Glass fiber reinforced plastic pipeline liquid leakage detection method, system and equipment and storage medium

Abstract

The invention provides a glass fiber reinforced plastic pipeline liquid leakage detection method, system and device and a storage medium, and relates to the technical field of glass fiber reinforced plastic pipeline safety monitoring and early warning. The method comprises the steps of obtaining structural data, process data and operation data of a pipeline, and inputting the structural data, the process data and the operation data into a preset liquid leakage risk early warning knowledge graph to obtain a leakage risk assessment result of the pipeline; acquiring a liquid level value in the pipeline liquid collection pool, acquiring real-time response signal data of each sensor in the pipeline when the liquid level value exceeds a preset liquid level threshold value and / or the leakage risk assessment result is a high risk level, and inputting the real-time response signal data into the multi-physical field model to obtain global sensor data; and performing leakage source inversion positioning based on the global sensor data to obtain leakage source position data. According to the scheme, the accuracy and timeliness of liquid leakage detection of the glass fiber reinforced plastic pipeline can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of safety monitoring and early warning of FRP pipelines, and in particular to a method, system, equipment and storage medium for detecting liquid leakage in FRP pipelines. Background Art

[0002] The pipe-in-pipe structure is widely used in liquid transportation systems, especially in areas that require high safety and environmental protection, such as oil and gas transportation, chemical raw material transportation, and urban water supply and drainage systems. This structure forms a double protective layer by setting an inner pipe inside an outer pipe, which can theoretically effectively prevent liquid leakage from directly polluting the environment. However, with the expansion of the scale of pipeline systems and the extension of their service life, the pipe-in-pipe structure is also facing the risk of leakage, especially the problem of small leakage of the inner pipe. The traditional method of detecting liquid leakage in pipe mainly relies on setting a simple liquid level sensor or pressure sensor in the space between the outer pipe and the inner pipe. Although this method can detect obvious leaks, it has significant limitations. First, since pipeline systems often span long distances, it is difficult to detect local small leaks in time by relying on only a small number of sensors set at intervals. Secondly, this method lacks the ability to predict leakage risks, and can often only detect leakage after it has occurred and accumulated to a certain extent, and cannot achieve early warning. Summary of the invention

[0003] The present application provides a method, system, device and storage medium for detecting liquid leakage in a FRP pipeline, which can improve the accuracy and timeliness of detecting liquid leakage in a FRP pipeline.

[0004] In a first aspect, the present application provides a method for detecting liquid leakage in a glass fiber reinforced plastic pipeline, the method comprising: Obtain the structural data, process data and operation data of the pipeline, and input the structural data, process data and operation data into the preset liquid leakage risk warning knowledge graph to obtain the leakage risk assessment result of the pipeline; Obtain the liquid level value in the pipeline collecting tank. When the liquid level value exceeds a preset liquid level threshold and / or the leakage risk assessment result is a high risk level, obtain the real-time response signal data of each sensor in the pipeline, and input the real-time response signal data into the multi-physics field model to obtain global sensor data; The leakage source is inverted and located based on the global sensor data to obtain the leakage source location data.

[0005] By adopting the above technical solution, the system can comprehensively evaluate the leakage risk of the pipeline by obtaining the structural data, process data and operation data of the pipeline and inputting these data into the preset liquid leakage risk warning knowledge graph. This risk assessment method based on knowledge graph can make full use of historical data and expert experience to improve the accuracy and timeliness of risk warning. Secondly, by monitoring the liquid level value of the pipeline sump and comparing it with the preset threshold, combined with the leakage risk assessment results, the system can timely trigger a more in-depth detection process. When an abnormality is detected, the system obtains the real-time response signal data of each sensor in the pipeline and inputs these data into the multi-physics field model to obtain global sensor data. This method can not only capture local anomalies, but also extend local information to the entire pipeline system through the multi-physics field model, providing a more comprehensive leakage status assessment. Finally, based on the obtained global sensor data, the system uses the leakage source inversion positioning technology to accurately determine the location of the leakage source. This inversion positioning method makes full use of the advantages of global data and overcomes the problem of inaccurate positioning caused by insufficient sensor coverage in traditional methods. Through this series of steps, this method realizes the intelligent management of the whole process from risk warning to precise positioning, greatly improving the efficiency and accuracy of FRP pipeline liquid leakage detection.

[0006] In a second aspect of the present application, a glass fiber reinforced plastic pipeline liquid leakage detection method system is provided, comprising: The risk assessment module is used to obtain the structural data, process data and operation data of the pipeline, and input the structural data, process data and operation data into the preset liquid leakage risk warning knowledge graph to obtain the leakage risk assessment result of the pipeline; A data processing module is used to obtain the liquid level value in the pipeline collecting tank. When the liquid level value exceeds a preset liquid level threshold and / or the leakage risk assessment result is a high risk level, the real-time response signal data of each sensor in the pipeline is obtained, and the real-time response signal data is input into the multi-physics field model to obtain global sensor data; The positioning module is used to perform leakage source inversion positioning based on global sensor data to obtain leakage source position data.

[0007] In a third aspect of the present application, a computer storage medium is provided, wherein the computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the above method steps.

[0008] In the fourth aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the above method.

[0009] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application obtains the structural data, process data and operation data of the pipeline, and inputs these data into the preset liquid leakage risk warning knowledge graph, so that the system can comprehensively evaluate the leakage risk of the pipeline. This risk assessment method based on knowledge graph can make full use of historical data and expert experience to improve the accuracy and timeliness of risk warning. Secondly, by monitoring the liquid level value of the pipeline sump and comparing it with the preset threshold, combined with the leakage risk assessment results, the system can trigger a more in-depth detection process in time. When an abnormality is detected, the system obtains the real-time response signal data of each sensor in the pipeline and inputs these data into the multi-physics field model to obtain global sensor data. This method can not only capture local anomalies, but also extend local information to the entire pipeline system through the multi-physics field model, providing a more comprehensive leakage status assessment. Finally, based on the obtained global sensor data, the system uses the leakage source inversion positioning technology to accurately determine the location of the leakage source. This inversion positioning method makes full use of the advantages of global data and overcomes the problem of inaccurate positioning caused by insufficient sensor coverage in traditional methods. Through this series of steps, this method realizes the intelligent management of the whole process from risk warning to precise positioning, greatly improving the efficiency and accuracy of FRP pipeline liquid leakage detection.

[0010] 2. During the operation of the pipeline, the distribution of physical quantities such as flow, temperature, and pressure of the internal medium presents complex spatiotemporal characteristics, forming a specific multi-physical field. By obtaining the multi-physical field data of the pipeline, such as the velocity field, pressure field, and temperature field of the fluid flow, the physical state inside the pipeline can be fully characterized.

[0011] Based on the acquired multi-physics data, a multi-physics model is constructed through numerical simulation, data-driven and other methods, which can establish the intrinsic connection between the distribution of pipeline physical quantities and the occurrence and location of leakage. The multi-physics model comprehensively considers factors such as the geometric structure, material properties, medium characteristics, boundary conditions and other factors of the pipeline, and can more accurately describe the distribution law of physical quantities under normal working conditions of the pipeline.

[0012] 3. After a pipeline leak occurs, the leaking medium gushes out from the leak point, forming abnormal physical fields such as pressure waves and temperature gradients inside the pipeline. These abnormal physical fields propagate in the pipeline according to certain spatiotemporal laws and are collected by sensor networks distributed at different locations in the pipeline to form global abnormal sensor data. In order to invert the location of the leak source from the abnormal sensor data, a leak inversion model needs to be established. The leak inversion model comprehensively considers factors such as the thermophysical parameters of the leaking medium, the geometric structure and material properties of the pipeline, and the response characteristics of the sensor, and quantitatively describes the functional relationship between the leak source parameters (such as leak location, leak amount, etc.) and the abnormal sensor data. The construction of the leak inversion model needs to rely on the aforementioned multi-physical field model, and establish the mapping relationship between the leak parameters, abnormal physical fields, and abnormal sensor data through numerical simulation, machine learning and other methods.

[0013] Based on the constructed leakage inversion model, the actually collected abnormal data (such as abnormal pressure, temperature, acoustic signal, etc.) are input into the inversion model, and the initial abnormal data is obtained through inversion operation. The initial abnormal data is the initial estimated value of the leakage source parameters inferred by the leakage inversion model based on the current abnormal sensor data, including key information such as leakage location and leakage amount. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of a flow chart of a method for detecting liquid leakage in a glass fiber reinforced plastic pipeline provided in an embodiment of the present application; Figure 2 An architectural diagram of a glass fiber reinforced plastic pipeline liquid leakage detection system provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0015] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0016] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0017] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0018] In order to facilitate understanding of the method and system provided by the embodiments of the present application, before introducing the embodiments of the present application, the background of the embodiments of the present application is first introduced.

[0019] As an innovative liquid transportation system design, the pipe-in-pipe structure has been widely used in fields that require high safety and environmental protection, such as oil and gas transportation, chemical raw material transportation, and urban water supply and drainage. This unique structure cleverly forms a double protection layer by setting an inner pipe inside an outer pipe, which theoretically provides a reliable guarantee for effectively preventing liquid leakage from directly polluting the environment. However, with the continuous expansion of the scale of pipeline systems and the continuous extension of their service life, the pipe-in-pipe structure is inevitably faced with potential leakage risks, among which the problem of small leakage of the inner pipe is the most difficult.

[0020] The traditional method of detecting liquid leakage in a pipe mainly relies on setting a simple liquid level sensor or pressure sensor in the space between the outer pipe and the inner pipe. Although this method has shown certain effectiveness in detecting obvious leaks, it has exposed significant limitations in practical applications. The first problem is that modern pipeline systems often span extremely long distances, and it is difficult to detect and locate localized small leaks in a timely manner by relying solely on a small number of sensors set at intervals. The existence of this detection blind spot greatly increases the safety risks of the system. Secondly, traditional detection methods have obvious deficiencies in the ability to predict leakage risks. This passive detection strategy can often only trigger an alarm after the leak has occurred and accumulated to a certain extent, which means that the best time for intervention has been missed. This lag not only fails to achieve early warning of leaks, but also leads to more serious environmental pollution and economic losses.

[0021] In addition, traditional methods are difficult to flexibly adjust and optimize in the face of complex and changing operating environments. Fixed detection thresholds and simple data analysis models cannot fully consider the dynamic changes of pipeline systems under different working conditions, which further limits their effectiveness in practical applications.

[0022] In summary, although the pipe-in-pipe structure theoretically provides double protection, the existing leakage detection technology has not yet fully utilized the advantages of this structure. Faced with increasingly complex pipeline systems and ever-increasing safety and environmental protection requirements, a more intelligent, comprehensive and accurate liquid leakage detection method is urgently needed to effectively respond to challenges such as small leak detection, early risk warning and precise positioning, so as to truly realize the safety potential of the pipe-in-pipe structure.

[0023] After the background introduction of the above content, those skilled in the art can understand the problems existing in the prior art. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0024] Based on the above background technology, please refer to Figure 1 , Figure 1 A flow chart of a method for detecting liquid leakage in a glass fiber reinforced plastic pipe provided in an embodiment of the present application. The system can be implemented by a computer program or can be run as an independent tool application. Specifically, in an embodiment of the present application, the method can be applied on a server, but can also be applied to electronic devices such as servers. A method for detecting liquid leakage in a glass fiber reinforced plastic pipe comprises the following steps: S101, obtaining structural data, process data and operation data of the pipeline, and inputting the structural data, process data and operation data into a preset liquid leakage risk warning knowledge graph to obtain a leakage risk assessment result of the pipeline; Specifically, first obtain the structural data, process data and operation data of the pipeline. The structural data includes the physical characteristics of the pipeline, such as material, size, connection method, etc.; the process data covers the process parameters such as pipeline design pressure, temperature, flow rate, etc.; the operation data includes dynamic operation information such as real-time pressure, temperature, flow, etc. The comprehensive collection of these multi-dimensional data is to fully reflect the current status and potential risk factors of the pipeline system.

[0025] After obtaining this data, the system inputs the structural data, process data and operation data into the pre-built liquid leakage risk warning knowledge graph. This knowledge graph is a complex network structure built based on a large amount of historical data, expert experience and machine learning algorithms, which can simulate and infer the relationship between various factors. The use of knowledge graphs enables the system to use existing knowledge to quickly analyze the similarities and differences between current data and historical patterns, thereby conducting more accurate risk assessments.

[0026] In specific operations, the system first maps the structural data to the corresponding nodes of the knowledge graph, such as the pipeline material node, pipe diameter node, etc. Then, the process data is associated with the corresponding process parameter node. Finally, the real-time operation data is input into the corresponding dynamic data node. The knowledge graph analyzes the relationship and impact of these data through preset reasoning rules and algorithms, and comprehensively evaluates the leakage risk of the current pipeline system.

[0027] For example, when the system detects abnormal fluctuations in the pressure data of a certain section of pipeline, the knowledge graph can quickly infer the leakage risk level in this case based on the material characteristics and service life of the pipeline. At the same time, the system can also predict the risk change trend in the future based on historical data analysis.

[0028] In this way, the system finally obtains the leakage risk assessment result of the pipeline. This result includes not only the current risk level (such as low, medium, and high risk), but also detailed information such as the specific type of risk, possible impact range, and risk evolution trend.

[0029] Based on the above embodiment, as an optional embodiment, before inputting the structural data, process data and operation data into a preset liquid leakage risk knowledge graph to obtain the pipeline leakage risk assessment result, the following steps are further included: S201, obtaining pipeline leakage sample data, and extracting the pipeline leakage sample data to obtain relationship entities related to leakage risks; Specifically, the pipeline leakage sample data is obtained and extracted to obtain the relational entities related to the leakage risk. The implementation of this step is crucial to building a comprehensive and accurate knowledge graph of liquid leakage risk, because it provides the system with a rich historical data and empirical knowledge base. The system first collects pipeline leakage sample data through various channels, including historical accident reports, maintenance records, laboratory simulation data, industry case analysis, and expert experience summary. These sample data cover pipeline leakage events of different types, scales, and causes, providing a broad data foundation for subsequent analysis. After obtaining the data, the system uses advanced data mining and natural language processing technologies to conduct in-depth analysis and extraction of these sample data. Specifically, the system first uses text preprocessing technology to clean and standardize unstructured data to remove irrelevant information and redundant data. Then, the system applies named entity recognition (NER) technology to identify and extract key entities in the text, such as leakage location, leakage cause, leakage scale, environmental conditions, etc. Next, the system uses a relation extraction algorithm to analyze the associations between these entities and identify important relationships such as "cause", "influence", and "association". In addition, the system also uses topic modeling techniques, such as Latent Dirichlet Allocation (LDA), to extract potential leakage risk topics and patterns from a large amount of text. In this process, the system pays special attention to those relationship entities that are directly related to leakage risk, such as the relationship between the degree of material corrosion and leakage probability, the relationship between ambient temperature changes and pipeline stress, and the relationship between operating pressure and leakage scale. The system also considers some indirect but important factors, such as pipeline age, maintenance frequency, geological conditions, etc., which, although not directly causing leakage, have an important impact on leakage risk. Through this comprehensive and in-depth data extraction and analysis process, the system successfully built a leakage risk knowledge base containing rich relationship entities.

[0030] S202, performing fusion processing on the relationship entities to build a consistent entity relationship pedigree; Specifically, the relational entities extracted in step S201 are first analyzed in depth to identify duplication, conflict, and complementary relationships between entities. This process uses entity alignment technology to identify entities that represent the same concept but have different expressions by comparing the attributes, context, and semantic similarity of the entities. For example, "pipeline corrosion" and "pipeline wall erosion" refer to the same risk factor. The system uses machine learning algorithms, such as Word2Vec and BERT, to calculate the semantic similarity between entities and make judgments based on expert knowledge to ensure the accuracy of alignment. Next, the system merges the identified identical or highly similar entities to create a unified standard entity. In this process, the system retains unique information from each source to ensure that valuable details are not lost. For conflicting entity information, the system uses multi-source data fusion technology, such as Dempster-Shafer evidence theory, to comprehensively consider the reliability and weight of each data source and draw the most reasonable conclusion. In addition, the system also identifies the hierarchical relationship between entities and constructs a hierarchical entity classification system. For example, "material corrosion" is used as the parent category, which contains subcategories such as "inner wall corrosion" and "outer wall corrosion". This hierarchical structure helps to better organize and understand risk factors. On the basis of entity fusion, the system further analyzes the relationship between entities and constructs a complete relationship network. This includes causal relationships, correlation relationships, temporal relationships, etc. The system uses relationship extraction algorithms such as remote supervised learning and neural network models to automatically extract the relationship between entities from a large amount of text data. At the same time, the system also integrates expert knowledge to ensure that the extracted relationship accurately reflects the cognition of domain experts. Through these steps, the system finally builds a consistent entity relationship spectrum.

[0031] Based on the above embodiment, as an optional embodiment, the relationship entities are fused to build a consistent entity relationship pedigree, including: S401, aligning the relationship entities to obtain a standard relationship entity representation; Specifically, the relational entities are aligned to obtain standard relational entity representations. The implementation of this step is crucial to eliminating differences between data sources and unifying knowledge representation, and is the basis for building a consistent entity relationship pedigree. The system first conducts a comprehensive analysis of the relational entities obtained from different sources to identify entities that express the same or similar concepts but use different terms. This process uses a variety of technical means. First, the system uses string matching algorithms, such as edit distance and Jaccard similarity, to identify entities with similar spellings or structures. For example, "pipeline corrosion" and "pipeline corrosion" may be identified as similar entities. Secondly, the system uses semantic similarity calculation methods, such as deep learning models such as Word2Vec and BERT, to calculate the degree of semantic similarity between entities. This method can capture entities with different expressions but similar meanings, such as "excessive pressure" and "overpressure". In addition, the system also introduces domain ontologies and synonym libraries, and uses pre-defined concept hierarchies and synonymous relationships to further improve the accuracy of alignment. After identifying potential aligned entities, the system performs detailed attribute comparisons. It analyzes the attribute information of each entity, such as definition, characteristics, and associations, to ensure semantic consistency. For entities with subtle differences, the system uses a weighted attribute matching algorithm to comprehensively consider the importance of each attribute and calculate the overall matching degree. In this process, the system pays special attention to handling homonyms and polysemy to avoid incorrect alignment. For example, "tube wall" may refer to different concepts in different contexts, and the system correctly identifies its meaning by analyzing contextual information. In order to handle complex alignment situations, the system also uses an ensemble learning method to combine the results of multiple alignment algorithms and make the final alignment decision through voting or weighted averaging. This method can effectively reduce the limitations of a single algorithm and improve the robustness of alignment. After completing the preliminary alignment, the system generates a standard relational entity representation. For aligned entities, the system selects the most representative representation as the standard representation or creates a new, more comprehensive representation. This process not only takes into account the accuracy and clarity of the language expression, but also ensures consistency with industry standards and specifications. In order to maintain the quality of the alignment results, the system introduces an expert review mechanism. Domain experts review and revise the standard entity representations generated by the system to ensure their accuracy and applicability.

[0032] S402, processing entity conflicts in the standard relationship entity representation to obtain a first standard relationship entity representation; Specifically, first, the standard relational entity representation obtained in step S401 is comprehensively scanned to identify entity conflicts. These conflicts mainly include attribute conflicts, relation conflicts, and logical conflicts. Attribute conflicts refer to the fact that the same entity has different attribute values ​​in different data sources, such as different values ​​of pipeline material strength. Relationship conflicts involve contradictions in the relationship between entities, such as one data source indicating that high pressure will cause pipeline rupture, while another data source denies this relationship. Logical conflicts refer to logical inconsistencies within the knowledge system, such as circular dependencies or self-contradictory statements. The system uses a variety of technical means to identify and handle these conflicts. First, for attribute conflicts, the system uses statistical analysis methods, such as median and mode, to determine the correct value. For numerical attributes, the system also considers the timeliness and source reliability of the data, giving higher weights to newer and more reliable data sources. When dealing with relationship conflicts, the system uses methods such as evidence theory and Bayesian networks to comprehensively consider the credibility of multiple data sources and the strength of supporting evidence to obtain the most reasonable relationship judgment. For example, if multiple high-credibility sources support the causal relationship between high pressure and pipeline rupture, the system will tend to retain this relationship. For logical conflicts, the system uses automatic reasoning and consistency checking algorithms to identify logical contradictions in the knowledge system. Once a contradiction is found, the system will try to eliminate it by adjusting the relationship strength or adding conditional restrictions. For example, "high pressure will inevitably lead to pipeline rupture" is modified to "high pressure will lead to pipeline rupture under certain conditions." In the process of dealing with conflicts, the system also introduces uncertainty representation methods, such as fuzzy set theory and probability theory, to express the uncertainty and diversity of knowledge.

[0033] S403, processing the relationship conflict in the first standard relationship entity expression to obtain a second standard relationship entity expression, and organizing and linking the second standard relationship entity expression according to a preset pedigree model to obtain a consistent entity relationship pedigree.

[0034] Specifically, the first standard relationship entity representation obtained in step S402 is first analyzed in depth to identify the relationship conflicts therein. These conflicts mainly include relationship strength conflicts, relationship direction conflicts, and relationship type conflicts. Relationship strength conflicts refer to the significant differences in the relationship strength between the same pair of entities in different data sources, such as one data source believes that high temperature has a great impact on pipeline materials, while another data source believes that the impact is small. Relationship direction conflicts involve contradictions in causal relationships between entities, such as one source indicates that pipeline corrosion leads to an increased risk of leakage, while another data source believes that the increased risk of leakage leads to accelerated corrosion of the pipeline. Relationship type conflicts refer to different understandings of the nature of the relationship between the same pair of entities, such as one data source describes the relationship between pipeline pressure and leakage risk as a linear relationship, while another data source believes that it is a nonlinear relationship.

[0035] To handle these relationship conflicts, the system uses a variety of advanced technologies. First, for relationship strength conflicts, the system uses a fusion method based on evidence weight. It considers the reliability, timeliness, and data volume of each data source, assigns different weights to the relationship strengths of different sources, and then obtains the comprehensive relationship strength through weighted average or other aggregation functions. For example, the relationship strength from recent experimental data will receive a higher weight, while the relationship strength based on old theories will receive a lower weight.

[0036] For relationship direction conflicts, the system uses causal reasoning and time series analysis techniques. It analyzes the time series data between entities and uses statistical methods such as Granger causality test to identify reliable causal directions. In some cases, the system also considers the nature of bidirectional causal relationships and uses models such as dynamic Bayesian networks to represent complex interactive relationships.

[0037] When dealing with relationship type conflicts, the system uses model comparison and validation technology. It builds different types of relationship models (such as linear models, nonlinear models, etc.) based on existing data, and then uses cross-validation, AIC (Akaike Information Criterion) and other methods to evaluate which model can most accurately describe the relationship between entities.

[0038] For complex relationship conflicts that are difficult to resolve through automatic methods, the system introduces an expert knowledge base and a decision support system. It submits the conflict situation to domain experts and provides relevant background information and data analysis results to assist experts in making judgments. The expert's decision is recorded and used to update the system's knowledge base for reference in similar situations in the future. After completing the relationship conflict processing, the system generates a second standard relationship entity representation. This representation not only contains the consistent relationship knowledge after conflict processing, but also retains the uncertainty information and diversity of the relationship. For example, for some complex relationships, the system retains multiple models and attaches their respective applicable conditions and confidence. Next, the system organizes and links the second standard relationship entity representation according to the preset pedigree model to obtain a consistent entity relationship pedigree. This preset pedigree model is built based on the ontology of the liquid leakage risk field, which defines the hierarchical structure of entities, relationship types, and attribute frameworks. The system uses graph database technology, such as Neo4j, to map entities and relationships into the pedigree model. In this process, the system not only establishes direct relationships between entities, but also discovers and establishes implicit indirect relationships through reasoning mechanisms.

[0039] S203, constructing a triple entity relationship representation based on the entity relationship pedigree, and constructing a preset liquid leakage risk knowledge graph based on the triple entity relationship representation.

[0040] Specifically, a triple entity relationship representation is constructed based on the entity relationship pedigree constructed in the previous step, and a preset liquid leakage risk knowledge graph is constructed based on this. The implementation of this step is crucial for converting complex entity relationship pedigrees into structured data that can be processed by computers, and finally forming an intuitive and efficient knowledge graph. The system first systematically analyzes the entity relationship pedigree and converts each entity relationship into a "subject-predicate-object" triple form. For example, "high pressure-leads to-pipeline cracks" is a typical triple representation. In this process, the system uses natural language processing techniques, such as semantic role labeling and dependency syntax analysis, to automatically extract subjects, predicates, and objects from the pedigree. For some complex relationships, the system uses relational decomposition technology to split them into multiple simple triples. For example, "under high temperature and high pressure environment, pipeline materials are prone to corrosion" can be decomposed into multiple triples such as "high temperature-affects-pipeline materials", "high pressure-affects-pipeline materials", and "pipeline materials-prone to-corrosion".

[0041] In the process of constructing triples, the system pays special attention to maintaining the accuracy and completeness of semantics. For some implicit relationships, the system supplements the corresponding triples through reasoning to ensure the integrity of knowledge. For example, if there are two triples of "corrosion-cause-thinning of pipe wall" and "thinning of pipe wall-increase-leakage risk", the system will infer and add the triple of "corrosion-indirect increase-leakage risk". In order to ensure the quality of triples, the system also introduces an expert review mechanism, and domain experts verify and correct the automatically generated triples. After completing the construction of the triple entity relationship representation, the system begins to build a preset liquid leakage risk knowledge graph based on these triples. The system uses graph database technology, such as Neo4j, to take the subject and object in each triple as nodes in the graph, and the predicate as the edge connecting the nodes. In this process, the system not only establishes direct relationships between entities, but also identifies and establishes indirect relationships and complex paths between entities through graph algorithms. For example, the system can find multiple relationship paths from "environmental factors" to "leakage accidents". In the process of constructing the graph, the system also pays attention to the organization and representation of knowledge. It sets different attributes and weights for different types of entities and relationships to reflect their importance in risk assessment. At the same time, the system also establishes a multi-level knowledge structure, allowing users to explore risk factors from macro to micro levels. In order to improve the practicality of the knowledge graph, the system also integrates a reasoning engine that can automatically reason based on existing knowledge and discover potential risk relationships. For example, by analyzing the combined effects of multiple related factors, new high-risk scenarios can be inferred. The implementation of this method has brought significant results. First, the knowledge representation in the form of triples greatly improves the structured degree of data, making complex domain knowledge easy to process and analyze by computers. Secondly, the knowledge graph built based on triples provides an intuitive and comprehensive knowledge framework for liquid leakage risk assessment, making the complex relationships between risk factors clear and visible. Furthermore, the graphical representation of the knowledge graph enables users to more easily understand and explore the associations between risk factors, which helps to discover non-intuitive risk patterns.

[0042] S102, obtaining a liquid level value in a pipeline liquid collecting tank, and when the liquid level value exceeds a preset liquid level threshold and / or the leakage risk assessment result is a high risk level, obtaining real-time response signal data of each sensor in the pipeline, and inputting the real-time response signal data into a multi-physics field model to obtain global sensor data; Specifically, the system first obtains the liquid level value in the pipeline collection tank. The pipeline collection tank is an important facility for collecting leaked liquid in the pipe-in-pipe structure. Its liquid level value directly reflects whether there is liquid leakage and its severity. The system monitors the liquid level changes in real time through the liquid level sensor set in the collection tank and transmits the data to the central control system.

[0043] The system compares the obtained liquid level value with the preset liquid level threshold. This preset liquid level threshold is determined based on the normal operating parameters, safety margin and historical data analysis of the pipeline system. At the same time, the system also takes into account the leakage risk assessment results obtained in step S101. When the liquid level value exceeds the preset liquid level threshold, or the leakage risk assessment result shows a high risk level, the system will trigger the next step of the in-depth detection process.

[0044] Once the in-depth inspection is triggered, the system immediately starts to obtain the real-time response signal data of each sensor in the pipeline. These sensors include but are not limited to pressure sensors, temperature sensors, acoustic sensors, and fiber optic sensors, which are distributed in key locations of the pipeline system. Each sensor can provide a specific type of data to reflect changes in the local environment. The system collects real-time data from all sensors simultaneously through high-speed data acquisition equipment and communication networks to ensure the time synchronization and integrity of the data.

[0045] The acquired real-time response signal data is then input into a pre-built multi-physics model. This multi-physics model is a complex mathematical model that comprehensively considers the interaction of multiple physical fields such as fluid mechanics, thermodynamics, acoustics and structural mechanics. The model is constructed based on the detailed physical parameters, material properties and operating conditions of the pipeline system, and is numerically simulated and solved by high-performance computers.

[0046] The core function of the multi-physics model is to expand local sensor data into global sensor data. Specifically, the model uses limited measured data points, combined with the propagation laws and interaction relationships of the physical field, to generate global state data of the entire pipeline system through interpolation, extrapolation, and iterative calculation. For example, through the pressure data of several key points, the model can calculate the pressure distribution of the entire pipeline system; through local temperature changes, the model can infer the heat propagation and distribution.

[0047] Based on the above embodiment, as an optional embodiment, before acquiring the real-time response signal data of each sensor in the pipeline, the method further includes: S301, acquiring multi-physics field data of the pipeline, and constructing a multi-physics field model according to the multi-physics field data; Specifically, first of all, the system comprehensively collects multi-physics data of the pipeline. These data cover the characteristics of the pipeline system in multiple physical fields, mainly including: fluid mechanics data (such as flow velocity distribution, pressure distribution, fluid density and viscosity, etc.), thermodynamics data (such as temperature distribution, thermal conductivity, etc.), structural mechanics data (such as elastic modulus, Poisson's ratio, stress-strain relationship, etc. of pipeline materials), acoustic data (such as sound wave propagation speed, attenuation characteristics, etc.) and electromagnetic data (such as dielectric constant, conductivity, etc.). These data are obtained in a variety of ways, including design parameter extraction, field measurement, laboratory analysis and historical operation data collection. After the data collection is completed, the system begins to build a multi-physics model.

[0048] S302, generating a target sensor layout plan based on the multi-physics field model, and performing sensor layout on the pipeline based on the target sensor layout plan.

[0049] Specifically, firstly, a large number of numerical simulations and sensitivity analyses are carried out using multi-physics field models to simulate various leakage situations and analyze the impact of leakages at different locations and scales on various physical fields. Focus on the physical quantities and locations that are most sensitive to leakage, such as the areas with the most obvious pressure fluctuations, the points with the most drastic temperature changes, and the locations of stress concentration, to provide a scientific basis for determining the optimal sensor layout. Then, the system uses multi-objective optimization technology to generate a target sensor layout plan, taking into account factors such as detection sensitivity, coverage, redundancy, cost-effectiveness, and installation feasibility. The system uses genetic algorithms combined with expert knowledge bases to iteratively generate multiple sensor layout plans. Each plan is verified by the multi-physics field model and its performance is evaluated by simulating various leakage situations. After multiple rounds of optimization and comparison, the system finally determined an optimal target sensor layout plan. This plan not only includes the specific location of the sensor, but also specifies in detail the type and specifications of the sensor to be installed at each location, including pressure sensors, temperature sensors, acoustic sensors, optical fiber strain sensors, and other types. Their combined use can comprehensively monitor various physical parameters of the pipeline system. Based on this target sensor layout plan, the system guides engineers to carry out actual sensor installation. The installation process was carried out strictly according to the plan, ensuring that each sensor was installed in the precise designated location.

[0050] S103, performing leakage source inversion positioning based on the global sensor data to obtain leakage source position data.

[0051] Specifically, the process of leak source inversion and location begins with the analysis of global sensor data. The system uses high-performance computing equipment to conduct in-depth data mining and pattern recognition on global sensor data. This includes comprehensive analysis of multi-dimensional data such as pressure fluctuations, temperature changes, acoustic signals, and optical fiber strain. By conducting in-depth research on the temporal and spatial distribution characteristics, change trends, and correlations of these data, the system preliminarily determines the leak area.

[0052] Subsequently, the system starts the inversion algorithm for precise positioning. The core idea of ​​the inversion algorithm is to deduce the unknown cause (i.e., the location of the leak source) from the known result (i.e., the observed global sensor data). In this embodiment, the system uses an inversion method based on Bayesian reasoning, combined with Monte Carlo simulation technology, to build a powerful leak source positioning model.

[0053] In specific operations, the system first establishes a series of hypothetical leak source locations based on the physical model and leakage characteristics of the pipeline. For each hypothetical location, the system calculates the sensor data that should be observed if the leak occurs at that location through forward simulation. Then, the system compares these simulated data with the actual observed global sensor data and calculates the similarity between them. Through repeated iterations and optimization, the system continuously adjusts the hypothetical location until it finds the simulation result that best matches the actual observed data, thereby determining the best leak source location.

[0054] Based on the above embodiment, as an optional embodiment, the leakage source is inverted and located based on the global sensor data to obtain the leakage source position data, including: The global sensor data is compared with a preset normal sensor threshold to determine abnormal data in the global sensor data, and the leakage source is inverted and located based on the abnormal data to obtain leakage source position data.

[0055] Specifically, the global sensor data is compared with the preset normal sensor threshold. The purpose of this step is to identify abnormal data to determine whether a leak has occurred and the potential leak area. The preset normal sensor threshold is based on historical data, expert knowledge and statistical analysis, and represents the expected parameter range of the system under normal operating conditions. The system uses real-time data processing technology to continuously monitor data streams from various sensors, including pressure sensors, flow sensors, temperature sensors, etc. For each sensor reading, the system will compare it with its corresponding normal threshold.

[0056] During the comparison process, the system not only considers the data at a single point in time, but also analyzes the time series changes of the data. This dynamic comparison can capture sudden changes and slowly accumulated anomalies. The system uses a variety of statistical methods and machine learning algorithms to perform this comparison, such as moving averages, exponential smoothing, ARIMA models, etc. These methods can effectively filter out short-term fluctuations and highlight persistent abnormal patterns.

[0057] After determining the abnormal data, the system enters the leak source inversion and positioning stage. The purpose of this stage is to infer the location of the largest leak source by analyzing the distribution and characteristics of the abnormal data. The system uses a variety of advanced inversion algorithms to achieve this goal. First, the system uses gradient analysis to determine the area with the strongest abnormal signal by calculating the spatial gradient of the abnormal data. This method is particularly suitable for the analysis of pressure sensor data, because the leak point usually manifests as a sharp change in pressure gradient. Secondly, the system applies spatiotemporal correlation analysis. It examines the time delay and correlation of sensor data at different locations and locates the leak source by tracing the propagation path of the abnormal signal. This method is particularly effective in utilizing the physical properties of liquid leaks, such as the propagation speed and attenuation law of pressure waves.

[0058] The system also uses an inversion method based on a physical model. It uses a fluid dynamics model to simulate the sensor response under different leakage scenarios, and then matches the actual observed abnormal data with the simulation results to find the best fitting leakage location. This method can take into account complex pipe network structures and fluid characteristics, providing more accurate positioning results.

[0059] Based on the above embodiment, as an optional embodiment, the leakage source is inverted and located according to the abnormal data to obtain the leakage source position data, including: S501, constructing a leakage inversion model according to global sensor data and a multi-physics field model, and performing an inversion operation based on the leakage inversion model and abnormal data to obtain initial abnormal data; Specifically, the leakage inversion model is constructed using global sensor data and multi-physics field models. The multi-physics field model is a complex model that comprehensively considers multiple physical fields such as fluid mechanics, thermodynamics and material science. It can describe the flow, pressure distribution, temperature change and interaction of liquids with pipe materials in the pipe network. The system uses the finite element method to discretize this multi-physics field model and create a numerical model that can simulate the entire pipe network system.

[0060] When building a leak inversion model, the system first parameterizes the multi-physics model. It uses global sensor data to calibrate and optimize model parameters to ensure that the model accurately reflects the current system status. This process involves large-scale data analysis and model fitting. The system uses machine learning algorithms, such as genetic algorithms or particle swarm optimization, to automatically adjust model parameters so that the model output matches the actual observed data to the greatest extent possible. Next, the system builds the core part of the leak inversion model - the sensitivity matrix. This matrix describes the impact of changes in the location and intensity of the leak on the readings of each sensor. The system constructs a complete sensitivity matrix by performing a large number of simulations on the multi-physics model, setting virtual leak points at different locations and intensities, and calculating the response of each sensor. This matrix is ​​a key bridge connecting the characteristics of the leak source with the observed data.

[0061] After the leakage inversion model is built, the system starts to perform inversion operations based on the model and the abnormal data. The inversion process uses an iterative method. The system first assumes an initial leakage location and intensity, and then uses the leakage inversion model to calculate the sensor readings under this assumption. The calculated results are compared with the actual observed abnormal data, and the system calculates the error.

[0062] Based on this error, the system uses advanced optimization algorithms, such as the conjugate gradient method or the Levenberg-Marquardt algorithm, to adjust the assumed leak location and strength. This process is repeated until the error between the calculated result and the observed data drops below a preset threshold, or the maximum number of iterations is reached.

[0063] In order to improve the accuracy and efficiency of inversion, the system adopts a multi-scale strategy. It first performs a quick inversion on a coarser grid to determine the approximate leakage area, and then uses a finer grid in this area for precise positioning. This method greatly reduces the amount of calculation while ensuring the accuracy of positioning.

[0064] The system also takes data uncertainty into account during the inversion process. It uses Bayesian inference methods to take into account the noise in the observed data and the uncertainty in the model to obtain a probability distribution of the leak location and intensity, rather than just an estimate of a single location.

[0065] Finally, the system obtains initial anomaly data through this inversion process. These data include the estimated leak location coordinates, leak intensity, and the uncertainty estimates associated with them. The system also generates a leak heat map that intuitively shows the probability of different locations in the entire pipeline network becoming leak sources.

[0066] S502, optimally iterating the initial abnormal data and the abnormal data to obtain an optimal leakage parameter, and determining leakage source location data according to the optimal leakage parameter.

[0067] Specifically, the initial abnormal data obtained in step S501 is used as the starting point to start the optimal iteration process. The goal of this process is to minimize the difference between the model prediction results and the actual observed abnormal data by continuously adjusting the leakage parameters. The system uses advanced optimization algorithms to perform this iteration process, including the improved Levenberg-Marquardt algorithm, genetic algorithm, and particle swarm optimization algorithm.

[0068] In the optimal iteration process, the system first defines an objective function. This function is usually the weighted sum of squared errors between the model predictions and the observed abnormal data. The weight setting takes into account the reliability and importance of different sensors, ensuring that more reliable data plays a greater role in the optimization process.

[0069] The system uses the Levenberg-Marquardt algorithm as the main optimization method. This algorithm combines the advantages of gradient descent and Gauss-Newton methods and can converge quickly in nonlinear least squares problems. In each iteration, the algorithm calculates the Jacobian matrix and the gradient of the objective function, and then decides whether to use gradient descent or Gauss-Newton steps based on the current state. This adaptive strategy allows the algorithm to converge stably when it is far from the optimal solution, and converge quickly when it is close to the optimal solution.

[0070] In order to avoid falling into the local optimal solution, the system also adopts a multi-starting point strategy. It starts the optimization from multiple different initial points, which are determined based on the results of step S501, combined with expert knowledge and historical data. The system runs multiple optimization instances simultaneously and uses parallel computing technology to improve efficiency.

[0071] During the iteration process, the system dynamically adjusts the optimization parameters. For example, it adjusts the learning rate according to the changing trend of the objective function. In the early stage of the iteration, a larger learning rate is used to quickly approach the optimal solution, and when it is close to convergence, the learning rate is reduced to improve accuracy. At the same time, the system also adopts an early stopping strategy. When the improvement of multiple consecutive iterations is less than the preset threshold, the iteration is terminated early to avoid overfitting.

[0072] In order to further improve the robustness of the optimization, the system introduces a regularization term. This regularization term takes into account the physical constraints of the leakage parameters, such as the non-negativity of the leakage intensity and the boundary conditions of the location. This not only ensures the physical meaning of the optimization results, but also helps to suppress the overfitting problem caused by data noise.

[0073] During the iteration process, the system also performs a sensitivity analysis. It calculates the sensitivity of the objective function to each leakage parameter and identifies the parameters that have the greatest impact on the result. These highly sensitive parameters receive more attention during the optimization process, and the system allocates more computing resources to them, thereby improving the optimization efficiency.

[0074] When the iteration process converges or reaches the preset maximum number of iterations, the system obtains the optimal leakage parameters. These parameters include the precise coordinates of the leakage source, the leakage intensity, and the uncertainty range of these estimates. The system uses these optimal leakage parameters, combined with the geometric information and operating status of the pipeline network, to determine the final leakage source location data.

[0075] Based on the above embodiment, as an optional embodiment, before inputting the real-time response signal data into the multi-physics field model to obtain the global sensor data, the method includes: S601, constructing a pipeline multi-physics field coupling model according to the structural data and the process data; Specifically, the system collects and organizes structural data. These data include the geometric parameters of the pipeline (such as pipe diameter, wall thickness, length), material properties (such as elastic modulus, Poisson's ratio, thermal expansion coefficient), connection methods (such as welding, flanges), and pipeline layout and direction. The system uses computer-aided design (CAD) software and geographic information system (GIS) technology to convert these structural data into accurate three-dimensional digital models.

[0076] Next, the system integrates process data. These data cover the operating parameters of the pipeline system, including fluid type, flow, pressure, temperature, and the working status of various control valves and pumps. The system obtains these process data in real time through the interface with the process control system (PCS) and distributed control system (DCS) and associates them with the structural model.

[0077] When building a multi-physics coupled model, the system takes into account the interaction of multiple physical fields. This includes fluid dynamics, thermodynamics, structural mechanics, and in some cases, electromagnetics. The system uses finite element method (FEM) and computational fluid dynamics (CFD) techniques to describe these physical fields.

[0078] For the fluid dynamics field, the system uses the Navier-Stokes equations to describe the flow of fluid in the pipe. It takes into account factors such as the viscosity, compressibility, and friction with the pipe wall of the fluid. The system uses high-order numerical formats, such as the WENO format, to handle the fluid shocks and discontinuities that occur.

[0079] The simulation of the thermodynamic field is based on the energy conservation equation. The system takes into account the heat conduction of the fluid, the heat transfer by convection, and the heat exchange with the environment. It also simulates the thermal expansion and stress changes of the material caused by temperature changes.

[0080] The simulation of the structural mechanics field is based on the theory of elasticity. The system calculates the stresses and deformations caused by internal pressure, temperature changes, and external loads. It pays special attention to stress concentration areas such as pipe elbows and joints, which are often high-risk areas for leakage.

[0081] In some special cases, such as electrochemical corrosion in metal pipes, the system also takes into account the influence of electromagnetic fields. It simulates the current distribution and potential differences to assess the corrosion risk.

[0082] In order to achieve the coupling of these physical fields, the system adopts a multi-physics coupling algorithm. It uses the alternating direction implicit (ADI) method to deal with the interaction between different physical fields. This method can effectively deal with rigid problems and ensure the stability of the numerical solution.

[0083] During the construction process, the system also takes into account boundary conditions and initial conditions. It sets appropriate inlet and outlet boundary conditions, as well as heat transfer and friction conditions of the pipe wall according to the actual operating conditions. The system also takes into account time-varying conditions, such as startup, shutdown, and load changes.

[0084] In order to improve the calculation efficiency, the system uses adaptive grid technology. In the area where the fluid changes violently or the stress is concentrated, the system automatically refines the grid, and uses a coarser grid in the area where the change is gentle. This strategy not only ensures the calculation accuracy of the key areas, but also reduces the overall calculation load.

[0085] During the model validation phase, the system uses historical operating data and laboratory test results to calibrate and optimize model parameters. It uses parameter identification techniques, such as genetic algorithms and particle swarm optimization, to adjust model parameters so that the model output matches the actual observed data to the greatest extent possible.

[0086] The implementation of this approach has brought significant results. First, by integrating structural data and process data, the system creates a highly accurate virtual pipeline system. This model can not only accurately reflect the static characteristics of the pipeline, but also simulate the system behavior under various dynamic conditions. Secondly, the establishment of a multi-physics field coupling model enables the system to fully consider the impact of various physical factors on pipeline integrity, thereby more accurately predicting potential leakage risks.

[0087] S602, obtaining historical global data of the pipeline, and constructing a training set and a validation set according to the historical global data of the pipeline; Specifically, the system begins to collect historical global data of the pipeline. These data include multiple aspects: operating parameters (such as pressure, flow, temperature), environmental factors (such as air temperature, humidity, geological conditions), equipment status (such as valve opening, pump operating status), maintenance records (such as maintenance time, replacement of parts), and detailed records of historical leakage events (such as leakage location, cause, scale). The system obtains this data through multiple channels, including the historical database of the real-time monitoring system, operation logs, maintenance reports, accident investigation reports, etc.

[0088] During the data collection process, the system pays special attention to the integrity and quality of the data. It uses data verification algorithms to identify and mark outliers, missing values, and inconsistent data. For detected problematic data, the system uses a variety of strategies to handle them. For missing values, depending on the data type and degree of missingness, the system uses interpolation methods (such as linear interpolation, spline interpolation), mean / median filling, or more complex machine learning methods (such as K nearest neighbor algorithm) to estimate missing values. For outliers, the system combines expert knowledge and statistical methods (such as Z-score method, IQR method) to determine whether it is a true anomaly or a measurement error, and retains or corrects these data accordingly.

[0089] Next, the system preprocesses the collected data. This includes data cleaning, standardization, and feature engineering. During data cleaning, the system removes duplicate records and handles inconsistent data formats. In the standardization stage, the system unifies data of different dimensions to the same scale, usually using Z-score standardization or Min-Max scaling methods. In the feature engineering stage, the system creates new features, such as pressure gradient, flow rate change rate, etc., to capture more information.

[0090] After data preprocessing is completed, the system begins to build training sets and validation sets. To ensure the representativeness and balance of the data set, the system uses a stratified sampling method. It first divides the data into multiple layers based on key features (such as pipe type, operating conditions, whether leakage occurs, etc.), and then randomly extracts a certain proportion of samples from each layer. This method ensures that both the training set and the validation set reflect the distribution characteristics of the overall data.

[0091] The system usually uses 70-80% of the data to build the training set and the remaining 20-30% to build the validation set. For relatively rare cases such as leakage events, the system uses sampling techniques (such as the SMOTE algorithm) to increase the minority class samples to solve the data imbalance problem.

[0092] The system also takes time correlation into account when constructing the training set. It ensures that the training set contains a long enough time span to capture seasonal changes and long-term trends. For the validation set, the system selects data from a recent period to better evaluate the model's ability to adapt to the latest situation.

[0093] In addition, the system also creates an independent test set. This test set is completely independent of the training and validation process and is used to finally evaluate the generalization ability of the model. The test set usually selects a small amount of recent data or particularly important historical event data.

[0094] In the process of building these datasets, the system strictly adheres to the principle of data isolation to ensure that information in the validation set and test set does not leak into the training process, avoiding overfitting and false high performance evaluation.

[0095] The implementation of this approach has brought significant results. First, by systematically collecting and processing historical global data, the system has established a comprehensive and accurate database, which provides a solid foundation for subsequent model training and verification. These data not only contain information about normal operating conditions, but also various anomalies and leakage situations, allowing the model to learn more diverse patterns.

[0096] S603, training the pipeline multi-physics field coupling model according to the training set and the verification set, and obtaining the multi-physics field model when the simulation accuracy of the pipeline multi-physics field coupling model reaches a preset threshold.

[0097] Specifically, the system began to train the pipeline multi-physics field coupling model. The training process adopted an iterative optimization method, which mainly included two aspects: parameter estimation and model structure optimization. In the parameter estimation stage, the system used the data in the training set to adjust the model parameters by minimizing the error between the model prediction value and the actual observation value. The system uses a variety of optimization algorithms, such as gradient descent, Newton's method, genetic algorithm, etc., to adapt to the characteristics of different types of parameters. For parameters with strong nonlinearity, the system also uses global optimization methods such as simulated annealing algorithm and particle swarm optimization algorithm to avoid falling into local optimal solutions.

[0098] In the model structure optimization stage, the system uses adaptive model structure technology. It analyzes the model residuals and sensitivity analysis results to identify problematic parts of the model, such as whether the coupling relationship between certain physical fields is correctly described or whether new physical mechanisms need to be introduced. The system will try different model structures, such as adding or reducing certain coupling terms, adjusting the interaction between physical fields, etc., and evaluate the impact of these changes on model performance.

[0099] During the training process, the system uses cross-validation technology to prevent overfitting. It uses the K-fold cross-validation method to divide the training set into K subsets, using K-1 subsets for training each time, and the remaining subset for validation. This process is repeated K times, and each subset is used as a validation set once. In this way, the system can better evaluate the generalization ability of the model.

[0100] In order to improve the training efficiency, the system adopts parallel computing technology. It uses high-performance computing clusters to distribute large-scale parameter optimization tasks to multiple computing nodes for simultaneous execution. The system also implements a dynamic load balancing strategy to fully utilize computing resources and accelerate the training process.

[0101] At each iteration of training, the system uses the validation set to evaluate the performance of the model. It calculates multiple performance metrics, including mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R²), etc. For classification problems such as leak detection, the system also calculates metrics such as precision, recall, F1 score, etc. The system pays special attention to the performance of the model in extreme cases and rare events (such as large-scale leaks), and sets higher weights for these cases.

[0102] The system sets a preset threshold as the target of model performance. This threshold is determined based on industry standards, expert experience and actual system requirements. For example, the system requires that the prediction error of the model under normal operation does not exceed 1%, the false positive rate in leak detection does not exceed 0.1%, and the true positive rate is not less than 99.9%.

[0103] When the simulation accuracy of the model reaches the preset threshold, the system considers that a multi-physics model that meets the requirements has been obtained. However, in order to further ensure the reliability of the model, the system will also use the previously retained independent test set for final verification. Only when the performance of the model on the test set also meets the preset standards will the system officially adopt this model.

[0104] See also Figure 2 , Figure 2 An architecture diagram of a glass fiber reinforced plastic pipeline liquid leakage detection system provided in an embodiment of the present application, the glass fiber reinforced plastic pipeline liquid leakage detection system may include: Data acquisition module 1, used to acquire the engineering project plan of the engineering project and determine the expected engineering cost of the current process according to the engineering project plan; The unfinished workload determination module 2 is used to obtain the completed workload of the current process, the engineering cost of the completed workload and the unfinished workload, and determine the manpower data, engineering cycle and raw material data required for the unfinished workload based on the unfinished workload and the engineering project plan; The cost prediction module 3 is used to obtain the labor price trend curve and the raw material price trend curve, and determine the predicted cost of the unfinished project volume according to the labor price trend curve, the raw material price trend curve, the required labor data, the project cycle and the raw material data; The early warning module 4 is used to determine the predicted cost of the current process based on the predicted cost of the unfinished project volume and the project cost of the completed project volume, and to compare the predicted cost of the current process with the expected project cost in real time to obtain a comparison difference. When the comparison difference exceeds a preset difference threshold, a risk early warning signal is generated.

[0105] It should be noted that: when the system provided in the above embodiment realizes its functions, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0106] Please refer to Figure 3 The application also discloses an electronic device. Figure 3 The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302 or end-to-end wireless communication.

[0107] The communication bus 302 is used to realize the connection and communication between these components.

[0108] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0109] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0110] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field~Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0111] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read~Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally be at least one storage system located away from the aforementioned processor 301. Refer to Figure 3 The memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module and an application program of a method for detecting liquid leakage in a glass fiber reinforced plastic pipeline.

[0112] exist Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call the application program of the glass fiber reinforced plastic pipeline liquid leakage detection method stored in the memory 305. When executed by one or more processors 301, the electronic device 300 executes one or more methods in the above-mentioned embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application. In the above-mentioned embodiments, the descriptions of each embodiment have their own emphasis. For the parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0113] In the several implementations provided in this application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are only schematic, such as the division of modules, which is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, indirect coupling or communication connection of systems or modules, which can be electrical or other forms.

[0114] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0115] The present application also provides a computer storage medium that can store multiple instructions, which are suitable for being loaded and executed by a processor as described above. Figure 1 The glass fiber reinforced plastic pipe liquid leakage detection method of the embodiment shown in the figure can be found in the specific implementation process. Figure 1 The specific description of the illustrated embodiment will not be repeated here.

[0116] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0117] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0118] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0119] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for detecting liquid leakage in a glass fiber reinforced plastic pipeline, characterized in that: The method comprises: Acquire structural data, process data and operation data of the pipeline, and input the structural data, process data and operation data into a preset liquid leakage risk warning knowledge graph to obtain a leakage risk assessment result of the pipeline; Obtaining a liquid level value in a pipeline liquid collecting tank, and when the liquid level value exceeds a preset liquid level threshold and / or the leakage risk assessment result is a high risk level, obtaining real-time response signal data of each sensor in the pipeline, and inputting the real-time response signal data into a multi-physics field model to obtain global sensor data; The leakage source is inverted and located based on the global sensor data to obtain leakage source position data.

2. The method according to claim 1, characterized in that Before acquiring the real-time response signal data of each sensor in the pipeline, the method further includes: Acquire multi-physics field data of the pipeline, and construct a multi-physics field model according to the multi-physics field data; A target sensor layout plan is generated based on the multi-physics field model, and the sensor layout is performed on the pipeline based on the target sensor layout plan.

3. The method according to claim 1, characterized in that Before inputting the structural data, the process data and the operating data into a preset liquid leakage risk knowledge graph to obtain a pipeline leakage risk assessment result, the method further includes: Acquire pipeline leakage sample data, and extract the pipeline leakage sample data to obtain relationship entities related to leakage risks; Performing fusion processing on the relationship entities to build a consistent entity relationship pedigree; A triple entity relationship representation is constructed based on the entity relationship pedigree, and the preset liquid leakage risk knowledge graph is constructed based on the triple entity relationship representation.

4. The method according to claim 3, characterized in that The fusing of the relationship entities to build a consistent entity relationship pedigree includes: Performing alignment processing on the relationship entities to obtain a standard relationship entity representation; Processing entity conflicts in the standard relationship entity representation to obtain a first standard relationship entity representation; The relationship conflict in the first standard relationship entity expression is processed to obtain a second standard relationship entity expression, and the second standard relationship entity expression is organized and linked according to a preset pedigree model to obtain the consistent entity relationship pedigree.

5. The method according to claim 1, characterized in that The inverse positioning of the leakage source based on the global sensor data to obtain the leakage source position data includes: The global sensor data is compared with a preset normal sensor threshold to determine abnormal data in the global sensor data, and the leakage source is inverted and located according to the abnormal data to obtain the leakage source position data.

6. The method according to claim 5, characterized in that The step of performing inverse positioning of the leakage source according to the abnormal data to obtain the leakage source position data includes: Constructing a leakage inversion model according to the global sensor data and the multi-physics field model, and performing an inversion operation based on the leakage inversion model and the abnormal data to obtain initial abnormal data; The initial abnormal data and the abnormal data are optimally iterated to obtain an optimal leakage parameter, and the leakage source position data is determined according to the optimal leakage parameter.

7. The method according to claim 1, characterized in that Before inputting the real-time response signal data into the multi-physics field model to obtain global sensor data, the method includes: Constructing a pipeline multi-physics field coupling model according to the structural data and the process data; Acquire historical global data of the pipeline, and construct a training set and a validation set according to the historical global data of the pipeline; The pipeline multi-physics field coupling model is trained according to the training set and the verification set, and when the simulation accuracy of the pipeline multi-physics field coupling model reaches a preset threshold, a multi-physics field model is obtained.

8. A glass fiber reinforced plastic pipeline liquid leakage detection system, characterized in that: The system comprises: A risk assessment module, used to obtain structural data, process data and operation data of the pipeline, and input the structural data, process data and operation data into a preset liquid leakage risk warning knowledge graph to obtain a leakage risk assessment result of the pipeline; A data processing module is used to obtain the liquid level value in the pipeline liquid collecting tank, and when the liquid level value exceeds a preset liquid level threshold and / or the leakage risk assessment result is a high risk level, obtain the real-time response signal data of each sensor in the pipeline, and input the real-time response signal data into the multi-physics field model to obtain global sensor data; The positioning module is used to perform leakage source inversion positioning based on the global sensor data to obtain leakage source position data.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a processor, a memory and a transceiver, the memory is used to store instructions, the transceiver is used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method according to any one of claims 1 to 7.

Images