Underground pipe network full life cycle data intelligent management method and system

Abstract

The invention discloses an underground pipe network full life cycle data intelligent management method and system, and relates to the data management related technical field, and the method comprises the steps: building a BIM-GIS fusion model; entity extraction is carried out, and multi-dimensional entity modeling is executed; performing entity alignment and fusion on the multi-dimensional entities, performing association analysis, and establishing an association knowledge graph; receiving data streams from design, construction, monitoring and inspection, and carrying out data time sequence mounting and updating through a dynamic attribute mapping engine; analyzing and configuring a retrieval template for the retrieval request based on semantic analysis; and executing multi-scale sliding window analysis of the candidate object set, and outputting an update result. The technical problems of data islanding, information staticization, management fragmentation and insufficient retrieval analysis capability in the prior art are solved, and the technical effects of realizing full-life-cycle data fusion, dynamic real-time perception, cross-entity association analysis and intelligent and accurate retrieval, and improving the safety and operation and maintenance management efficiency of the underground pipe network are achieved.

Description

Technical Field

[0001] This application relates to the field of data management technology, specifically to intelligent management methods and systems for the entire lifecycle of underground pipeline networks. Background Technology

[0002] Underground pipe networks, as the lifeline of urban operations, encompass various types of water supply, drainage, gas, heating, electricity, and communications. Widely distributed and structurally complex, they are often hidden, making management challenging. Traditional underground pipe network management relies heavily on drawings, manual inspections, and fragmented information systems, resulting in data silos, delayed information updates, weak dynamic sensing capabilities, and difficulties in cross-stage data fusion. This makes it difficult to meet the demands of modern cities for pipe network safety, reliability, and intelligent management. While Building Information Modeling (BIM) and Geographic Information System (GIS) provide support for the visualization and information-based management of underground pipe networks, BIM lacks macro-geographic context support, and GIS struggles to support detailed component-level information. Furthermore, the massive amounts of dynamic data generated during underground pipe network operation, such as pressure, flow, and temperature monitoring data, as well as inspection and maintenance records, are difficult to integrate effectively, impacting the level of precision and intelligence in underground pipe network management.

[0003] Therefore, current technologies suffer from technical problems such as data silos, static information, fragmented management, and insufficient retrieval and analysis capabilities. Summary of the Invention

[0004] This application provides an intelligent management method and system for the entire lifecycle data of underground pipeline networks, which solves the technical problems of data silos, static information, fragmented management, and insufficient retrieval and analysis capabilities in existing technologies. It achieves the technical effects of realizing full lifecycle data fusion, dynamic real-time perception, cross-entity correlation analysis, and intelligent and accurate retrieval, thereby improving the safety and operation and maintenance management efficiency of underground pipeline networks.

[0005] This application provides a method for intelligent management of underground pipeline network data throughout its entire lifecycle. The method includes: establishing a BIM-GIS fusion model, which contains a set of pipeline network objects, a set of graphic features, a set of static attributes, and a set of dynamic attribute fields; extracting entities from the BIM-GIS fusion model and performing multi-dimensional entity modeling, whereby the multi-dimensional entities include structural entities, operational entities, and environmental entities; aligning and fusing the multi-dimensional entities based on unique identifiers, and then performing correlation analysis on the fusion results through topological relationships, dependencies, influence relationships, and event relationships to establish a correlation knowledge graph; receiving data streams from design, construction, monitoring, and inspection, and performing time-series mounting and updating of the BIM-GIS fusion model and the correlation knowledge graph through a dynamic attribute mapping engine; upon receiving a retrieval request, parsing the retrieval request based on semantic parsing, configuring a retrieval template, and performing a joint retrieval in the BIM-GIS fusion model and the correlation knowledge graph to establish a candidate object set; and performing multi-scale sliding window analysis on the candidate object set and outputting updated results.

[0006] In a possible implementation, the intelligent management method for the entire lifecycle data of underground pipeline networks further performs the following processes: using the unique identifier to perform spatial alignment, topological alignment, and temporal alignment of multi-dimensional entities to establish a fusion result; establishing entity topological relationships within the fusion result based on the BIM model topology in the BIM-GIS fusion model; identifying dependencies based on the binding relationships of operating entities and structural entities in the fusion result to establish entity dependency relationships; performing spatial overlay analysis on environmental entities and structural entities in the fusion result to establish entity influence relationships; performing multi-dimensional entity operation and maintenance log analysis to establish entity event relationships; and establishing an associated knowledge graph using the entity topological relationships, entity dependency relationships, entity influence relationships, and entity event relationships.

[0007] In a possible implementation, the intelligent management method for the entire lifecycle data of underground pipeline networks further performs the following processing: constructing multiple semantic views for each entity within the associated knowledge graph, encoding the multiple semantic views into vector representations using a graph neural network encoder, and projecting them onto a common semantic space; introducing contrastive learning during the projection onto the common semantic space, the contrastive learning including: using different semantic views of the same entity as positive sample pairs and different entities as negative sample pairs, and performing standard contrastive learning using the InfoNCE loss function; configuring spatiotemporal conditional weights to perform contrastive learning, the spatiotemporal conditional weights including time decay weights and spatial proximity weights, the spatial proximity weights being constructed by modeling edge weights of entity topological relationships and entity dependencies; establishing embedded semantic features based on the common semantic space, and performing semantic enhancement of the associated knowledge graph through the embedded semantic features.

[0008] In a possible implementation, the intelligent management method for the entire life cycle data of underground pipeline networks also performs the following processing: sending the retrieval request to a multi-layer semantic parsing network, using the natural language parsing layer, domain ontology mapping layer and semantic rule matching layer in the multi-layer semantic parsing network to perform multi-layer semantic parsing respectively, and establishing a multi-dimensional semantic representation; performing interactive authentication fusion of the multi-dimensional semantic representation, establishing a parsing fusion result, and configuring a retrieval template according to the parsing fusion result.

[0009] In a possible implementation, the intelligent management method for the entire lifecycle data of underground pipeline networks further performs the following processing: extracting spatial index conditions based on the location data of the parsing fusion results using the spatial constraint generator; extracting attribute conditions based on the parameter types and numerical ranges of the parsing fusion results using the attribute constraint generator; generating temporal conditions based on the temporal entities and evolution models of the parsing fusion results using the temporal constraint generator; establishing logical constraint conditions based on the relational operators and ontology relations of the parsing fusion results using the semantic logic generator; and combining the extracted index conditions to establish a retrieval template.

[0010] In a possible implementation, the intelligent management method for the entire life cycle data of underground pipeline networks also performs the following processing: based on the spatial index conditions and attribute conditions in the retrieval template, the spatial index and attribute index are called in the BIM-GIS fusion model to establish a first retrieval result; based on the logical constraints and temporal conditions in the retrieval template, the graph database retrieval and graph reasoning are called in the associated knowledge graph to establish a second retrieval result; the first retrieval result and the second retrieval result are cross-source aligned to establish a candidate object set.

[0011] In a possible implementation, the intelligent management method for the entire life cycle data of underground pipeline networks also performs the following processing: one-to-one alignment based on the unique identifier of the entity; if any entity lacks a unique identifier, joint judgment and alignment are performed through semantic similarity calculation and spatial proximity calculation.

[0012] In a possible implementation, the intelligent management method for the entire life cycle data of underground pipeline networks further performs the following processing: configuring multiple time backtracking scales based on the stability and complexity characteristics of the candidate object set, activating a multi-scale sliding window using the multiple time backtracking scales; performing time-series trajectory anomaly verification of the candidate object set under the multi-scale sliding window; updating the risk score of the candidate object set using the time-series trajectory anomaly verification results, and generating an updated result using the risk score and the candidate object set.

[0013] In a possible implementation, the intelligent management method for the entire life cycle data of underground pipelines also performs the following processing: the dynamic attribute field set includes the circulating medium, direction, location, and burial depth, and each pipeline object in the pipeline object set is configured with a confidence label and a timestamp sequence to form a traceable pipeline object identifier.

[0014] This application also provides an intelligent management system for the entire lifecycle of underground pipeline networks. The system includes: a resume screening item segmentation module for receiving resume screening items for a target position and segmenting them into semantically explicit items and fuzzy skill items; a defuzzification analysis module for performing defuzzification analysis based on skill-related semantics for the fuzzy skill items, and constructing a defuzzification adaptation template library; a resume reconstruction module for receiving a resume dataset corresponding to the target position and reconstructing resumes according to a preset resume template to generate a reconstructed resume set; a reconstructed resume set screening module for performing a first screening on the reconstructed resume set based on the semantically explicit items to generate a first resume set; and a fit analysis module for performing a second screening on the first resume set based on the fuzzy adaptation template library to obtain a second resume set, performing context-aware core coefficient analysis on the second resume set, performing a third screening based on the core coefficients, and combining the semantically explicit items to analyze the fit between each resume in the third resume set and the target position, sorting the resumes and recommending them to the target HR terminal.

[0015] This application proposes a method and system for intelligent management of underground pipeline network data throughout its entire lifecycle. The system establishes a BIM-GIS fusion model; extracts entities and performs multi-dimensional entity modeling; aligns and merges these multi-dimensional entities and performs correlation analysis to establish a relational knowledge graph; receives data streams from design, construction, monitoring, and inspection, and performs time-series data mounting and updating through a dynamic attribute mapping engine; parses and configures retrieval templates based on semantic parsing; performs multi-scale sliding window analysis of candidate object sets, and outputs updated results. This addresses the technical problems of data silos, static information, fragmented management, and insufficient retrieval and analysis capabilities in existing technologies. It achieves the technical effects of full lifecycle data fusion, dynamic real-time perception, cross-entity correlation analysis, and intelligent and accurate retrieval, thereby improving the safety and operation and maintenance efficiency of underground pipeline networks. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 A schematic diagram of the intelligent management method for the entire life cycle data of underground pipelines provided in this application embodiment.

[0018] Figure 2 A schematic diagram of the structure of the intelligent management system for the entire life cycle data of underground pipelines provided in this application embodiment.

[0019] Figure labeling: Fusion model building module 10, entity modeling module 20, association analysis module 30, data update module 40, joint retrieval module 50, update result output module 60. Detailed Implementation

[0020] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides an intelligent management method for the entire lifecycle data of underground pipeline networks, such as... Figure 1 As shown, the method includes: Step S100: Establish a BIM-GIS fusion model, which includes a set of pipeline objects, a set of graphic features, a set of static attributes, and a set of dynamic attribute fields.

[0024] Step S100 further includes that the dynamic attribute field set includes the circulation medium, direction, location, and burial depth, and each pipeline object in the pipeline object set is configured with a confidence label and a timestamp sequence to form a traceable pipeline object identifier.

[0025] Preferably, a BIM-GIS fusion model is constructed using a set of pipeline objects, a set of graphic features, a set of static attributes, and a set of dynamic attribute fields. This model is used to simultaneously manage the internal attributes and external environment of pipeline facilities. Specifically, the set of pipeline objects refers to the collection of all managed pipeline facility instances. Each element in the set is a specific and uniquely identifiable component of the pipeline network, such as a specific pipe section, a particular valve, pump station, or manhole cover. The set of graphic features refers to data describing the geometric shape and spatial appearance of pipeline objects, including 3D models, 2D vector graphics, textures, and rendering styles on GIS maps, ensuring the model's visual representation. The set of static attributes refers to inherent attribute data that is relatively stable throughout the pipeline network's lifecycle and does not change frequently over time. This typically includes design parameters such as pipe diameter, wall thickness, and material, as well as information on water supply systems, drainage systems, model specifications, and construction details. The set of dynamic attribute fields refers to predefined data used to store data that is updated over time or changes in status. The data field structure includes the flow medium, direction, location, and burial depth. The flow medium records the substance currently flowing in the pipeline, such as water, natural gas, or sewage. The direction records the flow direction of the medium, which may be reversed due to pump and valve operation. The location records the spatial coordinates of the object, specifically referring to pipelines that may experience minor displacement or settlement due to geological subsidence, engineering modifications, etc., requiring periodic monitoring to update their precise location. The burial depth records the depth of the soil covering the top of the pipeline. Each dynamic attribute field is a sequence of timestamp-value pairs, enabling the recording and tracing of the attribute's state at any historical point in time. The confidence label attaches a quality assessment index to the attribute data of each pipeline object, indicating the reliability or accuracy of the data source; for example, design drawing data has a high confidence level, while inferred data from older pipelines has a medium confidence level. The timestamp sequence records a precise timestamp for each data update of each object, making all historical state changes of each object traceable.

[0026] Step S200: Extract entities from the BIM-GIS fusion model and perform multi-dimensional entity modeling. Multi-dimensional entities include structural entities, operational entities, and environmental entities.

[0027] Preferably, entity extraction refers to identifying, classifying, and extracting independent data objects with clear semantic and management significance from the BIM-GIS fusion model. This involves semantically recognizing and reorganizing elements in the fusion model based on predefined rules and classification systems to obtain multi-dimensional entities, including structural entities, operational entities, and environmental entities. Multi-dimensional entity modeling is then performed, assigning attributes beyond geometric shape to entities from different dimensions and establishing relationships. Structural entities refer to the static, tangible components representing the physical structure of the pipeline network system. These are primarily extracted from the components of the BIM model and their spatial location information from the GIS. Attributes include geometric attributes such as shape, size, material, and wall thickness, and spatial attributes such as coordinates, altitude, depth, and orientation. The inherent attributes include the number, design life, and installation date; the operating entity refers to the state of the flowing medium and the working state of the equipment in the pipeline system, which is obtained through binding sensors, real-time data streams from monitoring systems, and business records. Attributes include state parameters such as pressure, flow rate, temperature, and flow velocity, control states such as valve opening, pump start / stop status, and equipment running time, and media attributes such as water, gas, and oil; the environmental entity refers to the entity representing the external environment of the pipeline system, which is mainly extracted from GIS data, geological survey data, and urban planning data. Attributes include geological environment such as soil type, corrosivity, groundwater level, and geological faults, surface environment such as topography, roads, buildings, and vegetation cover, and social environment such as administrative divisions, population density, and important facilities.

[0028] Step S300: After performing entity alignment and fusion based on unique identifiers on the multidimensional entities, the fusion results are analyzed for association through topological relationships, dependency relationships, influence relationships, and event relationships to establish an association knowledge graph.

[0029] Step S300 further includes step S310, using the unique identifier to perform spatial alignment, topological alignment, and temporal alignment of multi-dimensional entities to establish a fusion result; step S320, establishing entity topological relationships within the fusion result based on the BIM model topology in the BIM-GIS fusion model; step S330, identifying dependencies based on the binding relationships of running entities and structural entities in the fusion result to establish entity dependency relationships; step S340, performing spatial overlay analysis on environmental entities and structural entities in the fusion result to establish entity influence relationships; step S350, performing multi-dimensional entity operation and maintenance log analysis to establish entity event relationships; and step S360, establishing an associated knowledge graph using the entity topological relationships, entity dependency relationships, entity influence relationships, and entity event relationships.

[0030] Preferably, unique identifiers are used for spatial, topological, and temporal alignment of multi-dimensional entities. This ensures that entities representing the same physical object from different dimensions have consistent coordinates in three-dimensional space, that the connections between entities from different sources are correct, and that dynamic data, static models, and external events are synchronized on the timeline to ensure consistent timestamps across all data. The aligned data is then fused to form a spatiotemporally consistent, high-quality multi-dimensional entity dataset as the fusion result. BIM model topology refers to entity connectivity. This topology is explicitly defined as relational edges in the fusion result, defining the physical connections between entities, determining the entity topology, and accurately describing the physical network structure of the pipeline. Then, dependency identification is performed based on the binding relationships between operating entities and structural entities in the fusion result. This involves analyzing the functional dependencies between operating entities and their attached structural entities, such as control and being controlled, monitoring and being monitored. For example, a pressure reading operating entity depends on a pipeline structural entity; a water pump structural entity depends on a power line environmental entity.

[0031] Preferably, spatial analysis algorithms such as buffer analysis and overlay analysis are used to perform spatial overlay analysis on environmental entities and structural entities in the fusion results. This quantifies the potential impact of environmental entities on structural entities, including calculating the spatial intersection of environmental entity polygons and structural entities, and establishing the potential risk impact relationship of the external environment on the pipeline network itself. The relationship attributes may include impact intensity and risk level. Natural language processing and text mining are used to perform multi-dimensional entity operation and maintenance log analysis, that is, extracting event information from unstructured operation and maintenance logs and associating it with entities, thereby establishing entity-event relationships to inject historical event knowledge into a knowledge graph. Finally, entity topology relationships, entity dependencies, entity impact relationships, and entity-event relationships are used as connecting edges, and multi-dimensional entities are integrated as nodes to form a directed attribute graph containing various types of nodes and relationships, thus determining the associated knowledge graph.

[0032] Furthermore, step S360 also includes step S361, constructing multiple semantic views for each entity in the associated knowledge graph, encoding the multiple semantic views into vector representations through a graph neural network encoder, and projecting them onto a common semantic space; step S362, introducing contrastive learning during the projection onto the common semantic space, the contrastive learning including: step S363, using different semantic views of the same entity as positive sample pairs and different entities as negative sample pairs, and performing standard contrastive learning using the InfoNCE loss function; step S364, configuring spatiotemporal conditional weights to perform contrastive learning, the spatiotemporal conditional weights including time decay weights and spatial proximity weights, the spatial proximity weights being constructed by modeling edge weights of entity topological relationships and entity dependencies; step S365, establishing embedded semantic features based on the common semantic space, and performing semantic enhancement of the associated knowledge graph through the embedded semantic features.

[0033] Preferably, a semantic view is constructed for each entity within the associated knowledge graph from different dimensions, including a structural view, a topological view, an operational view, and an environmental view. Multiple graph neural network encoders encode these semantic views into vector representations. The graph neural network encoder is a neural network used to process graph structure data, aggregating information about the node itself and its neighboring nodes to generate a vector representation for that node. The structural view is a feature vector based on its own attributes; the topological view is a feature vector based on its connectivity within the network; the operational view is a feature vector based on its historical operational data; and the environmental view is a feature vector based on its surrounding environment. Then, the vectors generated from all semantic views are projected into a common semantic space. Entities with similar semantics have geometrically close vectors, allowing for direct comparison and computation of information from different sources.

[0034] Preferably, contrastive learning is introduced during the projection onto the common semantic space. This includes using different semantic views of the same entity as positive sample pairs and different entities as negative sample pairs, and performing standard contrastive learning using the InfoNCE loss function. Specifically, a positive sample pair refers to a vector pair generated by different views of the same entity, such as the structure vector of pipe A and the running vector of pipe A. A negative sample pair refers to a vector pair composed of vectors from different entities, such as the structure vector of pipe A and the running vector of pipe B. The InfoNCE loss function is a mathematical function used for contrastive learning. Standard contrastive learning is performed by calculating the ratio of the similarity between positive sample pairs to the similarity between positive and negative sample pairs, that is, maximizing the consistency between positive sample pairs while minimizing the similarity between negative sample pairs. Spatiotemporal conditional weights are configured to perform contrastive learning. These weights include time decay weights and spatial proximity weights. Time decay weights indicate that when processing historical events or operational data, data closer to the current time has higher importance. Spatial proximity weights mean that when constructing negative samples, not all different entities are pushed away. Spatial proximity weights are constructed through the edge weights of topological and dependency relationships between entities. The closer two entities are in space or function, the more semantically similar their vectors are. This reduces the penalty for using spatially or functionally adjacent entities as negative samples in contrastive learning, preventing them from being pushed too far away, thus more accurately reflecting the semantic relationships in the real world. Embedded semantic features are established based on a common semantic space, which are the final vector representations of each entity obtained after comparative learning. Finally, semantic enhancement of the associated knowledge graph is performed through embedded semantic features, that is, using high-quality vectors to improve the ability of the entire associated knowledge graph, thereby quickly identifying all other pipelines that are most similar to a leaking pipeline in terms of structure, operation and environment. This is used for risk assessment, predicting whether there are undiscovered dependencies or influence relationships between two entities, or if the vector of an entity is far away from the vector cluster of its similar entities, it may indicate that the entity is in an abnormal state.

[0035] Step S400: Receive data streams from design, construction, monitoring, and inspection, and use the dynamic attribute mapping engine to perform time-series mounting and updating of the BIM-GIS fusion model and related knowledge graph.

[0036] Preferably, data streams for design, construction, monitoring, and inspection are received from design software, construction management platforms, IoT platforms, and inspection apps. Specifically, design data streams refer to incremental updates generated during the planning and design phase, such as design change notices, updated drawings, and modified model files; construction data streams refer to as-built data generated during the construction phase, such as actual installation coordinates, equipment models used, construction acceptance reports, and project records; monitoring data streams refer to real-time time-series data transmitted from IoT sensors installed on the pipeline network, such as pressure, flow, temperature, vibration, and corrosion monitoring data; and inspection data streams refer to records generated from manual inspections, drone or robot inspections, such as photos and videos taken on-site, manually filled inspection forms, and maintenance reports. The dynamic attribute mapping engine is a software middleware used for pattern mapping and entity recognition. It predefines the correspondence between external data sources and dynamic attribute field sets in the fusion model. When data is received, the dynamic attribute mapping engine quickly locates the structural entity bound to the sensor in the BIM-GIS model using a unique identifier. Then, the dynamic attribute mapping engine is used to mount and update the BIM-GIS fusion model and the associated knowledge graph with time series data. This includes adding each new data point as a data pair to the time series of the dynamic attributes corresponding to the entity, thereby forming traceable historical data and plotting the curve of pressure changing over time; updating the time series data in the dynamic attribute field set of the BIM-GIS fusion model; and updating the entity status and relationships in the associated indicator graph.

[0037] Step S500: After receiving the retrieval request, the retrieval request is parsed based on semantic parsing. After configuring the retrieval template, a joint retrieval is performed in the BIM-GIS fusion model and the associated knowledge graph to establish a candidate object set.

[0038] Step S500 further includes step S510, sending the retrieval request to a multi-layer semantic parsing network, and using the natural language parsing layer, domain ontology mapping layer and semantic rule matching layer in the multi-layer semantic parsing network to perform multi-layer semantic parsing and establish a multi-dimensional semantic representation; step S520, performing interactive authentication fusion of the multi-dimensional semantic representation, establishing a parsing fusion result, and configuring a retrieval template according to the parsing fusion result.

[0039] Preferably, the received retrieval request is parsed based on semantic parsing, i.e., it is input into a multi-layer semantic parsing network. This multi-layer semantic parsing network consists of a natural language parsing layer, a domain ontology mapping layer, and a semantic rule matching layer. Each layer is responsible for different levels of semantic understanding, gradually transforming the original query into a machine-readable format. Specifically, the natural language parsing layer performs basic natural language processing tasks, including splitting the query sentence into words / vocabulary units, identifying the part-of-speech of each word, identifying and classifying key entities in the query, and analyzing the grammatical relationships between words to construct a syntax tree. The domain ontology mapping layer maps general vocabulary to specific concepts in the underground pipeline network professional domain. The system uses a predefined domain ontology that defines all concepts, attributes, and relationships within the pipeline network domain. It receives the output from the natural language parsing layer and matches the identified entities and words with the concepts in the ontology. A semantic rule matching layer further enriches the query semantics using predefined business rules. For example, when a search request mentions pressure fluctuations, it defaults to calculating the difference between the maximum and minimum values ​​over the past 30 days divided by the average, ensuring that all searches only target pipelines in service unless the user explicitly specifies abandoned pipelines. The results of the three-layer parsing are then used to build a multi-dimensional semantic representation, including intent, entity, spatial, attribute, and time dimensions. Interactive authentication and fusion of the multi-dimensional semantic representation are then performed, including checking for contradictions or ambiguities within the representation, such as whether the time range and fluctuation calculation method match. A simple interactive confirmation is then conducted with the user, fusing all dimensions into a complete description of the query intent as the parsing and fusion result. Finally, a search template is configured based on the parsing and fusion result, generating an executable query framework containing all search conditions. This provides precise input for joint searches in the BIM database and knowledge graph, improving the intelligence and accuracy of the search.

[0040] Furthermore, step S520 also includes extracting index conditions from the parsing fusion results using a spatial constraint generator, an attribute constraint generator, a temporal constraint generator, and a semantic logic generator, respectively. This includes step S521, extracting spatial index conditions based on the location data of the parsing fusion results using the spatial constraint generator; step S522, extracting attribute conditions based on the parameter types and numerical ranges of the parsing fusion results using the attribute constraint generator; step S523, generating temporal conditions based on the temporal entities and evolutionary models of the parsing fusion results using the temporal constraint generator; step S524, establishing logical constraint conditions based on the relational operators and ontology relations of the parsing fusion results using the semantic logic generator; and step S525, combining the extracted index conditions to establish a retrieval template.

[0041] Preferably, the index conditions of the parsed fusion results are extracted using a spatial constraint generator, an attribute constraint generator, a temporal constraint generator, and a semantic logic generator, respectively. Specifically, the spatial constraint generator extracts spatial index conditions based on the location data of the parsed fusion results, describing the spatial location data relationships as precise GIS spatial query conditions for quickly filtering ranges using the spatial database index; the attribute constraint generator extracts attribute conditions based on the parameter types and numerical ranges of the parsed fusion results, converting abstract attribute requirements into specific field names and value range matching conditions for filtering in relational databases or attribute indexes; and the temporal constraint generator extracts index conditions based on the temporal entities and evolution of the parsed fusion results. The model generates time-series conditions, which involves converting time entities into absolute or relative time windows and evolutionary models into calls to time series analysis functions, such as determining whether a certain time series data satisfies a monotonically increasing trend, for filtering and analysis of time series databases or data with timestamps. A semantic logic generator is used to establish logical constraints based on relational operators and ontology relationships from the parsed fusion results, converting relational operators into Boolean logic and high-level relationships based on domain ontology into traversal paths in the knowledge graph. This defines the combination methods between various conditions and the paths that need to be traversed in the knowledge graph. Finally, the extracted index conditions are combined to create a retrieval template containing all extracted index conditions.

[0042] Furthermore, step S500 also includes step S530, which involves calling the spatial index and attribute index in the BIM-GIS fusion model based on the spatial index conditions and attribute conditions in the retrieval template to establish a first retrieval result; step S540, which involves calling graph database retrieval and graph reasoning in the associated knowledge graph based on the logical constraints and temporal conditions in the retrieval template to establish a second retrieval result; and step S550, which involves cross-source alignment of the first retrieval result and the second retrieval result to establish a candidate object set.

[0043] Preferably, based on the spatial index conditions and attribute conditions in the retrieval template, a retrieval is performed in the BIM-GIS fusion model. The spatial index is used to quickly locate candidate objects that meet the spatial index conditions, eliminating the need to check the coordinates of each object individually. Then, the attribute index is used to quickly find objects that meet the attribute conditions. The set of objects that simultaneously meet both spatial and attribute conditions is taken as the first retrieval result, primarily based on the inherent characteristics and spatial location of the objects. Next, based on the logical constraints and temporal conditions in the retrieval template, a retrieval is performed in the associated knowledge graph. This includes using a graph query language to traverse the relational network in the knowledge graph and utilizing predefined relational rules in the knowledge graph to discover implicit conclusions. Temporal conditions are then used to filter the temporal data attached to entities. Finally, the set of entities found through relational traversal and reasoning that meet the temporal conditions is taken as the second retrieval result, primarily based on the interrelationships and dynamic behaviors of objects in the network. Then, cross-source alignment is performed on the first and second retrieval results, i.e., identifying objects in both result sets that point to the same physical entity, thereby establishing a candidate object set.

[0044] Furthermore, step S550 also includes step S551, which performs one-to-one alignment based on the unique identifier of the entity; and step S552, which performs joint alignment by semantic similarity calculation and spatial proximity calculation if any entity lacks a unique identifier.

[0045] Preferably, the unique identifiers of entity A in the first search result and entity B in the second search result are compared. If the unique identifiers are exactly the same, entity A and entity B are determined to be the same object, and all information of the two is merged. If either entity lacks a unique identifier, a joint determination and alignment is performed through semantic similarity calculation and spatial proximity calculation. Specifically, semantic similarity calculation refers to using a string similarity algorithm to compare the descriptive attributes of two entities, such as name, type, material, and pipe diameter, to calculate their semantic matching degree and calculate a weighted semantic similarity score. Spatial proximity calculation refers to using GIS spatial calculation functions to compare the spatial coordinates of the geometric center points or feature points of two entities and calculate their distance in the physical world. The closer the distance, the higher the spatial proximity score. Then, the semantic similarity score and spatial proximity score are combined for judgment. If the semantic similarity score is higher than a set threshold and the spatial distance is less than a tolerance threshold, they are determined to be the same entity. If one of the conditions is not met, they are determined to be different entities, thereby constructing a high-quality, non-repeating, and information-complete candidate object set.

[0046] Step S600: Perform multi-scale sliding window analysis on the candidate object set and output the update results.

[0047] Step S600 further includes step S610, configuring multiple time backtracking scales based on the stability and complexity characteristics of the candidate object set, and activating a multi-scale sliding window using the multiple time backtracking scales; step S620, performing time-series trajectory anomaly verification of the candidate object set under the multi-scale sliding window; step S630, updating the risk score of the candidate object set using the time-series trajectory anomaly verification result, and generating an updated result using the risk score and the candidate object set.

[0048] Preferably, a multi-scale sliding window analysis is performed on the candidate object set to verify its anomalies and dynamically assess its risks. Specifically, stability characteristics refer to the frequency and magnitude of changes in the attributes of an entity object. For example, the material of a pipe is extremely stable, while the pressure fluctuates frequently. Complexity characteristics refer to the complexity of the object's behavior patterns. For example, the energy consumption curve of a pumping station is periodic, while the flow curve of a pipe located in a complex urban pipe network is complex. Based on the stability and complexity characteristics of the candidate object set, multiple time backtracking scales are dynamically configured for different objects or different analysis purposes. That is, different historical time window lengths are configured. For example, a short-scale window is configured for parameters with high-frequency fluctuations, a long-scale window is configured for slow evolution phenomena such as corrosion rate and sedimentation trend, and a periodic-scale window is configured for periodic behaviors such as water use, to compare the differences between the current period and the historical period. Then, multiple time backtracking scales are used to simultaneously activate multiple time windows of different lengths for each candidate object, and the historical time series data of the object is analyzed in parallel at different time granularities.

[0049] Preferably, a time-series trajectory refers to a curve formed by the change of specific dynamic attribute values ​​of an object over time within a certain time window. Anomaly verification of the time-series trajectories of candidate object sets is performed under a multi-scale sliding window. This involves using anomaly detection algorithms within each time-scale window, employing historical behavior as evidence, to analyze the time-series trajectory and determine the presence of abnormal patterns. Specifically, at short-scale, abrupt change detection algorithms are used to identify sudden peaks or drops; at long-scale, trend analysis algorithms are used to identify sustained, slow increases or decreases, such as a sustained, slow decrease in pressure potentially indicating a leak; and at periodic scales, periodic decomposition algorithms are used to break down the data into trends, periods, and residuals, then analyzing residual anomalies or variations in periodic patterns to obtain the time-series trajectory anomaly verification results. Then, the risk score of the candidate object set is updated using the time-series trajectory anomaly verification results. That is, the real-time risk score of the object is quantitatively adjusted and updated by comprehensively analyzing the results of multi-scale anomaly verification. Finally, the updated results are generated, including the candidate object set that has been confirmed as normal through historical verification, the latest risk score of each object, anomaly diagnosis information, and visualization charts such as time-series curves marked with anomaly points. This enables full life-cycle data fusion, dynamic real-time perception, cross-entity correlation analysis, and intelligent and accurate retrieval, thereby greatly improving the accuracy and efficiency of underground pipeline network operation and maintenance management decisions.

[0050] In the above text, refer to Figure 1 This paper describes in detail a method for intelligent management of underground pipeline network data throughout its entire lifecycle, according to embodiments of the present invention. Next, we will refer to... Figure 2 This invention describes an intelligent management system for the entire lifecycle data of underground pipeline networks according to embodiments of the present invention.

[0051] The intelligent management system for the entire lifecycle data of underground pipeline networks according to embodiments of the present invention addresses the technical problems of data silos, static information, fragmented management, and insufficient retrieval and analysis capabilities in existing technologies. It achieves the technical effects of realizing full lifecycle data fusion, dynamic real-time perception, cross-entity correlation analysis, and intelligent and accurate retrieval, thereby improving the safety and operation and maintenance efficiency of underground pipeline networks. Figure 2 As shown, the intelligent management system for the entire life cycle of underground pipeline network data includes: a fusion model establishment module 10, an entity modeling module 20, a correlation analysis module 30, a data update module 40, a joint retrieval module 50, and an update result output module 60.

[0052] The fusion model establishment module 10 is used to establish a BIM-GIS fusion model, which includes a set of pipeline objects, a set of graphic features, a set of static attributes, and a set of dynamic attribute fields. The entity modeling module 20 is used to extract entities from the BIM-GIS fusion model and perform multi-dimensional entity modeling, including structural entities, operational entities, and environmental entities. The association analysis module 30 is used to perform entity alignment and fusion based on unique identifiers on the multi-dimensional entities, and then perform association analysis on the fusion results through topological relationships, dependency relationships, influence relationships, and event relationships to establish an association knowledge graph. The data update module 40 is used to receive data streams from design, construction, monitoring, and inspection, and perform data time-series mounting and updating on the BIM-GIS fusion model and the association knowledge graph through a dynamic attribute mapping engine. The joint retrieval module 50 is used to parse the retrieval request based on semantic parsing after receiving the retrieval request, configure the retrieval template, and perform a joint retrieval on the BIM-GIS fusion model and the association knowledge graph to establish a candidate object set. The update result output module 60 is used to perform multi-scale sliding window analysis on the candidate object set and output the update results.

[0053] The specific configuration of the association analysis module 30 will be described in detail below. The association analysis module 30 further includes: using the unique identifier to perform spatial alignment, topological alignment, and temporal alignment of multi-dimensional entities to establish a fusion result; establishing entity topological relationships within the fusion result based on the BIM model topology in the BIM-GIS fusion model; identifying dependencies based on the binding relationships of running entities and structural entities in the fusion result to establish entity dependency relationships; performing spatial overlay analysis on environmental entities and structural entities in the fusion result to establish entity influence relationships; performing multi-dimensional entity operation and maintenance log analysis to establish entity event relationships; and establishing an association knowledge graph using the entity topological relationships, entity dependency relationships, entity influence relationships, and entity event relationships.

[0054] The specific configuration of the association analysis module 30 will be described in detail below. The association analysis module 30 further includes: constructing multiple semantic views for each entity within the association knowledge graph, encoding these multiple semantic views into vector representations using a graph neural network encoder, and projecting them onto a common semantic space; introducing contrastive learning during the projection onto the common semantic space, wherein the contrastive learning includes: using different semantic views of the same entity as positive sample pairs and different entities as negative sample pairs, and performing standard contrastive learning using the InfoNCE loss function; configuring spatiotemporal conditional weights to perform contrastive learning, wherein the spatiotemporal conditional weights include time decay weights and spatial proximity weights, wherein the spatial proximity weights are constructed by modeling edge weights of entity topological relationships and entity dependencies; establishing embedded semantic features based on the common semantic space, and performing semantic enhancement of the association knowledge graph through the embedded semantic features.

[0055] The specific configuration of the joint retrieval module 50 will be described in detail below. The joint retrieval module 50 further includes: sending the retrieval request to a multi-layer semantic parsing network, performing multi-layer semantic parsing using the natural language parsing layer, domain ontology mapping layer, and semantic rule matching layer in the multi-layer semantic parsing network to establish a multi-dimensional semantic representation; performing interactive authentication fusion of the multi-dimensional semantic representation to establish a parsing fusion result, and configuring a retrieval template based on the parsing fusion result.

[0056] The specific configuration of the joint retrieval module 50 will be described in detail below. The joint retrieval module 50 further includes: extracting index conditions from the parsing fusion results using a spatial constraint generator, an attribute constraint generator, a temporal constraint generator, and a semantic logic generator, respectively. This includes: extracting spatial index conditions based on the location data of the parsing fusion results using the spatial constraint generator; extracting attribute conditions based on the parameter types and numerical ranges of the parsing fusion results using the attribute constraint generator; generating temporal conditions based on the temporal entities and evolutionary models of the parsing fusion results using the temporal constraint generator; establishing logical constraint conditions based on the relational operators and ontology relations of the parsing fusion results using the semantic logic generator; and combining the extracted index conditions to establish a retrieval template.

[0057] The following will describe the specific configuration of the joint retrieval module 50 in detail. The joint retrieval module 50 further includes: based on the spatial index conditions and attribute conditions in the retrieval template, calling the spatial index and attribute index in the BIM-GIS fusion model to establish a first retrieval result; based on the logical constraints and temporal conditions in the retrieval template, calling graph database retrieval and graph reasoning in the associated knowledge graph to establish a second retrieval result; and performing cross-source alignment on the first and second retrieval results to establish a candidate object set.

[0058] The specific configuration of the joint retrieval module 50 will be described in detail below. The joint retrieval module 50 further includes: one-to-one alignment based on the unique identifier of the entity; if any entity lacks a unique identifier, joint alignment is determined through semantic similarity calculation and spatial proximity calculation.

[0059] The specific configuration of the update result output module 60 will be described in detail below. The update result output module 60 further includes: configuring multiple time backtracking scales based on the stability and complexity characteristics of the candidate object set; activating a multi-scale sliding window using the multiple time backtracking scales; performing temporal trajectory anomaly verification of the candidate object set under the multi-scale sliding window; updating the risk score of the candidate object set using the temporal trajectory anomaly verification results; and generating an update result using the risk score and the candidate object set.

[0060] The specific configuration of the fusion model establishment module 10 will be described in detail below. The fusion model establishment module 10 further includes: the dynamic attribute field set includes the flow medium, direction, location, and burial depth, and each pipeline object in the pipeline object set is configured with a confidence label and a timestamp sequence to form a traceable pipeline object identifier.

[0061] The intelligent management system for the entire life cycle data of underground pipelines provided in this embodiment of the invention can execute the intelligent management method for the entire life cycle data of underground pipelines provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0062] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0063] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent management of underground pipeline network data throughout its entire lifecycle, characterized in that: The method includes: A BIM-GIS fusion model is established, which includes a set of pipeline objects, a set of graphic features, a set of static attributes, and a set of dynamic attribute fields. In the BIM-GIS fusion model, entity extraction is performed, and multi-dimensional entity modeling is carried out. The multi-dimensional entities include structural entities, operational entities, and environmental entities. After the multidimensional entities are aligned and fused based on unique identifiers, the fusion results are analyzed for association through topological relationships, dependency relationships, influence relationships, and event relationships to establish an association knowledge graph. It receives data streams from design, construction, monitoring, and inspection, and uses a dynamic attribute mapping engine to perform time-series mounting and updating of data in the BIM-GIS fusion model and related knowledge graph; After receiving a retrieval request, the retrieval request is parsed based on semantic parsing. After configuring the retrieval template, a joint retrieval is performed on the BIM-GIS fusion model and the associated knowledge graph to establish a candidate object set. Perform a multi-scale sliding window analysis on the candidate object set and output the updated results.

2. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 1, characterized in that, After aligning and fusing the multidimensional entities based on unique identifiers, the fusion results are analyzed for association through topological relationships, dependency relationships, influence relationships, and event relationships to establish an association knowledge graph, including: The unique identifier is used to perform spatial alignment, topological alignment, and temporal alignment of multidimensional entities to establish a fusion result. Establish entity topology relationships within the fusion result based on the BIM model topology in the BIM-GIS fusion model; Dependency relationships are identified based on the binding relationships between runtime entities and structural entities in the fusion results, and entity dependency relationships are established. Spatial overlay analysis is performed on environmental and structural entities in the fusion results to establish entity influence relationships; Perform multi-dimensional entity operation and maintenance log analysis to establish entity event relationships; A knowledge graph is established using the entity topology, entity dependency, entity influence, and entity event relationships.

3. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 2, characterized in that, The process of establishing a knowledge graph using the entity topology, entity dependency, entity influence, and entity event relationships includes: For each entity in the associated knowledge graph, multiple semantic views are constructed, and the multiple semantic views are encoded into vector representations by a graph neural network encoder and projected into a common semantic space. Contrastive learning is introduced during the projection onto the common semantic space, and the contrastive learning includes: Different semantic views of the same entity are used as positive sample pairs, and different entities are used as negative sample pairs. Standard contrastive learning is performed using the InfoNCE loss function. Contrastive learning is performed by configuring spatiotemporal condition weights, which include time decay weights and spatial proximity weights. The spatial proximity weights are constructed by modeling the edge weights of entity topological relationships and entity dependencies. Embedded semantic features are established based on the common semantic space, and semantic enhancement of the associated knowledge graph is performed through the embedded semantic features.

4. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 1, characterized in that, The step of parsing the retrieval request based on semantic analysis and configuring the retrieval template after receiving the retrieval request includes: The retrieval request is sent to a multi-layer semantic parsing network, and the natural language parsing layer, domain ontology mapping layer and semantic rule matching layer in the multi-layer semantic parsing network are used to perform multi-layer semantic parsing to establish a multi-dimensional semantic representation. Perform interactive authentication fusion of the multidimensional semantic representation, establish a parsed fusion result, and configure a retrieval template based on the parsed fusion result.

5. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 4, characterized in that, The step of configuring the retrieval template based on the parsing and fusion results includes: The index conditions of the parsed fusion results are extracted using spatial constraint generators, attribute constraint generators, temporal constraint generators, and semantic logic generators, respectively, including: The spatial constraint generator is used to extract spatial index conditions based on the location data of the analytical fusion result; The attribute constraint generator is used to extract attribute conditions based on the parameter types and numerical ranges of the parsed fusion results; The temporal constraint generator is used to generate temporal conditions based on the temporal entities and evolutionary models of the analytical fusion results; A semantic logic generator is used to establish logical constraints based on relational operators and ontology relations in the parsed fusion results. Combine the results extracted from the index criteria to create a search template.

6. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 5, characterized in that, The joint retrieval of the BIM-GIS fusion model and the associated knowledge graph to establish a candidate object set includes: Based on the spatial index conditions and attribute conditions in the search template, the spatial index and attribute index are called in the BIM-GIS fusion model to establish the first search result; Based on the logical constraints and temporal conditions in the retrieval template, a second retrieval result is established by calling graph database retrieval and graph reasoning in the associated knowledge graph. Cross-source alignment is performed on the first and second search results to establish a candidate object set.

7. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 6, characterized in that, The cross-source alignment includes: Alignment is performed one-to-one based on the unique identifier of each entity; If any entity lacks a unique identifier, alignment is determined jointly through semantic similarity calculation and spatial proximity calculation.

8. The intelligent management method for the entire lifecycle data of underground pipeline networks as described in claim 1, characterized in that, The multi-scale sliding window analysis of the candidate object set is performed, and the updated results are output, including: Multiple time backtracking scales are configured based on the stability and complexity characteristics of the candidate object set, and a multi-scale sliding window is activated using the multiple time backtracking scales. Temporal trajectory anomaly verification of the candidate object set is performed under the multi-scale sliding window; The risk score of the candidate object set is updated using the time-series trajectory anomaly verification results, and the updated result is generated using the risk score and the candidate object set.

9. The intelligent management method for the entire life cycle data of underground pipeline networks as described in claim 1, characterized in that, The dynamic attribute field set includes the circulation medium, direction, location, and burial depth, and each pipeline object in the pipeline object set is configured with a confidence label and a timestamp sequence to form a traceable pipeline object identifier.

10. An intelligent management system for the entire lifecycle of underground pipeline networks, characterized in that: The system is used to implement the intelligent management method for the entire life cycle data of underground pipeline networks as described in any one of claims 1 to 9, and the system includes: The fusion model establishment module is used to establish a BIM-GIS fusion model, which includes a set of pipeline objects, a set of graphic features, a set of static attributes, and a set of dynamic attribute fields. The entity modeling module is used to extract entities from the BIM-GIS fusion model and perform multi-dimensional entity modeling. Multi-dimensional entities include structural entities, operational entities, and environmental entities. The association analysis module is used to align and fuse the multidimensional entities based on unique identifiers, and then perform association analysis on the fusion results through topological relationships, dependency relationships, influence relationships, and event relationships to establish an association knowledge graph. The data update module is used to receive data streams from design, construction, monitoring, and inspection, and to perform time-series mounting and updating of data in the BIM-GIS fusion model and related knowledge graph through the dynamic attribute mapping engine; The joint retrieval module is used to parse the retrieval request based on semantic parsing after receiving the retrieval request, configure the retrieval template, and perform joint retrieval in the BIM-GIS fusion model and the associated knowledge graph to establish a candidate object set. The update result output module is used to perform multi-scale sliding window analysis on the candidate object set and output the update results.

Images