Ethylene device exception handling system and equipment based on knowledge graph

Abstract

The invention relates to the field of chemical production, in particular to an ethylene device exception handling system and equipment based on a knowledge graph. The system carries out representation learning on a knowledge graph of the ethylene device through an embedded learning module and a plurality of preset embedded strategies so as to obtain vector representation corresponding to the knowledge graph, and carries out anomaly reasoning according to the vector representation through a reasoning prediction module so as to obtain an anomaly prediction result of the ethylene device and a corresponding solution. Thus, information such as equipment, faults and process parameters of the ethylene device is associated in a structured mode by combining a knowledge graph, embedded learning, reasoning and prediction algorithms, the equipment faults of the ethylene device are rapidly and accurately detected, an abnormal source is positioned, and an efficient fault solution is provided.

Description

Technical Field

[0001] The present application relates to the field of chemical production, and in particular to an ethylene plant abnormality handling system and equipment based on a knowledge graph. Background Art

[0002] In modern industrial production, ethylene, as a key chemical raw material, is widely used in fields such as plastics, synthetic fibers, and rubber. As crucial equipment in the chemical industry, ethylene plants are facing increasing complexity in the operating procedures and knowledge associated with their production processes, driven by the continuous development of ethylene production processes.

[0003] However, the operating manuals and related operating procedures of ethylene plants are usually updated slowly, and traditional methods of handling abnormalities in ethylene plants often rely on manual data search or expert consultation. This is not only inefficient but also easily limited by experience, resulting in a high error rate in abnormal handling, which seriously reduces the efficiency and accuracy of abnormal handling in ethylene plants. Summary of the Invention

[0004] The present invention provides an ethylene plant exception handling system and equipment based on a knowledge graph, which is used to improve the efficiency and accuracy of exception handling in ethylene plants, thereby improving the efficiency and safety of ethylene production.

[0005] In a first aspect, the present application provides an ethylene plant exception handling system based on a knowledge graph, the system comprising:

[0006] an embedding learning module for performing representation learning on the knowledge graph based on a plurality of preset embedding strategies to obtain a vector representation corresponding to the knowledge graph; the knowledge graph comprising a plurality of entities of an ethylene plant and relationships between the entities, wherein each entity corresponds to a device, a sensor, and control logic of the ethylene plant; the vector representation representing a feature vector of each entity and relationship in the knowledge graph;

[0007] The reasoning and prediction module is used to perform abnormal reasoning based on the vector representation to obtain abnormal prediction results of the ethylene device and corresponding solutions.

[0008] Optionally, the system further includes an interactive question-and-answer module for responding to questions about ethylene plants input by a target object and outputting corresponding question-and-answer results based on the knowledge graph.

[0009] Optionally, the system further includes a graph construction module for acquiring device data of the ethylene device and constructing the knowledge graph based on the device data.

[0010] Optionally, the graph construction module is specifically used to:

[0011] Based on the ethylene plant data, an ontology file corresponding to the knowledge graph is determined, wherein the ontology file includes custom classes, relationship attributes, and data attributes; the classes represent the physical entities, operation logic, and abnormal phenomena involved in the operation of the ethylene plant, the relationship attributes represent the association between the classes, and the data attributes represent the attribute information corresponding to each class;

[0012] The unstructured data in the ethylene plant data is structured to obtain triple data of the knowledge graph, and the knowledge graph is visualized.

[0013] Optionally, the multiple embedding strategies include a geometric model embedding strategy, a language model embedding strategy, and a graph neural network embedding strategy; wherein,

[0014] The geometric model embedding strategy representation: representing the triple data of the knowledge graph through a geometric model;

[0015] The language model embedding strategy representation: Based on natural language processing technology, combined with a pre-trained language model, the text information in the ethylene plant data is embedded to obtain corresponding semantic representation;

[0016] The graph neural network embedding strategy is characterized by: processing the entities and association relationships of the knowledge graph based on the graph neural network to determine the high-order relationships between the entities in the knowledge graph; and generating vector representations of each entity through the graph convolutional network.

[0017] Optionally, the inference prediction module is specifically used to:

[0018] Based on a preset anomaly detection training set and the vector representation, performing anomaly reasoning on the ethylene plant; the anomaly detection training set includes ethylene plant data labeled with normal and abnormal states;

[0019] When an abnormal situation occurs in the ethylene device, the abnormality prediction result and the solution are output based on the knowledge graph.

[0020] Optionally, the inference prediction module is further used to:

[0021] When the knowledge graph does not include the solution, generating a solution corresponding to the abnormal prediction result based on a preset decision algorithm;

[0022] Based on the current operating status of the ethylene plant, personalized recommendations are made for the solution.

[0023] Optionally, the ethylene plant data includes at least equipment information, sensor data, process parameters and control loop information of the ethylene plant.

[0024] Optionally, the interactive question-and-answer module is specifically used to:

[0025] Extract keywords from the ethylene plant problem and determine corresponding abnormality description information;

[0026] Based on the knowledge graph and the inference prediction module, the question and answer results corresponding to the abnormal description information are obtained and fed back to the target object; the question and answer results include the corresponding abnormal prediction results and solutions.

[0027] Optionally, the system further includes a visualization module for displaying the question-and-answer results and the equipment status of the ethylene plant to the target object based on a visualization interface.

[0028] In a second aspect, the present application provides a method for constructing an ethylene plant abnormality handling system based on a knowledge graph, the method comprising:

[0029] Obtaining demand analysis information of the ethylene device abnormality handling system, wherein the demand analysis information includes user demand information and system demand information;

[0030] Based on the demand analysis information, a system development environment is constructed, and based on the system development environment, a system architecture of the ethylene device abnormality handling system is constructed; the system architecture includes a data layer, an application service layer, and a user layer;

[0031] Based on a preset modular development strategy, the functional modules of the ethylene unit abnormality handling system are constructed in sequence.

[0032] In a third aspect, the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the memory is used to store a knowledge graph of an ethylene device; and when the processor executes the computer program, the steps of any one of the ethylene device exception handling systems described in the first aspect are implemented.

[0033] The beneficial effects of the present invention are as follows:

[0034] An embodiment of the present application provides an ethylene plant exception handling system and equipment based on a knowledge graph. The system performs representation learning on the knowledge graph through an embedded learning module and a plurality of preset embedding strategies to obtain a vector representation corresponding to the knowledge graph. The system performs exception reasoning based on the vector representation through an inference and prediction module to obtain an exception prediction result of the ethylene plant and a corresponding solution. Thus, the system combines the knowledge graph, embedded learning, reasoning and prediction algorithms, and associates the equipment, faults, process parameters and other information of the ethylene plant in a structured manner, thereby realizing rapid and accurate detection of equipment faults of the ethylene plant, locating the source of the abnormality, and providing efficient fault solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0036] Figure 1 A schematic diagram of the structure of an ethylene plant abnormality handling system based on a knowledge graph provided in an embodiment of the present application;

[0037] Figure 2 A schematic diagram of the processing flow of a knowledge graph construction module provided in this application;

[0038] Figure 3 A schematic diagram of an ontology file provided in an embodiment of the present application;

[0039] Figures 4(a) and 4(b) are schematic diagrams of entities of a knowledge graph provided in an embodiment of the present application;

[0040] Figure 5 A schematic diagram of a local representation of a knowledge graph provided in an embodiment of the present application;

[0041] Figure 6 A schematic diagram of a processing flow of an embedded learning module provided in an embodiment of the present application;

[0042] Figure 7 A schematic diagram of the architecture of an interactive question-and-answer module provided in an embodiment of the present application;

[0043] Figure 8 A schematic diagram of an abnormal reasoning process provided in an embodiment of the present application;

[0044] Figure 9 A schematic diagram of the system architecture of an ethylene plant abnormality handling system provided in an embodiment of the present application;

[0045] Figure 10 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present application clearer, 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, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Unless there is a conflict, the embodiments in the present application and the features in the embodiments can be combined with each other in any way. In addition, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in an order different from that here.

[0047] The terms "first" and "second" in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the term "comprising" and any of its variations are intended to cover non-exclusive protection. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices. "Multiple" in this application can mean at least two, for example, two, three or more, and the embodiments of this application are not limited thereto.

[0048] The term "and / or" in the embodiments of this application is simply a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0049] It is understood that in the following specific implementation methods of this application, data related to the chemical production process is involved. When the various embodiments of this application are applied to specific products or technologies, relevant licenses or consents need to be obtained, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions. For example, it is possible to recruit relevant volunteers and sign relevant agreements for volunteer authorization data, and then use the data of these volunteers for implementation; or, by implementing within the scope of an authorized organization, data management is carried out by implementing the following implementation methods using the data of internal members of the organization; or, the relevant data used in the specific implementation are all simulated data, such as simulated data generated in a virtual scene.

[0050] The following is a brief introduction to the design concept of the embodiments of this application.

[0051] In modern industrial production, ethylene, as a key chemical raw material, is widely used in fields such as plastics, synthetic fibers, and rubber. As crucial equipment in the chemical industry, ethylene plants are facing increasing complexity in the operating procedures and knowledge associated with their production processes, driven by the continuous development of ethylene production processes.

[0052] However, the operating manuals and related operating procedures of ethylene plants are usually updated slowly, and traditional methods of handling abnormalities in ethylene plants often rely on manual data search or consultation with experts. This is not only inefficient but also easily limited by experience, resulting in a high error rate in abnormal handling, which seriously reduces the efficiency of abnormal handling in ethylene plants.

[0053] In view of the above problems, an embodiment of the present application provides an ethylene plant abnormality handling system based on a knowledge graph. The system performs representation learning on the knowledge graph through an embedded learning module and a plurality of preset embedding strategies to obtain a vector representation corresponding to the knowledge graph. The system performs abnormality reasoning based on the vector representation through an inference and prediction module to obtain abnormality prediction results of the ethylene plant and corresponding solutions. Thus, the system combines knowledge graphs, embedded learning, reasoning and prediction algorithms, and associates information such as equipment, faults, and process parameters of the ethylene plant in a structured manner, thereby realizing rapid and accurate detection of equipment faults of the ethylene plant, locating the source of the abnormality, and providing efficient fault solutions.

[0054] Furthermore, the system also includes an interactive question-and-answer module and a visualization module, which realizes intelligent question-and-answer functions through natural language processing technology, thereby improving the intelligence level of ethylene plant operation.

[0055] Below, the system provided by the exemplary embodiment of the present application is described with reference to the accompanying drawings. It should be noted that the above application scenarios are only shown to facilitate understanding of the spirit and principles of the present application, and the implementation of the present application is not limited in this respect.

[0056] refer to Figure 1 FIG. 1 is a schematic diagram of a knowledge graph-based ethylene plant exception handling system provided in an embodiment of the present application. The system includes the following unit modules:

[0057] The embedding learning module is used to perform representation learning on the knowledge graph based on multiple preset embedding strategies to obtain the vector representation corresponding to each entity and relationship in the knowledge graph.

[0058] The reasoning and prediction module is used to perform knowledge reasoning based on vector representation to determine the abnormal prediction results and corresponding solutions corresponding to the ethylene plant.

[0059] The interactive question-and-answer module is used to respond to the question input operation of the target object, obtain the abnormal questions input by the target object for the ethylene plant, and output the corresponding question-and-answer results based on the knowledge graph.

[0060] In a possible embodiment, the system may further include a graph construction module for acquiring device data of the ethylene plant and constructing a knowledge graph of the ethylene plant based on the device data.

[0061] Specifically, the graph construction module can collect ethylene plant data such as equipment information, sensor data, process parameters, and control loop information from the ethylene plant to construct a knowledge graph containing entities such as equipment, sensors, and control logic. A graph database is used to store and manage these entities and their relationships, such as the causal relationships between equipment interlocking actions, the topological relationships between logistics flows, and the mapping relationships between sensors and control loops. The knowledge graph also defines the operating status of the equipment, the range of sensor readings, and their abnormality standards, providing basic data support for subsequent reasoning and question-answering modules. In this way, through ontology construction, classes, relationship attributes, and data attributes can be defined based on the basic concepts of the ethylene plant regulations, thereby providing semantic definitions and specifications for nodes and edges (i.e., entities and relationships) in the knowledge graph.

[0062] In one possible implementation, reference Figure 2 The figure shows a processing flow diagram of a knowledge graph construction module provided by this application. The graph construction module can construct an ontology file corresponding to the knowledge graph based on ethylene plant data. This ontology file includes custom classes, relationship attributes, and data attributes. Classes represent the physical entities, operational logic, and abnormal phenomena involved in the operation of the ethylene plant. Relationship attributes represent the associations between classes, that is, the relationship between two entities in the same knowledge representation, reflecting the abnormal propagation path and processing logic in the ethylene plant. Data attributes represent the attribute information corresponding to each class, that is, the attribute value and / or state description information of an entity data in the same knowledge representation.

[0063] Specifically, the classes of the ontology file in this application may include process parameters, process logistics, equipment, and instruments of an ethylene plant. For example, this application may divide the classes into equipment, pollutants, mixtures, process parameters, process logistics, operating units, abnormal phenomena, actuators, controllers, control loops, and chain action results based on process technical regulations and operating guidelines. The relationship attributes of the ethylene plant may include interlocking actions and causes between devices, the flow direction of logistics that describes the logistics topology, and the merge into that describes the equipment combination relationship, such as interlocking actions, from, including unit equipment, merge into, flow direction into, diversion into, pollutants into, etc. Data attributes may include the bit number of the equipment, the numerical range of the sensor, and the operating pressure unit of the process parameter, such as temperature, bit number, pressure, inner diameter, name, specifications and fixed pressure, description, alarm value, value, blind plate position, operating temperature, operating pressure, operating specifications, phase state, etc.

[0064] Specifically, refer to Figure 3 The figure shows a schematic diagram of an ontology file provided by this application. Figure 3 Each node in the ethylene plant covers the equipment entities (actuators, controllers, etc.), process parameters (NH3, SO2, NOX, etc.), control loops (layered control loops, three-impulse control, etc.), abnormal phenomena (out-of-control phenomena, exogenous phenomena, etc.), association relationships (logistics flow to, merger at, etc.) and data attributes.

[0065] Next, the graph construction module in this application can perform structured transformation on the unstructured data in the ethylene plant data, obtain triple data corresponding to the knowledge graph, realize knowledge extraction, and thus store the knowledge graph in different forms. Based on the above knowledge extraction, the knowledge graph can be visualized using the Neo4j graph database.

[0066] Specifically, as above Figure 2 As shown, for chart data from ethylene plant data, such as ethylene operating procedures, this application can extract knowledge using a Python-developed ethylene operating procedure extraction program and a knowledge graph extraction program. These data are then stored in the Resource Description Framework (RDF) format, conforming to knowledge ontology specifications. For unstructured text statements, this application can employ the Begin-Inside-Outside (BIO) annotation method, using Doccano software to obtain the corresponding triple data.

[0067] In a possible implementation, the present application can extract data based on the construction of a good knowledge ontology, and the data can mainly come from engineering and technical documents, including the "Technical Regulations for Ethylene Plants" and the "Process Flow Diagram (PFD) for Ethylene Plants". Among them, the "Technical Regulations for Ethylene Plants" contains basic data such as quality indicators of raw materials and chemical raw materials, equipment lists, device interlocking actions and causes, and control parameter alarm values, most of which are presented in tabular form. The data in the "PFD for Ethylene Plants" mainly connects the relationship between the equipment in series based on the flow direction of logistics and presents it in a graphical form. As shown in Table 1 below, the data of the boundary blind plate table in tabular form is data. For this type of tabular data, the present application can use Python's related table document library (such as Pandas) for extraction, and the graphical data is manually extracted, and the data is saved in RDF format.

[0068]

[0069] Table 1 Boundary blind plate table

[0070] Further, referring to FIG4 (a) and FIG4 (b), which are schematic diagrams of entities of a knowledge graph provided by an embodiment of the present application, FIG4 (a) and FIG4 (b) are entities of blind plate type and pipeline type, respectively, and the present application visualizes them through Neo4j. Figure 5 The figure shows a schematic diagram of a partial representation of a knowledge graph provided by this application, where burgundy represents the "system unit class" and blue represents the "instrument class", and the line between the two represents the association relationship. Figure 5 It also shows the measuring instruments of the "cracking gas compressor system".

[0071] In one possible embodiment, the embedding learning module in the present application performs representation learning on the knowledge graph through a variety of different embedding strategies, thereby extracting the semantic information of entities and relationships in the graph and representing it in the form of vectors. The various embedding strategies of the present application may include geometric model embedding strategies, language model embedding strategies and graph neural network embedding strategies. Among them, the geometric model embedding strategy represents: representing the triple data of the knowledge graph through a geometric model, the language model embedding strategy represents: based on natural language processing technology, combined with a pre-trained language model, the text information in the ethylene plant data is embedded to obtain the corresponding semantic representation; the graph neural network embedding strategy represents: processing the entities and association relationships of the knowledge graph based on the graph neural network, determining the high-order relationships between the entities in the knowledge graph, and generating vector representations of each entity through a graph convolutional network.

[0072] Specifically, the geometric model embedding strategy in this application can embed the relationship between nodes and edges through spatial geometry methods to process more complex spatial relationships between entities. The language model embedding strategy in this application can be combined with natural language processing technology, and use the bidirectional encoder representation transformation (Bidirectional Encoder Representations from Transformers, BERT) to pre-train the language model to embed text information such as process parameters and equipment descriptions that appear in the ethylene unit data to obtain a more semantic representation, thereby helping the system understand the complex relationships and semantics in the knowledge graph and provide high-quality input data for subsequent reasoning and prediction. The graph neural network (GNN) embedding strategy in this application can process the nodes and edges (i.e., entities and association relationships) in the knowledge graph through the graph neural network to learn the high-order relationships between entities in the graph, and generate a vector representation of each node (entity) through the graph convolutional network (Graph Convolutional Network, GCN). In summary, the embedding learning module in this application can convert entities such as equipment, sensors, process parameters and their relationships in the ethylene plant into low-dimensional vector representations through embedding learning technologies such as geometric models, language models and graph neural networks, thereby capturing the complex relationships and dependencies between devices. In this way, through multi-dimensional embedding learning, this system can more accurately model the normal and abnormal states of the equipment.

[0073] In one possible implementation, the inference and prediction module in this application can perform anomaly inference on an ethylene plant using a pre-set anomaly detection training set and vector representation. The anomaly detection training set includes ethylene plant data labeled with normal and abnormal states. When an anomaly occurs in the ethylene plant, the knowledge graph outputs anomaly prediction results and solutions.

[0074] Specifically, the reasoning and prediction module in this application can construct an anomaly detection training set for anomaly detection through ethylene plant data such as historical equipment operation data, sensor data and fault records, and mark normal and abnormal states. Then, the embedded learning module is used to train the equipment's operation data, learn the difference between the normal and abnormal states of the equipment, and combine the embedding results of the graph neural network, geometric model and language model to infer possible abnormal situations. For example, when the reading of a certain sensor deviates greatly from the historical data, it is inferred that there may be an equipment failure. And by inferring the equipment relationship in the graph, the source of the anomaly can be quickly located. For example, when the pressure of a valve is abnormal, it can be inferred based on the graph structure whether the equipment connected to the valve is affected, and after obtaining the inferred anomaly prediction result, the corresponding solution is output based on the repair measures and processing solutions related to the equipment failure in the query knowledge graph.

[0075] In one possible embodiment, the reasoning and prediction module in the present application is also used to: when the knowledge graph does not include a solution, generate a solution corresponding to the abnormal prediction result through a preset decision algorithm, and make personalized recommendations for the solution based on the current operating status of the ethylene unit.

[0076] Specifically, if the knowledge graph doesn't contain relevant solutions for a particular source, the application can also use a decision-making algorithm to generate corresponding solutions and make personalized recommendations based on the current operating status of the ethylene plant. For example, if the system detects a fault in device A, it will recommend replacement or adjustment.

[0077] In one possible implementation, reference Figure 6 The figure shows a processing flow diagram of an embedding learning module provided by an embodiment of the present application. After data preprocessing and knowledge graph construction, the present application can select three embedding methods: a distance-based translation embedding (Translating Embeddings, TransE) model, a language-based knowledge graph-BERT (KG-Bert) model, and a graph neural network model. The three models are trained separately using triple data to generate embedding vectors, and the node representation generated by the graph neural network is used for reasoning based on the query request submitted by the user, and the relevant entities and relationships are returned. Among them, the graph is one of the most expressive data structures and has been used to model various problems. In order to systematically utilize the entity-relationship combination operation in the knowledge graph embedding technology, the present application can achieve better embedding performance by jointly learning the vector representation of nodes and relationships in the graph.

[0078] Specifically, this application learns a low-dimensional embedding vector for each node and edge in the graph. Node embeddings represent the node's feature information, while edge embeddings represent the relationship information represented by the edge. These embedding vectors are randomly initialized and updated during training based on the optimization objective.

[0079] For example, this application may define a multi-relationship graph G = (V, R, E, X, Z),

[0080] Among them, V represents the node set, R represents the relationship set, and E represents the edge set;

[0081] Represents the d0-dimensional input features of each node;

[0082] Represents the initial relationship characteristics.

[0083] In traditional knowledge graphs, edges are directional. For example, (u, r, v) indicates that entity u points to entity v through relationship r. This application can use the symmetry of information propagation to explicitly add a reverse edge (v, r) to each edge (u, r, v). -1 ,u), and expand the relation set to R′=R∪{r -1}.

[0084] For example, in the ethylene plant operation anomaly dataset, there is a pair of inverse relationships: "causedBy" and "consequenceIn." By introducing inverse edges, the neighborhood N(v) of node v includes not only the nodes pointing to it (e.g., u→v) but also the nodes it points to (e.g., the reverse edge w→v from v→w), thus more comprehensively capturing the graph structure.

[0085] Furthermore, in order to incorporate relation embedding into the GCN formula, this application can utilize the entity-relation combination operation used in the knowledge graph embedding method, which is in the form of

[0086] e0=φ(e s ,e r )

[0087] Where, φ:R d ×R d →R d It is a combination operator, s, r and o represent the subject, relationship and object in the knowledge graph, e (.) ∈R d Represents their respective embeddings. For each edge (u,v,r)∈E, there is a reverse edge (v,u,r -1 ), the representation obtained by k-layer directed GCN is:

[0088]

[0089] The set N(v) is defined as the set of direct neighbors of node v connected through its outgoing edges, and W r is the relationship-specific parameter of the model. To avoid the over-parameterization problem of traditional methods, this application can also fuse neighbor nodes u with their corresponding relationships r by introducing a combination function φ, so that the model can perceive the relationship semantics while maintaining the linear complexity of the feature dimension (O(|R|d)). In addition, to distinguish between original edges, reverse edges, and self-loop edges, independent filters can be defined for each edge type. The update equation is as follows:

[0090]

[0091] Among them, x u ,z rare the initial features of node u and relationship r, respectively, h v is the updated representation of node v, is a relation type specific parameter. In this application, the direction specific weight is used in the GCN model, that is, λ(r) = dir(r), as shown below:

[0092]

[0093] Furthermore, after the node embedding is updated, the relationship embedding in this application is also transformed as follows:

[0094] h r =W rel z r

[0095] in, is a learnable transformation matrix that projects all relations as nodes into the same embedding space and allows them to be used in the next layer of GCN.

[0096] In one possible embodiment, the interactive question-and-answer module in the present application can extract keywords from ethylene plant questions, determine the corresponding abnormal description information, and use the knowledge graph and reasoning prediction module to obtain the abnormal prediction results and solutions corresponding to the abnormal description information, and feed back the question-and-answer results to the target object. In this way, the interactive question-and-answer module can efficiently handle users' abnormal questions about ethylene plants and provide accurate answers based on the reasoning results of the knowledge graph.

[0097] Specifically, the interactive question-and-answer module in this application can perform semantic analysis on questions raised by target objects such as users, and use regular expressions and keyword extraction to identify key entities (such as "temperature sensor") and operational behaviors (such as "fault handling"). Based on the identified abnormal description, the relevant information or reasoning results are determined from the knowledge graph through the knowledge graph query algorithm (such as graph search, path finding, etc.). In this way, the information in the knowledge graph is combined with the reasoning results to obtain a more accurate answer. For example, if the question involves "sensor failure", the query graph is combined with the reasoning module to generate possible fault analysis and solutions, and returned to the user through the question-and-answer module.

[0098] In one possible implementation, reference Figure 7Shown is a schematic diagram of the architecture of an interactive question-and-answer module provided in an embodiment of the present application. In the present application, the interactive question-and-answer module is respectively composed of a user input layer, a data preprocessing layer, a knowledge graph reasoning layer and an output result layer from bottom to top. Among them, in the user input layer, the user can interact with the system through the interface, input ethylene device problems such as fault description, equipment information or process parameters, and initiate a query request to the system. The input method can be free text or a historical input record. In the data preprocessing layer, the interactive question-and-answer module is mainly used to clean, format and structure the original data input by the user, and identify the user's query intention (for example: querying equipment status, requesting fault diagnosis, obtaining historical records, etc.) through natural language processing (Natural Language Processing, NLP) technology, and provide structured data for subsequent reasoning. The knowledge graph reasoning layer encapsulates the trained models of KG-BERT and GCN into two back-end interfaces based on Flask, so that after the user enters the content, the module will call the pre-trained model interface of the corresponding reasoning method in the back end to reason about the user input content. Finally, the output result layer is responsible for presenting the question-and-answer results returned by the inference layer to the user through a graphical interface or text. For example, the output content displayed will visualize the most likely answers in the form of a knowledge graph.

[0099] Specifically, refer to Figure 8 The figure shows a schematic diagram of an abnormal reasoning process provided in an embodiment of the present application. By selecting the reasoning type as KG-BERT and inputting the abnormal phenomenon "partial ice blockage, resulting in low-load production of the device" and the reasoning content "the cause is" respectively, after clicking the reasoning, 10 groups of answers will appear after a while, and the answers will be sorted according to the scores.

[0100] For the convenience of description, the above parts are divided into each unit module (or module) according to function and described separately. Of course, when implementing this application, the functions of each unit (or module) can be implemented in the same one or more software or hardware. Those skilled in the art will understand that various aspects of this application can be implemented as a system, method or program product. Therefore, various aspects of this application can be specifically implemented in the following forms, namely: a complete hardware implementation method, a complete software implementation method (including firmware, microcode, etc.), or an implementation method combining hardware and software, which can be collectively referred to as "circuit", "module" or "system" here.

[0101] In a possible implementation, the present application also provides a method for constructing an ethylene plant abnormality handling system, and the specific implementation steps of the method are as follows:

[0102] Step 901: Obtain demand analysis information of the ethylene plant abnormality handling system, where the demand analysis information includes user demand information and system demand information.

[0103] In an embodiment of the present application, user demand information of the ethylene device abnormality handling system may include: the ability to raise questions related to device operation abnormalities through abnormal phenomena in the operation of the ethylene device and obtain corresponding reasoning results; system demand information may include: system processing speed, stability, reliability and scalability, etc.

[0104] Specifically, users expect the system to understand the relationship between device anomalies and solutions, quickly return reliable query results, and enable interaction and result display through an intuitive interface. The user-side goal is to implement three functions: query input, inference algorithm selection, and visual display of inference results. Due to the large amount of data in the knowledge graph, users require selectable and searchable entities and relationships, allowing them to visually construct inference statements. Similarly, users can independently select the inference algorithm on the page. Visualization of inference results includes graph visualization and a table display of the top K most reliable inference entities. To meet system requirements, this system needs to utilize pre-trained models and call inference interfaces to improve the speed of geographic logic queries. Regarding stability, support for input validation and data verification is required, ensuring the reliability and validity of user input through front-end and back-end rule configuration. Reliability requires high usability and accessibility, and stable operation across multiple devices. Regarding scalability, the system needs to be able to efficiently manage and maintain the massive knowledge graph database and be extensible to support future knowledge updates and feature enhancements.

[0105] Step 902: Based on the demand analysis information, a system development environment is constructed, and based on the system development environment, a system architecture of the ethylene plant abnormality handling system is constructed.

[0106] In the embodiment of the present application, the system architecture of the ethylene plant abnormality handling system includes a data layer, an application service layer and a user layer.

[0107] In one possible implementation, the present application can create a system development environment based on a browser-server (BS) architecture, including front-end interface design, back-end logic processing, and data storage and management.

[0108] Specifically, the front-end interface design in this application can use the Vue2 progressive front-end development framework to build the user interface, which is not only ready to use and easy to develop, but also can provide a rich interactive experience. And through the desktop component library based on Vue2.0 such as Element, a rich UI component is provided to quickly build a beautiful and simple front-end page. As a powerful JavaScript library, D3.js can be used to create dynamic and interactive data visualizations on web pages. After obtaining data from Neo4j, this application can render graphics through D3.js and add interactions and styles to nodes and relationships, so that the front-end page can display a complete knowledge graph. It is worth mentioning that the above software tools are only exemplary. This application can select other suitable software tools according to actual needs, and this application does not make specific limitations on this.

[0109] Specifically, the backend logic processing for this application can be done in Python. Python has extensive library support for data processing and machine learning, making it one of the most widely used languages ​​in the field of artificial intelligence. Python can be used to build and deploy pre-trained deep learning models, and to parse and process geographic knowledge graph query and answer tasks. Data storage and management for this application can be done in the Neo4j graph database. Neo4j, currently the most popular high-performance NoSQL database, is used to store structured data on the network and uses the Cypher language for efficient querying.

[0110] Specifically, refer to Figure 9 The following is a schematic diagram of the system architecture of an ethylene plant exception handling system provided by an embodiment of the present application. The data layer is used for data storage and response for the ethylene plant exception handling system. It uses the Neo4j graph database to store the ethylene plant's knowledge graph data, supporting basic knowledge graph queries. It also employs distance-based, language-based, and graph neural network-based algorithms to support device exception reasoning and query functions. The application service layer parses query statements returned by the front end and implements logical queries for device operation exceptions. For general user-entered knowledge graph queries, the front end retrieves the query statement, parses it into a Cypher query statement, and retrieves the user's query content from the Neo4j database. For user-entered knowledge graph operation exception queries, the front end retrieves the query statement, parses the operation exception logical query statement, and uses pre-trained GCN, KG-Bert, TransE, and other models to reason about the user's query request, generating inference results and returning them to the user. The user layer is a web-based front-end interface built using the Vue2 front-end framework. It allows users to enter inference query statements and displays graph visualizations and attribute tables of inference results.

[0111] Step 903: Based on the preset modular development strategy, construct the functional modules of the ethylene plant abnormality handling system in sequence.

[0112] In the embodiment of the present application, the functional design of the ethylene unit abnormality handling system needs to revolve around core requirements, including algorithm selection, query parsing, knowledge graph matching, logical query and result display, etc., mainly including parsing the query input by the user by setting specified input rules and commands, converting it into corresponding reasoning statements and query statements, and searching for relevant information in the knowledge graph through the knowledge graph query engine, executing logical queries in the knowledge graph through the reasoning algorithm, and displaying the query results and reasoning paths through the user interface.

[0113] In a possible implementation, the present application may adopt a modular development strategy to sequentially design the contents of the front-end page design and back-end reasoning algorithm modules, thereby sequentially constructing the functional modules of the ethylene plant abnormality handling system.

[0114] Specifically, the user interface ensures the interactivity of the query process and the intuitiveness of the results. The server receives inference instructions from the front-end user, parses and converts them into corresponding queries or geographic inference tasks, calls the inference algorithm to generate the abnormal inference results and the top K entities with high credibility, and displays the graph and attributes through the front-end. When a user accesses the front-end page of the geographic knowledge graph inference system, the system loads the relevant CSS style sheet, JavaScript script, and graphics visualization library. The front-end page provides interactive functions such as user input and inference algorithm selection. The user interface mainly consists of two areas: the input area and the inference result display area.

[0115] Specifically, users can enter the triple information they wish to infer (head entity, relationship, and tail entity) through the input boxes in the front-end interface. Users can also select different inference algorithms and whether to enable feature aggregation. The inference algorithm selection box includes GCN, KG-Bert, and TransE, and users can select the corresponding inference algorithm through the drop-down box. The Generate Inference Triples area consists of three search selection boxes: head entity search box, relationship edge search box, and tail entity search box. Due to the large number of entities in the knowledge graph, the system also supports users to manually select entities through the drop-down box and search for entities through the search box. After selecting the head entity, relationship edge, and tail entity, click Generate Edge to preview the generated inference triples. Clicking Generate Inference Path adds the generated triples to the Inference Content input box, and can add multiple triples. The Inference Content input box supports manual addition and modification of rules. After the user has added the content to be inferred, click Start Inference to transmit the inference content to the backend, which then returns the inference results, which are displayed in the Inference Result Display Area.

[0116] Specifically, the inference result display area consists of two sections: the upper section displays the visualized knowledge graph and inference path, while the lower section displays the attribute table of the inference results. The inference results returned by the backend are sent to the frontend via HTTP responses. The D3.js library is used to visualize the query results, displaying the geographic logical query path of the knowledge graph and listing the attribute values ​​of the relevant query results in the attribute table. Furthermore, the system supports users adjusting the input content to perform multiple queries using different algorithms.

[0117] In one possible implementation, during the design of the back-end reasoning algorithm of the present application, after the user clicks the "Start Reasoning" button, the front-end can listen to the button click event through JavaScript. The user's input data is obtained and sent to the application layer on the server side through an Ajax request. The application layer on the server side receives the reasoning query and the user's setting options from the front-end. The application layer parses the data containing the user input into specific query commands, which are divided into basic queries of the knowledge graph and geographic logical queries. For basic queries of the knowledge graph such as a single triple, the system will generate Cypher statements, and the application layer will send query commands to the data layer. The data layer will use the Neo4j graph database to execute these commands, retrieve data, and generate query results. For reasoning of knowledge graphs such as multiple triples, the corresponding pre-trained model is called to perform logical query processing. The data layer returns the reasoning results to the application layer, and the application layer organizes and formats the results.

[0118] In a possible embodiment, the ethylene unit abnormality handling system in the present application also includes a general question-and-answer module and an operational abnormality question-and-answer module. After the server receives the user's query request, it can parse the query statement into a Cypher language form for general query questions and answers and an inference language form for operational abnormality query questions and answers according to different query rules. General query questions and answers refer to questions and answers that do not involve operational abnormality logical reasoning in knowledge graph query questions and answers, and only perform simple retrieval. General query questions and answers use Neo4j as a database, and use Cypher commands to retrieve the corresponding answers from the database for output. The operational abnormality question-and-answer module means that users can use pre-trained models for reasoning. When users need to perform geographic logical queries, they can select the corresponding algorithm and generate an inference path through inference triples. After the server obtains the inference requirements, it calls the pre-trained model to generate the inference results and attribute table.

[0119] See Figure 10 As shown, based on the same technical concept, the embodiment of the present application also provides a computer device 100. In one embodiment, the computer device can be a device dedicated to the system processing of the ethylene device, or a control device for the overall chemical production. Figure 10As shown, it includes a memory 1001 and one or more processors 1002.

[0120] Memory 1001 is used to store the knowledge graph of the ethylene plant and the computer program executed by processor 1002. Memory 1001 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required for running instant messaging functions, while the data storage area may store various instant messaging messages and operating instruction sets.

[0121] Memory 1001 may be a volatile memory, such as random-access memory (RAM); a non-volatile memory, such as read-only memory, flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 1001 may be a combination of the aforementioned memories.

[0122] The processor 1002 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 1002 is configured to implement the steps of the above-mentioned ethylene plant abnormality handling system when calling the computer program stored in the memory 1001 .

[0123] The specific connection medium between the memory 1001 and the processor 1002 is not limited in the embodiment of the present application. Figure 10 In the embodiment, the memory 1001 and the processor 1002 are connected via a bus 1003. The bus 1003 is connected to the processor 1002 via a bus 1003. Figure 10 The connections between the other components are shown in bold lines, which are only for illustration and are not intended to be limiting. The bus 1003 can be divided into an address bus, a data bus, a control bus, etc. For ease of description, Figure 10 The diagram shows a single thick line, but this does not indicate that there is only one bus or one type of bus.

Claims

1. An ethylene plant abnormality handling system based on knowledge graph, characterized in that: The system comprises: An embedding learning module is configured to perform representation learning on a knowledge graph of an ethylene plant based on a plurality of preset embedding strategies to obtain a vector representation corresponding to the knowledge graph; the knowledge graph includes multiple entities of the ethylene plant and relationships between the entities, and each entity corresponds to a device, a sensor, and control logic of the ethylene plant; the vector representation represents a feature vector of each entity and relationship in the knowledge graph; The reasoning and prediction module is used to perform abnormal reasoning based on the vector representation to obtain abnormal prediction results of the ethylene device and corresponding solutions.

2. The system according to claim 1, wherein The system further includes a graph construction module for: Acquiring device data of the ethylene device; Based on the device data, the knowledge graph is constructed.

3. The system according to claim 1, wherein: The system further comprises: The interactive question-answering module is used to respond to the ethylene plant questions input by the target object and output corresponding question-answering results based on the knowledge graph.

4. The system according to claim 2, wherein: The graph construction module is specifically used to: Based on the device data, an ontology file corresponding to the knowledge graph is determined, wherein the ontology file includes custom classes, relationship attributes, and data attributes; the classes represent the physical entities, operation logic, and abnormal phenomena involved in the operation of the ethylene device, the relationship attributes represent the association between the classes, and the data attributes represent the attribute information corresponding to each class; The unstructured data in the device data is structured to obtain triple data of the knowledge graph, and the knowledge graph is visualized.

5. The system according to claim 2, wherein: The multiple embedding strategies include geometric model embedding strategy, language model embedding strategy and graph neural network embedding strategy; wherein, The geometric model embedding strategy representation: representing the triple data of the knowledge graph through a geometric model; The language model embedding strategy representation: Based on natural language processing technology, combined with a pre-trained language model, the text information in the device data is embedded to obtain the corresponding semantic representation; The graph neural network embedding strategy is characterized by: processing the entities and association relationships of the knowledge graph based on the graph neural network to determine the high-order relationships between the entities in the knowledge graph; and generating vector representations of each entity through the graph convolutional network.

6. The system according to claim 1, wherein: The inference prediction module is specifically used to: Based on a preset anomaly detection training set and the vector representation, performing anomaly reasoning on the ethylene plant; the anomaly detection training set includes ethylene plant data labeled with normal and abnormal states; When an abnormal situation occurs in the ethylene device, the abnormality prediction result and the solution are output based on the knowledge graph.

7. The system according to claim 5, wherein: The inference prediction module is further used to: When the knowledge graph does not include the solution, generating a solution corresponding to the abnormal prediction result based on a preset decision algorithm; Based on the current operating status of the ethylene plant, personalized recommendations are made for the solution.

8. The system according to claim 1, wherein: The device data at least includes equipment information, sensor data, process parameters and control loop information of the ethylene device.

9. The system according to claim 3, wherein: The interactive question-answering module is specifically used to: Extract keywords from the ethylene plant problem and determine corresponding abnormality description information; Based on the knowledge graph and the inference prediction module, the question and answer results corresponding to the abnormal description information are obtained and fed back to the target object; the question and answer results include the corresponding abnormal prediction results and solutions.

10. The system according to claim 9, wherein The system further includes a visualization module for displaying the question-and-answer results and the equipment status of the ethylene plant to the target object based on a visualization interface.

11. The system according to claim 1, wherein: The method for constructing the system includes: Obtaining demand analysis information of the ethylene device abnormality handling system, wherein the demand analysis information includes user demand information and system demand information; Based on the demand analysis information, a system development environment is constructed, and based on the system development environment, a system architecture of the ethylene device abnormality handling system is constructed; the system architecture includes a data layer, an application service layer, and a user layer; Based on a preset modular development strategy, the functional modules of the ethylene unit abnormality handling system are constructed in sequence.

12. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The memory is used to store the knowledge graph of the ethylene device; When the processor executes the computer program, the steps of the system according to any one of claims 1 to 11 are implemented.

Images