A method for automatic construction and evolution of dam safety monitoring spatiotemporal knowledge graph

Abstract

The application discloses a kind of dam safety monitoring space-time knowledge graph automatic construction and evolution method, it is related to dam safety monitoring technical field, including: the three-dimensional coordinate data of acquisition monitoring point and the data of dam body structure model, the spatial adjacent relationship and spatial inclusion relation between monitoring point are automatically identified, and spatial relationship set is established;Combined with dam body structure model, the structural relationship set between monitoring point and dam body structure unit is automatically established;The time series data of monitoring point is acquired and preprocessed, and time series relationship set is established;Space-time knowledge graph is constructed;Real-time monitoring dam change event, dynamically evolving space-time knowledge graph, consistency verification and rollback are carried out to updated relationship, dam safety state reasoning and early warning are carried out, and change audit log is recorded.The application realizes the full-automatic construction and dynamic updating of dam safety monitoring space-time knowledge graph, provides efficient, accurate, dynamic knowledge graph construction method for dam safety monitoring.

Description

Technical Field

[0001] This invention belongs to the field of dam safety monitoring technology, and specifically relates to an automatic construction and evolution method for a spatiotemporal knowledge graph of dam safety monitoring. Background Technology

[0002] Dam safety monitoring is an important means of assessing the safety status of a dam by comprehensively processing and analyzing various monitoring data on the dam's operational status. Dam monitoring points are numerous and geographically dispersed but physically interconnected. The data exhibits periodicity and trends over time, and complex spatiotemporal correlations exist between data from different sensors, requiring comprehensive analysis.

[0003] Currently, dam safety monitoring mainly relies on physical methods and statistical models. Physical methods use structural design theory to calculate various quantities of the dam under current loads and compare them with measured values. This method has a clear objective and high interpretability, but it only stays at the level of independent analysis of single measuring points or simple aggregation analysis of multiple measuring points, lacking in-depth analysis of the correlation between monitoring points. Statistical models extract the developmental patterns and characteristics of monitoring data to construct dam safety early warning indicators and rules, which have the advantages of fast response and strong generalization ability. However, they also only stay at the level of independent analysis of single measuring points and lack a holistic study of dam structure monitoring and early warning.

[0004] In recent years, knowledge graph technology has been introduced into the field of dam safety management. A knowledge graph is a new technology that uses graph models to describe knowledge and model the relationships between everything, possessing capabilities such as multi-level relationship querying and knowledge analysis and reasoning. Existing dam safety monitoring methods based on knowledge graphs mainly suffer from the following technical problems: First, while existing methods can use knowledge graphs to describe the spatial relationships and data sequence similarity between monitoring points, structural relationships such as the structural affiliation and connection relationships between monitoring points and dam structural units need to be defined manually. This results in low construction efficiency and the manual definition is prone to omissions or errors.

[0005] Second, the discovery of time series relationships in existing methods mainly relies on manual analysis or simple correlation calculations, without using the structural relationships between monitoring points for constraints, which can easily lead to spurious correlations and inaccurate time series relationships.

[0006] Third, the knowledge graph constructed by existing methods is static. When the coordinates of monitoring points change (such as equipment relocation, coordinate correction, or the addition of new monitoring points), the relationships need to be redefined manually or the entire knowledge graph needs to be reconstructed, which takes several hours or even days and cannot meet the real-time requirements of dam safety monitoring.

[0007] In summary, existing technologies have significant shortcomings in the automatic construction, relationship accuracy, and dynamic updating of knowledge graphs. There is an urgent need for a method to construct a dam monitoring knowledge graph that can automatically build multidimensional relationships, including spatial, structural, and temporal relationships, and support dynamic updates. Summary of the Invention

[0008] The main objective of this invention is to disclose an automatic construction and evolution method for a spatiotemporal knowledge graph of dam safety monitoring, comprising: Collect 3D coordinate data of monitoring points and data of dam structure model, automatically identify spatial proximity and spatial inclusion relationships between monitoring points, generate spatial influence buffer, filter spatial relationships and establish a set of spatial relationships; Specifically, the three-dimensional coordinate data includes the X, Y, and Z coordinates of deformation monitoring points, seepage monitoring points, and stress-strain monitoring points; the dam structure model is generated by instantiating conceptual nodes based on the actual dam definition model; the data of the dam structure model includes the geometric data of dam segment polygons, joint lines, and foundation structural surfaces; the data preprocessing includes denoising using the moving average method, filling in missing values ​​using the linear interpolation method, and normalizing using the Z-score standardization method. Furthermore, when establishing the spatial proximity relationship identification: the Euclidean distance between monitoring points is calculated in batches, and monitoring point pairs with a distance less than a threshold are filtered to establish spatial proximity relationships; when establishing the spatial inclusion relationship identification: the coordinates of each monitoring point are spatially included with the polygon of the dam structure unit, the dam structure unit to which each monitoring point belongs is identified, and the spatial inclusion relationship between the monitoring point and the dam structure unit is established; when generating the spatial influence buffer: a spatial influence buffer is generated with each monitoring point as the center and the distance threshold of the corresponding type as the radius.

[0009] Based on the spatial relationship set and combined with the dam structure model, a set of structural relationships between monitoring points and dam structural units is automatically established; the set of structural relationships includes structural attribution relationships, structural connection relationships, spatial orientation relationships, and support relationships. Specifically, the set of structural relationships is established through spatial density clustering, structural mapping, and association derivation; The spatial density clustering refers to the use of the DBSCAN density clustering algorithm to cluster monitoring points into monitoring point groups based on spatial proximity. The association derivation refers to the structural affiliation between monitoring points within the same dam section, the structural connection between monitoring points in adjacent dam sections, the spatial orientation between upstream and downstream monitoring points, and the support relationship between the dam body and the foundation monitoring points.

[0010] Time-series data from monitoring points are acquired, preprocessed, and a set of time-series relationships is established. The set of time-series relationships includes mechanical coupling relationships, hydraulic propagation relationships, and environmental driving relationships between monitoring points. Specifically, establishing the time series relation set includes: preprocessing the time series data by denoising, imputing missing values ​​and normalizing, performing full correlation screening and time delay detection, and using structural relations for constraint verification and filtering out spurious correlations; The filtering of false correlations refers to: using structural relationships for constraint verification, retaining only monitoring point pairs that simultaneously meet the two conditions of "significant correlation" and "existence of structural association", and excluding false correlations between monitoring point pairs that are spatially and structurally unrelated due to data randomness; the time lag detection refers to: performing time lag detection on the monitoring point pairs filtered by structural constraints to determine whether there is a time lead and lag relationship between the changes of the two monitoring points.

[0011] A spatiotemporal knowledge graph is constructed based on the spatial relationship set, structural relationship set, and temporal relationship set, and stored in a graph database. The spatiotemporal knowledge graph is a knowledge network with dam monitoring points as nodes and three layers of relationships—spatial proximity, structural affiliation, and temporal causality—as edges.

[0012] Based on the spatiotemporal knowledge graph, dam change events are monitored in real time, and the types of change events and change coordinate data are recorded. The scope of impact is determined based on the magnitude of coordinate changes, and the spatiotemporal knowledge graph is dynamically evolved. Consistency verification and rollback are performed on the spatial, structural, and temporal relationships affected by the update, dam safety status reasoning and early warning are performed, and change audit logs are recorded. Specifically, the dynamically evolving spatiotemporal knowledge graph includes: using a cascading triggering mechanism to locally update the affected spatial relationships, structural relationships, and temporal relationships, thereby realizing the dynamic evolution of the spatiotemporal knowledge graph; The cascading triggering mechanism includes: triggering the re-derivation of structural relationships when spatial relationships change; triggering the recalculation of temporal relationships when structural relationships change; and keeping unaffected relationships unchanged. Furthermore, the determination of the impact range based on the coordinate change amplitude includes: pre-setting critical displacement thresholds for each type of monitoring point, comparing the coordinate change before and after the change with the critical displacement thresholds, and marking it as a major change if it exceeds the thresholds, and expanding the spatial neighborhood of a preset number of levels as the impact range with the monitoring point where the major change occurs as the center. Furthermore, the consistency verification and rollback refer to: verifying the consistency of relational data in the two types of storage by comparing the spatial relations in the spatial database with the relational edges in the graph database; when inconsistency is detected, automatically triggering the rollback mechanism to restore the state before the update, and marking the inconsistent relation as pending audit status; the dynamic evolution includes: monitoring change events, recording the change event type, involved monitoring points, coordinate values ​​before and after the change, and change timestamp; after the update is completed, recording the change audit log, including: relational state before and after the change, change timestamp, change type, and scope of impact.

[0013] According to this invention, a progressive multidimensional relationship automatic construction mechanism is used to achieve three-dimensional progressive automatic construction of spatial, structural, and temporal relationships. Automatic derivation of structural relationships is achieved through spatial density clustering, structural mapping, and association derivation, avoiding the inefficiency and omissions caused by manually defining structural relationships. Automatic discovery of temporal relationships is achieved through structural relationship constraint filtering, correlation analysis, time delay detection, and relationship type determination. The correlation of monitoring point pairs with structural associations is calculated, reducing the computational load from full combination to a scale comparable to the number of structural associations, effectively filtering out spurious correlations and improving the accuracy of temporal relationships. The cascading triggering mechanism only updates the affected relationship chains, eliminating the need for full reconstruction of the knowledge graph, thus meeting the real-time requirements of dam safety monitoring. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method for automatically constructing and dynamically evolving multidimensional relationships in a spatiotemporal knowledge graph for dam safety monitoring, provided by an embodiment of the present invention. Detailed Implementation

[0015] The specific implementation of the present invention will now be described in detail with reference to the accompanying drawings.

[0016] like Figure 1 As shown, this invention discloses an automatic construction and evolution method for a spatiotemporal knowledge graph of dam safety monitoring, comprising the following steps: Step S100: Collect the three-dimensional coordinate data of the monitoring points and the data of the dam structure model, automatically identify the spatial proximity and spatial inclusion relationships between the monitoring points, generate a spatial influence buffer, filter spatial relationships and establish a set of spatial relationships; The three-dimensional coordinate data and dam structure model data include: 1) Three-dimensional coordinate data: X, Y, and Z coordinates of deformation monitoring points, seepage monitoring points, and stress-strain monitoring points; 2) Dam structure model data: geometric data of dam segment polygons, joint lines, and foundation structure surfaces; The dam structure model is defined by abstracting conceptual nodes based on the actual dam body and instantiating the dam body model.

[0017] In this implementation step, coordinate preprocessing and benchmark unification are carried out by acquiring the three-dimensional coordinate data of all monitoring points, converting coordinate data from different sources and coordinate systems into the CGCS2000 national geodetic coordinate system, verifying coordinate accuracy, and eliminating coordinate outliers.

[0018] The spatial proximity relationship is identified using the ST_DWithin spatial function of the PostGIS spatial database. With a preset distance threshold as a condition, the Euclidean distance between monitoring points is calculated in batches, and monitoring point pairs with a distance less than the threshold are filtered to establish spatial proximity relationships. Different thresholds were set for monitoring points at different structural locations: the distance threshold for monitoring points on the dam surface was 50 meters, the distance threshold for monitoring points inside the corridor was 20 meters, and the distance threshold for monitoring points on the foundation was 30 meters.

[0019] The spatial inclusion relationship is identified using the ST_Contains spatial function of the PostGIS spatial database. The coordinates of each monitoring point are compared with the polygon of the dam structure unit to determine the spatial inclusion relationship, identify the dam structure unit (dam section) to which each monitoring point belongs, and establish the spatial inclusion relationship between the monitoring point and the dam structure unit.

[0020] The spatial impact buffer is generated using the ST_Buffer spatial function of the PostGIS spatial database. The buffer polygon is generated with each monitoring point as the center and the distance threshold of the corresponding type as the radius. This helps to determine the spatial impact range of the monitoring points and provides spatial reference for joint detection and cross-dam section association in the subsequent structural relationship derivation.

[0021] When filtering spatial relationships, spatial proximity relationships between monitoring points of different structural types are excluded (e.g., monitoring points on the dam surface and monitoring points on the slope are not established as proximity relationships even if the distance is less than the threshold, because their physical mechanisms are different), and the final set of spatial relationships S is established. The spatial relation set S includes: 1) Spatial proximity relationship set, including: a list of monitoring point pairs that meet the distance threshold condition and have the same structure type, each relationship contains two monitoring point IDs and a calculated distance value; 2) A set of spatial inclusion relationships, including: the mapping relationship between each monitoring point and its corresponding dam structural unit; 3) Spatial influence buffer set, including buffer geometry objects for each monitoring point, used for spatial reference in subsequent steps.

[0022] Step S110: Based on the spatial relationship set and combined with the dam structure model, automatically establish a set of structural relationships between monitoring points and dam structure units; the set of structural relationships includes structural attribution relationships, structural connection relationships, spatial orientation relationships, and support relationships.

[0023] In this implementation step, spatial density clustering is based on spatial proximity, and the DBSCAN density clustering algorithm is used to cluster the monitoring points into measurement point groups; The clustering parameters are set as follows: the clustering radius parameter eps is 50 meters (i.e., the distance between any two monitoring points in the same monitoring point group does not exceed 50 meters), and the minimum sample size parameter minPts is 2 (i.e., each monitoring point group contains at least 2 monitoring points). The clustering results organize the discrete monitoring points into monitoring point groups with spatial clustering characteristics.

[0024] The mapping from measuring point groups to structural units is based on the spatial distribution characteristics of the measuring point groups, mapping each measuring point group to a dam structural unit; The mapping rules include: if all monitoring points in a measurement point group are located within the same dam section polygon (based on spatial inclusion relationship judgment), then the measurement point group is mapped to that dam section; if the measurement point group is located within a preset range on both sides of the joint (using spatial influence buffer and joint line intersection judgment), then it is identified as a measurement point group near the joint, and candidate connection relationship of adjacent dam sections is established.

[0025] The structural correlation derivation is based on the mapping results from the measurement point group to the structural unit, and automatically derives four types of structural correlations according to preset derivation rules: 1) Structural attribution relationship (PART_OF): Between monitoring points within the same dam section, and between a monitoring point and its respective dam section; 2) Structural connection relationship (CONNECTED_BY): The monitoring points between adjacent dam sections connected by joints; 3) Spatial orientation relationship (UPSTREAM_DOWNSTREAM): Based on the upstream and downstream direction of the dam structure model, it is located between the monitoring points on the upstream and downstream sides; 4) Support relationship (SUPPORTED_BY): The relationship between the monitoring points of the dam body and the monitoring points of the foundation, reflecting the mechanical relationship between the dam body and the foundation.

[0026] Specifically, four types of structural relationships are summarized, and for each relationship, attributes such as source monitoring point ID, target monitoring point ID, relationship type, dam section to which it belongs, and confidence level are recorded to establish a structural relationship set R_struct; The structural relation set R_struct includes: 1) Clustering results of monitoring point groups: The monitoring point group number to which each monitoring point belongs; 2) Mapping table from measuring point group to dam section: the dam structural unit corresponding to each measuring point group; 3) List of four types of structural relationships: Each relationship includes source node, target node, relationship type (PART_OF / CONNECTED_BY / UPSTREAM_DOWNSTREAM / SUPPORTED_BY), dam section, and confidence level.

[0027] Step S120: Obtain time-series data of monitoring points, preprocess the time-series data, and establish a time-series relationship set R_temporal; the time-series relationship set includes the mechanical coupling relationship, hydraulic propagation relationship, and environmental driving relationship between monitoring points; The temporal relation set R_temporal includes: List of mechanical coupling relationships: Each relationship includes source node, target node, correlation coefficient, time delay value (lag=0), and the structural unit to which it belongs; List of hydraulic propagation relationships: Each relationship includes source node, target node, correlation coefficient, time delay value (lag>0), and propagation direction; Environment-driven relationship list: Each relationship includes a source node, a target node, a correlation coefficient, and an environmental association factor.

[0028] This implementation process includes the following specific steps: 1) Time series data preprocessing: Denoising, missing value imputation and normalization are performed on the time series data. Denoising is performed using the moving average method (window is 3 days), missing value imputation is performed using the linear interpolation method, and normalization is performed using the Z-score standardization method.

[0029] 2) Initial screening of full correlation: Calculate the Pearson correlation coefficient of all monitoring point pairs with structural relationship constraints, and screen monitoring point pairs with an absolute value of correlation coefficient greater than 0.7 and a significance level less than 0.05; Specifically, structural relationships are used for constraint filtering, and only monitoring point pairs with structural affiliation, structural connection, or spatial orientation relationships are retained for correlation calculation.

[0030] 3) False correlation filtering: For the selected highly correlated monitoring point pairs, further use structural relationships to perform constraint verification, and only retain the monitoring point pairs that simultaneously meet the two conditions of significant correlation and structural association, and exclude the false correlation between monitoring point pairs that are spatially and structurally unrelated due to data randomness.

[0031] 4) Time lag detection: Time lag detection is performed on the monitoring point pairs filtered by structural constraints. The Granger causality test method is used (with a significance level of 0.05 and a maximum time lag of 7 days) to determine whether there is a time lead-lag relationship between the changes of the two monitoring points.

[0032] 5) Determining the type of time series relationship: Based on correlation and time delay results, the type of time series relationship is determined as follows: Among them, the time-series relation types include: Monitoring point pairs that change synchronously and have structural correlations are determined to have a mechanical coupling relationship; Monitoring point pairs with lagging changes and upstream-downstream relationships are identified as having a hydraulic propagation relationship; monitoring point pairs with synchronous changes and environmental associations are identified as having an environmentally driven relationship.

[0033] Step S130: Construct a spatiotemporal knowledge graph based on the spatial relationship set, structural relationship set, and temporal relationship set, and store it in the graph database Neo4j; the graph database Neo4j is a database that stores data in a graph structure (nodes + edges); Specifically, the spatiotemporal knowledge graph is a knowledge network with dam monitoring points as nodes and three layers of relationships—spatial proximity, structural affiliation, and temporal causality—as edges. Step S140: Based on the spatiotemporal knowledge graph, monitor dam change events in real time, record change event types and change coordinate data, determine the scope of impact based on the magnitude of coordinate changes, and dynamically evolve the spatiotemporal knowledge graph; perform consistency verification and rollback on the affected spatial, structural, and temporal relationships after the update, perform dam safety status reasoning and early warning, and record change audit logs. The change events include: monitoring point addition events, coordinate modification events, and monitoring point deletion events; This implementation process includes the following specific steps: 1) Change event monitoring and recording: Real-time monitoring of change events in the dam monitoring system, recording the type of each change event, the ID of the monitoring point involved, the coordinate value before the change, the coordinate value after the change, and the change timestamp.

[0034] 2) Intelligent Determination of Impact Range: A three-level determination rule is adopted. When the coordinate change amplitude is less than the first threshold, it is determined to be a small amplitude change, and only the spatial proximity relationship of the monitoring point is updated; when the coordinate change amplitude is greater than the first threshold but less than the second threshold, it is determined to be a medium amplitude change, and the spatial proximity relationship and structural attribution relationship are updated, triggering the recalculation of the affected temporal relationship; when the coordinate change amplitude is greater than the second threshold or is a new or deleted event, it is determined to be a large amplitude change, and all relationships of the monitoring point are completely recalculated. The critical displacement thresholds differ for different monitoring point types and structures. The first threshold for dam surface deformation monitoring points is 10 meters, and the second threshold is 50 meters; the first threshold for gallery internal seepage monitoring points is 5 meters, and the second threshold is 20 meters; the first threshold for foundation stress monitoring points is 8 meters, and the second threshold is 30 meters.

[0035] 3) Cascade-triggered local update: Based on the determined scope of influence, the affected relationship chains are locally updated. Changes in spatial relationships trigger the re-derivation of structural relationships, changes in structural relationships trigger the recalculation of temporal relationships, and unaffected relationships remain unchanged.

[0036] 4) Consistency Verification and Rollback: After the update is completed, the spatial relationships in the spatial database are compared with the relation edges in the graph database to verify the consistency of relation data in the two types of storage. When inconsistency is detected, the rollback mechanism is automatically triggered to restore the state before the update, and the inconsistent relationship is marked as pending review.

[0037] 5) Dam safety status reasoning and early warning: Based on spatiotemporal knowledge graph, dam safety status reasoning and early warning are performed, including single-point anomaly detection, associated anomaly detection, and comprehensive early warning judgment; The single-point anomaly detection includes: marking an anomaly when the current value of a deformation monitoring point exceeds a preset threshold, marking an anomaly when the seepage pressure of a seepage monitoring point exceeds a specified ratio (e.g., 120%) of the design value, and marking an anomaly when the stress of a stress monitoring point exceeds the allowable stress. The associated anomaly detection includes: traversing the structural connection relationships and temporal relationships in the knowledge graph to identify associated anomaly points of the anomaly monitoring points; The comprehensive early warning determination includes: determining the early warning level based on the number of single-point anomalies, the degree of clustering of associated anomalies, and the scope of structural influence, in accordance with DL / T 5313-2014 "Guidelines for Safety Evaluation of Hydropower Station Dam Operation".

[0038] 6) Change audit log recording: Record the complete audit log for each change, including the relationship status before and after the change, the change timestamp, the reason for the change, the scope of impact, and the consistency verification results, forming a traceable change history.

[0039] This invention provides a specific example of an automatic construction and evolution method using a spatiotemporal knowledge graph for dam safety monitoring: This embodiment takes a concrete gravity dam as an example, which includes approximately 200 monitoring points, including 120 deformation monitoring points, 50 seepage monitoring points, and 30 stress-strain monitoring points. The specific implementation steps are as follows: Step S100: Obtain the three-dimensional coordinate data of 200 monitoring points in the CGCS2000 national geodetic coordinate system, as well as the dam structure model data (15 dam segment polygons, 3 main joint lines, and geometric data of the foundation structure surface), and perform coordinate accuracy verification to remove outlier coordinate values.

[0040] Using the ST_DWithin spatial function with a preset distance threshold, spatial proximity relationships between monitoring points were identified, and approximately 800 spatial proximity relationships were calculated and identified.

[0041] The ST_Contains spatial function was used to identify the spatial inclusion relationship between monitoring points and dam structural units. A total of 200 spatial inclusion relationships were calculated and identified, meaning that each monitoring point belongs to a dam structural unit.

[0042] The ST_Buffer spatial function is used to generate the influence area buffer for each monitoring point, generating 200 spatial influence buffer geometric objects; among them, the buffer radius of the monitoring point on the dam surface is 50 meters, the buffer radius of the monitoring point inside the corridor is 20 meters, and the buffer radius of the basic monitoring point is 30 meters.

[0043] Spatial proximity relationships between monitoring points of different structural types were excluded (for example, spatial proximity relationships were not established between monitoring points on the dam surface and monitoring points on the slope even if the distance was less than the threshold, because their physical mechanisms are different). After screening, about 50 spatial proximity relationships across structural types were excluded, and about 750 valid spatial proximity relationships were finally retained.

[0044] The output spatial relationship set S includes 750 spatial proximity relationships (each record contains two monitoring point IDs and the calculated distance), 200 spatial containment relationships (each monitoring point corresponds to a dam section), and 200 spatial influence buffer geometry objects.

[0045] Step S110: Based on 750 spatial proximity relationships, the DBSCAN density clustering algorithm is used to cluster 200 monitoring points into measurement point groups; The 18 measuring point groups were mapped to the dam structure units. All the monitoring points in the 15 dam segment measuring point groups were located within the same dam segment polygon and were directly mapped to the corresponding dam segment. The buffer zones of the 3 measuring point groups near the joints intersected with the joint lines and were identified as connection points of adjacent dam segments, thus establishing candidate connection relationships between adjacent dam segments. Specifically, all 18 measuring point groups were successfully mapped to the dam structure units.

[0046] Based on the mapping results, structural relationships were derived. Structural attribution relationships (PART_OF) were established between monitoring points within the same dam section, totaling 200 relationships; structural connection relationships (CONNECTED_BY) were established between monitoring points in adjacent dam sections, totaling 3 relationships; spatial orientation relationships (UPSTREAM_DOWNSTREAM) were established between upstream and downstream monitoring points, totaling 15 relationships; and support relationships (SUPPORTED_BY) were established between the dam body and foundation monitoring points, totaling 20 relationships. Establish a set of structural relationships R_struct, where each relationship records the source monitoring point ID, target monitoring point ID, relationship type, dam section to which it belongs, and confidence level; The structural relationship set R_struct includes: clustering results of 18 measurement point groups, mapping relationships from 18 measurement point groups to dam sections, and 238 structural association relationships (200 PART_OF relationships + 3 CONNECTED_BY relationships + 15 PART_OF relationships + 15 CONNECTED_BY relationships). UPSTREAM_DOWNSTREAM + 20 SUPPORTED_BY).

[0047] Step S120: The time series data is preprocessed, noise is removed by a 3-day window moving average, missing values ​​are filled by linear interpolation, and normalization is achieved by Z-score standardization. Structural relationships were constrained and filtered, and Pearson correlation coefficients were calculated only for approximately 2000 pairs of monitoring points with structural relationships (if structural constraints were not used, the correlation coefficients for C(200,2)=19900 pairs would need to be calculated). Monitoring point pairs with correlation coefficients greater than 0.7 and significance levels less than 0.05 were screened, and approximately 150 pairs passed the correlation test. Structural constraints were used to verify the 150 highly correlated monitoring point pairs, and spurious correlations that were spatially and structurally unrelated were excluded, leaving approximately 120 pairs. Granger causality tests were performed on the 120 monitoring point pairs (significance level 0.05, maximum time lag 7 days), identifying approximately 50 pairs of monitoring point pairs with significant time lag relationships.

[0048] Specifically, determining the type of temporal relationship involves: 100 pairs of monitoring points that change synchronously and are structurally related are identified as having a mechanical coupling relationship. 20 pairs of monitoring points with lagging changes and upstream-downstream relationships were identified as having a hydraulic propagation relationship. 30 pairs of monitoring points that changed synchronously and were associated with the environment were identified as having an environmental driving relationship.

[0049] Output a time series relation set R_temporal, including: 100 mechanical coupling relations, 20 hydraulic propagation relations, and 30 environmental driving relations; each relation contains a source node, a target node, a correlation coefficient, a time delay value, and a relation type.

[0050] Step S130: Based on the spatial relation set S, the structural relation set R_struct, and the temporal relation set R_temporal, construct a spatiotemporal knowledge graph and store it in the graph database Neo4j; the graph database Neo4j stores 750 spatial proximity relations, 200 spatial containment relations, 238 structural association relations, and 150 temporal relations.

[0051] Step S140: Add a new deformation monitoring point TP-201, located within dam section A; the system detects the new event, the coordinate change is greater than the second threshold, and recalculates all relationships of the monitoring point. Record the change event (type: new, monitoring point ID: TP-201, change timestamp), and determine the scope of impact as a full update (new event, change magnitude greater than the second threshold). Calculate the spatial relationship between TP-201 and the existing 200 monitoring points: TP-201 is 30 meters away from TP-5 (less than the threshold of 50 meters), so a spatial proximity relationship is established; TP-201 is located within the polygon of dam segment A, so a spatial inclusion relationship is established; TP-201 belongs to dam segment A, so a structural affiliation relationship (PART_OF) is established, which does not affect the structural connection relationships of other dam segments, and only updates the structural relationships within dam segment A; TP-201 has structural affiliation relationships with approximately 15 monitoring points within dam segment A, so only the temporal correlation of these 15 pairs is calculated; Calculations revealed a mechanical coupling relationship between TP-201 and TP-5 (correlation=0.82, lag=0).

[0052] A consistency verification was performed, comparing the spatial relationships in the PostGIS spatial database with the relationship edges in the Neo4j graph database. All relationships were consistent, and there were no pending audit statuses.

[0053] Record the change audit log (Status before change: no TP-201 related relationships; Status after change: 2 new spatial relationships, 1 new structural relationship, and 1 new temporal relationship; consistency verification passed).

[0054] The final output is the evolved knowledge graph (with 4 new relationships) and 1 change audit log.

[0055] This invention, based on a fusion architecture of spatial and graph databases, achieves fully automated construction and updating of a spatiotemporal knowledge graph for dam safety monitoring through automatic spatial relationship identification, automatic structural relationship derivation, automatic temporal relationship discovery, and dynamic evolution mechanisms. It solves the problems of low efficiency in manual relationship definition, large computational load and susceptibility to spurious correlation interference due to lack of structural constraints in temporal correlation calculation, and high cost of full reconstruction after monitoring point changes in traditional knowledge graph construction. It significantly improves the efficiency, accuracy, and dynamic maintenance capabilities of knowledge graph construction.

[0056] The above-disclosed embodiments are merely a few specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for the automatic construction and evolution of a spatiotemporal knowledge graph for dam safety monitoring, characterized in that, include: Collect 3D coordinate data of monitoring points and data of dam structure model, automatically identify spatial proximity and spatial inclusion relationships between monitoring points, generate spatial influence buffer, filter spatial relationships and establish a set of spatial relationships; Based on the spatial relationship set and combined with the dam structure model, a set of structural relationships between monitoring points and dam structural units is automatically established; the set of structural relationships includes structural attribution relationships, structural connection relationships, spatial orientation relationships, and support relationships. Acquire time-series data from monitoring points, preprocess the time-series data, and establish a set of time-series relationships; The set of temporal relationships includes: mechanical coupling relationships, hydraulic propagation relationships, and environmental driving relationships between monitoring points; A spatiotemporal knowledge graph is constructed based on the spatial relationship set, structural relationship set, and temporal relationship set, and stored in a graph database. The spatiotemporal knowledge graph is a knowledge network with dam monitoring points as nodes and three layers of relationships—spatial proximity, structural affiliation, and temporal causality—as edges. Based on a spatiotemporal knowledge graph, dam change events are monitored in real time, and the types of change events and change coordinate data are recorded. The scope of impact is determined based on the magnitude of coordinate changes, and the spatiotemporal knowledge graph is dynamically evolved. Consistency verification and rollback are performed on the affected spatial, structural, and temporal relationships after the update, dam safety status reasoning and early warning are carried out, and change audit logs are recorded.

2. The method for automatic construction and evolution of the spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, The three-dimensional coordinate data includes the X, Y, and Z coordinates of deformation monitoring points, seepage monitoring points, and stress-strain monitoring points; the dam structure model is generated by instantiating conceptual nodes based on the definition model of a real dam; the data of the dam structure model includes the geometric data of dam segment polygons, joint lines, and foundation structural surfaces; the data preprocessing includes denoising using the moving average method, filling missing values ​​using the linear interpolation method, and normalizing using the Z-score standardization method.

3. The method for automatic construction and evolution of the spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, When establishing the spatial proximity relationship identification: calculate the Euclidean distance between monitoring points in batches, filter monitoring point pairs with a distance less than the threshold, and establish spatial proximity relationships; when establishing the spatial inclusion relationship identification: perform spatial inclusion judgment between the coordinates of each monitoring point and the polygon of the dam structure unit, identify the dam structure unit to which each monitoring point belongs, and establish the spatial inclusion relationship between the monitoring point and the dam structure unit; when generating the spatial influence buffer: generate a spatial influence buffer with each monitoring point as the center and the corresponding type of distance threshold as the radius.

4. The method for automatic construction and evolution of the spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, The set of structural relationships is established through spatial density clustering, structural mapping, and association derivation.

5. The method for automatic construction and evolution of the spatiotemporal knowledge graph for dam safety monitoring according to claim 4, characterized in that, The spatial density clustering refers to the use of the DBSCAN density clustering algorithm to cluster monitoring points into monitoring point groups based on spatial proximity. The association derivation refers to the structural affiliation relationship between monitoring points within the same dam section, the structural connection relationship between monitoring points in adjacent dam sections, the spatial orientation relationship between upstream and downstream monitoring points, and the support relationship between the dam body and the foundation monitoring points.

6. The method for automatic construction and evolution of the spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, Establishing the time series relation set includes: preprocessing the time series data by denoising, imputing missing values ​​and normalizing, performing full correlation screening and time delay detection, and using structural relations for constraint verification and filtering out spurious correlations.

7. The method for automatic construction and evolution of a spatiotemporal knowledge graph for dam safety monitoring according to claim 6, characterized in that, The filtering of false correlations refers to: using structural relationships for constraint verification, retaining only monitoring point pairs that simultaneously meet the two conditions of "significant correlation" and "existence of structural association", and excluding false correlations between monitoring point pairs that are spatially and structurally unrelated due to data randomness; the time lag detection refers to: performing time lag detection on the monitoring point pairs filtered by structural constraints to determine whether there is a time lead and lag relationship between the changes of the two monitoring points.

8. The method for automatic construction and evolution of a spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, The dynamically evolving spatiotemporal knowledge graph includes: using a cascading triggering mechanism to locally update affected spatial relationships, structural relationships, and temporal relationships, thereby achieving dynamic evolution of the spatiotemporal knowledge graph; the cascading triggering mechanism includes: triggering the re-derivation of structural relationships when spatial relationships change, triggering the recalculation of temporal relationships when structural relationships change, and keeping unaffected relationships unchanged.

9. The method for automatic construction and evolution of a spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, The method of determining the impact range based on the magnitude of coordinate change includes: pre-setting critical displacement thresholds for each type of monitoring point, comparing the coordinate changes before and after the change with the critical displacement thresholds, and marking the change as a major change if it exceeds the threshold. The spatial neighborhood of the monitoring point where the major change occurs is expanded to a preset level as the impact range.

10. The method for automatic construction and evolution of a spatiotemporal knowledge graph for dam safety monitoring according to claim 1, characterized in that, The consistency verification and rollback refer to: verifying the consistency of relational data in the two types of storage by comparing the spatial relations in the spatial database with the relational edges in the graph database; when inconsistency is detected, automatically triggering the rollback mechanism to restore the state before the update, and marking the inconsistent relation as pending review. The dynamic evolution includes: monitoring change events, recording the change event type, the monitoring points involved, the coordinate values ​​before and after the change, and the change timestamp, and recording the change audit log after the update is completed, including: the relationship status before and after the change, the change timestamp, the change type, and the scope of impact.

Images