledge corresponding to the evolution stage from the real-time updated knowledge module to generate multiple rounds of preliminary Q&A content as the operation and maintenance problem evolves.

[0098] The system utilizes an operations and maintenance (O&M) Q&A generation module with knowledge iteration capabilities within the large model. This module, based on the Transformer architecture, includes an encoder and a decoder. Dynamic scene adaptation results, containing evolutionary stage matching information, are input into the encoder for feature extraction and scene modeling. Simultaneously, incremental O&M knowledge corresponding to each evolutionary stage, such as the latest fault handling cases and updated technical manuals, is imported from a real-time updated knowledge module as additional input to the decoder. Based on the input scene features and incremental knowledge, the decoder generates multiple rounds of preliminary Q&A content that evolves with the O&M issues. Each round of Q&A content corresponds to an evolutionary stage and includes fault analysis and handling suggestions for that stage.

[0099] Step S131: Start the operation and maintenance Q&A generation module with knowledge iteration function in the large model, and load the stage Q&A generation rules in the operation and maintenance Q&A generation module; the stage Q&A generation rules include the core question response logic of the initial stage Q&A, the detailed supplementary logic of the development stage Q&A, and the solution integration logic of the stable stage Q&A.

[0100] After starting the operation and maintenance Q&A generation module, the phased Q&A generation rules are loaded. These rules are stored in the module's configuration file in JSON format. The core issue response logic for the initial phase Q&A specifies the response method to core fault phenomena, such as first confirming the phenomenon and then providing a preliminary judgment. The detailed supplementary logic for the development phase Q&A requires a detailed analysis of parameter changes and derivative phenomena. The solution integration logic for the stable phase Q&A emphasizes the completeness and operability of the solution. During rule loading, syntax checks and semantic parsing are performed to ensure the rules' correctness.

[0101] Step S132: Input the initial stage knowledge subdirectory, the matching point corresponding to the initial stage knowledge subdirectory, and the initial segment of the stage knowledge adaptation link from the dynamic scene adaptation results containing evolution stage matching information into the initial stage processing unit of the operation and maintenance Q&A generation module.

[0102] The initial stage knowledge subdirectory, corresponding matching points, and the initial segment of the stage knowledge adaptation link (i.e., the part within the initial stage and the part connecting with the development stage) are extracted from the dynamic scene adaptation results and input into the initial stage processing unit of the operation and maintenance Q&A generation module. The initial stage processing unit preprocesses the input data, such as text cleaning and format conversion, to prepare for the subsequent generation of Q&A content.

[0103] Step S133: Import the incremental operation and maintenance knowledge marked with the initial stage increment from the real-time updated knowledge module. This incremental operation and maintenance knowledge marked with the initial stage increment includes the latest solutions to the operation and maintenance problems in the initial stage, tips on common misconceptions about the operation and maintenance problems in the initial stage, and related case references for the operation and maintenance problems in the initial stage.

[0104] Import incremental maintenance knowledge with initial stage incremental tags from the real-time updated knowledge module. This knowledge is obtained by filtering through the "Incremental Tag" field of the knowledge module and includes the latest solutions to initial stage maintenance problems (such as recently discovered new fault handling methods), common misconception tips (such as easily confused fault phenomena), and related case references (such as handling cases of similar faults). Convert the above incremental knowledge into structured text and integrate it with the input data of the initial stage processing unit.

[0105] Step S134: Through the initial stage processing unit, combined with the core question response logic of the stage Q&A generation rules, the knowledge entries of the initial stage knowledge subdirectory, the matching points corresponding to the initial stage knowledge subdirectory, and the incremental operation and maintenance knowledge marked with the initial stage increment are integrated to generate the first round of preliminary Q&A content for the initial stage of operation and maintenance issues.

[0106] The initial processing unit integrates the core question response logic of the phased Q&A generation rules with the knowledge entries, matching points, and incremental operation and maintenance knowledge in the initial stage knowledge subdirectory. The integration process employs an attention mechanism to ensure that the generated Q&A content focuses on the core issues. First, based on the core question response logic, the structure of the Q&A content is determined, such as phenomenon confirmation, preliminary analysis, and suggested steps. Then, relevant content is selected from the knowledge entries and incremental knowledge to populate the structure, with matching points used to ensure the relevance of the content to the current scenario. Finally, the first round of preliminary Q&A content for the initial stage is generated, in natural language paragraph format.

[0107] Step S1341: Sort the knowledge entries in the initial stage knowledge subdirectory according to their relevance to the matching points corresponding to the initial stage knowledge subdirectory. The relevance is determined based on the number of matching points corresponding to the initial stage knowledge subdirectory contained in the knowledge entry and the importance of the matching points corresponding to the initial stage knowledge subdirectory in the knowledge entry.

[0108] For all knowledge entries in the initial stage knowledge subdirectory, calculate the relevance score between each entry and its corresponding stage matching point. The score consists of two parts: first, the matching point quantity score, which is the proportion of initial stage matching points included in the knowledge entry to the total number of matching points, with a weight set to 0.5; second, the matching point importance score, which is based on preset matching point weights in the operation and maintenance domain, such as a core fault phenomenon matching point weight of 0.3, a preliminary parameter matching point weight of 0.2, and a basic requirement matching point weight of 0.1. The total weight of the matching points covered by the knowledge entry is calculated and set to 0.5. The relevance score = (number of matching points / total number of matching points) × 0.5 + total matching point weights × 0.5. The knowledge entries are then sorted from highest to lowest score.

[0109] Step S1342: Extract the top-ranked knowledge items as the core knowledge item group. The core knowledge item group includes basic solutions for the initial stage of operation and maintenance problems, explanations of the core phenomena in the initial stage of operation and maintenance problems, and common coping strategies for the initial stage of operation and maintenance problems.

[0110] Based on the ranking results, the top 3-5 highest-scoring knowledge items are selected to form a core knowledge item group. This group ensures that the items cover the three core categories of content required for the initial Q&A phase: basic solutions (such as preliminary troubleshooting steps for equipment malfunctions), explanations of core phenomena (such as potential causes of malfunctions), and common coping strategies (such as priority handling strategies in different scenarios). If any category of content is missing, it is supplemented from subsequent ranked items until the core knowledge item group fully covers all three categories.

[0111] Step S1343: Associate and annotate each knowledge entry in the core knowledge entry group with the matching point corresponding to the initial stage knowledge subdirectory, mark the text paragraph corresponding to the matching point in the description of the knowledge entry, and clarify the correspondence between the knowledge entry and the description of the operation and maintenance problem.

[0112] By using text highlighting, the text paragraphs corresponding to the initial stage matching points of each entry in the core knowledge item group are marked with a specific color (such as yellow), and labeled with annotations indicating the matching point number and the corresponding description. For example, in the knowledge item "Abnormal equipment noise may be caused by bearing wear," the paragraph "Abnormal equipment noise" is labeled "Matching point 1: Description of core fault phenomenon," forming a visual association annotation to ensure that the correspondence between the Q&A content and the maintenance issues is clear and traceable.

[0113] Step S1344: Import the incremental operation and maintenance knowledge marked with the initial stage incremental marker, and separate the latest solution text, common misconception prompt text, and related case reference text in the incremental operation and maintenance knowledge marked with the initial stage incremental marker. Each text part in the latest solution text, common misconception prompt text, and related case reference text corresponds to a set of incremental knowledge points.

[0114] Incremental maintenance knowledge marked with initial incremental tags was retrieved through the knowledge base retrieval interface. A topic segmentation algorithm was used to split this knowledge into three categories: latest solution text (troubleshooting methods and technical optimization solutions updated within the last three months), common misconception warning text (easily confused fault phenomena and incorrect operation alerts), and related case reference text (handling examples of similar equipment and similar faults). For each category, incremental knowledge points were generated through keyword extraction and semantic summarization. Each knowledge point corresponds to a specific text fragment and source identifier.

[0115] Step S1345: Compare the incremental knowledge points of the latest solution text with the basic solutions for the initial stage of operation and maintenance issues in the core knowledge item group, extract the new key points in the latest solution text that supplement the basic solutions for the initial stage of operation and maintenance issues, and integrate the new key points into the corresponding paragraphs of the basic solutions for the initial stage of operation and maintenance issues.

[0116] A semantic similarity comparison algorithm is used to compare the incremental knowledge points of the latest solution with the basic solutions in the core item group point by point. If the incremental points are content not covered by the basic solutions (such as newly added troubleshooting tools or optimized operation steps), they are identified as new points. Following the expression logic of the basic solutions, the new points are inserted into the corresponding paragraphs. For example, the usage method of the latest tools is added to the "Troubleshooting Steps" paragraph, and a "Supplementary Incremental Knowledge" label and the update time are marked.

[0117] Step S1346: Compare the incremental knowledge points of the common misconception prompt text with the common ways to deal with the initial stage of operation and maintenance issues in the core knowledge item group, and use the incremental knowledge points of the common misconception prompt text to supplement the explanation of the common ways to deal with the initial stage of operation and maintenance issues.

[0118] For common response directions in the core item groups, incremental key points of common misconceptions are matched one by one. If there are common misconceptions in a certain response direction, the misconception explanation and avoidance methods are added after the description of the direction. For example, after the response direction of "prioritize checking the circuit", add "Misconception reminder: Avoid directly measuring live parts. You must first disconnect the power and do a good job of insulation protection (incremental knowledge 2026XXXX)" to enhance the practicality and rigor of the Q&A content.

[0119] Step S1347: Compare the incremental knowledge points of the related case reference text with the core phenomenon explanations of the initial stage of the operation and maintenance problem in the core knowledge item group, select case fragments that are highly similar to the core phenomena, and supplement the phenomenon handling process in the case fragments into the core phenomenon explanations of the initial stage of the operation and maintenance problem.

[0120] Calculate the similarity between related cases and the core phenomenon, select case segments with a similarity of ≥80%, extract the processing procedures related to the core phenomenon (such as phenomenon observation and preliminary judgment basis), and add them to the core phenomenon explanation paragraph. When adding, retain key case information (such as equipment model and processing result), and mark "Case Reference" and case ID, for example, "Explanation of core phenomenon: intermittent abnormal noise of equipment (similar case ID: CASE202601XX, XX model equipment has abnormal noise, which was found to be due to insufficient bearing lubrication, and the abnormal noise was eliminated after processing)".

[0121] Step S1348: Invoke the core question response logic of the phased Q&A generation rules. This core question response logic includes the requirement that the core questions be answered first, the explanation of the phenomenon should follow the question, and the solution should be presented in a point-by-point manner.

[0122] The core issue response logic is called from the configuration file of the operation and maintenance Q&A generation module, and the order of expression is clearly defined: First, a direct response is given to the core fault phenomenon in the consultation request to confirm the rationality of the phenomenon and make a preliminary judgment; then, the core phenomenon is explained, explaining the potential causes and related logic; finally, the basic solution and the incremental supplementary points are expanded point by point, and the step numbers and key precautions are marked to ensure that the Q&A content is clearly structured and conforms to the reading habits of operation and maintenance personnel.

[0123] Step S1349: In accordance with the presentation order requirements, reorganize the core knowledge item group after association and annotation, the basic solutions for the initial stage of operation and maintenance problems after integrating the newly added key points, the common coping directions for the initial stage of operation and maintenance problems after supplementary description, and the explanation of the core phenomena of the initial stage of operation and maintenance problems with added case fragments.

[0124] The content is reorganized following the order of "Core Response - Phenomenon Explanation - Solution - Response Tips": The first part is the core response, concisely addressing the core fault phenomenon; the second part is the phenomenon explanation, integrating explanations of the core phenomenon with supplementary case studies; the third part is the basic solution, presenting the troubleshooting steps after integrating incremental key points; the fourth part is common response directions, supplementing information on common misconceptions. Transitional sentences are added during the reorganization process to ensure smooth connections between the parts.

[0125] Step S13410: During the reorganization process, first describe the core phenomena of the initial stage of the operation and maintenance problem corresponding to the core problem, then elaborate on the basic solutions for the initial stage of the operation and maintenance problem including the newly added key points, then explain the common ways to deal with the initial stage of the operation and maintenance problem after the supplementary description, and finally attach case fragments from the relevant case reference text.

[0126] The content is organized strictly according to a predetermined order. The explanation of the core phenomenon highlights the matching points with the consultation request. The basic solutions are clearly divided into points and steps. Common response directions are prioritized according to different scenarios. Case excerpts are placed at the end as supplementary explanations for easy reference by maintenance personnel. For example: "Explanation of core phenomenon: The abnormal noise you mentioned in the equipment (matching point 1) is commonly caused by bearing wear and loose parts (case reference: CASE202601XX); Solution: 1. Power off and check the tightness of the parts (incremental supplement: use a torque wrench to confirm)...; Common response directions: prioritize checking vulnerable parts and avoid operating with electricity (common mistakes:...)."

[0127] Step S13411: Improve the fluency of the reorganized content, adjust the sentence connections, standardize the terminology, and remove repetitive statements to form the first round of preliminary Q&A content for the initial stage of operation and maintenance issues.

[0128] The text employs natural language processing algorithms to optimize fluency, adding transition words such as "firstly," "secondly," and "in addition" to adjust sentence connections; it standardizes terminology based on a dictionary of standard terms in the operations and maintenance field, such as unifying "machine" as "equipment" and "repair" as "maintenance"; and it removes duplicate content (such as repeated misconception prompts and case fragments) using text deduplication algorithms, ultimately generating logically coherent, standardized, and non-redundant preliminary Q&A content for the first round, in paragraph-style natural language format to adapt to subsequent terminal display requirements.

[0129] Step S135: After generating the first round of preliminary Q&A content for the initial stage of operation and maintenance issues, input the development stage knowledge subdirectory, the matching point corresponding to the development stage knowledge subdirectory, and the development segment of the stage knowledge adaptation link from the dynamic scenario adaptation results containing evolution stage matching information into the development stage processing unit of the operation and maintenance Q&A generation module.

[0130] After generating the initial Q&A content, the development stage knowledge subdirectory, corresponding matching points, and development segments of the stage knowledge adaptation link are extracted from the dynamic scenario adaptation results and input into the development stage processing unit of the operation and maintenance Q&A generation module. The development stage processing unit preprocesses the input data, similar to the preprocessing steps of the initial stage processing unit.

[0131] Step S136: Import incremental operation and maintenance knowledge with development stage increment tags from the real-time updated knowledge module. This incremental operation and maintenance knowledge with development stage increment tags includes parameter adjustment schemes for operation and maintenance problems in the development stage, phenomenon analysis methods for operation and maintenance problems in the development stage, and risk avoidance suggestions for operation and maintenance problems in the development stage.

[0132] Import incremental operation and maintenance knowledge marked with development stage increments from the real-time updated knowledge module, including parameter adjustment schemes (such as specific adjustment values ​​proposed based on parameter changes), phenomenon analysis methods (such as fault tree analysis, fishbone diagram analysis, etc.), and risk avoidance suggestions (such as safety precautions during operation). Integrate the above incremental knowledge with the input data from the development stage processing unit.

[0133] Step S137: Through the development stage processing unit, based on the detailed supplementary logic of the stage Q&A generation rules, the knowledge entries of the development stage knowledge subdirectory, the matching points corresponding to the development stage knowledge subdirectory, and the incremental operation and maintenance knowledge marked with the development stage increment are integrated to generate the second round of preliminary Q&A content for the development stage of operation and maintenance issues.

[0134] The development stage processing unit integrates development stage knowledge items, matching points, and incremental knowledge based on the detailed supplementation logic. This detailed supplementation logic requires a thorough analysis of parameter changes and derivative phenomena in the answer content; therefore, the integration process focuses on these details. The generated second-round preliminary answer content includes interpretations of parameter changes, causal analysis of derivative phenomena, and detailed inspection steps, with a more refined text structure than the initial stage.

[0135] Step S138: After generating the second round of preliminary Q&A content for the development stage of operation and maintenance issues, input the stable stage knowledge subdirectory, the matching point corresponding to the stable stage knowledge subdirectory, and the stable segment of the stage knowledge adaptation link from the dynamic scenario adaptation results containing evolution stage matching information into the stable stage processing unit of the operation and maintenance Q&A generation module.

[0136] After generating the second round of preliminary Q&A content, extract the stable phase knowledge subdirectory, matching points, and stable segments of the phase knowledge adaptation link, and input them into the stable phase processing unit for preprocessing.

[0137] Step S139: Import the incremental operation and maintenance knowledge marked with stable phase increments from the real-time updated knowledge module. This incremental operation and maintenance knowledge marked with stable phase increments includes comprehensive solutions to stable phase operation and maintenance problems, methods for verifying the effectiveness of stable phase operation and maintenance problems, and subsequent maintenance suggestions for stable phase operation and maintenance problems.

[0138] Import incremental maintenance knowledge marked with stable phase increments, including comprehensive solutions (such as complete maintenance procedures), effectiveness verification methods (such as how to detect whether a fault has been resolved), and subsequent maintenance recommendations (such as regular inspection items and maintenance cycles). Integrate this incremental knowledge with the input data of the stable phase processing unit.

[0139] Step S1310: Through the stable phase processing unit, according to the scheme integration logic of the phase Q&A generation rules, the knowledge entries of the stable phase knowledge subdirectory, the matching points corresponding to the stable phase knowledge subdirectory, and the incremental operation and maintenance knowledge with stable phase incremental tags are integrated to generate the third round of preliminary Q&A content for the stable phase of operation and maintenance issues.

[0140] The stabilization phase processing unit integrates stabilization phase knowledge items, matching points, and incremental knowledge according to the solution integration logic. The solution integration logic requires that the Q&A content provide complete and actionable solutions; therefore, the integrated content includes detailed operational steps, effect verification indicators, and subsequent maintenance plans. A third round of preliminary Q&A content is generated, serving as the final round of multiple preliminary Q&A sessions.

[0141] Step S1311: Link the first round of preliminary Q&A content for the initial stage of the operation and maintenance problem, the second round of preliminary Q&A content for the development stage of the operation and maintenance problem, and the third round of preliminary Q&A content for the stable stage of the operation and maintenance problem in the order of the stages, add stage jump markers and knowledge iteration markers, and form multiple rounds of preliminary Q&A content as the operation and maintenance problem evolves.

[0142] The three rounds of initial Q&A content are linked in the order of initial stage, development stage, and stable stage. Stage transition markers are added at the beginning and end of each round of Q&A content, such as "Previous Round: Initial Stage Q&A" and "Next Round: Development Stage Q&A," to facilitate users switching between Q&A content at different stages. Simultaneously, a knowledge iteration marker is added to each round of Q&A content to record the incremental knowledge version and update time referenced in that round. The linked content is then combined to form multiple rounds of initial Q&A content that evolves with operational issues, and stored as an HTML page.

[0143] Step S140: Based on the initial Q&A content from multiple rounds and the feedback from the initiator of the consultation request, dynamically optimize and adjust the initial Q&A content from multiple rounds, and combine it with the incremental knowledge supplementation of the knowledge module in real time to obtain dynamically optimized Q&A content that is synchronized with the evolution of operation and maintenance issues.

[0144] The system receives feedback from the initiator of the consultation request (such as an on-site engineer) regarding multiple rounds of initial Q&A. This feedback includes acceptance of the answers, questions, and additional needs. Sentiment analysis and intent recognition are performed on the feedback to identify areas for optimization. Combined with the latest incremental knowledge from the real-time updated knowledge module, the initial Q&A content is dynamically optimized and adjusted, such as supplementing missing information, correcting errors, and refining operational steps. The final result is dynamically optimized Q&A content that evolves in sync with the operational issues.

[0145] Step S141: Receive feedback from the initiator of the consultation request regarding the content of the multiple rounds of preliminary Q&A, break down the approval statements, question statements, and supplementary requirement statements in the feedback information, and distinguish the specific stages of the multiple rounds of preliminary Q&A content corresponding to each part of the approval statements, question statements, and supplementary requirement statements.

[0146] After receiving feedback, a text classification model is used to break it down into three parts: expressions of approval, expressions of inquiry, and expressions of supplementary needs. The text classification model is based on the SVM algorithm and trained on the labeled feedback corpus. Then, through keyword matching and semantic similarity calculation, the specific stage of the multi-round preliminary Q&A content corresponding to each part is determined. For example, if the feedback mentions "the second round of analysis on temperature changes was very helpful," then the expression of approval corresponds to the development stage.

[0147] Step S142: Extract the preliminary Q&A content paragraphs corresponding to the approved statements, retain the expression logic and knowledge references of the preliminary Q&A content paragraphs, and use the retained preliminary Q&A content paragraphs as the basic retained paragraphs for dynamically optimizing the Q&A content.

[0148] Extract the initial Q&A paragraphs corresponding to the approved statements, retaining their logical structure (such as causal relationships and progressive relationships) and knowledge references (such as cited knowledge modules, cases, etc.). These retained paragraphs will serve as the foundation for dynamically optimizing the Q&A content, requiring no modification.

[0149] Step S143: For the preliminary answer paragraph corresponding to the question statement, analyze the core question points in the question statement and locate the knowledge entries in the real-time updated knowledge module referenced by the preliminary answer paragraph.

[0150] For the preliminary answer paragraphs corresponding to the question statements, interrogative word recognition and dependency parsing are used to extract the core question points. For example, the core question point of "Why does the temperature continue to rise?" is "The reason for the continuous rise in temperature". Then, by analyzing the knowledge reference tags of the paragraph, the knowledge entry ID in the real-time updated knowledge module it references is located.

[0151] Step S144: If the knowledge item is not fully associated with the core question, retrieve the latest incremental knowledge related to the core question from the real-time updated knowledge module, supplement the corresponding preliminary Q&A content paragraph with the retrieved latest incremental knowledge, and adjust the expression logic of the preliminary Q&A content paragraph.

[0152] To determine the completeness of the connection between knowledge entries and core questions, check whether the knowledge entries contain answers or related explanations for the core questions. If the connection is incomplete, retrieve the latest incremental knowledge from the incremental knowledge index of the real-time updated knowledge module using keywords related to the core questions. Add the retrieved incremental knowledge to the corresponding preliminary Q&A paragraphs, and adjust the paragraph's expression logic to ensure the supplementary content is organically integrated with the original content and logically coherent.

[0153] For example, step S1441: Locate the preliminary Q&A content paragraph corresponding to the core question, extract the knowledge entry identifier of the real-time updated knowledge module referenced in the preliminary Q&A content paragraph, and retrieve the complete text content of the knowledge entry based on the knowledge entry identifier.

[0154] Locate the preliminary answer paragraphs corresponding to the core questions, and extract the knowledge entry identifiers (such as knowledge entry IDs) from the paragraph's citation annotations. Based on the knowledge entry identifiers, retrieve the complete text content of the knowledge entry through the knowledge base interface, including the title, abstract, main text, and references.

[0155] Step S1442: Compare the text describing the core question with the complete text of the knowledge entry sentence by sentence to find the missing text segments in the knowledge entry text that do not cover the description of the core question.

[0156] The text describing the core questions is compared sentence by sentence with the complete text of the knowledge entries. During the comparison, semantic similarity at the sentence level is calculated using a Siamese network model. Sentences with similarity scores below a preset threshold are identified as not covering the core questions, and these sentences constitute the missing text segments.

[0157] Step S1443: Record the details of the core question points corresponding to the missing text segments. The details of the core question points include the types of operation and maintenance parameters involved in the core question points, the characteristics of the changes in the phenomena involved in the core question points, and the direction of the requirements involved in the core question points.

[0158] Record the details of the core questions corresponding to the missing text segments. The operation and maintenance parameter type is the specific parameter involved in the question, such as temperature, pressure, etc.; the phenomenon change characteristics are the trend of the phenomenon described by the question, such as "continuously rising" or "suddenly falling"; the demand direction is the user demand behind the question, such as "understanding the cause" or "seeking a solution".

[0159] Step S1444: Query the incremental knowledge index of the real-time updated knowledge module. The incremental knowledge index of the real-time updated knowledge module includes the keywords of the incremental knowledge, the question type corresponding to the incremental knowledge, and the knowledge update time of the incremental knowledge.

[0160] Query the incremental knowledge index of the real-time updated knowledge module. The incremental knowledge index is an inverted index that records information such as keywords of incremental knowledge, corresponding question types (e.g., cause-based, method-based), and knowledge update time. Search the index using keywords from the details of the core question (e.g., "reasons for temperature increase").

[0161] Step S1445: Construct query conditions using keywords from the details of the core question points, retrieve incremental knowledge entries that match the keywords in the incremental knowledge index of the real-time updated knowledge module, and select the incremental knowledge entries with the latest knowledge update time as target incremental knowledge.

[0162] Build query conditions using keywords from the core points of inquiry (such as the type of operation and maintenance parameters and the characteristics of phenomenon changes), and perform a Boolean search in the incremental knowledge index. Sort the search results in descending order of knowledge update time, and select the incremental knowledge entry with the most recent update time as the target incremental knowledge.

[0163] Step S1446: Retrieve the complete content of the target incremental knowledge, and break it down into the core answer text, the supporting explanation text, and the related knowledge reference text. Each text part in the core answer text, the supporting explanation text, and the related knowledge reference text corresponds to a set of incremental knowledge points.

[0164] The complete content of the target incremental knowledge is retrieved, and through text segmentation and topic recognition, it is broken down into core answer text (content that directly answers the question), supporting explanation text (evidence and reasons supporting the answer), and related knowledge reference text (other knowledge resources cited). For each text part, keyword extraction and entity recognition are used to extract the key points of incremental knowledge. For example, the key point of the core answer text is "the reason for the temperature increase is poor heat dissipation".

[0165] Step S1447: Extract the answer content corresponding to the missing text segment of the core question from the core answer text, and insert the answer content into the missing position of the preliminary question content paragraph according to the expression logic.

[0166] Extract the corresponding answer content from the core answer text that corresponds to the missing text segment of the core question, thus filling in the missing text segment. Following the logical structure of the initial answer paragraphs (e.g., phenomenon first, then cause), insert the answer content into the missing position to ensure contextual coherence.

[0167] Step S1448: Add the reasonable explanation from the explanatory text to the position following the inserted answer content, describing the technical basis and applicable scenarios of the answer content.

[0168] Add reasonable explanations (such as experimental data, theoretical analysis, etc.) from the explanatory text after the inserted answer to explain the technical basis and applicable scenarios of the answer, thereby enhancing the credibility of the answer.

[0169] Step S1449: Add the supplementary information from the related knowledge reference text to the corresponding sentence of the inserted answer content in the form of related annotations, and mark the related knowledge identifier and the reference paragraph of the related knowledge.

[0170] Supplementary information from related knowledge citations can be added as related annotations (such as footnotes) next to the corresponding sentences in the inserted answer content, along with the identifier of the related knowledge (such as the knowledge entry ID) and the cited paragraph (such as "Chapter 3, Section 2"), to facilitate further user reference.

[0171] Step S14410: Adjust the expression logic of the preliminary Q&A paragraph, changing the original expression order around the knowledge items to the expression order around the answers to the core questions, forming the adjusted Q&A paragraph.

[0172] The logic of the initial Q&A paragraphs was adjusted to shift the structure from organizing around knowledge items to addressing the core questions, such as in the sequence of "Question—Answer—Basis—Explanation." This was achieved by rearranging sentence order and adding transition words, resulting in the revised Q&A paragraphs.

[0173] Step S145: For the preliminary Q&A content paragraphs corresponding to the supplementary requirements statement, extract the description of the new operation and maintenance requirements in the supplementary requirements statement, query the incremental knowledge in the real-time updated knowledge module that matches the description of the new operation and maintenance requirements, and integrate the queried incremental knowledge into the corresponding preliminary Q&A content paragraphs.

[0174] Extract the description of the new maintenance requirements from the supplementary requirements section, such as "more detailed disassembly steps are needed." Use keyword searches to find incremental knowledge in the real-time updated knowledge module that matches the new requirements (such as detailed disassembly step instructions). Integrate this incremental knowledge into the corresponding preliminary Q&A paragraphs, supplementing the relevant content.

[0175] Step S146: Reorganize the basic retention section, the adjusted Q&A section, and the expanded supplementary requirement section according to the evolutionary stages of the operation and maintenance problem. The reorganization process ensures that the reorganized content conforms to the evolutionary logic from the initial stage to the development stage and then to the stable stage.

[0176] Following the evolutionary sequence of initial stage, development stage, and stable stage, the basic retention section, the adjusted Q&A section, and the expanded supplementary requirements section are reorganized. During the reorganization, it is ensured that the internal logic of each stage is coherent, the transition between stages is natural, and it conforms to the evolutionary logic of operation and maintenance issues from phenomenon description to cause analysis to solution.

[0177] Step S147: Perform consistency processing on the reorganized content, unify the terminology of each stage of the content in the initial stage, development stage, and stable stage, and adjust the sentence structure of each stage.

[0178] The reorganized content underwent consistency processing. A standardized terminology dictionary was used to unify synonyms across different stages, such as unifying "excessively high temperature" and "exceeding body temperature" as "abnormally high temperature." Simultaneously, sentence structure was adjusted to maintain a consistent style across all stages, such as using declarative sentences and avoiding overly long complex sentences.

[0179] Step S148: If there are unanswered questions or missing supplementary needs in the reorganized content, repeatedly retrieve incremental knowledge from the real-time updated knowledge module to supplement the reorganized content, forming dynamically optimized Q&A content that is synchronized with the evolution of operation and maintenance issues.

[0180] The reorganized content is examined and compared with the question and supplementary requirement statements in the feedback information to determine if there are any unanswered questions or uncovered supplementary requirements. If so, steps S143-S147 are repeated to retrieve incremental knowledge for supplementation until all questions and requirements are covered, resulting in dynamically optimized Q&A content.

[0181] Step S150: Feedback the dynamically optimized Q&A content to the party initiating the consultation request. At the same time, associate the dynamic evolution characteristics of the operation and maintenance problem, the multiple rounds of preliminary Q&A content, feedback information and dynamically optimized Q&A content with the knowledge iteration module of the large model knowledge base, drive the incremental update of the knowledge module in real time, and complete the intelligent dialogue Q&A process of operation and maintenance technical services.

[0182] The dynamically optimized Q&A content is fed back to the mobile operations and maintenance app of the requester via the reverse channel of the original consultation request. The format is rich text, supporting mixed text and images. Simultaneously, the dynamic evolution characteristics of the operations and maintenance issue, multiple rounds of preliminary Q&A content, feedback information, and dynamically optimized Q&A content are packaged and sent to the knowledge iteration module of the large model knowledge base via API. The knowledge iteration module analyzes the above data, extracts new knowledge points and relationships, and updates the content of the real-time knowledge module, such as adding new cases and correcting knowledge entries. After the knowledge update is completed, the entire intelligent dialogue Q&A process for operations and maintenance technical services ends.

[0183] Step S151: Identify the terminal type of the party initiating the consultation request who receives the Q&A content. The terminal type includes desktop operation and maintenance management platform, mobile operation and maintenance APP and web operation and maintenance consultation interface.

[0184] By parsing the "Terminal Type" field in the consultation request metadata, the type of terminal used by the initiator to receive the Q&A content can be identified. The value of the Terminal Type field may be "desktop", "mobile", or "web", corresponding to a desktop-based operations and maintenance management platform, a mobile operations and maintenance app, and a web-based operations and maintenance consultation interface, respectively.

[0185] Step S152: Adjust the format of the dynamically optimized Q&A content according to the content display requirements corresponding to the terminal type. The desktop maintenance management platform corresponds to a multi-stage column display format, the mobile maintenance APP corresponds to a stage-folded display format, and the web-based maintenance consultation interface corresponds to a combined text and image display format.

[0186] Based on the identified terminal type, the corresponding format conversion template is invoked to adjust the format of the dynamically optimized Q&A content. The desktop multi-stage column display format shows the Q&A content for different stages in left and right columns, with the left column serving as stage navigation and the right column as content. The mobile stage-folded display format folds the content for each stage by default, allowing users to expand it with a click. The web-based combined text and image display format supports the insertion of visual elements such as images and charts. Format adjustments are achieved through XSLT stylesheet conversion.

[0187] Step S153: Send the dynamically optimized Q&A content after adjusting the format to the terminal of the party initiating the consultation request, and wait for the terminal to return a content reception confirmation signal; the content reception confirmation signal includes the reception time, content integrity identifier and terminal display status.

[0188] The dynamically optimized Q&A content, after adjustment of its format, will be sent to the terminal of the party initiating the consultation request via HTTP protocol. A timeout period will be set; if no confirmation signal is received within the timeout period, the content will be resent. The received confirmation signal will be in JSON format, including the reception time (timestamp), content integrity indicator (such as "complete" or "missing"), and terminal display status (such as "displayed" or "not displayed").

[0189] Step S154: After receiving the content reception confirmation signal, the knowledge iteration module of the large model knowledge base is retrieved. The knowledge iteration module includes a knowledge association unit, an incremental marking unit, and a module update unit.

[0190] After receiving the content reception confirmation signal, the knowledge iteration module is invoked through the knowledge base management interface. The knowledge iteration module is the core component of the large model knowledge base, which includes a knowledge association unit (responsible for establishing associations between knowledge), an incremental marking unit (responsible for marking newly added and updated knowledge), and a module update unit (responsible for updating the content of knowledge modules).

[0191] Step S155: Input the dynamic evolution characteristics of the operation and maintenance problem into the knowledge association unit, and locate the stage knowledge catalog of the corresponding real-time updated knowledge module in the large model knowledge base according to the stage identifier of the dynamic evolution characteristics.

[0192] The dynamic evolution characteristics of operation and maintenance issues are input into the knowledge association unit. The knowledge association unit parses the stage identifiers of the dynamic evolution characteristics, such as "development stage", and locates the corresponding stage knowledge directory of the real-time updated knowledge module (such as "development stage fault analysis directory") through the directory index.

[0193] Step S156: Simultaneously input the preliminary Q&A content, feedback information, and dynamically optimized Q&A content from multiple rounds into the knowledge association unit. The knowledge association unit establishes the association relationship between the preliminary Q&A content, feedback information, and dynamically optimized Q&A content from multiple rounds and the corresponding knowledge items in the stage knowledge catalog, and generates an association mapping record.

[0194] The initial Q&A content, feedback information, and dynamically optimized Q&A content from multiple rounds are input into the knowledge association unit. The knowledge association unit establishes associations between this content and knowledge entries in the stage knowledge catalog using text similarity calculation and entity linking technology. For example, newly added solutions in the dynamically optimized Q&A content are associated with the "Solutions" section in the knowledge entries. The generated association mapping record includes fields such as content ID, knowledge entry ID, and association type.

[0195] Step S157: Input the association mapping record into the incremental marking unit. The incremental marking unit adds incremental update marks to the knowledge items in the association mapping record based on the questions and supplementary needs in the feedback information. The incremental update marks include the update priority and the direction of the update content.

[0196] The association mapping record is input into the incremental tagging unit. The incremental tagging unit analyzes the questions and supplementary needs in the feedback information to determine the update priority (e.g., high, medium, low) and update content direction (e.g., supplementing content, correcting errors, refining descriptions) of the knowledge item. For example, if multiple users have questions about a certain knowledge item in the feedback information, the update priority is high, and the update content direction is to provide supplementary explanations.

[0197] Step S158: Input the association mapping record with incremental update mark into the module update unit. The module update unit integrates the newly added knowledge in the dynamically optimized Q&A content into the corresponding stage knowledge catalog of the real-time updated knowledge module according to the update priority, supplements the missing content of the original knowledge entries in the corresponding stage knowledge catalog of the real-time updated knowledge module, and adjusts the expression logic of the original knowledge entries in the corresponding stage knowledge catalog of the real-time updated knowledge module.

[0198] The module update unit processes associated mapping records marked with incremental updates in descending order of update priority. For each record, new knowledge points are extracted from the dynamically optimized Q&A content and integrated into the corresponding stage's knowledge catalog of the real-time updated knowledge module. Missing content in existing knowledge entries is supplemented, such as adding new cases or data; the expression logic is adjusted, such as optimizing sentence order and enhancing coherence. During the update process, version control is implemented for knowledge entries, retaining historical versions.

[0199] Step S159: After completing the incremental update of the real-time knowledge module, generate an update log. The update log includes the updated knowledge item identifier, update time, associated Q&A process identifier, and a summary of the updated content.

[0200] After completing the incremental update, an update log is generated. The update log records the identifier (ID) of each updated knowledge entry, the update time (timestamp), the associated Q&A process identifier (a unique ID for this Q&A session), and a summary of the updated content (e.g., "Supplementary analysis of the causes of temperature anomalies"). The update log is stored in the knowledge base's log database for auditing and backtracking.

[0201] Step S1510: Associate and store the update log with all the information in this Q&A process, mark the completion of the intelligent dialogue Q&A process for this operation and maintenance technical service, and enter the preparation state for receiving the next consultation request.

[0202] The update log, along with all information from this Q&A process, including consultation requests, dynamic evolution characteristics, multi-round Q&A content, and feedback information, is stored in the database using a Q&A process identifier. The status of this Q&A process is marked as "completed" in the system status table. Then, relevant caches and temporary variables are reset, and the system enters a state ready to receive the next consultation request, listening for new requests from the API interface.

[0203] During the data collection phase, privacy-sensitive data, such as user identity information and device serial numbers, is processed using federated learning technology. Each terminal device trains the data locally, generates model parameter updates, and uploads them to the central server via an encrypted channel. The central server aggregates these parameter updates without accessing the original data. Homomorphic encryption is used to encrypt the transmitted model parameters to ensure they are not leaked during transmission. When storing data, differential privacy technology is applied to privacy-sensitive fields, adding noise that meets a preset privacy budget to prevent data from being associated with specific individuals, thus achieving data anonymization and preventing privacy leaks.

[0204] Figure 2 This application illustrates an intelligent dialogue and Q&A system 100 for operation and maintenance technical services based on a large model knowledge base, comprising a processor 1001, a memory 1003, and program code stored in the memory 1003. The processor 1001 executes the program code to implement the steps of the intelligent dialogue and Q&A method for operation and maintenance technical services based on a large model knowledge base. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the intelligent dialogue and Q&A system 100 for operation and maintenance technical services based on a large model knowledge base may further include a transceiver 1004. The transceiver 1004 can be used for data interaction between this intelligent dialogue and Q&A system for operation and maintenance technical services based on a large model knowledge base and other intelligent dialogue and Q&A systems for operation and maintenance technical services based on a large model knowledge base, such as sending and / or receiving data. It should be noted that in actual scheduling, the transceiver 1004 is not limited to one, and the structure of this intelligent dialogue and Q&A system 100 for operation and maintenance technical services based on a large model knowledge base does not constitute a limitation on the embodiments of this application.

[0205] The memory 1003 is used to store program code for executing the embodiments of this application, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing method embodiments.

[0206] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application, without departing from the technical concept of this application, also fall within the protection scope of the embodiments of this application.

Claims

1. An intelligent dialogue answering method for operation and maintenance technical service based on a large model knowledge base, characterized in that, The method includes: Receive consultation requests for operation and maintenance technical services, synchronously capture the dynamic evolution characteristics of operation and maintenance issues in the consultation requests, associate them with real-time updated knowledge modules in the large model knowledge base that match the dynamic evolution characteristics, and obtain the correspondence between consultation requests marked with dynamic evolution characteristics and real-time updated knowledge modules. Based on the correspondence with dynamic evolution feature tags, and combined with the phased description information of operation and maintenance issues in the consultation request, the operation and maintenance knowledge in the real-time updated knowledge module is dynamically adapted to the scenario, and dynamic scenario adaptation results containing evolution stage matching information are generated. Call the operation and maintenance Q&A generation module with knowledge iteration function in the large model, input the dynamic scenario adaptation result containing evolution stage matching information into the operation and maintenance Q&A generation module, and import the incremental operation and maintenance knowledge corresponding to the evolution stage from the real-time updated knowledge module to generate multiple rounds of preliminary Q&A content as the operation and maintenance problem evolves. Based on the initial Q&A content from multiple rounds and the feedback from the parties initiating the consultation requests, the initial Q&A content from multiple rounds was dynamically optimized and adjusted. Combined with the incremental knowledge supplementation of the knowledge module in real time, the dynamically optimized Q&A content was obtained in sync with the evolution of operation and maintenance issues. The dynamic optimization of the Q&A content is fed back to the party that initiated the consultation request. At the same time, the dynamic evolution characteristics of the operation and maintenance problem, the content of multiple rounds of preliminary Q&A, the feedback information and the dynamic optimization of the Q&A content are associated with the knowledge iteration module of the large model knowledge base, driving the incremental update of the knowledge module in real time, and completing the intelligent dialogue Q&A process of operation and maintenance technical services. The process of receiving consultation requests for operation and maintenance technical services involves simultaneously capturing the dynamic evolution characteristics of the operation and maintenance issues within the consultation requests, associating them with real-time updated knowledge modules in the large model knowledge base that match the dynamic evolution characteristics, and obtaining the correspondence between consultation requests marked with dynamic evolution characteristics and real-time updated knowledge modules, including: Extract the text describing the operation and maintenance issues from the consultation request, split the core issue description section and the supplementary explanation description section from the text, and distinguish between the fixed description section of the core issue description section and the variable description section of the supplementary explanation description section. A time-series analysis was performed on the variable representation part to extract the descriptions of operational phenomena, parameter changes, and requirement adjustments that were added as the representation progressed. The order and correlation of the newly added descriptions of operational phenomena, parameter changes, and requirement adjustments were then determined. Based on the order and correlation of the newly added descriptions of operation and maintenance phenomena, parameter changes, and demand adjustments, an evolution path model of operation and maintenance problems is constructed. The evolution path model is divided into an initial stage, a development stage, and a stable stage. Each stage in the initial stage, development stage, and stable stage corresponds to a set of aggregated features of the newly added descriptions. The aggregated features of each stage in the initial stage, development stage and stable stage are used as the dynamic evolution features of the operation and maintenance problem. A stage identifier and an occurrence sequence identifier are added to each dynamic evolution feature to form a set of identified dynamic evolution features. Query the knowledge update logs of all real-time updated knowledge modules in the large model knowledge base, and extract the operation and maintenance problem stage adaptation information corresponding to the update content of each real-time updated knowledge module. This operation and maintenance problem stage adaptation information includes the evolution stage that the real-time updated knowledge module can match and the corresponding aggregation features. The set of labeled dynamic evolution features is compared with the operation and maintenance problem stage adaptation information of each real-time updated knowledge module. Real-time updated knowledge modules whose operation and maintenance problem stage adaptation information contains at least one dynamic evolution feature are selected and selected as candidate real-time updated knowledge modules. Add a matching dynamic evolution feature identifier to each candidate real-time updated knowledge module, and construct an association mapping table between consultation requests and candidate real-time updated knowledge modules. The association mapping table contains consultation request identifier, candidate real-time updated knowledge module identifier, matching dynamic evolution feature, and corresponding stage identifier. The candidate real-time updated knowledge modules in the association mapping table are prioritized and sorted according to the number of matches between each candidate real-time updated knowledge module and the dynamic evolution feature, and the time difference between the update time of the candidate real-time updated knowledge module and the consultation request reception time. Select the top-priority candidate real-time updated knowledge modules, retain the associated mapping relationship corresponding to the candidate real-time updated knowledge modules, and form a correspondence between consultation requests with dynamic evolution feature tags and real-time updated knowledge modules; The correspondence based on dynamically evolving feature tags, combined with the phased description information of the operation and maintenance issues in the consultation request, performs dynamic scenario adaptation processing on the operation and maintenance knowledge in the real-time updated knowledge module, generating dynamic scenario adaptation results containing evolutionary stage matching information, including: Extract the identifier of the real-time updated knowledge module and the matching dynamic evolution feature stage identifier from the correspondence with the dynamic evolution feature label. Based on the identifier, retrieve the operation and maintenance knowledge structure in the real-time updated knowledge module. The operation and maintenance knowledge structure includes a knowledge classification directory, stage adaptation tags and knowledge association links. The information describing the operational issues in the consultation request is broken down into stages. Based on the stage identifiers of dynamic evolution characteristics, the initial stage description text, the development stage description text, and the stable stage description text are divided. Each stage description text in the initial stage description text, the development stage description text, and the stable stage description text corresponds to a set of detailed descriptions of the operational issues. The detailed description of the operation and maintenance issues in the initial stage description text is compared with the operation and maintenance knowledge with the initial stage adaptation tag in the real-time updated knowledge module. The initial stage knowledge subdirectory that matches the detailed description of the operation and maintenance issues is extracted from the knowledge classification directory, and the matching points between the knowledge entries in the initial stage knowledge subdirectory and the detailed description of the operation and maintenance issues are recorded. Following the same approach as the initial stage description text, the detailed descriptions of operation and maintenance issues in the development stage description text are compared with the operation and maintenance knowledge with development stage adaptation tags in the real-time updated knowledge module. The development stage knowledge subdirectories that match the detailed descriptions of operation and maintenance issues are extracted from the knowledge classification directory, and the matching points between the knowledge entries in the development stage knowledge subdirectories and the detailed descriptions of operation and maintenance issues are recorded. Following the same approach as the initial stage description text and the development stage description text, the detailed description of operation and maintenance issues in the stable stage description text is compared with the operation and maintenance knowledge with stable stage adaptation tags in the real-time updated knowledge module. The stable stage knowledge subdirectories that match the detailed description of operation and maintenance issues in the knowledge classification directory are extracted, and the matching points between the knowledge entries in the stable stage knowledge subdirectories and the detailed description of operation and maintenance issues are recorded. Based on the knowledge subdirectories of the initial stage, development stage, and stable stage, and their corresponding matching points, a stage knowledge adaptation link is constructed. The stage knowledge adaptation link includes the relationship between each knowledge subdirectory in the knowledge subdirectories of the initial stage, development stage, and stable stage, the cross-reference information of the matching points of each knowledge subdirectory, and the connection logic of operation and maintenance knowledge between different knowledge subdirectories. Supplement the missing knowledge connection nodes in the knowledge adaptation chain of the supplementary stage, and improve the association and reference of knowledge entries in each knowledge subdirectory of the initial stage knowledge subdirectory, the development stage knowledge subdirectory, and the stable stage knowledge subdirectory. Add evolution stage matching identifiers to the stage knowledge adaptation link. The evolution stage matching identifiers include the matching degree of each stage in the initial stage, development stage, and stable stage, the coverage of knowledge items in each stage, and the completeness information of the connecting nodes of each stage. The stage knowledge adaptation links with evolution stage matching identifiers are integrated with the initial stage knowledge subdirectory, the development stage knowledge subdirectory, the stable stage knowledge subdirectory and their corresponding matching points to form a dynamic scene adaptation result containing evolution stage matching information.

2. The method according to claim 1, wherein, The aforementioned call to the maintenance Q&A generation module with knowledge iteration functionality within the large model involves inputting dynamic scenario adaptation results containing evolution stage matching information into the maintenance Q&A generation module. Simultaneously, it imports incremental maintenance knowledge corresponding to the evolution stage from the real-time updated knowledge module, generating multiple rounds of preliminary Q&A content that evolves with the maintenance issues, including: Start the operation and maintenance Q&A generation module with knowledge iteration function in the large model, and load the stage Q&A generation rules in the operation and maintenance Q&A generation module. The stage Q&A generation rules include the core question response logic of the initial stage Q&A, the detailed supplementary logic of the development stage Q&A, and the solution integration logic of the stable stage Q&A. The initial stage knowledge subdirectory, the matching point corresponding to the initial stage knowledge subdirectory, and the initial segment of the stage knowledge adaptation link in the dynamic scene adaptation result containing the evolution stage matching information are input into the initial stage processing unit of the operation and maintenance Q&A generation module. Import incremental operation and maintenance knowledge marked with the initial stage increment from the real-time updated knowledge module. This incremental operation and maintenance knowledge marked with the initial stage increment includes the latest solutions to operation and maintenance problems in the initial stage, tips on common misconceptions about operation and maintenance problems in the initial stage, and related case references for operation and maintenance problems in the initial stage. Through the initial stage processing unit, combined with the core question response logic of the stage Q&A generation rules, the knowledge items of the initial stage knowledge subdirectory, the matching points corresponding to the initial stage knowledge subdirectory, and the incremental operation and maintenance knowledge marked with the initial stage incremental mark are integrated to generate the first round of preliminary Q&A content for the initial stage of operation and maintenance issues. After generating the first round of preliminary Q&A content for the initial stage of operation and maintenance issues, the development stage knowledge subdirectory, the matching point corresponding to the development stage knowledge subdirectory, and the development segment of the stage knowledge adaptation link from the dynamic scenario adaptation results containing evolution stage matching information are input into the development stage processing unit of the operation and maintenance Q&A generation module. Import incremental operation and maintenance knowledge marked with development stage increments from the real-time updated knowledge module. This incremental operation and maintenance knowledge marked with development stage increments includes parameter adjustment schemes for operation and maintenance problems in the development stage, phenomenon analysis methods for operation and maintenance problems in the development stage, and risk avoidance suggestions for operation and maintenance problems in the development stage. Through the development stage processing unit, based on the detailed supplementary logic of the stage Q&A generation rules, the knowledge entries of the development stage knowledge subdirectory, the matching points corresponding to the development stage knowledge subdirectory, and the incremental operation and maintenance knowledge marked with the development stage increment are integrated to generate the second round of preliminary Q&A content for the development stage of operation and maintenance issues. After generating the second round of preliminary Q&A content for the development stage of operation and maintenance issues, the stable stage knowledge subdirectory, the matching point corresponding to the stable stage knowledge subdirectory, and the stable segment of the stage knowledge adaptation link from the dynamic scenario adaptation results containing evolution stage matching information are input into the stable stage processing unit of the operation and maintenance Q&A generation module. Import incremental operation and maintenance knowledge marked with stable phase increments from the real-time updated knowledge module. This incremental operation and maintenance knowledge marked with stable phase increments includes comprehensive solutions to operation and maintenance problems in the stable phase, methods for verifying the effectiveness of operation and maintenance problems in the stable phase, and suggestions for subsequent maintenance of operation and maintenance problems in the stable phase. Through the stable phase processing unit, the knowledge entries of the stable phase knowledge subdirectory, the matching points corresponding to the stable phase knowledge subdirectory, and the incremental operation and maintenance knowledge marked with the stable phase increment are integrated according to the scheme integration logic of the phase Q&A generation rules to generate the third round of preliminary Q&A content for the stable phase of operation and maintenance issues. The initial Q&A content for the initial stage of the operation and maintenance problem, the second initial Q&A content for the development stage of the operation and maintenance problem, and the third initial Q&A content for the stable stage of the operation and maintenance problem are linked in the order of the stages. Stage jump markers and knowledge iteration markers are added to form multiple rounds of initial Q&A content as the operation and maintenance problem evolves.

3. The method of claim 1, wherein the method further comprises: The process involves dynamically optimizing and adjusting the initial Q&A content based on multiple rounds of preliminary answers and feedback from the initiator of the consultation request. This is combined with real-time updates to the knowledge module to supplement incremental knowledge, resulting in dynamically optimized Q&A content that evolves in sync with the operational issues. This includes: The system receives feedback from the initiator of the consultation request regarding the content of multiple rounds of preliminary Q&A. It breaks down the feedback information into the parts of acceptance statements, questions, and supplementary needs statements, and distinguishes the specific stages of the multiple rounds of preliminary Q&A content corresponding to each part of the acceptance statements, questions, and supplementary needs statements. Extract the preliminary Q&A content paragraphs corresponding to the approved statements, retain the expression logic and knowledge references of the preliminary Q&A content paragraphs, and use the retained preliminary Q&A content paragraphs as the base retained paragraphs for dynamically optimizing the Q&A content; For the preliminary answer paragraphs corresponding to the question statements, analyze the core questions in the question statements and locate the knowledge entries in the real-time updated knowledge module referenced by the preliminary answer paragraphs. If the knowledge item is not fully associated with the core question, retrieve the latest incremental knowledge related to the core question from the real-time updated knowledge module, add the latest incremental knowledge to the corresponding preliminary Q&A content paragraph, and adjust the expression logic of the preliminary Q&A content paragraph. For the preliminary Q&A content paragraphs corresponding to the supplementary requirements statement, extract the description of the new operation and maintenance requirements in the supplementary requirements statement, query the incremental knowledge in the real-time updated knowledge module that matches the description of the new operation and maintenance requirements, and integrate the queried incremental knowledge into the corresponding preliminary Q&A content paragraphs. The basic retention section, the adjusted Q&A section, and the expanded supplementary requirements section are reorganized according to the evolutionary stages of the operation and maintenance issues. The reorganization process ensures that the reorganized content conforms to the evolutionary logic from the initial stage to the development stage and then to the stable stage. The reorganized content was made consistent in its expression, and the terminology used in the initial stage, development stage, and stable stage was standardized. The sentence structure of each stage was also adjusted. If there are unanswered questions or missing supplementary needs in the reorganized content, the incremental knowledge from the real-time updated knowledge module will be repeatedly retrieved to supplement the reorganized content, forming dynamically optimized Q&A content that is synchronized with the evolution of operation and maintenance issues.

4. The method of claim 1, wherein the method further comprises: The process of dynamically optimizing Q&A content and feeding it back to the party initiating the consultation request, while simultaneously associating the dynamic evolution characteristics of the operation and maintenance problem, multiple rounds of preliminary Q&A content, feedback information, and dynamically optimized Q&A content with the knowledge iteration module of the large model knowledge base, drives the incremental updates of the knowledge module in real time, and completes the intelligent dialogue Q&A process for operation and maintenance technical services, including: Identify the type of terminal from which the initiator of the consultation request receives the Q&A content. This type of terminal includes desktop operation and maintenance management platform, mobile operation and maintenance APP and web operation and maintenance consultation interface. Based on the content display requirements corresponding to the terminal type, the format of the dynamically optimized Q&A content is adjusted. The desktop operation and maintenance management platform corresponds to a multi-stage column display format, the mobile operation and maintenance APP corresponds to a stage-folded display format, and the web-based operation and maintenance consultation interface corresponds to a combined text and image display format. The dynamically optimized Q&A content with adjusted format is sent to the terminal of the party initiating the consultation request, and the terminal returns a content reception confirmation signal, which includes the reception time, content integrity identifier and terminal display status. After receiving the content reception confirmation signal, the knowledge iteration module of the large model knowledge base is retrieved. This knowledge iteration module includes a knowledge association unit, an incremental marking unit, and a module update unit. The dynamic evolution characteristics of operation and maintenance issues are input into the knowledge association unit. Based on the stage identifier of the dynamic evolution characteristics, the stage knowledge catalog of the corresponding real-time updated knowledge module in the large model knowledge base is located. The initial Q&A content, feedback information, and dynamically optimized Q&A content from multiple rounds are simultaneously input into the knowledge association unit. The knowledge association unit establishes the association relationship between the initial Q&A content, feedback information, and dynamically optimized Q&A content from multiple rounds and the corresponding knowledge items in the stage knowledge catalog, and generates association mapping records. The association mapping record is input into the incremental marking unit. The incremental marking unit adds incremental update marks to the knowledge items in the association mapping record based on the questions and supplementary needs in the feedback information. The incremental update marks include the update priority and the direction of the update content. The associated mapping record with incremental update mark is input into the module update unit. The module update unit integrates the newly added knowledge in the dynamically optimized Q&A content into the corresponding stage knowledge catalog of the real-time updated knowledge module according to the update priority, supplements the missing content of the original knowledge entries in the corresponding stage knowledge catalog of the real-time updated knowledge module, and adjusts the expression logic of the original knowledge entries in the corresponding stage knowledge catalog of the real-time updated knowledge module. After completing the incremental update of the real-time knowledge module, an update log is generated. The update log includes the updated knowledge item identifier, update time, associated Q&A process identifier, and a summary of the updated content. The update log is associated with and stored along with all information from this Q&A process, marking the completion of the intelligent dialogue Q&A process for this operation and maintenance technical service, and entering the preparation state for receiving the next consultation request.

5. The method of claim 1, wherein the method further comprises: Based on the order and correlation of the newly added descriptions of operational phenomena, parameter changes, and demand adjustments, an evolutionary path model for operational problems is constructed. This model is divided into an initial stage, a development stage, and a stable stage. Each stage corresponds to a set of aggregated features of the newly added descriptions, including: The order in which new descriptions appear is assigned a time sequence number, and the new descriptions are arranged in order of number to form a time sequence of new descriptions. Each new description in the time sequence of new descriptions corresponds to a time sequence node. Analyze the relationship between the new descriptions of adjacent time series nodes in the new description sequence, and determine whether the new description of the next time series node is based on the new description of the previous time series node, or whether the new description of the next time series node is a supplementary description of the new description of the previous time series node. Adjacent time-series nodes with direct relationships are grouped into a node group. Each node group contains at least two time-series nodes. New descriptions within a node group focus on the details of the same operational issue. The number of time-series nodes in each node group is counted, and the information density of newly added descriptions in each node group is calculated. The information density is determined based on the ratio of the number of valid operation and maintenance information entries in the newly added description to the length of the description text of the newly added description. Based on the temporal sequence of node groups and the trend of information density change, the stage division boundary of the evolution path is determined: the first node group with information density increasing from low to high is divided into the initial stage, and the new description of the initial stage is mainly based on the preliminary description of core operation and maintenance phenomena. The node group whose information density is consistently stable and higher than that of the initial stage, and whose newly added descriptions are detailed around the core phenomenon, is divided into the development stage. The newly added descriptions in the development stage include detailed information such as changes in operation and maintenance parameters and extensions of related phenomena. The nodes whose information density is stable and whose new descriptions no longer introduce new details, and which only confirm or adjust existing information, are classified as the stable stage. New descriptions in the stable stage are mainly based on clear needs and expected effects. Extract the common features of the newly added descriptions in all node groups in the initial stage. The common features of the newly added descriptions in all node groups in the initial stage include core phenomenon keywords, preliminary parameter range descriptions and basic requirements. Use the common features of the newly added descriptions in all node groups in the initial stage as the aggregated features of the initial stage. Following the same method as the initial stage aggregated feature extraction, common features of newly added descriptions in all node groups during the development stage are extracted. These common features include keywords of detailed parameter changes, descriptions of related phenomena, and directions of extended needs. These common features of newly added descriptions in all node groups during the development stage are used as the aggregated features of the development stage. Following the same method as extracting aggregated features in the initial stage and the development stage, common features of newly added descriptions in all node groups in the stable stage are extracted. The common features of newly added descriptions in all node groups in the stable stage include keywords for demand confirmation, descriptions of expected effects, and statements of adjustment direction. The common features of newly added descriptions in all node groups in the stable stage are used as aggregated features of the stable stage. By integrating the node group sequence, information density change curves, and aggregation characteristics of the initial, development, and stable stages, an evolution path model for operation and maintenance problems is constructed.

6. The intelligent dialogue Q&A method for operation and maintenance technical services based on a large model knowledge base according to claim 1, characterized in that, The construction of a stage-specific knowledge adaptation link based on the initial stage knowledge subdirectory, the development stage knowledge subdirectory, the stable stage knowledge subdirectory, and their corresponding matching points includes: Arrange the knowledge subdirectories of the initial stage, development stage, and stable stage in the order of the stages, and determine the core knowledge items of each knowledge subdirectory of the initial stage, development stage, and stable stage knowledge subdirectories. The core knowledge items of each knowledge subdirectory are determined based on the number of matching points between the knowledge items in that knowledge subdirectory and the corresponding stage description text. Establish the relationship between the core entries of the knowledge subdirectory in the initial stage and the core entries of the knowledge subdirectory in the development stage. Analyze the dependency between the operation and maintenance solutions contained in the core entries of the knowledge subdirectory in the initial stage and the detailed supplementary solutions contained in the core entries of the knowledge subdirectory in the development stage. Determine that the operation and maintenance solutions of the core entries of the knowledge subdirectory in the initial stage are the basic solutions for the detailed supplementary solutions of the core entries of the knowledge subdirectory in the development stage, and the detailed supplementary solutions of the core entries of the knowledge subdirectory in the development stage are the extension solutions for the operation and maintenance solutions of the core entries of the knowledge subdirectory in the initial stage. Following the same approach as establishing the association between the core entries of the knowledge subdirectory in the initial stage and the core entries in the knowledge subdirectory in the development stage, establish the association between the core entries of the knowledge subdirectory in the development stage and the core entries of the knowledge subdirectory in the stable stage. Analyze the integration relationship between the detailed schemes contained in the core entries of the knowledge subdirectory in the development stage and the comprehensive schemes contained in the core entries of the knowledge subdirectory in the stable stage. Determine that the detailed schemes of the core entries of the knowledge subdirectory in the development stage are components of the comprehensive schemes of the core entries of the knowledge subdirectory in the stable stage, and that the comprehensive schemes of the core entries of the knowledge subdirectory in the stable stage are a summary of the detailed schemes of the core entries of the knowledge subdirectory in the development stage. Extract the overlapping descriptions from the matching points of the initial stage knowledge subdirectory and the matching points of the development stage knowledge subdirectory. Use the overlapping descriptions from the matching points of the initial stage knowledge subdirectory and the matching points of the development stage knowledge subdirectory as cross-reference identifiers. Establish the cross-reference relationship between the matching points of the initial stage knowledge subdirectory and the matching points of the development stage knowledge subdirectory. Record the reference direction and reference basis from the matching points of the initial stage knowledge subdirectory to the matching points of the development stage knowledge subdirectory. Extract the overlapping descriptions from the matching points of the knowledge subdirectories of the development stage and the matching points of the knowledge subdirectories of the stable stage. Use the overlapping descriptions from the matching points of the knowledge subdirectories of the development stage and the matching points of the knowledge subdirectories of the stable stage as cross-reference identifiers. Establish the cross-reference relationship between the matching points of the knowledge subdirectories of the development stage and the matching points of the knowledge subdirectories of the stable stage. Record the reference direction and reference basis from the matching points of the knowledge subdirectories of the development stage to the matching points of the knowledge subdirectories of the stable stage. The logic for connecting the initial stage knowledge subdirectory to the development stage knowledge subdirectory is constructed. The construction process is based on the basic scheme of the core entries of the initial stage knowledge subdirectory, and supplements the detailed scheme of the core entries of the development stage knowledge subdirectory. This expands the basic scheme into an intermediate scheme that includes detailed parameters. The source and order of the supplemented detailed parameters are recorded in the connection logic. The logic for connecting the knowledge subdirectories of the development stage to the knowledge subdirectories of the stable stage is constructed. The construction process is based on the intermediate solutions of the core items of the knowledge subdirectories of the development stage, and integrates the comprehensive solutions of the core items of the knowledge subdirectories of the stable stage. The intermediate solutions are summarized into a complete solution that includes effect verification and subsequent maintenance. The integrated solution modules and the summarization order are recorded in the connection logic. The relationships between each knowledge subdirectory in the initial stage, development stage, and stable stage knowledge subdirectories, the cross-references between the corresponding matching points in the initial stage and development stage knowledge subdirectories, the cross-references between the corresponding matching points in the development stage and stable stage knowledge subdirectories, and the connection logic from the initial stage to the development stage and from the development stage to the stable stage knowledge subdirectories are combined in sequence, and stage knowledge adaptation link node identifiers are added. Each node identifier corresponds to a core knowledge item or a cross-reference point, forming a stage knowledge adaptation link.

7. The intelligent dialogue Q&A method for operation and maintenance technical services based on a large model knowledge base according to claim 2, characterized in that, The initial stage processing unit, combined with the core question response logic of the stage Q&A generation rules, integrates the knowledge entries of the initial stage knowledge subdirectory, the corresponding matching points of the initial stage knowledge subdirectory, and the incremental operation and maintenance knowledge marked with the initial stage incremental tag, to generate the first round of preliminary Q&A content for the initial stage of operation and maintenance issues, including: The knowledge entries in the initial stage knowledge subdirectory are sorted according to their relevance to the matching points corresponding to the initial stage knowledge subdirectory. The relevance is determined based on the number of matching points corresponding to the initial stage knowledge subdirectory contained in the knowledge entry and the importance of the matching points corresponding to the initial stage knowledge subdirectory in the knowledge entry. The top-ranked knowledge items are extracted as the core knowledge item group. The core knowledge item group includes basic solutions for the initial stage of operation and maintenance problems, explanations of the core phenomena in the initial stage of operation and maintenance problems, and common ways to deal with the initial stage of operation and maintenance problems. Each knowledge item in the core knowledge item group is associated with the matching point corresponding to the initial stage knowledge subdirectory. The text paragraphs corresponding to the matching points in the description of the knowledge item are marked to clarify the correspondence between the knowledge item and the description of the operation and maintenance problem. Import incremental operation and maintenance knowledge marked with initial stage increments, and separate the latest solution text, common misconception prompt text, and related case reference text from the incremental operation and maintenance knowledge marked with initial stage increments. Each text part in the latest solution text, common misconception prompt text, and related case reference text corresponds to a set of incremental knowledge points. Compare the incremental knowledge points of the latest solution text with the basic solutions for the initial stage of operation and maintenance issues in the core knowledge item group, extract the new key points in the latest solution text that supplement the basic solutions for the initial stage of operation and maintenance issues, and integrate the new key points into the corresponding paragraphs of the basic solutions for the initial stage of operation and maintenance issues. Compare the incremental knowledge points of the common misconception prompt text with the common ways to deal with the initial stage of operation and maintenance problems in the core knowledge item group, and use the incremental knowledge points of the common misconception prompt text to supplement the explanation of the common ways to deal with the initial stage of operation and maintenance problems. Compare the incremental knowledge points of the reference text of the related cases with the explanation of the core phenomena in the initial stage of the operation and maintenance problem in the core knowledge item group, select case fragments that are highly similar to the core phenomena, and add the phenomenon handling process in the case fragments to the explanation of the core phenomena in the initial stage of the operation and maintenance problem. The core question response logic of the Q&A generation rules in the call phase includes the following requirements for the order of expression: answering core questions first, explaining phenomena closely following questions, and expanding solutions point by point. In accordance with the requirements of the order of presentation, the core knowledge item group after association and annotation, the basic solutions for the initial stage of operation and maintenance problems after integrating the newly added key points, the common coping directions for the initial stage of operation and maintenance problems after supplementary description, and the explanation of the core phenomena of the initial stage of operation and maintenance problems with added case fragments are reorganized. During the reorganization process, the core phenomena of the initial stage of the operation and maintenance problem corresponding to the core issue are explained first, then the basic solutions for the initial stage of the operation and maintenance problem, including the newly added key points, are elaborated, then the common ways to deal with the initial stage of the operation and maintenance problem after the supplementary description are explained, and finally, case fragments from the relevant case reference text are attached. The reorganized content was processed to improve fluency, sentence transitions were adjusted, terminology was standardized, and repetitive statements were removed to form the first round of preliminary Q&A content for the initial stage of operation and maintenance issues.

8. A smart dialogue and Q&A system for operation and maintenance technical services based on a large model knowledge base, characterized in that, The method includes a processor and a computer-readable storage medium storing machine-executable instructions, which, when executed by the processor, implement the intelligent dialogue and Q&A method for operation and maintenance technical services based on a large model knowledge base as described in any one of claims 1-7.

Images