Metallurgical informatization production line operation and maintenance management system based on dynamic knowledge graph

Abstract

The application discloses a metallurgical informatization production line operation and maintenance management system based on a dynamic knowledge graph, belongs to the technical field of metallurgical production operation and maintenance management, and specifically comprises a management cloud platform, a metallurgical subsequence processing unit, a quality evaluation unit and a quality prediction optimization unit. The application aims to solve the problems of hysteresis and process isolation in traditional metallurgical production line quality control. The core lies in constructing a full-process dynamic knowledge graph serving the metallurgical production line, creating a "quality evaluation model" and a "quality prediction model" based on the knowledge graph, constituting a "two-stage, cross-process" continuous process intelligent quality control system, realizing "prediction and waste prevention in advance" based on virtual verification and optimization of pre-production parameters, realizing "production compensation after the event" based on cross-process quality prediction and optimization adjustment, and realizing accurate perception, causal reasoning and collaborative decision-making of the whole-process quality through the dynamic knowledge graph.

Description

Technical Field

[0001] This invention relates to the field of metallurgical production operation and maintenance management technology, and more specifically, to a metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs. Background Technology

[0002] Metallurgical production is a complex system with a long process and many steps. From raw material pretreatment, smelting, refining to finished product processing, changes in parameters at each node can affect each other. Currently, mainstream metallurgical enterprises have achieved automated monitoring and basic control of key process quality indicators within a single process.

[0003] Current quality assessment standards are often based on fixed thresholds within a single process, evaluating the quality of semi-finished or finished products after a single process to determine whether they are qualified and to eliminate unqualified products. However, for slightly unqualified products, there is no proactive adjustment of downstream process parameters to "compensate" for the deficiencies in their initial state. In other words, it is impossible to comprehensively predict the final quality of downstream output based on the semi-finished product status of upstream processes and the production parameters of downstream processes, and it is also impossible to provide early warning and intervention before quality problems occur. This leads to the problems of lag and process isolation in the existing metallurgical quality control.

[0004] Therefore, in response to practical metallurgical production problems, a metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs is proposed. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of lag and process isolation in traditional metallurgical quality control. It provides a metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graph.

[0006] The objective of this invention can be achieved through the following technical solution: a metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs, including a management cloud platform, a metallurgical process segmentation unit, a quality assessment unit, and a quality prediction and optimization unit. Within the management cloud platform, a dynamic knowledge graph serving the entire metallurgical production line is constructed, integrating quality assessment and quality prediction models. The metallurgical process segmentation unit is used to divide the metallurgical production line into sequential upstream and downstream processes, setting original production parameters for each process and sending them to the quality prediction and optimization unit.

[0007] The quality prediction and optimization unit acquires the quality prediction model, optimizes the original production parameters of each process based on the model prediction, generates actual production parameters and sends them to the metallurgical sequence processing unit to obtain the upstream output based on the upstream process and the downstream output based on the downstream process.

[0008] The quality assessment unit is used to collect the actual status information of upstream products, input it into the acquired quality assessment model, judge the quality of upstream products, divide them into qualified upstream products and products to be optimized upstream products, send qualified upstream products to the downstream process of the metallurgical sequence processing unit, and send products to be optimized upstream products to the quality prediction and optimization unit.

[0009] The quality prediction and optimization unit obtains the actual production parameters of the downstream process and inputs them together with the actual status information of the upstream product to be optimized into the obtained quality prediction model to obtain the future status information of the downstream product. Based on the future status information, it determines whether the predicted status of the downstream product meets the standard and decides whether to further optimize the actual production parameters of the downstream process, generates downstream optimization parameters, and sends them to the downstream process of the metallurgical sequence processing unit for continuous metallurgical production.

[0010] Furthermore, the full-process dynamic knowledge graph is used to store the entity nodes (upstream status nodes, downstream quality nodes), relationships, standard index ranges, and rule bases of multiple processes in the metallurgical production line. The metallurgical sequential processing unit includes iron pretreatment module, steelmaking processing module, steel rolling processing module, and finishing processing module. Each processing module corresponds to a process and sets the corresponding original production parameters for the process content of each process.

[0011] Furthermore, the quality assessment model is a machine learning classification model trained based on historical production data and a dynamic knowledge graph of the entire process. Its input is the actual status information of the upstream products, and its output is the quality judgment result. The judgment rule is: when all status parameters in the actual status information are within the standard index range of the knowledge graph, it is judged as an upstream qualified product; otherwise, it is judged as an upstream product that needs to be optimized.

[0012] The quality prediction model is a time-series prediction model trained based on historical production data and a dynamic knowledge graph of the entire process. It adopts an architecture that combines time-series neural networks and graph neural networks to predict the output status of future processes in advance and across processes. When making advance predictions, its input is the original production parameters set for each process, and the output is the predicted output status information. When making cross-process predictions, its input is the actual status information of the upstream product to be optimized and the production parameters of the downstream process, and the output is the future status information of the downstream product.

[0013] Furthermore, the process of optimizing the original production parameters for each process includes:

[0014] The quality prediction model for each process is obtained. The original production parameters of each process are input into the quality prediction model of the corresponding process for graph matching to obtain the output status prediction information of upstream products and the output status prediction information of downstream products.

[0015] The parameters of the output status prediction information of each process are compared with the parameters of the standard status information corresponding to the rule base. When all parameters are within the standard index range, the standard is determined to be met, and the original production parameters are used as the actual production parameters for metallurgical production. Otherwise, the standard is determined to be unmet, and parameter optimization signals are generated. Based on the unmet index and the degree of deviation of the index comparison, the compensation rules associated with the unmet index stored in the knowledge graph are queried. Based on the compensation rules, the direction and magnitude of parameter adjustment are determined, the original production parameters are adjusted and optimized, and the generated optimized production parameters are used as the actual production parameters for metallurgical production.

[0016] Furthermore, the process of quality judgment for upstream products is as follows: The actual status information of upstream products is input into the quality assessment model. After receiving the actual status information of upstream products, the quality assessment model classifies upstream products as qualified upstream products when all status parameters in the actual status information are within the standard index range, and otherwise classifies them as upstream products to be optimized.

[0017] Furthermore, the process of determining whether the predicted state of downstream output meets the standard based on future state information analysis includes: obtaining the future state information of downstream output; if all state parameters of the future state information are within the standard index range of the corresponding standard state information, then the future state information is determined to match the standard state information of downstream output and meet the standard. The actual production parameters are then directly used as downstream optimization parameters for metallurgical production in downstream processes. Otherwise, it is determined to be substandard, and the actual production parameters of downstream processes are further optimized to generate downstream optimization parameters.

[0018] Furthermore, the process of further optimizing the actual production parameters of downstream processes includes:

[0019] When a product is deemed non-compliant, the upstream state node of the product to be optimized is located in the graph using a quality prediction model. Along the relationship edges in the graph, the downstream quality nodes of the downstream processes affected by the upstream state node are found. Based on the downstream quality nodes of the affected downstream processes, the compensation strategies of the known downstream processes in the dynamic knowledge graph of the entire process are automatically associated. The downstream optimization parameters are obtained by optimizing the actual production parameters based on the compensation strategies.

[0020] This invention also proposes a method for operation and maintenance management of metallurgical information production lines based on dynamic knowledge graphs, including the following steps:

[0021] Step 1: Divide the metallurgical production line into upstream and downstream processes that are connected one after another, and set the original production parameters for each process;

[0022] Step 2: Based on the quality prediction model, optimize the original production parameters of each process to generate actual production parameters;

[0023] Step 3: Collect the actual status information of upstream products, input the actual status information into the quality assessment model, judge the quality of upstream products, and classify upstream qualified products and upstream products that need to be optimized.

[0024] Step 4: When the product is classified as an upstream qualified product, it will be processed in the downstream process.

[0025] When an upstream product is classified as an output to be optimized, the actual status information of the upstream product to be optimized and the actual production parameters of the downstream process are input into the quality prediction model to obtain the future status information of the downstream product. It is then determined whether the predicted status of the downstream product meets the standard. If it does, the actual production parameters are directly used as the downstream optimization parameters. If it does not meet the standard, the actual production parameters of the downstream process are further optimized to generate downstream optimization parameters. The upstream product to be optimized is then produced in the downstream process based on the downstream optimization parameters.

[0026] Compared with the prior art, the advantages of this invention are:

[0027] This invention aims to solve the problems of lag and process isolation in traditional metallurgical production line quality control. The core of the invention is to construct a dynamic knowledge graph that serves the entire metallurgical production line. Based on the knowledge graph, a "quality assessment model" and a "quality prediction model" are created, forming a "two-stage, cross-process" continuous intelligent quality control system. The system achieves "pre-production prediction and waste prevention" based on virtual verification and optimization of pre-production parameters, and achieves "post-production compensation" based on cross-process quality prediction and optimization. Through the dynamic knowledge graph, the system achieves accurate perception, causal reasoning, and collaborative decision-making of the entire process quality.

[0028] Based on a dual-model evaluation and verification framework, intelligent decision-making is achieved throughout the entire process, from parameter optimization to production compensation. Specifically, before actual production begins, the original production parameters of each process are input into the quality prediction model for virtual production simulation. The original production parameters of each process are optimized to eliminate the generation of defective products at the source. During actual production, the actual status information of the upstream products to be optimized and the actual production parameters of the downstream processes are input into the quality prediction model to predict the future status of the downstream products. By iteratively optimizing the production parameters of the downstream processes, customized downstream optimization parameters are generated to compensate for the upstream products to be optimized, so that they ultimately meet the quality requirements. Attached Figure Description

[0029] Figure 1 This is a system principle block diagram of the present invention;

[0030] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0032] Example 1: This invention discloses a metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs. Please refer to [link / reference]. Figure 1 , Figure 2 It includes a management cloud platform, a metallurgical sequence processing unit, a quality assessment unit, and a quality prediction and optimization unit. Within the operation and maintenance management cloud platform, a dynamic knowledge graph serving the entire metallurgical production line is constructed, integrating quality assessment models and quality prediction models.

[0033] Among them, the full-process dynamic knowledge graph is used to store the entity nodes (upstream status nodes, downstream quality nodes), relationships, standard indicator ranges, and rule bases of multiple processes in the metallurgical production line. Each process has different production parameters and status information types. Multimodal data is uniformly extracted and integrated into the knowledge graph. The power of the knowledge graph lies in revealing the implicit relationships across processes, continuously learning from operation and maintenance feedback and optimizing the knowledge graph, and realizing a mechanism for the self-evolution of decision knowledge.

[0034] The metallurgical sequential processing unit includes an iron pretreatment module, a steelmaking processing module, a steel rolling processing module, and a finishing processing module. Each processing module corresponds to a process and sets corresponding original production parameters for the process content of each process.

[0035] The quality assessment model is a machine learning classification model trained on historical production data and a dynamic knowledge graph of the entire process. It prioritizes the use of the random forest classification algorithm to diagnose and classify actual state information. Its input is the actual state information of the upstream products, and its output is the quality judgment result. The judgment rule is: when all state parameters in the actual state information are within the standard index range of the knowledge graph, it is judged as an upstream qualified product; otherwise, it is judged as an upstream product that needs to be optimized. The quality assessment model is dynamically updated through the following mechanism: during the training process, the actual production parameters of different processes, the actual state information of the corresponding products, and the manual quality judgment results are used as training samples to continuously iterate and optimize in order to improve the accuracy of quality assessment of the products produced in each process.

[0036] The quality prediction model is a time-series prediction model trained based on historical production data and a dynamic knowledge graph of the entire process. It adopts an architecture combining temporal neural networks and graph neural networks to predict the output status of future processes in advance and across processes. When making advance predictions, the input is the original production parameters set for each process, and the output is the predicted output status information. When making cross-process predictions, the input is the actual status information of the upstream product to be optimized and the production parameters of the downstream process, and the output is the future status information of the downstream product. The specific explanation for obtaining the future status information is as follows: the input information is mapped to the entity feature vector in the knowledge graph, the affected downstream quality nodes are traversed along the causal relationship edges in the graph, and the historical compensation case data stored in the graph is combined to perform time-series deduction to obtain the future status information of the downstream product.

[0037] The quality prediction model is dynamically updated through the following mechanism: using historical production information of the metallurgical production line, actual output status information of each process, and production parameters of downstream processes as training samples, and after simulation and training, it can accurately predict the output quality of future processes.

[0038] Regular incremental training of the model is performed. Both the quality assessment model and the quality prediction model can use newly generated production data (such as quality inspection results and actual status information) for incremental training or regular retraining to adapt to changes in working conditions. New production modes and quality cases are integrated into the knowledge graph. The two models are constantly updated and work together through the whole process quality knowledge graph to achieve intelligent decision-making throughout the entire process, from parameter optimization to production compensation.

[0039] The metallurgical process division unit is used to divide the metallurgical production line into multiple processes, including upstream processes and downstream processes for the next process after the upstream processes. It sets the original production parameters (equipment parameters and process parameters) for each process and sends them to the quality prediction and optimization unit.

[0040] The quality prediction and optimization unit acquires the quality prediction model, optimizes the original production parameters of each process based on the model prediction, and generates actual production parameters which are then sent to the metallurgical sequence processing unit.

[0041] The process of optimizing the original production parameters of each process to generate actual production parameters includes:

[0042] The quality prediction model for each process is obtained. The original production parameters of each process are input into the quality prediction model of the corresponding process for graph matching to obtain the output status prediction information of upstream products and downstream products. The output status prediction information includes the index information corresponding to different processes, specifically including the output temperature field distribution, chemical composition, physical and mechanical properties, as well as appearance, size, and microstructure information.

[0043] The parameters of the output status prediction information of each process are compared with the parameters of the standard status information corresponding to the rule base. When all parameters are within the standard index range, the standard is determined to be met, and the original production parameters are used as the actual production parameters for metallurgical production. Otherwise, the standard is determined to be unmet, and parameter optimization signals are generated. Based on the unmet index and the degree of deviation of the index comparison, the compensation rules associated with the unmet index stored in the knowledge graph are queried. Based on the compensation rules, the direction and magnitude of parameter adjustment are determined, and the original production parameters are adjusted and optimized. The generated optimized production parameters are used as the actual production parameters for metallurgical production. The quality prediction model is continuously changed and optimized. The quality prediction model based on dynamic knowledge graph is used to achieve "pre-emptive prediction and waste prevention".

[0044] The metallurgical sequence processing unit receives actual production parameters for multi-process continuous metallurgical production. The metallurgical sequence processing unit obtains upstream outputs based on upstream processes and downstream outputs obtained by continuing production of upstream outputs in downstream processes.

[0045] The quality assessment unit is used to collect the actual status information of the upstream output obtained from the production based on the actual production parameters in the first upstream process, obtain the quality assessment model, input the actual status information of the upstream output into the quality assessment model, make quality judgment on the upstream output of the first process, divide the upstream output into qualified upstream output and upstream output to be optimized according to the judgment result, send the qualified upstream output directly to the downstream process of the metallurgical sequence processing unit to continue metallurgical production, and send the actual status information of the upstream output to be optimized to the quality prediction and optimization unit.

[0046] The process of quality assessment of upstream products:

[0047] The actual status information of the upstream products is input into the quality assessment model. After receiving the actual status information of the upstream products, the quality assessment model classifies the upstream products as qualified upstream products when all status parameters in the actual status information are within the standard index range, and otherwise classifies the upstream products as products to be optimized upstream products.

[0048] The quality prediction and optimization unit acquires the quality prediction model and the actual production parameters of the downstream process. It inputs the actual state information of the upstream product to be optimized and the actual production parameters of the downstream process into the quality prediction model to obtain the future state information of the downstream product. It compares the future state information with the standard state information of the downstream product to determine whether the predicted state of the downstream product meets the standard. If it does not meet the standard, it further optimizes the actual production parameters of the downstream process through graph matching to generate downstream optimization parameters. It then sends the upstream product to be optimized and the downstream optimization parameters to the next process of the metallurgical sequence processing unit to continue metallurgical production. This process is repeated to carry out "dual-stage, cross-process" continuous intelligent monitoring until the product produced by the last process meets the qualification standard.

[0049] The process of determining whether the downstream output forecast status meets the target based on future status information includes:

[0050] If the future state information of the downstream product is obtained, and all state parameters of the future state information are within the standard index range of the corresponding standard state information, then the future state information is determined to match the standard state information of the downstream product. The actual production parameters are then directly used as the downstream optimization parameters for the metallurgical production of the downstream process. Otherwise, it is determined to be substandard, and the actual production parameters of the downstream process are further optimized to generate downstream optimization parameters.

[0051] The process of further optimizing the actual production parameters of downstream processes includes:

[0052] When a product is deemed non-compliant, the upstream status node of the product to be optimized is located in the graph using the quality prediction model. Along the relationship edges in the graph, the downstream quality nodes of the downstream processes affected by the upstream status node are found. Based on the downstream quality nodes of the affected downstream processes, the compensation strategies of the known downstream processes in the dynamic knowledge graph of the entire process are automatically associated.

[0053] For example, taking the first process (pre-treatment module for ironmaking) as the upstream process, the upstream output is liquid molten iron. Its actual state information includes the chemical composition and physical temperature of the molten iron. The actual state information of the upstream output (chemical composition and physical temperature of molten iron) and the actual production parameters of the downstream process (converter production parameters) are substituted into the quality prediction model to conduct the first virtual steelmaking and output the future state information. When it is identified that the "physical temperature of molten iron" exceeds the standard, it is predicted that if production is carried out according to the original plan, "over-oxidized steel" will be obtained, which has an excessive oxygen content and the mechanical properties of the steel will not meet the standards. At the same time, it will also predict that the equipment wear (furnace lining erosion) will be aggravated. Since the prediction results are not up to standard, the parameter optimization engine is activated, and the graph will automatically be associated with the compensation strategies of the known downstream processes in the knowledge graph.

[0054] For example, taking the process corresponding to the steelmaking module as the upstream process, the upstream state node of the product to be optimized in the graph is located by the quality prediction model. For example, the upstream state node "continuous casting billet - core surface temperature difference - too large" is used as the upstream state node of the steelmaking module. Along the relationship edge in the graph, the downstream quality node of the downstream process affected by the upstream state node is found. For example, the downstream quality node "affects - hot rolled plate - microstructure uniformity" is used as the downstream quality node of the rolling module. Based on the downstream quality node of the affected downstream process, the compensation strategy of the known downstream process in the dynamic knowledge graph of the whole process is automatically associated. Based on the compensation strategy, the actual production parameters are optimized to obtain the downstream optimization parameters.

[0055] In actual production, the actual status information of the upstream products to be optimized and the actual production parameters of the downstream processes are input into the quality prediction model to predict the future status of the downstream products. By iteratively optimizing the production parameters of the downstream processes, customized downstream optimization parameters are generated to compensate the upstream products to be optimized, so that they ultimately meet the quality requirements.

[0056] Example 2: This invention also proposes a method for the operation and maintenance management of metallurgical information-based production lines based on dynamic knowledge graphs. Please refer to [link / reference]. Figure 2 It includes the following steps:

[0057] Step 1: Divide the metallurgical production line into upstream and downstream processes that are connected one after another, and set the original production parameters for each process;

[0058] Step 2: Based on the quality prediction model, obtain the output status prediction information of each process, and conduct quality prediction evaluation and judgment. When the judgment fails to meet the standard, optimize the original production parameters of each process to generate actual production parameters.

[0059] Step 3: Collect the actual status information of upstream products obtained from production based on actual production parameters in the upstream process, input the actual status information into the quality assessment model, judge the quality of upstream products, and classify upstream qualified products and upstream products that need to be optimized.

[0060] Step 4: When the product is classified as an upstream qualified product, it will be directly processed in the downstream process.

[0061] When an upstream product is classified as an output to be optimized, the actual status information of the upstream product to be optimized and the actual production parameters of the downstream process are input into the quality prediction model to obtain the future status information of the downstream product. It is then determined whether the predicted status of the downstream product meets the standard. If it does, the actual production parameters are directly used as the downstream optimization parameters. If it does not meet the standard, the actual production parameters of the downstream process are further optimized to generate downstream optimization parameters. The upstream product to be optimized is then produced in the downstream process based on the downstream optimization parameters.

[0062] In summary, this invention aims to solve the problems of lag and process isolation in traditional metallurgical production line quality control. The core of this invention is to construct a dynamic knowledge graph that serves the entire metallurgical production line and create a "quality assessment model" and a "quality prediction model" based on the knowledge graph. This constitutes a "two-stage, cross-process" continuous intelligent quality control system, enabling intelligent decision-making throughout the entire process, from parameter optimization to production compensation.

[0063] "Pre-production prediction and waste prevention" is achieved through virtual verification and optimization of pre-production parameters, and "post-production compensation" is achieved through cross-process quality prediction, optimization and adjustment. Accurate perception, causal reasoning and collaborative decision-making of the entire process quality are realized through dynamic knowledge graphs.

[0064] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto; any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and its improved concept, should be covered within the scope of protection of the present invention.

Claims

1. A metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs, characterized by: It includes a management cloud platform, a metallurgical sequence processing unit, a quality assessment unit, and a quality prediction and optimization unit. Within the management cloud platform, a dynamic knowledge graph for the entire metallurgical production line is constructed. The dynamic knowledge graph is used to store entity nodes, relationships, standard indicator ranges, and rule bases for multiple processes in the metallurgical production line. Entity nodes include upstream status nodes and downstream quality nodes, and integrates quality assessment models and quality prediction models. The quality assessment model is a machine learning classification model trained on historical production data and a dynamic knowledge graph of the entire process. Its input is the actual status information of the upstream products and the output is the quality judgment result. The judgment rule is: when all the status parameters in the actual status information are within the standard index range of the knowledge graph, it is judged as an upstream qualified product; otherwise, it is judged as an upstream product that needs to be optimized. The quality prediction model is a time-series prediction model trained based on historical production data and a dynamic knowledge graph of the entire process. It adopts an architecture that combines time-series neural networks and graph neural networks to predict the output status of future processes in advance and across processes. When making advance predictions, its input is the original production parameters set for each process, and the output is the predicted output status information. When making cross-process predictions, its input is the actual status information of the upstream product to be optimized and the production parameters of the downstream process, and the output is the future status information of the downstream product. The metallurgical sequence processing unit is used to divide the metallurgical production line into upstream and downstream processes that are connected one after another, and to set the original production parameters for each process and send them to the quality prediction and optimization unit. The quality prediction and optimization unit acquires the quality prediction model, optimizes the original production parameters of each process based on the model prediction, generates actual production parameters and sends them to the metallurgical sequence processing unit to obtain the upstream output based on the upstream process and the downstream output based on the downstream process. The quality assessment unit is used to collect the actual status information of upstream products, input it into the acquired quality assessment model, judge the quality of upstream products, divide them into qualified upstream products and products to be optimized upstream products, send qualified upstream products to the downstream process of the metallurgical sequence processing unit, and send products to be optimized upstream products to the quality prediction and optimization unit. The quality prediction and optimization unit obtains the actual production parameters of the downstream process and inputs them together with the actual status information of the upstream product to be optimized into the obtained quality prediction model to obtain the future status information of the downstream product. Based on the future status information, it determines whether the predicted status of the downstream product meets the standard and decides whether to further optimize the actual production parameters of the downstream process. It then generates downstream optimization parameters and sends them to the downstream process of the metallurgical sequence processing unit.

2. The metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs according to claim 1, characterized in that: The metallurgical sequential processing unit includes an iron pretreatment module, a steelmaking processing module, a steel rolling processing module, and a finishing processing module. Each processing module corresponds to a process and sets corresponding original production parameters for the process content of each process.

3. The metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs according to claim 2, characterized in that: The process of optimizing the original production parameters for each process includes: The quality prediction model for each process is obtained. The original production parameters of each process are input into the quality prediction model of the corresponding process for graph matching to obtain the output status prediction information of upstream products and the output status prediction information of downstream products. The parameters of the output status prediction information of each process are compared with the parameters of the standard status information corresponding to the rule base. When all parameters are within the standard index range, the standard is determined to be met, and the original production parameters are used as the actual production parameters for metallurgical production. Otherwise, the standard is determined to be unmet, and parameter optimization signals are generated. Based on the unmet index and the degree of deviation of the index comparison, the compensation rules associated with the unmet index stored in the knowledge graph are queried. Based on the compensation rules, the direction and magnitude of parameter adjustment are determined, the original production parameters are adjusted and optimized, and the generated optimized production parameters are used as the actual production parameters for metallurgical production.

4. The metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs according to claim 3, characterized in that: The process of quality assessment of upstream products is as follows: The actual status information of upstream products is input into the quality assessment model. After receiving the actual status information of upstream products, the quality assessment model classifies upstream products as qualified upstream products when all status parameters in the actual status information are within the standard index range, and otherwise classifies them as upstream products to be optimized.

5. The metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs according to claim 4, characterized in that: The process of determining whether the predicted state of downstream output meets the standard based on future state information analysis includes: obtaining the future state information of downstream output; if all state parameters of the future state information are within the standard index range of the corresponding standard state information, then the future state information is determined to match the standard state information of downstream output and meets the standard. The actual production parameters are then directly used as downstream optimization parameters for the metallurgical production of downstream processes. Otherwise, it is determined to be non-compliant, and the actual production parameters of downstream processes are further optimized to generate downstream optimization parameters.

6. The metallurgical information-based production line operation and maintenance management system based on dynamic knowledge graphs according to claim 5, characterized in that: The process of further optimizing the actual production parameters of downstream processes includes: When a product is deemed non-compliant, the upstream state node of the product to be optimized is located in the graph using a quality prediction model. Along the relationship edges in the graph, the downstream quality nodes of the downstream processes affected by the upstream state node are found. Based on the downstream quality nodes of the affected downstream processes, the compensation strategies of the known downstream processes in the dynamic knowledge graph of the entire process are automatically associated. The downstream optimization parameters are obtained by optimizing the actual production parameters based on the compensation strategies.

7. The metallurgical information-based production line operation and maintenance management method based on dynamic knowledge graphs according to any one of claims 1-6, characterized in that, Includes the following steps: Step 1: Divide the metallurgical production line into upstream and downstream processes that are connected one after another, and set the original production parameters for each process; Step 2: Based on the quality prediction model, optimize the original production parameters of each process to generate actual production parameters; Step 3: Collect the actual status information of upstream products, input it into the quality assessment model, judge the quality of upstream products, and classify upstream qualified products and upstream products that need to be optimized. Step 4: Execute qualified upstream products into downstream processes. Input the actual status information of the upstream product to be optimized and the actual production parameters of the downstream process into the quality prediction model to obtain the future status information of the downstream product. Determine whether the predicted status of the downstream product meets the standard. If it does, use the actual production parameters directly as the downstream optimization parameters. If it does not meet the standard, further optimize the actual production parameters of the downstream process to generate downstream optimization parameters and execute the downstream process.

Images