Steel performance prediction method and device, electronic equipment and storage medium

Abstract

The invention provides a steel performance prediction method and device, electronic equipment and a storage medium, and relates to the field of industrial digitization. Comprising the steps that production process information of target steel is obtained, and the production process information comprises set parameters of production equipment of the target steel in the production process; the production process information of the target steel is input into a trained performance prediction model, performance prediction is carried out through the trained performance prediction model, a performance prediction value of the target steel is obtained, the trained performance prediction model is a model trained based on a target knowledge graph, and the performance prediction value of the target steel is obtained. The target knowledge graph comprises historical production process information and historical performance values, the historical production process information and the historical performance values are in one-to-one correspondence, and the performance prediction value of the target steel comes from the historical performance values.

Description

Technical Field

[0001] The present application relates to the field of industrial digitalization, and in particular to a method, device, electronic device and storage medium for predicting properties of steel. Background Art

[0002] In the steel industry, performance prediction of steel materials to be produced (such as cold-rolled steel and / or hot-rolled steel) is a crucial part, however, this prediction process faces many challenges and difficulties. The primary difficulty is that the production process of steel materials is extremely complex. For example, the production process of cold-rolled steel spans multiple key links such as ironmaking, steelmaking, and hot rolling. The process parameters in each link, such as temperature, pressure, speed, etc., have a profound impact on the performance of the final product. Therefore, in order to achieve accurate performance prediction, it is necessary to fully and accurately master the process parameters of these production links, which is undoubtedly a huge and meticulous task.

[0003] In the related art, the performance of steel is usually predicted through the algorithm in the deep learning model, and the simulation data calculated by the algorithm is used as the prediction result, which has the problem of low accuracy. Summary of the invention

[0004] The present application provides a steel performance prediction method, device, electronic device and storage medium, which are used to solve the problem of low accuracy in the prior art of predicting the performance of steel through algorithms in deep learning models and using simulated data calculated by the algorithms as prediction results.

[0005] In a first aspect, the present application provides a method for predicting properties of steel, comprising:

[0006] Acquiring production process information of the target steel, wherein the production process information includes setting parameters of production equipment during the production process of the target steel;

[0007] The production process information of the target steel is input into a trained performance prediction model, and performance prediction is performed through the trained performance prediction model to obtain a performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, and the target knowledge graph includes historical production process information and historical performance values, and the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values.

[0008] In a second aspect, the present application also provides a steel performance prediction device, comprising:

[0009] An acquisition module, for acquiring production process information of target steel, wherein the production process information includes setting parameters of production equipment during the production process of the target steel;

[0010] A processing module is provided for inputting the production process information of the target steel into a trained performance prediction model, performing performance prediction through the trained performance prediction model, and obtaining a performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, the target knowledge graph includes historical production process information and historical performance values, the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values.

[0011] In a third aspect, the present application further provides an electronic device, including:

[0012] processor;

[0013] a memory for storing processor-executable instructions;

[0014] The processor is configured to execute instructions to implement the steel performance prediction method provided in the first aspect of the present application.

[0015] In a fourth aspect, the present application further provides a storage medium, when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the steel performance prediction method provided in the first aspect of the present application.

[0016] In a fifth aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, performs the steel performance prediction method provided in the first aspect of the present application.

[0017] The present application provides a method, device, electronic device and storage medium for predicting the performance of steel. Specifically, the production process information of the target steel is obtained, and the production process information includes the setting parameters of the production equipment of the target steel during the production process; the production process information of the target steel is input into a trained performance prediction model, and the performance prediction is performed through the trained performance prediction model to obtain the performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, and the target knowledge graph includes historical production process information and historical performance values, and the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values. In this way, the performance prediction value obtained by the performance prediction model based on the target knowledge graph is associated with the historical performance value in the target knowledge graph. Since the historical performance value is real data, the performance prediction value of the target steel also has a high authenticity. Compared with the simulation data calculated by the deep learning model as the output result in the related art, the performance prediction value provided by the embodiment of the present application has a higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0019] Figure 1 A flow chart of a method for predicting properties of steel provided in an embodiment of the present application;

[0020] Figure 2 A flow chart of a method for predicting properties of steel provided in an embodiment of the present application;

[0021] Figure 3 A flow chart of a method for predicting properties of steel provided in an embodiment of the present application;

[0022] Figure 4 A conceptual diagram of a steel performance prediction method provided in an embodiment of the present application;

[0023] Figure 5 A structural block diagram of a steel performance prediction device provided in an embodiment of the present application;

[0024] Figure 6 A structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings, but it should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present disclosure.

[0026] As described in the background technology, in the steel industry, performance prediction of steel materials (such as cold-rolled steel and / or hot-rolled steel) to be produced is a crucial link. However, this prediction process faces many challenges and difficulties. The primary difficulty is that the production process of steel materials is extremely complex. For example, the production process of cold-rolled steel spans multiple key links such as ironmaking, steelmaking, and hot rolling. The process parameters in each link, such as temperature, pressure, speed, etc., have a profound impact on the performance of the final product. Therefore, in order to achieve accurate performance prediction, it is necessary to fully and accurately master the process parameters of these production links, which is undoubtedly a huge and meticulous task.

[0027] In the related art, the performance of steel is usually predicted through the algorithm in the deep learning model, and the simulation data calculated by the algorithm is used as the prediction result, which has the problem of low accuracy.

[0028] In this regard, an embodiment of the present application provides a method for predicting the performance of steel. Specifically, the production process information of the target steel is obtained, and the production process information includes the setting parameters of the production equipment of the target steel during the production process; the production process information of the target steel is input into a trained performance prediction model, and the performance prediction is performed through the trained performance prediction model to obtain the performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, and the target knowledge graph includes historical production process information and historical performance values, and the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values. In this way, the performance prediction value obtained by the performance prediction model based on the target knowledge graph is associated with the historical performance value in the target knowledge graph. Since the historical performance value is real data, the performance prediction value of the target steel also has a high degree of authenticity. Compared with the related art in which the simulated data calculated by the deep learning model is used as the output result, the performance prediction value provided by the embodiment of the present application has a higher accuracy.

[0029] In addition, in the related art, predictions are made through deep learning models, and the selection and application of models is also a major challenge. Different prediction models have their own advantages and are suitable for different prediction scenarios and data characteristics, and the parameter adjustment of the model is a complex and time-consuming task. Different parameter configurations often lead to significant differences in prediction results, so it is necessary to optimize the parameter settings through a large number of experiments and verifications. With the innovation of production processes and fluctuations in raw material quality, the prediction model needs to be frequently updated and optimized to adapt to new prediction needs, which increases the cost and difficulty of prediction work. However, in an embodiment of the present application, the performance prediction model is a model based on a target knowledge graph, and the effect of updating the performance prediction model can be achieved by constructing a target knowledge graph, which simplifies the update process of the performance prediction model to a certain extent.

[0030] During the use of the deep learning model used in the related technology, the state of the production equipment will also change. The accuracy, stability and service life of the equipment will directly affect the product quality. Equipment aging, wear or failure are more likely to cause performance fluctuations, adding more uncertainty to the prediction work. In this regard, in the embodiment of the present application, the target knowledge graph includes historical production process information and historical performance values ​​as historical real data. In other words, the target knowledge graph already includes production process information and corresponding historical performance values ​​in various situations, which can provide more comprehensive and accurate performance prediction values.

[0031] It should be understood that the steel performance prediction method provided in the embodiment of the present application can be executed by a target device. The target device can be one electronic device or multiple electronic devices working together to execute the method. The electronic device can be a server, such as an independent physical server, a server cluster composed of multiple servers, and a cloud server capable of cloud computing.

[0032] The following is a detailed description of the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0033] Figure 1 This is a flow chart of a method for predicting the performance of steel provided in an embodiment of the present application. Figure 1 As shown, the steel performance prediction method provided in the embodiment of the present application includes the following steps:

[0034] Step 110, obtaining production process information of the target steel, wherein the production process information includes setting parameters of production equipment during the production process of the target steel.

[0035] In the embodiment of the present application, the target steel can be steel in any production process, and the embodiment of the present application does not impose specific restrictions on this. For example, in the process of producing steel plated steel coils, the production process involves hot rolling process, acid rolling process, etc. The target steel can be steel in the hot rolling process step, and the production process information can be the setting parameters of the production equipment in the hot rolling process design process, such as intermediate air cooling time, intermediate air cooling temperature, etc. The production process information may include at least one setting parameter of the production equipment of the target steel in the production process, and the embodiment of the present application does not impose specific restrictions on the number and type of setting parameters in the production process information.

[0036] In an embodiment of the present application, the target device can obtain the production process information of the target steel by receiving input from a third-party device, for example, the target device obtains intermediate air cooling through a detection device. After the target device obtains the production process information of the target steel, the target device can execute step 120 to predict the final performance effect of the target steel based on the production process information. In the process of the target device executing step 120, the performance prediction model can perform performance prediction based on all the production process information, and can also perform performance prediction based on part of the production process information. For example, as described below, key process parameters are preset in advance, and the performance prediction model performs performance prediction based on the key process parameters included in the production process information.

[0037] Step 120: input the production process information of the target steel into a trained performance prediction model, perform performance prediction through the trained performance prediction model, and obtain a performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, the target knowledge graph includes historical production process information and historical performance values, the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values.

[0038] In an embodiment of the present application, the target knowledge graph can provide data support to the trained performance prediction model in the form of a database, so as to facilitate the use of the performance prediction model. The target knowledge graph includes historical production process information, which can be the actual setting parameters of the production equipment in the production process that has been completed, and can also be the setting parameters of the production equipment pre-designed for the production process that has been completed. The historical performance value is the endpoint performance value after the production process to which the historical production process information belongs is completed. For example, when the target steel is an IF steel galvanized coil, the historical production process information can be information such as the furnace entry temperature and the total furnace time in any hot rolling process step in the past three months, and can also be information such as the intermediate air cooling time designed for this hot rolling process step. The historical performance value can be the endpoint performance value of the IF steel galvanized coil this time.

[0039] In the embodiment of the present application, the target knowledge graph may include ontology, attributes and relationships. The ontology may be the process steps of the target steel in the production process, the attributes may be the design parameters and / or actual parameters of the production equipment of the target steel in different process steps, and the relationship is used to associate the ontology with the attributes. For example, when the target steel is galvanized coil of IF steel, the ontology may be "hot rolling process design" and the attributes may be "intermediate air cooling time". It should be noted that the "design parameters" here refer to the setting parameters of the production equipment preset in advance according to the contract or experience, and the "actual parameters" refer to the setting parameters of the production equipment in the actual process steps. The same applies to similar expressions below.

[0040] In an embodiment of the present application, the trained performance prediction model can be any model that can generate data interaction with the target knowledge graph. For example, the trained performance prediction model can query and match the performance prediction value of the target steel from the target knowledge graph, and can also train the performance prediction value of the target steel through the historical real data in the target knowledge graph. For example, the trained performance prediction model can query the target knowledge graph for historical production process information that is the same or similar to the production process information of the target steel, and use the target historical performance value corresponding to the historical production process information obtained by the query as the performance prediction value of the target steel.

[0041] The embodiment of the present application provides a method for predicting the performance of steel, which can associate the performance prediction value obtained based on the performance prediction model of the target knowledge graph with the historical performance value in the target knowledge graph. Since the historical performance value is real data, the performance prediction value of the target steel also has a high degree of authenticity. Compared with the related technology that uses the simulated data calculated by the deep learning model as the output result, the performance prediction value provided by the embodiment of the present application has higher accuracy.

[0042] In the embodiment of the present application, the production process of the target steel includes N process steps, and the N process steps are performed in sequence, where N is a natural number. A specific implementation method of step 110 provided in the embodiment of the present application may be: obtaining the result information of the i-th process step, where the result information represents the performance information of the target steel after the i-th process step, where i is a natural number not greater than N; using the result information of the i-th process step and the design information of the (Ni)th process step after the i-th process step as the production process information of the target steel, where the design information represents the setting parameters of the production equipment of the target steel in the (Ni)th process step after the i-th process step.

[0043] In an embodiment of the present application, the production process of the target steel includes N process steps. For example, when the target steel is an IF steel galvanized coil, the production process of the target steel can be analyzed, which mainly includes contract standards, steelmaking process, hot rolling process, pickling process, galvanizing process, room temperature tensile properties and other process steps. At this time, N is equal to 6.

[0044] In the embodiment of the present application, when the production process of the target steel goes through the "steelmaking process", the performance information of the target steel after the "steelmaking process" can be obtained as the result information, and the result information is used as the production process information of the target steel along with the design information of the subsequent process steps, and the performance prediction is performed through the trained performance prediction model to obtain the endpoint performance value (i.e., performance prediction value) of the target steel based on the production process information at this time. In this way, in the production process of the target steel, the performance prediction can be performed based on the result information of the current process and the design information of the subsequent process, and the design information of the subsequent process can be adjusted in time to improve the qualified rate of the target steel.

[0045] In an embodiment of the present application, after obtaining the predicted performance value of the target steel, the next step of judgment can be performed based on the predicted performance value of the target steel. If the predicted performance value is not less than a first threshold value, the target steel is considered qualified. In an embodiment of the present application, the first threshold value may represent a preset performance value for the target steel, such as a qualified threshold value specified in a contract. If the predicted performance value is not less than the first threshold value, the target steel is considered qualified, and no other processing is required, allowing the production process of the target steel to end naturally. If the predicted performance value is less than the first threshold value, the target steel is considered unqualified.

[0046] In the embodiment of the present application, if the performance prediction value is less than the first threshold value, the target steel is considered to be unqualified or not in compliance with the contract provisions, and the production process of the target steel can be manually interfered with in a timely manner to improve the performance value of the target steel, or reduce the quality expectation of the target steel, and reduce the impact of unqualified products. If the target steel is unqualified, the setting parameters of the production equipment in the (i+1)th process step are adjusted. For example, when the production process of the target steel passes through the "steelmaking process", the result information of this time is used as the production process information of the target steel along with the design information of the subsequent process steps, and the performance prediction is performed through the trained performance prediction model. If the target steel is considered to be unqualified, the actual parameters of the hot rolling process after the steelmaking process can be adjusted. In addition, the setting parameters of the production equipment of all process steps after the i-th process step can also be adjusted. For example, after the "steelmaking process", the performance prediction value of the target steel indicates that the target steel is unqualified, and the actual parameters of the process steps such as the hot rolling process, the acid rolling process, the galvanizing process, and the room temperature stretching process can be adjusted to improve the qualified rate of the target steel.

[0047] In an embodiment of the present application, a specific implementation method of step 120 may be: obtaining a comparison step length, wherein the comparison step length is used to characterize the difference between the production process information of the target steel and the historical production process information; based on the production process information of the target steel and the comparison step length, the trained performance prediction model generates a target structured query language, wherein the target structured query language is used for matching in the target knowledge graph; the trained performance prediction model queries in the target knowledge graph whether there is historical process information that matches the target structured query language; if there is historical process information that matches the target structured query language in the target knowledge graph, the historical performance value corresponding to the historical process information that matches the target structured query language is used as the performance prediction value of the target steel.

[0048] In the embodiment of the present application, the comparison step length can be any value. When the production process information is the design value of "final rolling temperature", "furnace entry temperature" and "coiling temperature" (at this time, the attributes corresponding to the target knowledge graph are "final rolling temperature", "furnace entry temperature" and "coiling temperature"), after the performance prediction model obtains the comparison step length, the generated target structured query language can be as follows:

[0049] filter((ABS(final rolling temperature-"design value")<"comparison step"))

[0050] Based on the above target structured query language, the performance prediction model can search for historical production process information with the attributes of "finishing rolling temperature", "furnace entry temperature" and "coiling temperature" in the target knowledge graph, and the values ​​are the same as the production process information, and determine the historical performance value corresponding to the historical production process information as the performance prediction value of the target steel. If there is no historical production process information that is the same as the production process information in the target knowledge graph, the historical production process information can be used as the center of the numerical range and the comparison step size can be used as the numerical range. If the production process information of the target steel falls within the numerical range of the historical production process information, it can also be considered that there is historical production process information matching the production process in the target knowledge graph.

[0051] In the embodiment of the present application, based on the above query method, if the target knowledge graph includes multiple historical production process information that matches the production process information, the multiple historical production process information can be sorted according to the correlation, and the sorting result can be output to the user for the user to view. In the embodiment of the present application, by comparing the step size, the performance prediction model can query more historical production process information in the target knowledge graph, and obtain the performance prediction value of the target steel therefrom, so that the performance prediction value of the target steel is more accurate.

[0052] In the embodiment of the present application, the target knowledge graph is constructed, and the target knowledge graph is constructed in advance through the following steps:

[0053] Step 210, obtain historical production process information, the historical process production information includes at least N process steps and process parameters corresponding to each process step, take the process steps as the ontology of the knowledge graph, and take the process parameters as attributes of the knowledge graph.

[0054] In the embodiment of the present application, a schema can be designed first. In the knowledge graph, the schema is used to limit the format of the data to be added to the knowledge graph. It is the metadata of the knowledge graph and describes the entities, attributes and relationships in the knowledge graph. According to the actual production processes such as production contracts, production lines, steelmaking process design, hot rolling process design, pickling process design, continuous annealing process design, galvanizing process design, room temperature tensile performance, etc., the ontology, relationship and attributes are abstracted to form the performance and process knowledge graph schema.

[0055] Among them, data mapping includes ontology and attribute mapping and relationship mapping, which are introduced below:

[0056] Ontology and attribute mapping: This process focuses on accurately mapping the entities and their attributes defined in the knowledge graph schema to the corresponding tables and fields in the actual production database. It requires careful matching of the table structure of the entity represented by each ontology in the database, while ensuring that the value of each attribute can be accurately extracted or stored from the database field. Data mapping not only ensures the accuracy of the data, but also promotes the accessibility and consistency of the data, laying a solid foundation for subsequent data analysis, query and knowledge reasoning.

[0057] Relationship mapping: As a supplement to data mapping, relationship mapping focuses on parsing and expressing the relationships between tables in the database, which are usually implemented through association fields (such as foreign keys). In the knowledge graph, these relationships are abstracted as edges connecting different ontologies (entities), thereby building a complex and rich knowledge network. Relationship mapping requires a deep understanding of the design logic of the database, accurately identifying which tables are associated with each other, and how these associations affect the overall structure and meaning of the data. Through relationship mapping, we can reproduce the inherent logic and hierarchical structure of the data in the knowledge graph, providing strong support for complex data analysis and knowledge discovery.

[0058] Step 220, obtain the mapping relationship between the ontology of the knowledge graph and the attributes of the knowledge graph.

[0059] In the embodiment of the present application, the target device may obtain the mapping relationship by receiving manual input. For example, the target device may provide a visual interface for the user to perform the following operations:

[0060] (1) Connect to the production data source (which can be understood as the database to which historical production process information belongs) and select the table to be mapped;

[0061] (2) Map the business table where the ontology is located and select the unique primary key of the ontology;

[0062] (3) Map the attributes of the ontology to the fields in the table and confirm the source of the attributes;

[0063] (4) Mapping the relationship between ontologies through the relationship between tables

[0064] Step 230, based on the mapping relationship, extract the tabular data associated with the mapping relationship in the historical database, and construct a target knowledge graph based on the tabular data, wherein the historical database includes numerical values ​​corresponding to historical production process information and historical performance values.

[0065] In an embodiment of the present application, after receiving manual input through a visual interface, the knowledge graph platform on the target device can automatically generate SQL statements to query data in the data source, extract knowledge to form an NT file, and finally import the NT file into the RDF graph database to construct the target knowledge graph.

[0066] In an embodiment of the present application, after constructing the target knowledge graph, the target knowledge graph can be used to train and adjust the performance prediction model (or the original model mentioned below) to update the target knowledge graph. For example, after multiple prediction processes, the historical process information in the target knowledge graph can be adjusted to make the target knowledge graph more in line with actual use, thereby improving the correlation between the target knowledge graph and the performance value of the target steel, and further improving the accuracy of the performance prediction value.

[0067] In the embodiments of this application, the key process parameter factors that affect the performance in the production process can be deeply mined and associated, and a detailed performance and process knowledge map (i.e., target knowledge map) can be constructed by combining rich historical production data. With the powerful analysis and reasoning capabilities of the target knowledge map, accurate prediction of cold-hardened coil performance can be achieved in a complex production environment, providing strong support for production decision-making in the steel industry.

[0068] In the embodiment of the present application, the performance prediction model is constructed, and the performance prediction model is constructed in advance through the following steps:

[0069] Step 310: Acquire second training data, where the second training data at least includes a training step length, training process information, and a training performance value.

[0070] In the embodiment of the present application, the second training data can be obtained by receiving manual input, or can be obtained through a knowledge graph or a target knowledge graph, and the embodiment of the present application does not make specific restrictions on this. The training process information can be some process steps and some process parameters in the steel production process, and the embodiment of the present application does not make specific restrictions on this.

[0071] In step 320, the original model queries the target knowledge graph based on the training step size and the training process information, and obtains the query results.

[0072] In an embodiment of the present application, the steel performance prediction model can also be queried in the knowledge graph. The query result is the target knowledge graph or the historical performance value included in the knowledge graph.

[0073] Step 330: Acquire a performance prediction model based on the query result and the training performance value.

[0074] In the embodiment of the present application, after the steel performance prediction model obtains the query result, it can be compared with the training performance value, so as to adjust the training step size and training process information in the steel performance prediction model. For example, the size of the training step size is adjusted, and the process steps and / or process parameters in the training process information are adjusted so that the training process information has a higher correlation with the final performance of the steel. By training the steel prediction model, the accuracy of the performance prediction model can be made higher.

[0075] Figure 4 The conceptual diagram of the steel performance prediction method provided in the embodiment of the present application is as follows: Figure 4 As shown, the steel performance prediction method provided in the embodiment of the present application includes the following processing:

[0076] S1. Knowledge graph construction

[0077] S11.Schema design: In the knowledge graph, Schema is used to limit the format of the data to be added to the knowledge graph. It is the metadata of the knowledge graph and describes the entities, attributes and relationships in the knowledge graph. Based on the actual production processes such as production contracts, production lines, steelmaking process design, hot rolling process design, pickling process design, continuous annealing process design, galvanizing process design, room temperature tensile properties, etc., abstract the ontology, relationships and attributes to form the performance and process knowledge graph schema;

[0078] S12. Data mapping includes ontology and attribute mapping and relationship mapping

[0079] Ontology and attribute mapping: This process focuses on accurately mapping the entities and their attributes defined in the knowledge graph schema to the corresponding tables and fields in the actual production database. It requires careful matching of the table structure of the entity represented by each ontology in the database, while ensuring that the value of each attribute can be accurately extracted or stored from the database field. Data mapping not only ensures the accuracy of the data, but also promotes the accessibility and consistency of the data, laying a solid foundation for subsequent data analysis, query and knowledge reasoning.

[0080] Relationship mapping: As a supplement to data mapping, relationship mapping focuses on parsing and expressing the relationships between tables in the database, which are usually implemented through association fields (such as foreign keys). In the knowledge graph, these relationships are abstracted as edges connecting different ontologies (entities), thereby building a complex and rich knowledge network. Relationship mapping requires a deep understanding of the design logic of the database, accurately identifying which tables are associated with each other, and how these associations affect the overall structure and meaning of the data. Through relationship mapping, we can reproduce the inherent logic and hierarchical structure of the data in the knowledge graph, providing strong support for complex data analysis and knowledge discovery;

[0081] S13. By configuring the data source, according to the actual situation of the data mapping in the previous step, dynamically generate and extract SQL statements to extract entities, relationships, attributes and other information from each data source to form an NT file;

[0082] S14. Import NT into the RDF graph database to form a knowledge graph.

[0083] S2. Performance prediction model

[0084] S21. Set the key process parameters that affect the performance indicators. The specific steps are as follows:

[0085] (1) Set the key process parameters that affect each performance index under different steel grades, different production lines, different test directions and different gauge lengths;

[0086] (2) When analyzing and predicting in the knowledge graph, the probability of finding exactly the same data in historical data is extremely small, so it is necessary to set the comparison step of key process parameters so that there is a reasonable comparison range when comparing with historical process designs. This comparison range can be adjusted according to actual conditions;

[0087] S22. These parameters in step S21 are clearly identified as key factors with significant impact on the performance indicators. When building the model, we not only consider the current values ​​of these parameters, but also carefully incorporate their step changes to capture the dynamic evolution of performance during process adjustment.

[0088] By comprehensively applying these key process parameters and their comparative steps, we have built an efficient and accurate performance prediction model. This model can comprehensively capture the key variables in the production process and scientifically predict the performance of future products based on them, providing strong support for production decisions, process optimization and quality control.

[0089] S23. SPARQL query statements are finally generated through the input of prediction model and process parameters. SPARQL will match qualified data in the knowledge graph.

[0090] S3. After the knowledge graph is constructed and the model is designed, the next step is to perform performance prediction, which is divided into the following test steps:

[0091] S31. Preliminary performance forecast (production has not started)

[0092] This step is before the production line is started. After the engineers have designed the process for the steel coils to be produced, they use the prediction model based on the process design parameters, generate SPARQL through the performance prediction model, and perform performance prediction by matching the qualified data in the knowledge graph.

[0093] S32. Prediction of slab properties after production

[0094] With the smooth generation of slabs, we entered a more accurate performance prediction stage. At this point, we not only rely on the previous process design, but also combine the actual data of the steelmaking process, as well as the design parameters of subsequent processes such as hot rolling, pickling, annealing or galvanizing. Through the performance prediction model, SPARQL is regenerated to comprehensively predict the subsequent processing performance of the slab. This comprehensive prediction method effectively improves the accuracy of the prediction.

[0095] S33. Prediction of properties after hot rolling

[0096] After the hot rolling process is completed, we further refine the performance prediction work. Based on the actual production data of steelmaking and hot rolling processes, combined with the design requirements of subsequent processes such as pickling, annealing or galvanizing, we regenerate SPARQL through the performance prediction model to make a more accurate prediction of the final performance of the product. This process fully considers the mutual influence between the processes at each stage, ensuring the scientificity and reliability of the prediction results.

[0097] S34. Performance prediction after pickling

[0098] After the product has been pickled, we perform another performance prediction. At this point, the prediction basis has been expanded to the actual production data of the three processes of steelmaking, hot rolling and pickling. At the same time, combined with the design considerations of subsequent processes such as annealing or galvanizing, the performance prediction model is regenerated into SPARQL to comprehensively evaluate and predict the final performance of the product.

[0099] S4. The accuracy of performance prediction results is crucial for actual risk control, so prediction results need to be verified. We verify the results in two ways:

[0100] S41. Prediction verification based on historical data

[0101] In order to preliminarily evaluate the accuracy of the performance prediction model, a historical data sampling method was adopted. By carefully selecting representative sample data from a large historical database, using the powerful capabilities of the knowledge graph and combining it with the constructed performance prediction model, we conducted a predictive analysis on these samples. Subsequently, we conducted a detailed comparison of the prediction results with the corresponding historical actual results, and preliminarily verified the prediction performance of the model by calculating key indicators such as error rate and accuracy.

[0102] S42. Actual production inspection and verification

[0103] In order to further confirm the effectiveness of the performance prediction model in the actual production environment, we directly applied it to the performance prediction in the actual production process. During the production process, we strictly arranged the production according to the prediction results of the model and paid close attention to the actual performance of the product. Subsequently, we collected the inspection and test results in actual production and compared them with the prediction results of the model. Through the verification of this link, we can not only intuitively evaluate the accuracy of the model in actual application, but also promptly discover and correct possible deviations, providing valuable data support for the continuous optimization of the model.

[0104] In order to better understand the steel performance prediction method provided in the embodiment of the present application, an example is given, taking the production of IF steel galvanized coils, the test direction is horizontal, and the gauge length is 80 mm as an example.

[0105] Step 1: Construct a knowledge graph, analyze the production process of the steel coil, and then design the schema of the performance process design knowledge graph, which mainly includes the subject matter of contract standards, production line steelmaking process design, hot rolling process design, pickling process design, galvanizing process design, actual steelmaking process, actual hot rolling process, actual pickling process, actual galvanizing process, actual room temperature tensile performance, and the attributes of the subject matter. Some of the content can be referred to in the following Tables 1 and 2:

[0106] Ontology name Ontology attributes Hot rolling process design Intermediate air cooling time Pickling process design The whole process sequence number

[0107] Table 1 Ontology attribute table The relationship between ontologies. Some of the contents can be referred to in Table 2 below:

[0108]

[0109]

[0110] Table 2 Ontology relationship table

[0111] After the Schema design is completed, it is connected to the production data source. Through the knowledge graph construction platform visual interface, it can be operated manually to map the ontology, attributes and relationships with the tables and fields in the data source. Data mapping is divided into four steps:

[0112] (1) Connect to the production data source and select the table to be mapped;

[0113] (2) Map the business table where the ontology is located and select the unique primary key of the ontology;

[0114] (3) Map the attributes of the ontology to the fields in the table and confirm the source of the attributes;

[0115] (4) Map the relationship between ontologies through the relationship between tables.

[0116] After performing the data mapping in the above four steps, the knowledge graph construction platform will automatically generate SQL statements to query data in the data source, extract knowledge to form NT files, and finally import the NT files into the RDF graph database to form a knowledge graph.

[0117] Step 2: Build a prediction model

[0118] First, set the process parameter factors and comparison step length that affect the 90-degree n value (hereinafter referred to as N90) of the room temperature tensile mechanical properties of IF steel galvanized coils when the test direction is horizontal and the gauge length is 80 mm. They are mainly used to predict the source of model parameters, and the step length is used for query and comparison. Some of the content can be referred to in Table 3 below:

[0119] Process parameter type Process parameter name Comparison step length Steelmaking process Alt Aluminum Element 2 Hot rolling process Finishing temperature 20 Acid bonding process 5# rack outlet thickness 0.2 Galvanizing process Thickness gauge 0.2

[0120] Table 3 Performance influencing factors

[0121] Step 3: Performance prediction

[0122] First, according to the performance parameters required by the customer's contract, process design is carried out to determine the design values ​​of the production line and steelmaking, hot rolling, pickling and galvanizing process parameters.

[0123] After the process design is completed, SPARQL is generated through the prediction model. The process of generating SPARQL from the prediction model is as follows:

[0124] The production process information input into the performance prediction model may include the key process information mentioned above, and the performance prediction model performs performance prediction based on the key process information in the production process information.

[0125] (1) The process parameter design values ​​of the current contract are passed into the model as parameters;

[0126] (2) Since the key process parameters of performance indicators will be different in different situations, the key process parameters affecting the performance set in step 2 are used as the basis for searching in the knowledge graph, and SPARQL is dynamically generated. The generated SPARQL needs to meet the Schema design of building the graph in step 1;

[0127] (3) Since the probability of finding exactly the same process parameter design in historical contract production data is low, the generation of SPARQL requires incorporating the comparison step size of each process parameter into SPARQL.

[0128] (4) Based on the generated SPARQL, a path query is performed in the RDF graph database to find the path that meets the conditions. Because the query is compared through a range, there will generally be multiple results that meet the conditions. The next step is to perform performance judgment and accuracy evaluation on the query results.

[0129] Performance prediction S31 step has not started production. Through the steelmaking, hot rolling, pickling, and galvanizing process design parameters, the prediction model is generated, and some of the SPARQL content is as follows. It should be understood that the following SPARQL language is abridged and is for reference only:

[0130] select performance name performance actual value where {

[0131] Contract <Contract Production Line> Contract Production Line.

[0132] Contract production line <Contract hot rolling process> hot rolling process design.

[0133] filter(

[0134] (ABS(final rolling temperature-"design value")<"comparison step"))

[0135] Contract production line <Contract pickling process> Pickling process design.

[0136] filter(

[0137] (ABS(Upper limit of deformation rate - "design value") < "comparison step"))

[0138] Contract production line <Contract galvanizing process> galvanizing process design.

[0139] filter(

[0140] (ABS(flatness rate-"design value")<"comparison step length"))}

[0141] After the steelmaking production is completed, the performance prediction step S32 has the actual value of the furnace composition. Then, the above parameters are input into the prediction model through the hot rolling, pickling and galvanizing process design, and part of the SPARQL content is generated as follows:

[0142] select performance name performance actual value where {

[0143] Contract <Contract Production Line> Contract Production Line.

[0144] Steelmaking materials <actual composition of each furnace> composition of each furnace.

[0145] filter(

[0146] (ABS(Alt-"Actual Value")<"Comparison Step"))

[0147] Contract production line <Contract hot rolling process> hot rolling process design.

[0148] filter(

[0149] (ABS(final rolling temperature-"design value")<"comparison step"))

[0150] Contract production line <Contract pickling process> Pickling process design.

[0151] filter(

[0152] (ABS(Deformation rate upper limit - "actual value") < "comparison step"))

[0153] Contract production line <Contract galvanizing process> galvanizing process design.

[0154] filter(

[0155] (ABS(flatness rate-"design value")<"comparison step length"))}

[0156] The performance prediction step S33 is completed for the hot rolling process. The actual values ​​of the steelmaking furnace composition and the hot rolling process are combined with the pickling and galvanizing process design. The above parameters are passed into the prediction model to generate part of the SPARQL content as follows:

[0157] select performance name performance actual value where {

[0158] Contract <Contract Production Line> Contract Production Line.

[0159] Steelmaking materials <actual composition of each furnace> composition of each furnace.

[0160] filter(

[0161] (ABS(Alt-"Actual Value")<"Comparison Step"))

[0162] Hot rolled material <actual hot rolling process> actual hot rolling process.

[0163] filter(

[0164] (ABS(final rolling temperature-"design value")<"comparison step"))

[0165] Contract production line <Contract pickling process> Pickling process design.

[0166] filter(

[0167] (ABS(Upper limit of deformation rate - "design value") < "comparison step"))

[0168] Contract production line <Contract galvanizing process> galvanizing process design.

[0169] filter(

[0170] (ABS(flatness rate-"design value")<"comparison step length"))}

[0171] The performance prediction step S34 is completed for the pickling process. Through the steelmaking furnace composition, hot rolling process, and actual values ​​of the pickling process, combined with the galvanizing process design, the above parameters are passed into the prediction model to generate part of the SPARQL content as follows:

[0172] select performance name performance actual value where {

[0173] Contract <Contract Production Line> Contract Production Line.

[0174] Steelmaking materials <actual composition of each furnace> composition of each furnace.

[0175] filter(

[0176] (ABS(Alt-"Actual Value")<"Comparison Step"))

[0177] Hot rolled material <actual hot rolling process> actual hot rolling process.

[0178] filter(

[0179] (ABS(final rolling temperature-"actual value")<"comparison step"))

[0180] Contract production line <Actual pickling process> Actual pickling process.

[0181] filter(

[0182] (ABS(1# rack pressure rate-"actual value")<"comparison step length"))

[0183] Contract production line <Contract galvanizing process> galvanizing process design.

[0184] filter(

[0185] (ABS(flatness rate-"design value")<"comparison step length"))}

[0186] Step 4: Determine the performance prediction results. There are two ways to determine the prediction results, as follows:

[0187] (1) In the historical production process, a large amount of production data has been accumulated, including the full process data from quality design to production delivery. The production data of different periods, different steel grades, different test directions, and different gauge lengths are sampled, and SPARQL is generated through the prediction model. SPARQL is used to query in the knowledge graph. Each SPARQL can match multiple results, and the matched results are compared with the actual production results to determine whether the prediction model is accurate. Therefore, there is no need to perform new steps, no need to generate new materials, and no cost is incurred, so a large number of predictions and parameter adjustments can be performed to improve the prediction model.

[0188] (2) For newly produced materials, prediction is performed through step 3. After each step of prediction, if the final predicted performance result does not meet the acceptance criteria, the production personnel should be reminded of the risk and whether the subsequent process parameters need to be adjusted. The prediction results of each step need to be recorded. After the steel coil is generated, the actual performance data is collected and compared with the predicted results to determine whether the prediction model is accurate. The prediction results can also be used as the basis for model optimization.

[0189] The steel performance prediction method provided in the embodiment of the present application is mainly used for steel coil production risk identification, and the performance prediction results are used to determine whether the final steel coil meets the delivery standards. At the same time, multiple performance predictions can be performed in various links of steel coil production, and the process parameters of subsequent production links can be adjusted in time according to the prediction results to meet the final delivery standards. If the steel coil produced by the final prediction cannot meet the customer's delivery standards, it can be downgraded or discarded in time, and the production plan can be adjusted in advance to avoid delayed delivery of orders.

[0190] In addition, the steel performance prediction method provided in the embodiment of the present application also has the following advantages:

[0191] Data integration and fusion: Collect a wide range of heterogeneous data from multiple sources in the production process, including raw material characteristics, production process parameters, etc., and integrate and fuse them effectively through knowledge graphs to break data silos. For example, associate different types of data such as chemical composition data, hot rolling process, pickling process, continuous annealing, and galvanizing process to form a comprehensive knowledge system. Use the semantic relationships of the knowledge graph to clean and preprocess the data to improve data quality and availability.

[0192] Real-time pre-control: Use knowledge graph technology to achieve real-time monitoring of the cold-hardened coil production process, and timely capture abnormal conditions and performance changes in the production process. Through real-time monitoring, the latest data support is provided for performance prediction to ensure the timeliness and accuracy of the prediction results.

[0193] Feedback optimization: Compare and analyze the performance prediction results with the actual situation to find deviations and deficiencies in the prediction. Use the knowledge graph update mechanism to adjust and optimize the prediction model in a timely manner to improve the accuracy and reliability of the prediction model.

[0194] Figure 5 This is a structural block diagram of a steel performance prediction device provided in an embodiment of the present application. Figure 5 As shown, the steel performance prediction device 500 provided in the embodiment of the present application includes: an acquisition module 510 and a processing module 520.

[0195] An acquisition module 510 is used to acquire production process information of the target steel, wherein the production process information includes setting parameters of production equipment during the production process of the target steel;

[0196] The processing module 520 is used to input the production process information of the target steel into the trained performance prediction model, perform performance prediction through the trained performance prediction model, and obtain the performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on the target knowledge graph, the target knowledge graph includes historical production process information and historical performance values, the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values.

[0197] It should be noted that the embodiment of the steel performance prediction device in this specification and the embodiment of the steel performance prediction method in this specification are based on the same inventive concept. Therefore, the specific implementation method of this embodiment can refer to the corresponding embodiment of the steel performance prediction method in the previous text, and the repeated parts will not be repeated.

[0198] Figure 6 A structural block diagram of an electronic device provided for the implementation of this application. Figure 6 As shown, the electronic device provided in the embodiment of the present application includes a processor 610 and a memory 620, wherein the memory is used to store instructions executable by the processor; wherein the processor is configured to execute the instructions to implement as shown in Figure 1-3 Method for predicting the properties of steel shown in any of the figures.

[0199] In an exemplary embodiment, the electronic device may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above methods.

[0200] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions, and the instructions can be executed by a processor of the device to complete the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device. The non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by the processor of the electronic device, enables the electronic device to perform Figure 1 The data display method shown.

[0201] The present application also provides a computer program product, including a computer program, which, when executed by a processor, Figure 1-3 Method for predicting the properties of steel shown in any of the figures.

[0202] In the above description, the technical details such as the patterning of each layer are not described in detail. However, those skilled in the art should understand that various technical means can be used to form layers, regions, etc. of desired shapes. In addition, in order to form the same structure, those skilled in the art can also design methods that are not completely the same as the methods described above. In addition, although the various embodiments are described above separately, this does not mean that the measures in the various embodiments cannot be used in combination to advantage.

[0203] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0204] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for predicting properties of steel, characterized in that: The method comprises: Acquiring production process information of the target steel, wherein the production process information includes setting parameters of production equipment during the production process of the target steel; The production process information of the target steel is input into a trained performance prediction model, and performance prediction is performed through the trained performance prediction model to obtain a performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, and the target knowledge graph includes historical production process information and historical performance values, and the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values.

2. The method according to claim 1, characterized in that The production process of the target steel includes N process steps, and the N process steps are performed in sequence, where N is a natural number. The step of obtaining the production process information of the target steel includes: Obtaining result information of the i-th process step, wherein the result information includes performance information characterizing the target steel after the i-th process step, wherein i is a natural number not greater than N; The result information of the i-th process step and the design information of the (Ni)th process step after the i-th process step are used as the production process information of the target steel, and the design information represents the setting parameters of the production equipment of the target steel in the (Ni)th process step after the i-th process step.

3. The method according to claim 2, characterized in that After obtaining the predicted value of the property of the target steel, the method further comprises: If the performance prediction value is not less than a first threshold, determining that the target steel is qualified; If the performance prediction value is less than the first threshold, the target steel is determined to be unqualified, and the setting parameters of the production equipment in the (i+1)th process step are adjusted.

4. The method according to claim 1, characterized in that: The method of performing performance prediction based on the production process information of the target steel by using a trained performance prediction model to obtain a performance prediction value of the target steel includes: Acquire a comparison step length, wherein the comparison step length is used to characterize the difference between the production process information of the target steel and the historical production process information; Based on the production process information and comparison step length of the target steel, the trained performance prediction model generates a target structured query language, and the target structured query language is used for matching in the target knowledge graph; The trained performance prediction model queries the target knowledge graph for historical process information that matches the target structured query language; If there is historical process information matching the target structured query language in the target knowledge graph, the historical performance value corresponding to the historical process information matching the target structured query language is used as the performance prediction value of the target steel.

5. The method according to claim 1, characterized in that The target knowledge graph is pre-constructed through the following steps: Acquire historical production process information, where the historical process production information includes at least N process steps and process parameters corresponding to each process step, use the process steps as the ontology of the knowledge graph, and use the process parameters as attributes of the knowledge graph; Obtain the mapping relationship between the ontology of the knowledge graph and the attributes of the knowledge graph; Based on the mapping relationship, tabular data associated with the mapping relationship is extracted from a historical database, and a target knowledge graph is constructed based on the tabular data. The historical database includes numerical values ​​corresponding to historical production process information and historical performance values.

6. The method according to claim 1, characterized in that The performance prediction model is pre-built through the following steps: Acquire second training data, where the second training data at least includes a training step length, training process information, and a training performance value; The original model queries the target knowledge graph based on the training step size and training process information, and obtains the query results; A performance prediction model is obtained based on the query result and the training performance value.

7. A device for predicting properties of steel, characterized in that: The device comprises: An acquisition module, for acquiring production process information of target steel, wherein the production process information includes setting parameters of production equipment during the production process of the target steel; A processing module is provided for inputting the production process information of the target steel into a trained performance prediction model, performing performance prediction through the trained performance prediction model, and obtaining a performance prediction value of the target steel, wherein the trained performance prediction model is a model trained based on a target knowledge graph, the target knowledge graph includes historical production process information and historical performance values, the historical production process information corresponds to the historical performance values ​​one by one, and the performance prediction value of the target steel comes from the historical performance values.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the steel performance prediction method according to any one of claims 1 to 6.

9. A storage medium, when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the steel property prediction method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the steel property prediction method according to any one of claims 1 to 6.

Images