Dam monitoring data semantic fusion and intelligent query method based on knowledge graph

Abstract

The invention relates to a dam monitoring data semantic fusion and intelligent query method based on a knowledge graph, and the method comprises the steps: carrying out the unified semantic modeling of a dam structure, a monitoring object, a monitoring index, an operation condition and an abnormal event through constructing a knowledge graph ontology model facing a dam monitoring service; according to the method, standardized preprocessing is carried out on multi-source monitoring data, the multi-source monitoring data are mapped into entities, relations and time sequence attributes in a knowledge graph, semantic alignment and fusion of cross-data-source monitoring objects are achieved, and on the basis, an incremental updating mechanism is adopted to carry out dynamic maintenance on the knowledge graph; according to the method, the intelligent query and association reasoning of the monitoring data are realized under the semantic constraint condition, and meanwhile, the interpretable semantic association path is output, so that the understandability, association analysis capability and engineering application value of the dam monitoring data are improved, and the method is suitable for the operation safety monitoring and management of different types of reservoir dams.

Description

Technical Field

[0001] This application relates to the field of water conservancy project safety monitoring and intelligent information processing technology, and in particular to a method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs. Background Technology

[0002] To ensure the long-term safe operation of the dam, multiple types of monitoring systems are usually deployed in the project to continuously observe the state parameters of the dam body and its attached slopes, such as deformation, seepage, stress, strain, and temperature, and generate a large amount of monitoring data.

[0003] With the development of sensing and automated data acquisition technologies, dam safety monitoring is increasingly characterized by diversified monitoring types, large data volumes, and complex data sources. Existing dam monitoring systems typically include structural displacement monitoring data, seepage monitoring data, environmental condition data, and unstructured information such as inspection records and design documents. These data differ significantly in data format, time scale, spatial scale, and semantic expression, making direct correlation and comprehensive utilization difficult.

[0004] Currently, the management and analysis of dam monitoring data in engineering practice mainly relies on relational databases or monitoring platforms for storage and display. The analysis methods are mostly focused on threshold judgment and trend analysis for single measuring points or single indicators. While these methods can reflect local operating conditions to some extent, they lack the ability to systematically express and analyze the inherent relationships between different monitoring objects, the overall laws governing deformation evolution, and their influence on operating conditions.

[0005] On the other hand, dam operation and management also involve a large amount of implicit engineering experience and professional knowledge, such as dam segment division, structural stress characteristics, typical deformation modes and their inducing conditions. This kind of knowledge often exists in the form of textual specifications, design instructions or the experience of engineering personnel, and has not yet formed a unified expression system with real-time monitoring data, making it difficult for computer systems to directly understand and utilize it.

[0006] In terms of data querying and information retrieval, existing systems typically employ retrieval methods based on fields or table structures. Engineers need to clearly define the data source, data table, and query conditions, making it difficult to quickly obtain the required information through semantic methods. When analyzing the relationship between a specific abnormal deformation and related structural parts, historical operating conditions, or past events, manual comparison and experience-based judgment are often relied upon, resulting in low efficiency and difficulty in ensuring consistent results.

[0007] In recent years, knowledge graph technology has shown certain advantages in the organization of knowledge in complex systems, semantic modeling, and associative reasoning. It can perform unified modeling and semantic association of multi-source heterogeneous data through entity-relationship-attribute structures. However, in the field of dam safety monitoring, existing research focuses mainly on the application exploration of single data types. There is a lack of systematic knowledge graph models built specifically for the characteristics of dam monitoring operations, as well as a lack of mature technical solutions for deep semantic fusion of monitoring data and engineering knowledge for intelligent querying and analysis.

[0008] Therefore, how to perform unified semantic modeling of multi-source monitoring data and engineering knowledge of dams on the basis of the existing monitoring system, realize the correlation expression between data, and support intelligent query and analysis for engineering applications remains a technical problem that urgently needs to be solved in the field of dam safety monitoring and information management. Summary of the Invention

[0009] This application provides a knowledge graph-based semantic fusion and intelligent query method for dam monitoring data to address the problems in existing dam safety monitoring systems, such as complex multi-source monitoring data types, inconsistent semantic expressions, implicit data relationships that are difficult to utilize effectively, lack of systematic integration between monitoring data and engineering knowledge, reliance on manual experience for query methods, and insufficient intelligent analysis capabilities.

[0010] Firstly, a method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs is provided, which includes the following steps: S1, Multi-source monitoring data acquisition and standardization preprocessing steps: collect dam operation monitoring data from different monitoring systems, and sequentially perform data quality control, time scale alignment, unit unification and semantic labeling on the monitoring data to form a standardized monitoring dataset containing spatial attributes of monitoring points, unified time labels and semantic categories of monitoring indicators. S2, the steps for constructing the knowledge graph ontology model for dam monitoring: construct a knowledge graph ontology model for dam safety monitoring business. The ontology model defines at least the dam structure entity, the monitoring object entity, the monitoring indicator entity, and the operating condition or abnormal event entity. It also predefines the structural membership relationship, monitoring observation relationship, and operating condition influence relationship between the entities, and sets engineering semantic constraint rules to constrain the legality of entity relationships. S3, based on the multi-constraint judgment rule, the monitoring data semantic mapping and entity alignment fusion step, according to the knowledge graph ontology model, maps the standardized monitoring data into entities, relations and time series attributes in the knowledge graph; When data from different monitoring systems point to the same candidate monitoring object, entity alignment and fusion processing is performed according to the following joint judgment rules: a) The spatial distance between the candidate entity and the preset spatial distance threshold is less than the target spatial distance threshold; b) The similarity of the monitoring indicator time series corresponding to the candidate entity within the preset time window is greater than the preset time series similarity threshold; c) The candidate entities have the same structural affiliation in the dam's structural hierarchy; Only when all of the above determination rules are met simultaneously, the candidate entities are determined to be the same actual monitoring object, and entity merging and attribute fusion processing are performed. S4, Incremental storage and versioned dynamic update steps for knowledge graphs. The completed knowledge graph is stored and managed. When new monitoring data is accessed, incremental updates are performed only on newly added or changed entities, relationships and time series attributes. At the same time, a corresponding version record is generated for each update to achieve traceable management of the knowledge graph evolution process. S5, Semantic subgraph construction and constrained association query steps based on ontology constraints, parses the query requirements of engineers into a semantic query expression containing target entity type, relation type constraints and time or operating condition conditions, constructs a query subgraph that satisfies the semantic constraints in the knowledge graph, performs entity retrieval, relation traversal and association reasoning only within the scope of the query subgraph, and outputs the query results and the semantic association paths supporting the query results.

[0011] In some embodiments, the multi-source monitoring data collected in step S1 includes at least one or more of the following: structural deformation monitoring data, seepage monitoring data, environmental condition monitoring data, and engineering inspection or operation management text data. The different types of monitoring data come from independently set monitoring subsystems or management systems, and the monitoring subsystems differ in terms of data acquisition frequency, data organization format, time stamping method, and semantic description method. The structural deformation monitoring data is used to characterize the displacement, settlement, stress, or strain changes of the dam body or foundation during operation; the seepage monitoring data is used to characterize the seepage pressure, head, or seepage status information at the dam body, foundation, or contact surface; the environmental condition monitoring data is used to characterize external conditions such as water level, rainfall, and temperature during reservoir operation; and the engineering inspection or operation management text data is used to describe abnormal phenomena, changes in operating status, or historical event information recorded during manual inspections. By unifying the data from different sources, with different structures and different semantic expressions into the data collection and preprocessing process in step S1, a complete and fusionable data foundation is provided for subsequent semantic mapping, entity alignment and association analysis based on knowledge graphs, thereby avoiding the problem of fragmented monitoring information caused by heterogeneous data sources.

[0012] In some embodiments, during the data quality control process in step S1, outlier identification processing based on historical statistical characteristics is performed on the numerical monitoring data. Specifically, this includes: for the same monitoring indicator of the same monitoring object, statistically analyzing the distribution characteristics of its observation data during the historical operation phase, and calculating the corresponding statistical parameters; comparing the currently collected monitoring value with the historical statistical characteristics, and when the deviation of the current monitoring value from the historical statistical characteristics exceeds a preset statistical threshold, the monitoring value is determined to be abnormal monitoring data. Monitoring data that is determined to be abnormal is not deleted directly, but is retained by setting an abnormality label and is processed as monitoring data with abnormal attributes in subsequent semantic mapping, entity alignment and association reasoning processes, so that abnormal information can be explicitly expressed and identified and utilized by subsequent analysis processes. Through the above-mentioned outlier identification and labeling mechanism, the impact of sensor errors, environmental interference, or occasional failures on the semantic fusion results of knowledge graphs can be reduced without disrupting the continuity of the original monitoring data, thereby improving the reliability of the semantic expression of monitoring data and the credibility of engineering analysis.

[0013] In some embodiments, during the time scale alignment process in step S1, a unified time reference axis is constructed for monitoring data with inconsistent sampling frequencies or incomplete overlap of sampling time points, and monitoring data from different monitoring systems are mapped onto the unified time reference axis through interpolation or resampling. Specifically, for monitoring data with a high sampling frequency, the time resolution is reduced by resampling to keep it consistent with the unified time axis; for monitoring data with a low sampling frequency or missing measurements, the monitoring value at the corresponding time point on the unified time axis is estimated by interpolation. By aligning the time scales as described above, different monitoring indicators can have a unified time benchmark during semantic fusion and correlation analysis. This ensures that data across monitoring systems and indicators can be compared, correlated, and inferred in the same time dimension, avoiding semantic correlation bias caused by time inconsistency.

[0014] In some embodiments, the knowledge graph ontology model constructed in step S2 includes at least the dam structure entity, dam section, slope, gallery, or basic structure unit, used to describe the compositional relationship of the dam project in terms of spatial structure and functional hierarchy; the monitoring object entity includes at least monitoring points, sensors, or survey lines, used to describe the spatial carrying unit of monitoring data and its collection source; and the monitoring index entity includes at least displacement, settlement, seepage pressure, stress, strain, or temperature, used to uniformly describe the data types collected by different monitoring systems. By uniformly defining the above entity types in the ontology model and clarifying the structural affiliation, monitoring and observation relationships, and operational condition influence relationships among the entities, the monitoring data can accurately correspond to its engineering semantic role when mapped to the knowledge graph. This ensures the consistency between the knowledge graph structure and the actual business logic of the dam project and provides a stable semantic foundation for subsequent semantic-based queries and analyses.

[0015] In some embodiments, in step S3, the time series data corresponding to the monitoring indicator is attached as an attribute to the observation relationship node between the monitoring object entity and the monitoring indicator entity. The time series data is used to explicitly describe the evolution process of the monitoring indicator over time and includes timestamp information and corresponding monitoring value information. By associating time series attributes with observation relationship nodes, the knowledge graph can not only express the static structural relationships between monitored objects, but also the dynamic change characteristics of monitoring data in the time dimension, thereby supporting the time series analysis and trend judgment of dam structural deformation, seepage status or environmental response. This approach avoids the semantic fragmentation problem caused by storing time-series data in isolation in external databases, making time evolution information an organic part of the semantic structure of the knowledge graph and improving the efficiency of utilizing time features in subsequent associative reasoning processes.

[0016] In some embodiments, during the entity alignment and fusion process in step S3, when multiple data sources provide different attribute information for the same monitoring object, a type-based attribute fusion strategy is adopted for the fused monitoring object entity. Specifically, for deterministic attributes such as spatial location and structural affiliation, the data source credibility is prioritized; for numerical monitoring attributes, a weighted fusion process is performed according to preset weights to generate fused attribute values. By distinguishing different types of attributes and adopting corresponding fusion strategies, the attribute distortion problem caused by simple coverage or averaging is avoided. This ensures the uniqueness of entities while maximizing the retention of effective information contained in multi-source monitoring data, thereby improving the accuracy and engineering applicability of entity attribute expression after fusion.

[0017] In some embodiments, during the incremental update process in step S4, the time series attributes corresponding to the monitoring data are updated in an append-only manner, and the newly collected monitoring data are appended to the existing time series in chronological order. During the appending process, the timestamps are deduplicated and the order is verified. By adopting the above-mentioned append-only update method, the overall reconstruction of existing time series data is avoided, thereby reducing the computational overhead during the system update process. At the same time, the continuity and consistency of time series data are guaranteed, providing stable data support for subsequent trend analysis and backtracking analysis based on historical data.

[0018] In some embodiments, in step S5, when constructing the query subgraph based on the semantic query expression, only entity nodes and relation edges that simultaneously satisfy entity type constraints, relation type constraints, and time conditions or operating conditions are retained, and entities and relations that do not satisfy the constraints are pruned during the graph traversal process. By constructing constrained query subgraphs in advance during the query phase, indiscriminate traversal across the entire knowledge graph is avoided, thereby reducing the computational complexity of related queries and improving the matching degree between query results and engineering semantic requirements. This makes the query process more efficient and the results more focused on engineering-related objects.

[0019] In some embodiments, while outputting the query results, a semantic association path consisting of the monitored object, monitoring indicators, operating conditions or abnormal events is further output. The semantic association path is used to describe the entities involved in the formation of the query results and the semantic connection relationships between them. The semantic association path is not only used to display query results, but also to support engineers in understanding and judging the formation mechanism of anomaly monitoring results. This enables engineers to trace the related factors that cause anomalies based on the semantic association path, thereby improving the interpretability and engineering credibility when conducting anomaly analysis and operational decisions based on monitoring data.

[0020] The beneficial effects of the technical solution provided in this application include: 1. By constructing a knowledge graph ontology model for dam monitoring operations, a unified semantic definition is made for elements such as structural information, monitoring objects, monitoring indicators, operating conditions and abnormal events. This enables monitoring data from different systems and formats to be organized and associated under the same semantic framework, solving the problems of scattered storage, semantic inconsistency and difficulty in comprehensive utilization of monitoring data in existing technologies. 2. By mapping monitoring data to entities, relationships, and attributes, and explicitly establishing semantic relationships between monitoring points and structural parts, monitoring indicators, operating conditions, and abnormal events, the originally implicit data relationships are clearly expressed, thereby improving the ability to trace and analyze the deformation evolution process and its influencing factors. 3. By introducing incremental update mechanisms, consistency checks, and version control strategies, the knowledge graph can maintain semantic consistency while continuously receiving monitoring data, and supports the retrospective and comparative analysis of historical states, overcoming the problem of difficulty in tracing historical states after data updates in existing systems. 4. Based on the semantic structure and ontology constraint rules of knowledge graphs, semantic queries of monitoring data are realized, enabling engineers to obtain the required information based on high-level semantic conditions such as structural parts, time conditions, operating conditions or abnormal characteristics, reducing reliance on data table structure and manual experience, and improving query efficiency and accuracy. 5. By performing semantically constrained associative reasoning on the knowledge graph, a comprehensive analysis is conducted on the relationship between abnormal deformation and related monitoring indicators, structural parts and operating conditions, and a clear semantic association path is output, making the analysis results interpretable and improving engineers' ability to understand the causes of anomalies. 6. The method described in this invention does not rely on specific monitoring equipment or software platforms. It can be deployed on the basis of existing dam monitoring systems and supports the expansion of new monitoring indicators, new structural units and new business scenarios. It is applicable to different types of reservoir dams and slope engineering and has strong engineering promotion value. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the overall process of the knowledge graph-based dam monitoring data semantic fusion and intelligent query method of the present invention; Figure 2 A schematic diagram of the data acquisition and standardized preprocessing process for multi-source dam monitoring data; Figure 3 A schematic diagram of entity types and hierarchical structure for the ontology model of the dam monitoring knowledge graph; Figure 4 A schematic diagram illustrating the semantic relationships between monitoring objects, structural parts, and monitoring indicators; Figure 5 A schematic diagram illustrating the process of integrating semantic mapping and knowledge graph into monitoring data; Figure 6 This is a schematic diagram of multi-source monitoring data fusion based on entity alignment and relationship fusion; Figure 7 This is a schematic diagram of the knowledge graph storage structure and dynamic update mechanism. Figure 8 This is a schematic diagram of an intelligent query process based on semantic constraints. Figure 9 This is a schematic diagram of abnormal deformation association reasoning and semantic path output; Figure 10 This is a schematic diagram illustrating the application scenario of the present invention in dam operation safety monitoring. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] This application provides a knowledge graph-based semantic fusion and intelligent query method for dam monitoring data. It can solve the problems in existing dam safety monitoring systems, such as complex multi-source monitoring data types, inconsistent semantic expressions, implicit data relationships that are difficult to utilize effectively, lack of systematic integration between monitoring data and engineering knowledge, reliance on manual experience for query methods, and insufficient intelligent analysis capabilities.

[0025] This invention aims to construct a knowledge graph model for dam safety monitoring operations, enabling unified semantic modeling and relational expression of monitoring data and engineering information from different sources and structural forms. This allows for explicit description and management of the relationships between monitoring objects, monitoring indicators, structural parts, operating conditions, and abnormal events. Based on this, semantic fusion and dynamic updating of monitoring data are achieved, providing structured support for the comprehensive analysis of monitoring information.

[0026] Meanwhile, by introducing semantic query and associative reasoning mechanisms, this invention enables engineers to intelligently query and analyze the dam's operating status, deformation behavior, and related factors based on monitoring semantics and engineering cognitive needs, thereby improving the understandability and usability of monitoring data and enhancing the information support capabilities for dam operation safety management.

[0027] Please see Figures 1-10 This invention proposes a semantic fusion and intelligent query method for dam monitoring data based on knowledge graphs. This method constructs a structured knowledge expression system by performing unified semantic modeling on multi-source dam monitoring data and engineering knowledge, and on this basis, realizes semantic query and association analysis for engineering applications.

[0028] (I) Overall framework and process of the methodology The present invention proposes a knowledge graph-based method for semantic fusion and intelligent querying of dam monitoring data. This method addresses the actual business needs of dam operation safety monitoring and constructs an overall technical framework based on monitoring data, supported by engineering knowledge, and aimed at semantic association and intelligent querying. This method achieves step-by-step processing of monitoring information from "data layer acquisition" to "semantic layer understanding" and then to "application layer query analysis" by performing unified semantic modeling and association organization on multi-source dam monitoring data.

[0029] From an overall process perspective, the method includes multiple functional stages such as data acquisition, semantic modeling, fusion and updating, and query analysis. These stages are logically interconnected and form a closed-loop operating system through a unified semantic expression mechanism to support the continuous access and dynamic evolution of dam monitoring data.

[0030] Specifically, the method includes the following steps: S1: Multi-source monitoring data acquisition and standardized preprocessing steps are used to uniformly collect, quality control and format standardize dam operation monitoring data from different monitoring systems and data sources, so as to provide a consistent data foundation for subsequent semantic modeling; S2: Steps for constructing a knowledge graph ontology model in the field of dam monitoring, which is used to define core semantic elements such as dam structure, monitoring objects, monitoring indicators, operating conditions and abnormal events, and to establish a semantic relationship framework between the elements, forming a unified semantic model for monitoring business. S3: The step of semantic mapping and knowledge graph fusion of monitoring data is used to map the preprocessed monitoring data to the knowledge graph ontology model according to the preset semantic rules, so that the multi-source monitoring data can be associated and fused at the semantic level and form a structured knowledge representation. S4: Knowledge graph storage and dynamic update steps, used to store and manage the completed knowledge graph, and automatically perform incremental updates and consistency maintenance of the graph content when new monitoring data arrives or engineering information changes, so as to ensure the synchronization of semantic information with the actual operating status of the dam; S5: Intelligent query and association analysis steps based on semantic understanding, used to retrieve, combine and analyze monitoring data and its associated semantic relationships based on the knowledge graph and according to the query needs of engineers or system analysis needs, and output result information and its association paths that match the query conditions.

[0031] Through the coordinated operation of the above steps, the method can achieve unified semantic organization and intelligent utilization of multi-source monitoring data and engineering knowledge of dams without changing the existing monitoring system hardware architecture, thereby providing systematic information support for dam operation safety monitoring, status analysis and auxiliary decision-making.

[0032] (II) Multi-source monitoring data acquisition and standardized preprocessing (S1) In step S1, the multi-source monitoring data generated during the dam operation are collected and preprocessed in a unified and standardized manner to eliminate the differences in data structure, time scale and quality between different monitoring systems and data sources, so as to provide a consistent data foundation for subsequent semantic modeling and fusion.

[0033] 1. Multi-source monitoring data acquisition The multi-source monitoring data includes at least one or more of the following types: (1) Structural response monitoring data, including continuous time series data such as displacement, settlement, strain, and stress of the dam body and slope; (2) Seepage and environmental monitoring data, including data reflecting operating conditions such as seepage pressure, water level, rainfall, and temperature; (3) Spatial attribute data, including the spatial coordinates of the monitoring point, the dam section to which it belongs, the structural part and the section number; (4) Engineering management and text data, including inspection records, operation logs, anomaly descriptions and historical event information.

[0034] Various types of data may come from different monitoring subsystems or management systems, and their sampling frequency, data format and time base may differ.

[0035] 2. Standardized Data Representation To achieve unified processing, any monitoring point In time The observed data are represented in vector form: in, : indicates the first Each monitoring point at time The monitoring data vector; : Monitoring point number, used to distinguish monitoring objects in different spatial locations; : Time variable, corresponding to the sampling time of the monitoring data; : No. Each monitoring point at time The Observed values ​​of various monitoring indicators; : Monitoring indicator serial number, such as displacement, seepage pressure, stress, etc.; : Monitoring indicator dimensions.

[0036] 3. Data quality control and outlier handling To address potential outliers, noise, or sensor drift issues in the monitoring data, anomaly detection and quality control are implemented for numerical monitoring data.

[0037] In one implementation, anomaly identification rules based on statistical characteristics are used for any monitoring indicator. Calculate its historical statistical characteristics: in, : No. The first monitoring point Historical averages of the monitoring indicators; : No. The first monitoring point Historical standard deviation of the monitoring indicators; : The number of samples in historical monitoring data.

[0038] When the monitored value meets: in, When the preset threshold parameter is not used, the data point is marked as abnormal data and is removed, corrected or labeled according to engineering rules.

[0039] 4. Time Scale Alignment and Resampling Since the sampling frequencies of different monitoring indicators may be inconsistent, time scale alignment processing is performed on the monitoring data to ensure the temporal consistency of subsequent semantic fusion.

[0040] Assume the system's unified timeline is: in, A unified timeline set; The first in the unified timeline Each time point; The number of time nodes in a unified timeline.

[0041] For data that was not sampled on a uniform time axis, interpolation or resampling methods are used to calculate its time signature. The corresponding value on: in, Mapping to a unified timeline Estimated values ​​of monitoring indicators; : Time mapping function, used for interpolation or resampling, to transform the original time series to a uniform time scale.

[0042] 5. Data normalization and dimensional unification To eliminate the differences in dimensions and numerical ranges among different monitoring indicators, the numerical data are normalized.

[0043] In one implementation, a linear normalization method is used: in, : Normalized monitoring indicator values; : No. Historical minimum values ​​of the monitoring indicators; : No. The historical maximum value of the monitoring indicators.

[0044] 6. Structured preprocessing of text data For text-based data such as inspection records and operation logs, structured processing is performed using predefined semantic rules. Specifically, this includes: Extract descriptive objects (such as dam section number, structural parts); Extract behavioral or state descriptions (such as "abnormal leakage" or "increased displacement"); Extract time information and align it with a unified timeline.

[0045] The processed text data is converted into structured semantic units, providing a foundation for subsequent semantic mapping and knowledge graph fusion.

[0046] 7. Preprocessing result output After the above processing, a standardized monitoring dataset is formed: Among them, spatial attributes include the spatial location information and structural affiliation of the monitoring points; Time stamps: Time markers on a unified timeline; Semantic tags: Semantic classification information used for subsequent knowledge graph mapping.

[0047] This dataset serves as input for subsequent steps S2 and S3, enabling unified semantic modeling and fusion of multi-source monitoring data.

[0048] (III) Construction of the ontology model for the knowledge graph of dam monitoring (S2) In step S2, in response to the problems of diverse monitoring objects, complex data semantics, and implicit engineering knowledge in dam safety monitoring operations, a knowledge graph ontology model for the dam monitoring field is constructed to perform unified semantic modeling and structured expression of monitoring data and related engineering information.

[0049] The knowledge graph ontology model is used to define the types, attributes, and interrelationships of various semantic elements in the field of dam monitoring, providing a unified semantic framework for the semantic mapping and fusion of monitoring data.

[0050] 1. Principles of Knowledge Graph Ontology Modeling The construction of the dam monitoring knowledge graph ontology model follows these principles: (1) Engineering semantic consistency principle: The entity types and relation definitions in the ontology should be consistent with the actual business semantics of the dam project to avoid the disconnect between abstract concepts and engineering practice; (2) Hierarchical organization principle: The hierarchical entity classification structure is used to describe the subordinate relationship between the dam structure and the monitoring objects; (3) Scalability principle: Support the continuous expansion of new monitoring indicators, new structural units and new business scenarios; (4) Decoupling principle from data: The ontology model only describes the semantic structure and does not depend on specific monitoring data instances, which facilitates reuse and evolution.

[0051] 2. Overall formal representation of the ontology model In one implementation, the dam monitoring knowledge graph ontology is represented as a quintuple structure: in, : Knowledge graph ontology model for dam monitoring; : Entity set, used to represent various objects in the field of dam monitoring; : A set of relations used to represent semantic relationships between entities; : A collection of attributes used to describe the attribute information of an entity; : A set of hierarchical and inheritance relationships used to describe the hierarchical structure between entities; : A set of semantic constraint rules used to ensure the consistency and effectiveness of the ontology model.

[0052] 3. Definition of Entity Sets ( ) The entity set It includes at least the following core entities: (1) Dam structural entity Engineering components used to describe dams, including but not limited to: in, : A collection of dam structural entities; "Dam": refers to an overall hydraulic structure; "Dam section": refers to the segmented structural unit of the dam; "Slope": refers to the slope structure related to the dam body or foundation; "Corridor": refers to a passageway inside or attached to a dam; "Foundation": refers to the foundation or basic structural unit of a dam.

[0053] The structural entities are organized through hierarchical relationships, such as "dam section belongs to dam" and "slope belongs to dam section".

[0054] (2) Entity of the monitored object The monitoring object entity is used to describe various object units participating in the safety monitoring of water conservancy projects. It is the basic entity set for monitoring data collection, semantic fusion, and correlation reasoning, and its definition is as follows: in, The monitored entity set includes, but is not limited to, the following three core entities; "Monitoring point": This refers to the specific monitoring location deployed on the dam body, dam foundation, bank slope, or ancillary structures. It serves as the spatial unit for carrying various types of monitoring data. Monitoring points are used to characterize the geometric location, structural attributes, and corresponding monitoring effect types of the monitored objects within the engineering structure. They can be associated with spatial semantic attributes such as dam section, structural location, elevation, and coordinate information. "Sensor": refers to the sensing device used to collect monitoring data, which is the direct acquisition unit of the monitoring effect quantity.

[0055] Based on the type of bulk effect quantity in water conservancy project safety monitoring, sensor entities can be further subdivided into: - Deformation monitoring sensors: including but not limited to GNSS, surveying robots, displacement gauges, inclinometers, plumb lines, tension lines, hydrostatic levels, etc., used to acquire displacement, settlement and tilt information of dam body and foundation; - Seepage monitoring sensors: including piezometers, manometers, weirs, temperature fiber optics, etc., used to acquire information related to seepage pressure, water head, and seepage channels; - Stress, strain and temperature monitoring sensors: including stress gauges, strain gauges, thermometers, etc., used to reflect the stress state and thermal effects of a structure; - Vibration monitoring sensors: including strong-motion meters, accelerometers, speedometers, etc., used to acquire dynamic response characteristics under earthquake or operational conditions. "Measurement line": refers to a measurement line composed of multiple monitoring points arranged according to a certain spatial or structural logical relationship. It is used to describe the continuous monitoring relationship along the dam axis, dam height direction or specific structural parts, and supports deformation trend analysis, distribution feature identification and cross-point correlation reasoning.

[0056] In the process of modeling the monitored objects, a configuration relationship or binding relationship is established between the monitoring point entities and the sensor entities to clarify the installation location of the sensors in space, the type of monitoring effect quantity, and the corresponding relationship of their data sources. At the same time, the monitoring point entities can be further assigned to the survey line entities to express the spatial organization structure of multi-point collaborative monitoring.

[0057] In addition, to comprehensively depict the operational status of water conservancy projects, this invention also supports the following extended types of data association within the monitoring object system: Environmental monitoring information for project operation, including upstream and downstream water levels, rainfall, temperature, and other environmental factors; Unstructured data generated from inspections and video surveillance includes text records, images, video, and audio information. The aforementioned unstructured data can be associated with monitoring points, sensors, or engineering event entities through semantic annotation to assist in anomaly identification and comprehensive analysis.

[0058] (3) Monitoring indicator entities Used to describe the type of monitoring quantity, including but not limited to: in, : The set of entities for monitoring indicators; "Displacement": indicates the amount of change in the position of a structure in space; "Settlement": refers to the vertical displacement component of a structure; "Seepage pressure": refers to the seepage pressure inside a structure or foundation; "Stress": refers to the stress state inside a structure; "Strain": Indicates the degree of deformation of a structure under stress; "Temperature": refers to the temperature parameter of a structure or environment.

[0059] The monitoring indicator entity is used to uniformly describe the data types collected from different monitoring points.

[0060] (4) Operating conditions and event entities Used to describe the dam's operational status and abnormal situations, including: in, : A collection of operating conditions and event entities; "Water level conditions": indicates the water level status during reservoir operation; "Rainfall event": refers to a rainfall event that affects the operation of the dam; "Dispatch Status": Indicates the reservoir's dispatch and operation status; “Abnormal Deformation Event”: refers to abnormal deformation behavior identified during the monitoring process.

[0061] This type of entity is used to express the impact of external conditions and abnormal behavior on the dam's condition.

[0062] 4. Definition of relation sets ( ) Based on the entity set, construct a set of semantic relationships between entities. including but not limited to: Structural relationships: Belongs to (monitoring point → structural part) Includes (dam → dam section) Monitoring Relationship: Observation (monitoring points → monitoring indicators) Data collected from (sensor → monitoring point) Influence relationship: Impact (Operating conditions → Deformation behavior) Correlation (Abnormal Events → Monitoring Indicators) Timing relationships: Occurred within (Event → Time Range) The above relationship definition enables the formation of a traceable semantic network between different entities.

[0063] 5. Definition of entity attribute set ( ) To enhance the expressive power of entities, attribute sets are defined for various types of entities: Structural entity attributes: structural type, spatial location, design parameters; Attributes of the monitored object: ID, coordinates, installation time; Monitoring indicator attributes: unit, dimension, threshold range; Event entity attributes: occurrence time, duration, and description.

[0064] Entity attributes are used to describe the intrinsic characteristics of an entity and are not involved in the definition of relationships between entities.

[0065] 6. Entity hierarchy and inheritance relationship ( ) A hierarchical organization of the ontology is achieved by defining inheritance and hierarchical relationships between entities. For example: Among them, "monitoring object" is the higher-level abstract category of the entities to be monitored; "Monitoring point": The specific type of monitoring object; "Displacement monitoring point": A more detailed category of monitoring points used for displacement monitoring.

[0066] This hierarchical relationship allows different categories of monitored objects to inherit common attributes while retaining their unique attributes.

[0067] 7. Definition of semantic constraint rules ( ) To ensure semantic consistency of the knowledge graph, semantic constraint rules are defined in the ontology model, including: The monitoring point must belong to a single structural location; Monitoring indicators must be linked to the monitoring points for observation. An abnormal event must be associated with at least one monitoring indicator and a time interval.

[0068] Constraint rules are used to prevent semantic conflicts and data redundancy.

[0069] 8. Ontology Model Output After completing the above ontology construction, a unified semantic model for dam monitoring operations is formed, providing a standardized semantic foundation for the semantic mapping and knowledge graph fusion of monitoring data in the subsequent step S3.

[0070] (iv) Monitoring data semantic mapping and fusion algorithm (S3) In step S3, the standardized monitoring dataset output in step S1 is... Mapped to the dam monitoring knowledge graph ontology model constructed in step S2 It also performs semantic fusion and consistency maintenance on multi-source data to form a knowledge graph instance graph that can be used for querying, reasoning, and analysis. .

[0071] The core of step S3 is to explicitly express the business semantics of "monitoring object-structural part-indicator-operating condition-event" implicit in the monitoring data based on the entity type, relation type and constraint rules in the ontology, and to achieve cross-source data fusion through mechanisms such as entity alignment, relation merging and conflict resolution.

[0072] 1. Input and Output Input: Standardized monitoring dataset in, Standardized multi-source monitoring dataset; : No. Each monitoring point at time The monitoring indicator vector; Spatial attributes: spatial coordinates of monitoring points, dam section number, and structural component information; Time label: The sampling time or time interval corresponding to the monitoring data; Semantic tags: monitoring indicator type, data source identifier, quality markers and other semantic information.

[0073] Output: Knowledge graph instance diagram in, : A completed knowledge graph instance diagram; : A set of facts relating entities; Head Entity: Represents the initial entity of the relationship; Relation type: Represents the semantic relationship between entities; Tail Entity: Represents the entity to which a relation points. Entity set The set of attribute information contained therein (including time series attributes).

[0074] 2. Overall Semantic Mapping Strategy In one embodiment, S3 includes the following sub-steps: S31: Entity recognition and instantiation; S32: Relationship creation and binding; S33: Attributes are linked to time series; S34: Cross-source entity alignment and fusion; S35: Relationship conflict resolution and consistency verification; S36: Incremental Updates and Version History.

[0075] 3. Entity Recognition and Instantiation (S31) Based on ontology model Entity type collection in From the dataset Entity instances are identified and written into the graph.

[0076] (1) Structural entity instantiation: Based on the dam section number, structural part code, and spatial range definition, the structural entity is represented as follows: in, Entity type identifier; : The unique identifier of a structural entity; : A description of the geometric or spatial extent of a structural entity; : A set of design or management attributes.

[0077] (2) Instantiation of monitoring objects for any monitoring point Generate monitoring point entities: in, : Unique identification number for each monitoring point; : Three-dimensional spatial coordinates of the monitoring point.

[0078] (3) Instantiation of indicator entities For indicator types (e.g., displacement, seepage pressure, stress), generate index entities: in, : Name of monitoring indicator; : Unit of monitoring indicator; Reasonable range of values ​​for monitoring indicators.

[0079] (4) Instantiation of working conditions and event entities When monitoring data or text tags trigger condition / event recognition (e.g., water level condition, rainfall event, abnormal deformation event), generate a condition or event entity: in, : Operating condition entity; : Unique identification number for operating conditions; : Start time of operating condition; : End time of the working condition; Operating condition attribute information; : Exception or event entity; Event number; : Textual description of the event.

[0080] 4. Relationship generation and binding (S32) Based on the set of ontology relations Generate relation triples for the identified entities.

[0081] (1) Subordination (monitoring point - structural part) For monitoring points With structural entities Establish a subordinate relationship: in It can be determined by a combination of "code matching + spatial location": if the coordinates of the monitoring point fall within the spatial range of the structural entity, then a membership relationship is established.

[0082] (2) Observational relationship (monitoring point - indicator) Establish observational relationships between monitoring points and indicators: (3) Configuration relationship (sensor-monitoring point) If sensor information exists, then establish: (4) Influence Relationship (Operating Condition - Indicator / Event) Establish impact relationships between operating entities and indicators or events: 5. Linking attributes to time series (S33) Numerical time-series data can be attached as time-series attributes of monitoring point entities or "monitoring point-indicator" relationships.

[0083] In one implementation, for monitoring points Indicators Constructing time series: in, : timestamp; Monitoring point In time The Class indicator observations; : The quality marker of the data at this moment (generated by processes such as S1 anomaly detection and missing data completion).

[0084] And Write it as an attribute: Relationship attribute method: Attributes; Alternatively, use entity attribute method: attach to the entity at the monitoring point. In the indicator attributes.

[0085] 6. Cross-source entity alignment and fusion (S34) When multiple data sources generate duplicate entities for the same object (e.g., the same monitoring point is encoded differently in different systems), entity alignment and fusion are required.

[0086] 6.1 Similarity Calculation For candidate entity pairs Calculate the overall similarity: in, , , , The weighting coefficients of each similarity component; :entity and The overall similarity; ::Code / name similarity (e.g., edit distance or rule matching); Spatial distance similarity can be represented as: in, Attribute consistency similarity; Time series similarity; , :entity and The spatial coordinate vector; Euclidean distance calculation symbol; Spatial scale adjustment parameters.

[0087] When the following conditions are met: in If the threshold for determining semantic equivalence of entities is set, then the two entities are determined to be semantically equivalent, and fusion is performed.

[0088] 6.2 Integration Strategy The merged entity is denoted as Its attributes are combined according to the following rules: For deterministic attributes (such as coordinates, dam section), a priority rule is applied: select data sources with higher confidence. Weighted fusion is applied to numerical attributes: in, The attribute values ​​after fusion; : From data source The attribute value; Data source The weighting coefficients.

[0089] The time series attributes are handled using a strategy of "union + deduplication + quality mark retention".

[0090] 7. Relationship Conflict Resolution and Consistency Verification (S35) After merging, relationship conflicts or violations of ontology constraints may occur (e.g., the same monitoring point belongs to two structural parts simultaneously). Therefore, a set of consistency verification and conflict resolution rules is introduced. .

[0091] 7.1 Relationship Confidence relation triples Calculate the confidence level: in, Data source trust level; : Rule matching strength (e.g., "spatial location + encoding consistency" is higher than "encoding consistency only"); Support of observed data (such as time series stability, outlier rate, etc.); : Confidence component weights.

[0092] 7.2 Conflict Resolution Rules When a mutual exclusion relationship exists (e.g., only one monitoring point is allowed per monitoring point), When considering the relationship with the highest confidence level, retain the relationship with the highest confidence level. in, The target entity of the relationship with the highest confidence level; : Take the parameters that make the objective function reach its maximum value.

[0093] The remaining relationships are marked as candidate or invalid relationships, and the reasons for the conflict are recorded.

[0094] 7.3 Ontology Constraint Verification Based on the constraint set of S2 examine: Does each monitoring point possess a unique structural affiliation? Whether each monitoring point is associated with at least one indicator; Each event is bound to a time interval and associated with at least one metric or structural entity.

[0095] If the constraints are not met, a verification log is generated and a completion strategy is triggered (e.g., backtracking the data source to complete the encoding or time information).

[0096] 8. Incremental Updates and Version History (S36) To support continuous monitoring data access, S3 employs an incremental fusion mechanism. This applies to newly added data batches. ,implement: (1) Entity retrieval: Locate candidate entities; (2) Alignment fusion: Based on similarity threshold (3) Relationship update: add or update the relationship and calculate the confidence score; (4) Sequence append: append to the sequence. (5) Version record: Record the version number, source and difference set of each fusion update.

[0097] Version history can be represented as: in, : No. Sub-map updated version; Update time; : Entity change set; : Set of relationship changes; : Attribute change set; Data source identifier.

[0098] Used for subsequent audit follow-up and project review.

[0099] 9. Outputs and Interfaces After completing step S3, output a knowledge graph instance containing entity, relation, attribute, and version information. It provides the data structure required for graph storage and updating to step S4, and provides the index and relation path information required for semantic query and reasoning to step S5.

[0100] (v) Knowledge graph storage and dynamic update mechanism (S4) In step S4, the knowledge graph instance graph constructed in step S3 is processed. Storage management is implemented, and the knowledge graph is dynamically updated and maintained for consistency as new monitoring data is continuously added or engineering information changes, to ensure that the graph content can accurately reflect the evolution of the dam's operational status.

[0101] The core objective of step S4 is to achieve incremental access, traceable updates, and consistency management of monitoring data and engineering knowledge without disrupting the existing semantic structure and historical information.

[0102] 1. Knowledge Graph Storage Structure Design In one implementation, the knowledge graph is represented as a directed graph structure with attributes: in, : Represents the overall structure of the dam monitoring knowledge graph; : A collection of entity nodes used to store structural entities, monitoring object entities, indicator entities, operating condition entities, and event entities; : A set of relation edges used to describe the semantic relationships between entities; : A collection of attributes used to describe the attribute information of entities and relationships; Time stamps and time series sets are used to describe the temporal characteristics of monitored data and events; Version and update history collection, used to store the update history information of the knowledge graph.

[0103] Both entity nodes and relation edges can carry attribute information and timestamps to support time series analysis and historical backtracking.

[0104] 2. Knowledge Graph Indexing and Fast Access Mechanism To improve query and update efficiency, a multi-dimensional index structure is established in the knowledge graph storage layer. The index includes at least: (1) Entity Identifier Index is based on the unique identifier of the entity. Create an index for quickly locating entity nodes.

[0105] (2) Spatial Index A spatial index is established based on the spatial coordinates of monitoring points or structural entities to support rapid retrieval by dam section, structural part, or spatial range.

[0106] (3) Time Index Based on time tags or time interval Create an index to support time-based queries and time-series updates.

[0107] (4) Relation type index based on relation type Create an index to accelerate the search for related paths and semantic reasoning.

[0108] The aforementioned index structure provides efficient access support for intelligent querying and analysis in step S5.

[0109] 3. Incremental Update Trigger Mechanism The dynamic update process of the knowledge graph is triggered when any of the following conditions occur: New batches of monitoring data have arrived; Add or change monitoring points, sensors, or structural units; Changes occur in project management information, operating conditions, or abnormal events.

[0110] Let the newly added dataset be... The corresponding set of newly added or updated graph elements is: in, : Represents the set of incremental changes to the knowledge graph involved in this update; The set of newly added or updated entity nodes; : The set of relationships that are newly added or updated; : A set of newly added or updated attributes; : A collection of newly added or updated time tags or time series.

[0111] 4. Incremental update rules for entities and relations 4.1 Entity Update Rules For entity instances involved in the newly added data, the following rules apply: If the entity already exists If the time series is in the middle, then update its attribute set or time series attribute; If the entity does not exist, create a new entity node and add it to the collection. .

[0112] Entity updates can be represented as mapping functions: in, Entity update mapping function; The set of entities before the update; The incremental set of entities to be updated; : The updated set of entities.

[0113] 4.2 Relationship Update Rules For newly added or updated relation instances: If the relation triplet If it already exists, update its attributes or confidence level; If it does not exist, add it to the relation set. .

[0114] The relation update function can be expressed as: in, : Relationship update mapping function; The set of relationships before the update; : The set of incremental relationships to be updated; : The updated set of relationships.

[0115] 5. Additional updates to time series and attributes For the time series corresponding to monitoring points and indicators It adopts an append-only update mechanism: in, Updated monitoring points Corresponding indicators Time series; : Time series before update; : timestamp; Monitoring point In time Lowering the target Observed values; :time Quality labeling information for the observed data; The set union operation is used to represent the appending of time series data.

[0116] During the appending process, timestamp deduplication and sorting are performed to ensure the continuity and consistency of the time series.

[0117] 6. Consistency verification and constraint maintenance mechanism After updating entities, relations, and attributes, a consistency check is performed on the knowledge graph to ensure that it meets the ontology constraint set defined in step S2. .

[0118] Consistency checks should include at least the following: Entity integrity verification: Does each monitoring point have a unique structural affiliation? Relation constraint validation: Whether the relation type conforms to the ontology definition; Time consistency check: Whether the time interval of the event is reasonable and without conflict.

[0119] If a violation of the constraints is found, it will be corrected or marked according to the preset rules, and a verification log will be generated.

[0120] 7. Version control and change log mechanism To support historical retrospective and audit analysis, a version record is established for each map update.

[0121] In one implementation, each update is recorded as follows: in, : No. The version record corresponding to each knowledge graph update; Version number; The time when this update occurred; : No. The set of entity changes involved in this update; : No. The set of relationship changes involved in this update; : No. The set of attribute changes involved in this update; The data source or trigger source identifier for this update.

[0122] Version control mechanisms enable the historical tracing and comparison of knowledge graph states.

[0123] 8. Summary of Dynamic Update Process In summary, the dynamic update process in step S4 includes: Add new data access → Entity and relationship identification → Incremental update → Consistency check → Version record → Storage commit.

[0124] This process ensures that the knowledge graph is stable, scalable, and traceable during long-term operation, providing a reliable data foundation for subsequent semantic-based intelligent queries and association analysis.

[0125] (vi) Intelligent query and related reasoning based on semantic constraints (S5) In step S5, based on the knowledge graph instance graph stored and dynamically maintained in step S4... Under the premise of adhering to ontology constraints, it realizes intelligent query and correlation reasoning for engineering applications, which can be used to support dam operation status analysis, anomaly tracing and auxiliary decision-making.

[0126] The core of step S5 is to transform the query requirements or system analysis requirements of engineers into semantic constraints, perform constrained entity retrieval, relation traversal and association reasoning in the knowledge graph, and output interpretable query and analysis results.

[0127] 1. Overall Framework for Intelligent Query In one embodiment, the intelligent query and association reasoning process includes the following sub-steps: S51: Query request parsing and semantic modeling; S52: Query constraint generation and consistency verification; S53: Semantic Constraint-Based Graph Traversal and Path Search; S54: Related Reasoning and Result Filtering; S55: Output of query results and semantic path explanation.

[0128] 2. Query Request Parsing and Semantic Modeling (S51) User query requests can originate from interactive input by engineers or from analysis tasks automatically triggered by the system. Query requests are uniformly converted into semantic query expressions.

[0129] In one implementation, the query request is represented as a set of triple constraints: in, The overall representation of a semantic query request; : Target entity set, used to limit the type of entity or specific entity (such as dam section, monitoring point, abnormal event) of interest in the query. : A set of relational constraints used to limit the types of semantic relationships (such as "belong to", "observation", "influence") that are allowed to participate in queries between entities; : A set of conditional constraints used to limit the time conditions, numerical range conditions, or operating conditions of a query.

[0130] For example, the query "whether there is abnormal deformation in a certain dam section under high water level conditions" can be formally represented as: 3. Query constraint generation and consistency verification (S52) Based on the ontology constraint set defined in step S2 The system performs a validity check on the query conditions to prevent invalid or conflicting queries.

[0131] Consistency checks should include at least the following: Check if an entity belongs to the entity type defined in the ontology; Check if the relationship conforms to a valid relationship between entity types; Are the time interval conditions reasonable (the start and end times do not conflict)?

[0132] If the query conditions violate ontology constraints, an error message will be returned or a constraint correction strategy will be triggered.

[0133] 4. Semantic Constraint-Based Graph Traversal and Path Search (S53) After passing the consistency check, the query constraints are applied to the knowledge graph. In execution graph traversal and path search.

[0134] 4.1 Constraint Diagram Construction Map query conditions to constraint subgraphs: in, A complete example diagram of a dam monitoring knowledge graph; : Subgraphs that satisfy the query constraints; Subset relation: indicates that the constrained subgraph consists of entities and relations in the original graph that satisfy the conditions.

[0135] Only those that satisfy the condition are retained. and Entities and relationships.

[0136] 4.2 Path Search Rules In one implementation, a constrained path search is performed on the target entity, and the path is represented as: in, Semantic path: used to represent the link between entities; : No. One entity node; : Connecting the first With the The semantic relationships between entities; The number of relationships in the path.

[0137] Path search satisfies the following conditions: The path length does not exceed the preset maximum number of hops ; Path relationship type belongs to the allowed set ; The path nodes satisfy the query conditions and constraints.

[0138] 5. Relational Reasoning and Result Filtering (S54) After obtaining the set of candidate paths, perform association reasoning to filter and sort the query results.

[0139] 5.1 Calculation of Inference Confidence For any path Calculate its inference confidence: in, The confidence level of the relationship defined in step S3.

[0140] 5.2 Examples of Inference Rules In one implementation, the following associative reasoning rule is defined: If an abnormal event is significantly associated with a monitoring point, and the indicator of that monitoring point exceeds the threshold under the specified operating conditions, then the abnormal event is determined to be associated with that operating condition. If multiple monitoring points show similar deformation trends in the same structural part, it is inferred that there is a potential risk of overall deformation in that structural part.

[0141] 5.3 Results Screening An entity or path will be included in the final result set if one of the following conditions is met: in This is the confidence threshold for inference, used to determine whether the inference result is valid.

[0142] 6. Query results and semantic path explanation output (S55) The filtered query results are used to generate structured output and semantic explanation information.

[0143] The output should include at least: Retrieve the set of entities or events that match; Monitoring indicators and time intervals related to the results; The semantic path supporting this result ; Confidence information for each relation in the path.

[0144] By outputting semantic paths, engineers can clearly understand the source of query results and their logical basis.

[0145] 7. Application of query and inference results Step S5 enables semantic querying and correlation reasoning of dam monitoring data, allowing engineers to comprehensively analyze the dam's operational status from multiple dimensions such as structure, time, and operating conditions, providing interpretable information support for anomaly diagnosis, risk assessment, and operational decision-making.

[0146] (vii) Comprehensive description of the technical solution In summary, the knowledge graph-based semantic fusion and intelligent query method for dam monitoring data proposed in this invention addresses the problems of complex multi-source data types, scattered semantics, and difficulties in correlation analysis in dam operation safety monitoring. It constructs a complete technical solution system from data acquisition and semantic modeling to intelligent query and reasoning.

[0147] In terms of technical implementation, the solution is based on the standardized collection and preprocessing of multi-source monitoring data. By constructing a knowledge graph ontology model for dam monitoring business, it unifies the semantic definition of core elements such as dam structure, monitoring objects, monitoring indicators, operating conditions and abnormal events. On this basis, it semantically maps and integrates real-time monitoring data with engineering knowledge to form a structured and associative knowledge expression form.

[0148] By introducing mechanisms such as entity alignment, relation fusion, consistency verification, and version control, the knowledge graph can maintain semantic consistency and structural stability during the continuous ingestion of monitoring data, thereby supporting the dynamic evolution and historical backtracking of dam operation status information. Simultaneously, based on ontology constraints and semantic relationships, constrained intelligent querying and associative reasoning are implemented on the knowledge graph. This ensures that the query process not only returns result information but also provides clear semantic association paths, offering interpretable support for engineering analysis.

[0149] Overall, the technical solution described in this invention forms a closed-loop process of "data-semantics-knowledge-application" in terms of logical structure. Each step cooperates with each other and progresses step by step. Without relying on specific hardware devices or specific software, it realizes the systematic organization and intelligent utilization of dam monitoring data and engineering knowledge, and has good engineering adaptability and scalability.

[0150] Example 1: An Example of a Semantic Fusion Method for Multi-Source Monitoring Data of Dams Based on Knowledge Graphs This embodiment takes a concrete gravity dam project as the application object. During the operation of the dam, multiple types of monitoring systems are deployed, including dam deformation monitoring system, seepage monitoring system, environmental condition monitoring system and inspection management system. Each system operates independently and forms multi-source heterogeneous monitoring data.

[0151] (I) Implementation Method of Multi-Source Monitoring Data Acquisition and Preprocessing like Figure 2As shown, the raw data from different monitoring systems were first collected uniformly. The collected data included: Displacement monitoring data: Time series of displacements at monitoring points on and inside the dam body, collected by equipment such as surveying robots and GNSS; Seepage monitoring data: seepage pressure and head data collected by piezometers and piezometers; Environmental data includes reservoir water level, rainfall, and temperature. Project management data includes inspection records, operation logs, and anomaly description texts.

[0152] For the above data, perform standardized preprocessing according to step S1. Specifically, this includes: Monitoring data from different sources are mapped to a unified time axis, and data with inconsistent sampling frequencies are resampled or interpolated. Outlier identification is performed on numerical monitoring data, and when a monitored value deviates from historical statistical characteristics by more than a preset threshold, it is marked. Normalization is performed on monitoring indicators with different dimensions to eliminate dimensional differences; For text data such as inspection records, extract structural parts, time information, and keywords describing anomalies, and convert them into structured semantic tags.

[0153] After the above processing, a standardized monitoring dataset containing spatial attributes, time labels, and semantic labels is formed, providing input for subsequent semantic modeling.

[0154] (II) Implementation Method for Constructing the Knowledge Graph Ontology Model for Dam Monitoring like Figure 3 As shown, a knowledge graph ontology model for the field of dam monitoring is constructed based on the business needs of dam safety monitoring.

[0155] In this embodiment, the ontology model includes at least the following entity types: Dam structural entities: including "dam", "dam section", "gallery", "foundation", etc.; The entities being monitored include "monitoring points", "sensors", and "measuring lines". Monitoring indicators include entities such as "displacement", "settlement", "seepage pressure", "stress", and "temperature". Operating conditions and event entities: including "water level conditions", "rainfall events", "abnormal deformation events", etc.

[0156] At the same time, define the semantic relationships between entities, for example: "The monitoring point belongs to the dam section"; "Displacement index observed at monitoring points"; "The seepage pressure index is affected by the water level and operating conditions"; "The abnormal event occurred within a specific time interval."

[0157] Through the above entity and relationship definitions, a unified semantic structure framework is formed, providing a consistent semantic carrying space for monitoring data from different sources.

[0158] (III) Implementation of Semantic Mapping and Fusion of Monitoring Data like Figure 5 and Figure 6 As shown, the standardized monitoring data obtained in step (1) is mapped to the knowledge graph ontology model.

[0159] In this embodiment, the following operations are performed for each monitoring point: Based on the monitoring point number and spatial coordinates, generate the corresponding monitoring point entity in the knowledge graph; Based on the dam section coding and spatial affiliation, establish the affiliation between monitoring points and dam sections; For various monitoring indicators collected from monitoring points, corresponding monitoring indicator entities are generated, and "observation" relationships are established; The time series of monitoring data is attached as an attribute to the "monitoring point-monitoring indicator" relationship node.

[0160] When different data sources provide duplicate descriptions of the same monitoring point, entity alignment is performed based on the monitoring point number, spatial coordinates, and historical time series similarity, and attribute information is merged to form a unified monitoring point entity.

[0161] Through the above semantic mapping and fusion process, the monitoring data that were originally scattered in different systems are uniformly organized into knowledge graph instances with clear semantic relationships.

[0162] (iv) Implementation of Knowledge Graph Storage and Dynamic Update like Figure 7 As shown, the completed knowledge graph instance is stored as a directed graph structure with attributes.

[0163] In this embodiment, when new monitoring data arrives, incremental updates are performed only on the entities, relationships, and time series corresponding to the new data, without reconstructing the overall graph structure. Each update generates a corresponding version record to support historical state retrospection and engineering auditing.

[0164] Meanwhile, by verifying the uniqueness of entities, the legality of relationships, and the consistency of time, we ensure that the graph structure always meets the ontology constraint rules.

[0165] Example 2: Intelligent Query and Association Analysis of Dam Monitoring Data Based on Knowledge Graph This embodiment, based on the knowledge graph constructed in Embodiment 1, realizes semantic intelligent query for engineering applications.

[0166] (a) Implementation method for semantic query request processing like Figure 8 As shown, when engineers submit a query request, such as "whether there is abnormal deformation in a certain dam section under high water level conditions", the system first performs semantic parsing on the request and transforms it into a semantic query expression that includes entity type, relation constraints and time conditions.

[0167] (II) Implementation of Graph Traversal and Reasoning Based on Semantic Constraints Based on the query conditions, the system performs a constrained graph traversal operation in the knowledge graph, retaining only entities and relationships that satisfy the dam section range, water level conditions, and time conditions.

[0168] During the traversal, if an abnormal deformation event is found to be correlated with the displacement index of multiple monitoring points under the same working conditions, the abnormal event is determined to be causally related to the working conditions by using preset inference rules.

[0169] (III) Implementation of Query Result Output like Figure 9 As shown, the system not only outputs the query result entity (such as abnormal deformation event), but also outputs the semantic path supporting the result, including the monitoring points, monitoring indicators and operating conditions involved, so that engineers can intuitively understand the logical basis for the occurrence of the anomaly.

[0170] Example 3: Explanation of the Implementation Effect of Engineering Application like Figure 10 As shown, after the method of the present invention is deployed in the dam operation and management system, engineers can quickly obtain the required monitoring information through high-level semantic conditions such as structural parts, time intervals and operating conditions without directly operating the underlying database.

[0171] Compared to traditional table-based query methods, this implementation significantly improves the understandability of monitoring data, query efficiency, and the systematic nature of anomaly analysis, providing effective technical support for dam operation safety management.

[0172] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs, characterized in that: It includes the following steps: S1, Multi-source monitoring data acquisition and standardization preprocessing steps: collect dam operation monitoring data from different monitoring systems, and sequentially perform data quality control, time scale alignment, unit unification and semantic labeling on the monitoring data to form a standardized monitoring dataset containing spatial attributes of monitoring points, unified time labels and semantic categories of monitoring indicators. S2, the steps for constructing the knowledge graph ontology model for dam monitoring: construct a knowledge graph ontology model for dam safety monitoring business. The ontology model defines at least the dam structure entity, the monitoring object entity, the monitoring indicator entity, and the operating condition or abnormal event entity. It also predefines the structural membership relationship, monitoring observation relationship, and operating condition influence relationship between the entities, and sets engineering semantic constraint rules to constrain the legality of entity relationships. S3, based on the multi-constraint decision rule, the monitoring data semantic mapping and entity alignment fusion step, according to the knowledge graph ontology model, maps the standardized monitoring data into entities, relations and time series attributes in the knowledge graph; When data from different monitoring systems point to the same candidate monitoring object, entity alignment and fusion processing is performed according to the following joint judgment rules: a) The spatial distance between the candidate entity and the preset spatial distance threshold is less than the target spatial distance threshold; b) The similarity of the monitoring indicator time series corresponding to the candidate entity within the preset time window is greater than the preset time series similarity threshold; c) The candidate entities have the same structural affiliation in the dam's structural hierarchy; Only when all of the above determination rules are met simultaneously, the candidate entities are determined to be the same actual monitoring object, and entity merging and attribute fusion processing are performed. S4, Incremental storage and versioned dynamic update steps for knowledge graphs. The completed knowledge graph is stored and managed. When new monitoring data is accessed, incremental updates are performed only on newly added or changed entities, relationships and time series attributes. At the same time, a corresponding version record is generated for each update to achieve traceable management of the knowledge graph evolution process. S5, Semantic subgraph construction and constrained association query steps based on ontology constraints, parses the query requirements of engineers into a semantic query expression containing target entity type, relation type constraints and time or operating condition conditions, constructs a query subgraph that satisfies the semantic constraints in the knowledge graph, performs entity retrieval, relation traversal and association reasoning only within the scope of the query subgraph, and outputs the query results and the semantic association paths supporting the query results.

2. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: The multi-source monitoring data collected in step S1 includes at least one or more of the following: structural deformation monitoring data, seepage monitoring data, environmental condition monitoring data, and engineering inspection or operation management text data. The different types of monitoring data come from independently set monitoring subsystems or management systems, and the monitoring subsystems differ in terms of data acquisition frequency, data organization format, time stamping method, and semantic description method. The structural deformation monitoring data is used to characterize the displacement, settlement, stress, or strain changes of the dam body or foundation during operation; the seepage monitoring data is used to characterize the seepage pressure, head, or seepage status information at the dam body, foundation, or contact surface; the environmental condition monitoring data is used to characterize external conditions such as water level, rainfall, and temperature during reservoir operation; and the engineering inspection or operation management text data is used to describe abnormal phenomena, changes in operating status, or historical event information recorded during manual inspections. By unifying the data from different sources, with different structures and different semantic expressions into the data collection and preprocessing process in step S1, a complete and fusionable data foundation is provided for subsequent semantic mapping, entity alignment and association analysis based on knowledge graphs, thereby avoiding the problem of fragmented monitoring information caused by heterogeneous data sources.

3. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In the data quality control process of step S1, outlier identification processing based on historical statistical characteristics is performed on the numerical monitoring data. Specifically, this includes: for the same monitoring indicator of the same monitoring object, statistically analyzing the distribution characteristics of its observation data in the historical operation phase and calculating the corresponding statistical parameters; comparing the currently collected monitoring value with the historical statistical characteristics, and when the deviation of the current monitoring value from the historical statistical characteristics exceeds a preset statistical threshold, the monitoring value is determined to be abnormal monitoring data. Monitoring data that is determined to be abnormal is not deleted directly, but is retained by setting an abnormality label and is processed as monitoring data with abnormal attributes in subsequent semantic mapping, entity alignment and association reasoning processes, so that abnormal information can be explicitly expressed and identified and utilized by subsequent analysis processes. Through the above-mentioned outlier identification and labeling mechanism, the impact of sensor errors, environmental interference, or occasional failures on the semantic fusion results of knowledge graphs can be reduced without disrupting the continuity of the original monitoring data, thereby improving the reliability of the semantic expression of monitoring data and the credibility of engineering analysis.

4. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In the time scale alignment process of step S1, a unified time reference axis is constructed for monitoring data with inconsistent sampling frequencies or whose sampling time points do not completely overlap, and monitoring data from different monitoring systems are mapped to the unified time reference axis through interpolation or resampling. Specifically, for monitoring data with a high sampling frequency, the time resolution is reduced by resampling to keep it consistent with the unified time axis; for monitoring data with a low sampling frequency or missing measurements, the monitoring value at the corresponding time point on the unified time axis is estimated by interpolation. By aligning the time scales as described above, different monitoring indicators can have a unified time benchmark during semantic fusion and correlation analysis. This ensures that data across monitoring systems and indicators can be compared, correlated, and inferred in the same time dimension, avoiding semantic correlation bias caused by time inconsistency.

5. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In the knowledge graph ontology model constructed in step S2, the dam structure entity includes at least the dam as a whole, dam sections, slopes, galleries, or foundation structure units, used to describe the compositional relationship of the dam project in terms of spatial structure and functional hierarchy; the monitoring object entity includes at least monitoring points, sensors, or survey lines, used to describe the spatial carrying units of monitoring data and their collection sources; the monitoring index entity includes at least displacement, settlement, seepage pressure, stress, strain, or temperature, used to uniformly describe the data types collected by different monitoring systems. By uniformly defining the above entity types in the ontology model and clarifying the structural affiliation, monitoring and observation relationships, and operational condition influence relationships among the entities, the monitoring data can accurately correspond to its engineering semantic role when mapped to the knowledge graph. This ensures the consistency between the knowledge graph structure and the actual business logic of the dam project and provides a stable semantic foundation for subsequent semantic-based queries and analyses.

6. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In step S3, the time series data corresponding to the monitoring indicator is attached as an attribute to the observation relationship node between the monitoring object entity and the monitoring indicator entity. The time series data is used to explicitly describe the evolution process of the monitoring indicator over time and includes timestamp information and corresponding monitoring value information. By associating time series attributes with observation relationship nodes, the knowledge graph can not only express the static structural relationships between monitored objects, but also the dynamic change characteristics of monitoring data in the time dimension, thereby supporting the time series analysis and trend judgment of dam structural deformation, seepage status or environmental response. This approach avoids the semantic fragmentation problem caused by storing time-series data in isolation in external databases, making time evolution information an organic part of the semantic structure of the knowledge graph and improving the efficiency of utilizing time features in subsequent associative reasoning processes.

7. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In the entity alignment and fusion process of step S3, when multiple data sources provide different attribute information for the same monitoring object, a type-based attribute fusion strategy is adopted for the fused monitoring object entity. Specifically, for deterministic attributes such as spatial location and structural affiliation, the data source credibility is prioritized; for numerical monitoring attributes, a weighted fusion process is performed according to preset weights to generate fused attribute values. By distinguishing different types of attributes and adopting corresponding fusion strategies, the attribute distortion problem caused by simple coverage or averaging is avoided. This ensures the uniqueness of entities while maximizing the retention of effective information contained in multi-source monitoring data, thereby improving the accuracy and engineering applicability of entity attribute expression after fusion.

8. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In the incremental update process of step S4, the time series attributes corresponding to the monitoring data are updated by appending the newly collected monitoring data to the existing time series in chronological order, and the timestamps are deduplicated and the order is checked during the appending process. By using the above-mentioned append-only update method, the overall reconstruction of existing time series data is avoided, thereby reducing the computational overhead during the system update process. At the same time, the continuity and consistency of time series data are guaranteed, providing stable data support for subsequent trend analysis and backtracking analysis based on historical data.

9. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: In step S5, when constructing the query subgraph based on the semantic query expression, only entity nodes and relation edges that simultaneously satisfy entity type constraints, relation type constraints, and time conditions or operating conditions are retained, and entities and relations that do not satisfy the constraints are pruned during the graph traversal process. By constructing constrained query subgraphs in advance during the query phase, indiscriminate traversal across the entire knowledge graph is avoided, thereby reducing the computational complexity of related queries and improving the matching degree between query results and engineering semantic requirements. This makes the query process more efficient and the results more focused on engineering-related objects.

10. The method for semantic fusion and intelligent querying of dam monitoring data based on knowledge graphs as described in claim 1, characterized in that: While outputting the query results, a semantic association path consisting of the monitoring object, monitoring indicator, operating condition or abnormal event is further output. The semantic association path is used to describe the entities involved in the formation of the query results and the semantic connection relationship between them. The semantic association path is not only used to display query results, but also to support engineers in understanding and judging the formation mechanism of anomaly monitoring results. This enables engineers to trace the related factors that cause anomalies based on the semantic association path, thereby improving the interpretability and engineering credibility when conducting anomaly analysis and operational decisions based on monitoring data.

Images