Sintering dynamic knowledge graph construction method

Abstract

The invention provides a sintering dynamic knowledge graph construction method, and relates to the technical field of knowledge graphs, and the method comprises the steps: collecting full-chain data in a real-time sintering process, and carrying out the parameter classification, and obtaining a sintering target parameter and a sintering process parameter; carrying out trend relation analysis based on a preset data analysis method so as to establish a sintering trend influence knowledge graph; performing parameter classification on the sintering process parameters based on the target ore blending cycle, and matching a corresponding machine learning algorithm to construct a sintering target parameter prediction model; obtaining a sintering trend influence knowledge graph corresponding to a first historical data point having the highest goodness of fit with the real-time ore blending structure, and obtaining a second historical data point having the highest goodness of fit with the current production state based on the sintering target parameter prediction model to obtain a sintering quantization influence knowledge graph; and determining a sintering trend corresponding to the real-time ore blending structure and a quantitative influence knowledge graph result. Fusion of mechanism data can be realized, and system processing efficiency and precision are improved.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and in particular to a method for constructing a sintering dynamic knowledge graph. Background Art

[0002] At present, sintering, as an important link in the ironmaking process, has the characteristics of many influencing factors, strong parameter coupling, and high frequency of changes. In the actual production process, the sintering site engineers rely on their own experience to control the sintering process. The stability of the sintering process quality depends entirely on the experience level of the engineers. New engineers also often need to study for a long time, and the long learning cycle is not conducive to the long-term stability of the sintering production line. Therefore, it is of far-reaching significance to be able to construct a knowledge graph to realize the application of the blast furnace ironmaking sintering process.

[0003] However, in the field of sintering technology, there is currently no knowledge graph for the sintering process, and the knowledge sources of the existing knowledge graphs are not closely related to actual production. In addition, the construction of the existing blast furnace ironmaking process ontology knowledge graph generally adopts entity relationship triples and lacks quantitative relationships, which is not conducive to rapid decision-making and review analysis by on-site production personnel and managers.

[0004] Therefore, the present invention provides a method for constructing a sintering dynamic knowledge graph. Summary of the invention

[0005] The present invention provides a method for constructing a sintering dynamic knowledge graph, which is used to solve the problems in the prior art of incomplete collection of industrial process data, low processing efficiency, incomplete consideration of actual production data information in the construction process; the knowledge graph construction process is difficult to understand, and it is difficult to promote on-site applications.

[0006] The present invention provides a method for constructing a sintering dynamic knowledge graph, comprising: Step 1: Collect the full chain data of the real-time sintering process, and classify the parameters to obtain the sintering target parameters and sintering process parameters; Step 2: Analyze the trend relationship between sintering target parameters and sintering process parameters based on the preset data analysis method, so as to establish a sintering trend influence knowledge map; Step 3: Classify the sintering process parameters based on the target ore blending cycle, and match the corresponding machine learning algorithm based on each classification result to build a sintering target parameter prediction model for the sintering process; Step 4: Obtain the sintering trend influence knowledge graph corresponding to the first historical data point with the highest degree of consistency with the real-time ore blending structure, and at the same time obtain the second historical data point with the highest degree of consistency with the current production status based on the sintering target parameter prediction model, obtain the sintering quantitative influence knowledge graph, and determine the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore blending structure; Among them, based on matching the corresponding machine learning algorithm with each classification result, a sintering target parameter prediction model for the sintering process is constructed, including: Extract the machine learning algorithm with the highest matching degree with the target sintering process from the machine learning algorithms corresponding to each classification result as the optimal algorithm; The historical sintering process parameters and historical sintering target parameters of the target sintering process are extracted from the sintering database, and the initial sintering target parameter prediction model is obtained by combining the optimal algorithm, so as to perform model verification and obtain the sintering target parameter prediction model of the target sintering process.

[0007] According to the present invention, the whole chain data of the real-time sintering process is collected and parameter classification is performed to obtain the sintering target parameters and sintering process parameters, including: Step 11: collecting the whole chain data of the real-time sintering process, and processing the whole chain data to obtain the first processed data; Step 12: Obtain the data type of the first processed data, and classify the first processed data according to the data type and sintering production experience to obtain sintering target parameters and sintering process parameters.

[0008] The method for establishing a knowledge graph of sintering trend influence provided by the present invention includes: Step 21: extracting a data analysis method matching the sintering target parameters and the sintering process parameters from a statistical analysis database based on the sintering characteristics of the ironmaking process to obtain a first data analysis method; Step 22: Analyze the parameter correlation between the sintering target parameters and the sintering process parameters based on the first data analysis method to obtain a first analysis result; Step 23: Obtain the data type of each data in the first processed data as a node of the sintering trend dynamic impact knowledge graph, and establish the sintering trend impact knowledge graph in combination with the first analysis result.

[0009] The sintering target parameter prediction model for constructing a sintering process provided by the present invention includes: Step 31: obtaining an ore blending cycle of a target sintering process, and classifying sintering process parameters and sintering target parameters based on the ore blending cycle to obtain a first cycle parameter set; Step 32: extracting a corresponding machine learning algorithm from an algorithm database based on parameter characteristics in each first cycle parameter subset in the first cycle parameter set, obtaining a first machine learning algorithm corresponding to each first cycle parameter subset, thereby obtaining a first machine learning algorithm set for the target sintering process; Step 33: Compare and analyze each first machine learning algorithm in the first machine learning algorithm set to determine the first machine learning algorithm with the highest matching degree with the target sintering process as the optimal algorithm, thereby obtaining a sintering target parameter prediction model for the target sintering process.

[0010] The first machine learning algorithm that is determined to have the highest matching degree with the target sintering process as the optimal algorithm according to the present invention includes: Step 331: different ore blending structures are distinguished based on sintering production experience, and the first cycle parameters corresponding to each ore blending cycle are obtained, so as to obtain the trend influence between the sintering target parameters and the sintering process parameters determined by different machine learning algorithms in different ore blending cycles, and obtain a trend influence set; Step 332: determining the trend influence result corresponding to each sintering process parameter based on the trend influence set; If the trend influence of the target sintering process parameters and the corresponding sintering target parameters are consistent with the mechanism knowledge graph, the mechanism data matching index is increased by 1; If the trend influence of the target sintering process parameters and the corresponding sintering target parameters are inconsistent with the mechanism knowledge graph, the mechanism data matching index is reduced by 1, and at the same time, the method with a smaller mechanism data matching index is deleted; Step 333: determining a corresponding historical mechanism knowledge graph based on historical sintering process parameters and historical sintering target parameters of the target sintering process; Step 334: Determine the mechanism knowledge graph of the target sintering process based on the mean of the historical mechanism knowledge graph, so as to obtain the preset group of sintering process parameters that have the greatest impact on the sintering target parameters, and determine the optimal algorithm of the target sintering process based on the relative ranking result of each sintering process parameter corresponding to the first machine learning algorithm.

[0011] According to the present invention, an optimal algorithm for determining a target sintering process based on a relative ranking result of each sintering process parameter corresponding to a first machine learning algorithm includes: Obtaining the sequence number corresponding to the relative ranking result of each first machine learning algorithm in the first machine learning algorithm set of the preset group of sintering process parameters that have the greatest impact on the sintering target parameters, to obtain a first sequence number set; Each serial number in the first serial number set is added together, and combined with the corresponding mechanism data matching index to obtain a corresponding mechanism data matching coefficient; The first machine learning algorithm with the smallest mechanism data matching coefficient is taken as the optimal algorithm.

[0012] The sintering target parameter prediction model for obtaining a target sintering process provided by the present invention includes: Extracting historical sintering process parameters and historical sintering target parameters of a target sintering process from a sintering database, and randomly extracting historical sintering process parameters and corresponding historical sintering target parameters as a first training set of the target sintering process, and taking the remaining historical sintering process parameters and corresponding historical sintering target parameters in the historical sintering process as a first validation set; The first training set is trained based on the optimal algorithm to obtain an initial sintering target parameter prediction model; The initial sintering target parameter prediction model is verified based on the first verification set to obtain the sintering target parameter prediction model.

[0013] The sintering trend and quantitative impact knowledge graph results corresponding to the real-time ore blending structure are comprehensively determined according to the present invention, including: Step 41: obtaining a real-time ore blending structure of a real-time sintering process and a set of historical ore blending structures of historical sintering processes, and performing a first comparison between the real-time ore blending structure and each historical ore blending structure in the set of historical ore blending structures; Step 42: extracting the historical ore matching structure with the highest degree of consistency with the real-time ore matching structure from the historical ore matching structure set based on the first comparison result as the first historical ore matching structure, and obtaining the first historical data point corresponding to the first historical ore matching, thereby obtaining the sintering trend influence knowledge graph corresponding to the first historical data point; Step 43: predicting the quantitative influence of the midpoint data corresponding to each sintering process parameter on the sintering target parameter based on the sintering target parameter prediction model, thereby establishing a sintering quantitative influence knowledge graph based on the quantitative influence of the midpoint data corresponding to the sintering process parameter data interval on the sintering target parameter as a relationship; Step 44: acquiring the real-time production status of the real-time sintering process and the historical production status set of the historical sintering process based on the sintering quantitative impact knowledge graph, and performing a second comparison between the real-time production status and each historical production status in the historical production status set; Step 45: extracting the historical production state with the highest degree of consistency with the real-time production state from the historical production state set based on the second comparison result as the first historical production state, and obtaining a second historical data point corresponding to the first historical production state; Step 46: obtaining a sintering trend influence knowledge graph corresponding to the first historical data point, and at the same time, obtaining a sintering quantification influence knowledge graph corresponding to the second historical data point; Step 47: Obtain the first influence weight of the sintering trend influence knowledge graph on the construction of the sintering dynamic knowledge graph, and at the same time obtain the second influence weight of the sintering quantitative influence knowledge graph on the construction of the sintering dynamic knowledge graph, combine the first influence weight with the sintering trend influence knowledge graph, and integrate the second influence weight with the sintering quantitative influence knowledge graph, so as to realize the integration of the sintering trend influence knowledge graph and the sintering quantitative influence knowledge graph as the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore matching structure.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: a method for constructing a sintering dynamic knowledge graph provided by the present invention collects, processes and analyzes sintering process parameters and sintering target parameters in real time through big data technology, thereby constructing a sintering trend knowledge graph and a sintering quantitative influence knowledge graph, which can realize the fusion of mechanism data and improve the system processing efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces 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 invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a flow chart of a method for constructing a sintering dynamic knowledge graph provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1: The embodiment of the present invention provides a method for constructing a sintering dynamic knowledge graph, such as Figure 1 As shown, including: Step 1: Collect the full chain data of the real-time sintering process, and classify the parameters to obtain the sintering target parameters and sintering process parameters; Step 2: Analyze the trend relationship between sintering target parameters and sintering process parameters based on the preset data analysis method, so as to establish a sintering trend influence knowledge map; Step 3: Classify the sintering process parameters based on the target ore blending cycle, and match the corresponding machine learning algorithm based on each classification result to build a sintering target parameter prediction model for the sintering process; Step 4: Obtain the sintering trend influence knowledge graph corresponding to the first historical data point with the highest degree of consistency with the real-time ore blending structure, and at the same time obtain the second historical data point with the highest degree of consistency with the current production status based on the sintering target parameter prediction model, obtain the sintering quantitative influence knowledge graph, and determine the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore blending structure; Among them, based on matching the corresponding machine learning algorithm with each classification result, a sintering target parameter prediction model for the sintering process is constructed, including: Extract the machine learning algorithm with the highest matching degree with the target sintering process from the machine learning algorithms corresponding to each classification result as the optimal algorithm; The historical sintering process parameters and historical sintering target parameters of the target sintering process are extracted from the sintering database, and the initial sintering target parameter prediction model is obtained by combining the optimal algorithm, so as to perform model verification and obtain the sintering target parameter prediction model of the target sintering process.

[0019] In this embodiment, sintering is a heat treatment process, which is usually used to heat powdered or granular materials (such as ore, ceramic raw materials, etc.) to a certain temperature and keep the temperature for a certain period of time, so that the particles are bonded or melted, thereby forming a solid whole. In the steel industry, sintering is the process of mixing raw materials such as iron ore, fuel, flux, etc. in a certain proportion and sintering them into blocks at high temperature.

[0020] In this embodiment, the full chain data refers to the data generated in all links of the entire sintering process, from raw material preparation to finished product output, including but not limited to the composition, ratio, temperature, pressure, flow rate and other parameters of the raw materials, as well as the operating status and energy consumption of the equipment.

[0021] In this embodiment, the sintering target parameters are target values ​​expected to be achieved through the sintering process, such as quality indicators of the sintered ore (such as strength, reducibility, particle size distribution, etc.), output, energy consumption, etc.

[0022] In this embodiment, the sintering process parameters refer to parameters that can be adjusted or monitored during the sintering process, such as temperature, time, atmosphere, pressure, raw material ratio, etc. These parameters directly affect the quality and performance of the sintered product.

[0023] In this embodiment, the trend relationship refers to a regular relationship between two or more variables that changes over time or space. During the sintering process, there may be a trend relationship between the sintering target parameters and the sintering process parameters, such as the strength of the sintered ore may increase when the temperature rises.

[0024] In this embodiment, the sintering trend influence knowledge graph is a graphical representation method for showing the trend relationship between various parameters in the sintering process. It can help people intuitively understand which factors have a significant impact on the sintering target parameters, as well as the influence mode and degree of these factors.

[0025] In this embodiment, the target ore blending cycle refers to a raw material blending plan within a period (such as one day, one week or one month) determined according to the production plan and raw material supply conditions. This plan is intended to ensure the stability of the sintering process and the quality of the product.

[0026] In this embodiment, the machine learning algorithm is a type of algorithm that can automatically learn from data and improve its performance. During the sintering process, the machine learning algorithm can be used to predict sintering target parameters, optimize raw material ratios, etc.

[0027] In this embodiment, the sintering target parameter prediction model is a mathematical model that uses a machine learning algorithm to predict the sintering target parameters according to the sintering process parameters. It can help people understand the possible results of the sintering process in advance, so as to make production adjustments and optimizations.

[0028] In this embodiment, the first historical data point refers to the historical data point that has the highest degree of consistency with the real-time ore blending structure. This data point may contain information such as raw material proportions and sintering process parameters similar to the current production conditions.

[0029] In this embodiment, the second historical data point refers to the historical data point that is obtained based on the sintering target parameter prediction model and has the highest degree of consistency with the current production status. This data point may provide a sintering target parameter prediction result similar to the current production conditions.

[0030] In this embodiment, the sintering quantitative impact knowledge graph is a knowledge graph that further quantifies the influence of each parameter on the sintering target parameter based on the sintering trend impact knowledge graph and combines the specific sintering target parameter prediction results. It can help people understand the influence of each parameter on the sintering process more accurately and make targeted production adjustments and optimizations.

[0031] In this embodiment, the optimal algorithm refers to the machine learning algorithm that has the highest matching degree with the target sintering process in the comparison results of the machine learning algorithms.

[0032] In this embodiment, the sintering database refers to a database storing historical data and experimental results of various parameters in the sintering process.

[0033] In this embodiment, the historical sintering process parameters and the historical sintering target parameters refer to the sintering parameters recorded and measured in the past sintering processes and stored in the sintering database.

[0034] In this embodiment, the initial sintering target parameter prediction model refers to a preliminary model obtained after training based on the optimal algorithm and the first training set, and its purpose is to predict the sintering target parameters based on given sintering process parameters.

[0035] In this embodiment, model validation refers to the process of evaluating the initial sintering target parameter prediction model using the first validation set. By comparing the results predicted by the model with the actual data in the validation set, the prediction performance of the model, such as prediction accuracy, stability, etc., can be evaluated.

[0036] In this embodiment, the sintering target parameter prediction model is an initial sintering target parameter prediction model that is confirmed to have good prediction performance after model verification.

[0037] The beneficial effect of the above technical solution is: through big data technology, the sintering process parameters and sintering target parameters are collected, processed and analyzed in real time, so as to construct a sintering trend knowledge graph and a sintering quantitative influence knowledge graph, which can realize the fusion of mechanism data and improve the system processing efficiency and accuracy.

[0038] Embodiment 2: Based on Example 1, the sintering target parameters and sintering process parameters are obtained, including: Step 11: collecting the whole chain data of the real-time sintering process, and processing the whole chain data to obtain the first processed data; Step 12: Obtain the data type of the first processed data, and classify the first processed data according to the data type and sintering production experience to obtain sintering target parameters and sintering process parameters.

[0039] In this embodiment, sintering is a heat treatment process, which is usually used to heat powdered or granular materials (such as ore, ceramic raw materials, etc.) to a certain temperature and keep the temperature for a certain period of time, so that the particles are bonded or melted, thereby forming a solid whole. In the steel industry, sintering is the process of mixing raw materials such as iron ore, fuel, flux, etc. in a certain proportion and sintering them into blocks at high temperature.

[0040] In this embodiment, the full chain data refers to the data generated in all links of the entire sintering process, from raw material preparation to finished product output, including but not limited to the composition, ratio, temperature, pressure, flow rate and other parameters of the raw materials, as well as the operating status and energy consumption of the equipment.

[0041] In this embodiment, the first processed data refers to the processed data obtained by performing steps such as data cleaning and data standardization on the collected full-chain data.

[0042] In this embodiment, data may be classified according to nature, format or purpose, and data types may include numerical data (such as specific values ​​of temperature and pressure), text data (such as device status description), time series data (such as changes in temperature over time), etc.

[0043] In this embodiment, the sintering production experience refers to the actual operating knowledge and experience accumulated during the sintering process. For example, the sintering production experience comes from long-term practice, expert knowledge or historical data.

[0044] In this embodiment, data classification refers to classifying the first processed data into different categories according to data type and sintering production experience.

[0045] In this embodiment, the sintering target parameters are target values ​​expected to be achieved through the sintering process, such as quality indicators of the sintered ore (such as strength, reducibility, particle size distribution, etc.), output, energy consumption, etc.

[0046] In this embodiment, the sintering process parameters refer to parameters that can be adjusted or monitored during the sintering process, such as temperature, time, atmosphere, pressure, raw material ratio, etc. These parameters directly affect the quality and performance of the sintered product.

[0047] The beneficial effect of the above technical solution is: through the real-time collection, processing and classification of big data technology, the sintering process parameters and sintering target parameters are obtained, so as to perform parameter analysis, build a sintering trend knowledge graph and a sintering quantitative influence knowledge graph, which can improve the system processing efficiency and accuracy.

[0048] Embodiment 3: Based on Example 2, a knowledge graph of sintering trend influence is established, including: Step 21: extracting a data analysis method matching the sintering target parameters and the sintering process parameters from a statistical analysis database based on the sintering characteristics of the ironmaking process to obtain a first data analysis method; Step 22: Analyze the parameter correlation between the sintering target parameters and the sintering process parameters based on the first data analysis method to obtain a first analysis result; Step 23: Obtain the data type of each data in the first processed data as a node of the sintering trend dynamic impact knowledge graph, and establish the sintering trend impact knowledge graph in combination with the first analysis result.

[0049] In this embodiment, the sintering characteristics refer to the changes in the physical and chemical properties of the material during this process, such as sintering temperature, sintering speed, sintering density, etc.

[0050] In this embodiment, the statistical analysis database is a database that stores a large amount of data, which is usually collected and processed by statistical analysis methods. In the field of ironmaking and sintering, the statistical analysis database contains historical data, experimental results, production records, etc. of various sintering parameters to support data analysis, model building and decision making.

[0051] In this embodiment, data analysis methods refer to a series of techniques and tools for analyzing data, extracting information, discovering patterns and making decisions. In the sintering process, data analysis methods may include statistical analysis, machine learning, data mining, etc., which are used to understand the relationship between sintering parameters, predict sintering results and optimize the sintering process.

[0052] In this embodiment, parameter correlation refers to the mutual influence and relationship between the sintering target parameters and the sintering process parameters. By analyzing the correlation of these parameters, the sintering process can be better understood, the sintering conditions can be optimized, and the product quality can be improved.

[0053] In this embodiment, the first data analysis method refers to a data analysis method that matches the sintering target parameters and sintering process parameters extracted from the statistical analysis database based on the sintering characteristics of the ironmaking process. In this embodiment, the first analysis result refers to the result obtained by analyzing the parameter correlation between the sintering target parameters and the sintering process parameters using the first data analysis method. For example, the first analysis result includes the correlation coefficient, significance level, trend analysis, etc. between the parameters.

[0054] In this embodiment, the sintering trend influence knowledge graph is a graphical representation method for showing the trend relationship between various parameters in the sintering process. It can help people intuitively understand which factors have a significant impact on the sintering target parameters, as well as the influence mode and degree of these factors.

[0055] The beneficial effect of the above technical solution is: by constructing a sintering trend knowledge graph, and then combining it with the sintering quantitative impact knowledge graph for comprehensive graph display, it is possible to achieve the fusion of mechanism data and improve display accuracy.

[0056] Embodiment 4: Based on Example 3, a sintering target parameter prediction model for the sintering process is constructed, including: Step 31: obtaining an ore blending cycle of a target sintering process, and classifying sintering process parameters and sintering target parameters based on the ore blending cycle to obtain a first cycle parameter set; Step 32: extracting a corresponding machine learning algorithm from an algorithm database based on parameter characteristics in each first cycle parameter subset in the first cycle parameter set, obtaining a first machine learning algorithm corresponding to each first cycle parameter subset, thereby obtaining a first machine learning algorithm set for the target sintering process; Step 33: Compare and analyze each first machine learning algorithm in the first machine learning algorithm set to determine the first machine learning algorithm with the highest matching degree with the target sintering process as the optimal algorithm, thereby obtaining a sintering target parameter prediction model for the target sintering process.

[0057] In this embodiment, the target ore blending cycle refers to a raw material blending plan within a period (such as one day, one week or one month) determined according to the production plan and raw material supply conditions. This plan is intended to ensure the stability of the sintering process and the quality of the product.

[0058] In this embodiment, the first cycle parameter set is a parameter set obtained by classifying the sintering process parameters and the sintering target parameters based on the ore blending cycle, and includes all relevant sintering parameters in a specific ore blending cycle.

[0059] In this embodiment, the algorithm database refers to a database that stores a variety of machine learning algorithms, which can be used for various tasks such as data analysis, pattern recognition, and prediction.

[0060] In this embodiment, the machine learning algorithm is a type of algorithm that can automatically learn from data and improve its performance. During the sintering process, the machine learning algorithm can be used to predict sintering target parameters, optimize raw material ratios, etc.

[0061] In this embodiment, the first machine learning algorithm set refers to an algorithm set consisting of machine learning algorithms extracted from an algorithm database according to parameter characteristics of each subset in the first period parameter set.

[0062] In this embodiment, the optimal algorithm refers to the machine learning algorithm in the first machine learning algorithm set that has the highest matching degree with the target sintering process.

[0063] In this embodiment, the sintering database refers to a database storing historical data and experimental results of various parameters in the sintering process.

[0064] In this embodiment, the historical sintering process parameters and the historical sintering target parameters refer to the sintering parameters recorded and measured in the past sintering processes and stored in the sintering database.

[0065] In this embodiment, the sintering target parameter prediction model is a mathematical model that uses a machine learning algorithm to predict the sintering target parameters according to the sintering process parameters. It can help people understand the possible results of the sintering process in advance, so as to make production adjustments and optimizations.

[0066] The beneficial effect of the above technical solution is: by constructing a sintering target parameter prediction model, the knowledge graph of sintering quantitative influence is determined, and the graph display of the target sintering process is comprehensively determined by combining the sintering trend knowledge graph, which can realize the fusion of mechanism data and improve data utilization, while improving the efficiency and accuracy of the knowledge graph.

[0067] Embodiment 5: Based on Example 4, the first machine learning algorithm with the highest matching degree with the target sintering process is determined as the optimal algorithm, including: Step 331: different ore blending structures are distinguished based on sintering production experience, and the first cycle parameters corresponding to each ore blending cycle are obtained, so as to obtain the trend influence between the sintering target parameters and the sintering process parameters determined by different machine learning algorithms in different ore blending cycles, and obtain a trend influence set; Step 332: determining the trend influence result corresponding to each sintering process parameter based on the trend influence set; If the trend influence of the target sintering process parameters and the corresponding sintering target parameters are consistent with the mechanism knowledge graph, the mechanism data matching index is increased by 1; If the trend influence of the target sintering process parameters and the corresponding sintering target parameters are inconsistent with the mechanism knowledge graph, the mechanism data matching index is reduced by 1, and at the same time, the method with a smaller mechanism data matching index is deleted; Step 333: determining a corresponding historical mechanism knowledge graph based on historical sintering process parameters and historical sintering target parameters of the target sintering process; Step 334: Determine the mechanism knowledge graph of the target sintering process based on the mean of the historical mechanism knowledge graph, so as to obtain the preset group of sintering process parameters that have the greatest impact on the sintering target parameters, and determine the optimal algorithm of the target sintering process based on the relative ranking result of each sintering process parameter corresponding to the first machine learning algorithm.

[0068] In this embodiment, the ore blending structure refers to the ratio and mixing method of the raw materials. Different ore blending structures will affect the results of the sintering process.

[0069] In this embodiment, the trend influence refers to the trend relationship between the sintering process parameters and the sintering target parameters that change over time or the ore blending cycle.

[0070] In this embodiment, the trend influence set includes a set of trend influences between sintering target parameters and sintering process parameters determined by different machine learning algorithms in different ore blending cycles.

[0071] In this embodiment, the mechanism knowledge graph is constructed based on the physical and chemical principles of the ironmaking and sintering processes, and is used to describe the relationship between parameters.

[0072] In this embodiment, the mechanism data matching index is an indicator used to measure the consistency between the trend impact results and the mechanism knowledge graph. When the two are consistent, the index increases; when they are inconsistent, the index decreases.

[0073] In this embodiment, the historical mechanism knowledge graph is a mechanism knowledge graph determined based on the historical sintering process parameters and the historical sintering target parameters of the target sintering process.

[0074] In this embodiment, the preset group of sintering process parameters refers to several groups of sintering process parameters that have the greatest impact on the sintering target parameters in the mechanism knowledge graph, wherein the preset groups can be three groups or five groups.

[0075] In this embodiment, the relative ranking result refers to the ranking result of each machine learning algorithm in the first machine learning algorithm set according to its performance in the prediction or classification task.

[0076] The beneficial effects of the above technical solution are: by determining the optimal algorithm of the target sintering process, a sintering target parameter prediction model is constructed, the sintering quantitative influence knowledge graph is determined, and the graph display of the target sintering process is comprehensively determined in combination with the sintering trend knowledge graph, which can improve the efficiency and accuracy of the knowledge graph.

[0077] Embodiment 6: Based on Example 5, the optimal algorithm for the target sintering process is determined based on the relative ranking result of each sintering process parameter corresponding to the first machine learning algorithm, including: Obtaining the sequence number corresponding to the relative ranking result of each first machine learning algorithm in the first machine learning algorithm set of the preset group of sintering process parameters that have the greatest impact on the sintering target parameters, to obtain a first sequence number set; Each serial number in the first serial number set is added together, and combined with the corresponding mechanism data matching index to obtain a corresponding mechanism data matching coefficient; The first machine learning algorithm with the smallest mechanism data matching coefficient is taken as the optimal algorithm.

[0078] In this embodiment, the first serial number set refers to a serial number set obtained by the serial numbers corresponding to the relative ranking results of each first machine learning algorithm in the first machine learning algorithm set for the preset group of sintering process parameters that have the greatest impact on the sintering target parameters. The serial number represents the position of the algorithm in the ranking. The smaller the serial number, the better the algorithm performs under the corresponding evaluation index.

[0079] In this embodiment, the mechanism data matching coefficient is a coefficient obtained by combining each sequence number in the first sequence number set (i.e., the position of the algorithm in the relative ranking) and the corresponding mechanism data matching index. This coefficient comprehensively considers the position of the algorithm in the ranking and its consistency with the mechanism knowledge graph, and is used to evaluate the comprehensive performance of the algorithm.

[0080] In this embodiment, the optimal algorithm refers to the machine learning algorithm with the best performance determined in the first machine learning algorithm set by comprehensively considering factors such as relative ranking results and mechanism data matching index.

[0081] The beneficial effects of the above technical solution are: by determining the optimal algorithm of the target sintering process, a sintering target parameter prediction model is constructed, the sintering quantitative influence knowledge graph is determined, and the graph display of the target sintering process is comprehensively determined in combination with the sintering trend knowledge graph, which can improve the efficiency and accuracy of the knowledge graph.

[0082] Embodiment 7: Based on Example 1, a sintering target parameter prediction model for a target sintering process is obtained, including: Extracting historical sintering process parameters and historical sintering target parameters of a target sintering process from a sintering database, and randomly extracting historical sintering process parameters and corresponding historical sintering target parameters as a first training set of the target sintering process, and taking the remaining historical sintering process parameters and corresponding historical sintering target parameters in the historical sintering process as a first validation set; The first training set is trained based on the optimal algorithm to obtain an initial sintering target parameter prediction model; The initial sintering target parameter prediction model is verified based on the first verification set to obtain the sintering target parameter prediction model.

[0083] In this embodiment, the first training set is a portion of historical sintering process parameters and corresponding historical sintering target parameters randomly extracted from the sintering database, and is used to train the initial sintering target parameter prediction model.

[0084] In this embodiment, the first validation set refers to the remaining historical sintering process parameters and corresponding historical sintering target parameters after extracting the first training set, which are used to verify the initial sintering target parameter prediction model to evaluate the prediction performance of the model.

[0085] In this embodiment, the initial sintering target parameter prediction model refers to a preliminary model obtained after training based on the optimal algorithm and the first training set, and its purpose is to predict the sintering target parameters based on given sintering process parameters.

[0086] In this embodiment, model validation refers to the process of evaluating the initial sintering target parameter prediction model using the first validation set. By comparing the results predicted by the model with the actual data in the validation set, the prediction performance of the model, such as prediction accuracy, stability, etc., can be evaluated.

[0087] In this embodiment, the sintering target parameter prediction model is an initial sintering target parameter prediction model that is confirmed to have good prediction performance after model verification.

[0088] The beneficial effect of the above technical solution is: by constructing a sintering target parameter prediction model, the knowledge graph of sintering quantitative influence is determined, and the graph display of the target sintering process is comprehensively determined by combining the sintering trend knowledge graph, which can realize the fusion of mechanism data and improve data utilization, while improving the efficiency and accuracy of the knowledge graph.

[0089] Embodiment 8: Based on Example 7, the sintering trend and quantitative impact knowledge graph results corresponding to the real-time ore matching structure are comprehensively determined, including: Step 41: obtaining a real-time ore blending structure of a real-time sintering process and a set of historical ore blending structures of historical sintering processes, and performing a first comparison between the real-time ore blending structure and each historical ore blending structure in the set of historical ore blending structures; Step 42: extracting the historical ore matching structure with the highest degree of consistency with the real-time ore matching structure from the historical ore matching structure set based on the first comparison result as the first historical ore matching structure, and obtaining the first historical data point corresponding to the first historical ore matching, thereby obtaining the sintering trend influence knowledge graph corresponding to the first historical data point; Step 43: predicting the quantitative influence of the midpoint data corresponding to each sintering process parameter on the sintering target parameter based on the sintering target parameter prediction model, thereby establishing a sintering quantitative influence knowledge graph based on the quantitative influence of the midpoint data corresponding to the sintering process parameter data interval on the sintering target parameter as a relationship; Step 44: acquiring the real-time production status of the real-time sintering process and the historical production status set of the historical sintering process based on the sintering quantitative impact knowledge graph, and performing a second comparison between the real-time production status and each historical production status in the historical production status set; Step 45: extracting the historical production state with the highest degree of consistency with the real-time production state from the historical production state set based on the second comparison result as the first historical production state, and obtaining a second historical data point corresponding to the first historical production state; Step 46: obtaining a sintering trend influence knowledge graph corresponding to the first historical data point, and at the same time, obtaining a sintering quantification influence knowledge graph corresponding to the second historical data point; Step 47: Obtain the first influence weight of the sintering trend influence knowledge graph on the construction of the sintering dynamic knowledge graph, and at the same time obtain the second influence weight of the sintering quantitative influence knowledge graph on the construction of the sintering dynamic knowledge graph, combine the first influence weight with the sintering trend influence knowledge graph, and integrate the second influence weight with the sintering quantitative influence knowledge graph, so as to realize the integration of the sintering trend influence knowledge graph and the sintering quantitative influence knowledge graph as the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore matching structure.

[0090] In this embodiment, the real-time ore blending structure is to mix different types of ores and additives in a certain proportion and manner during the sintering process according to the properties of the raw materials and the target product requirements. This real-time mixing proportion and method is called the real-time ore blending structure.

[0091] In this embodiment, the historical sintering process refers to the sintering process that has been completed in the past.

[0092] In this embodiment, the historical ore blending structure set is a set of ore blending structures used in each sintering process in the past multiple sintering processes.

[0093] In this embodiment, the first comparison is to compare the real-time ore distribution structure with each historical ore distribution structure in the historical ore distribution structure set to find out the historical ore distribution structure that is similar to the real-time ore distribution structure.

[0094] In this embodiment, the first historical ore distribution structure refers to a historical ore distribution structure in the set of historical ore distribution structures that has the highest degree of consistency with the real-time ore distribution structure.

[0095] In this embodiment, the first historical data point is associated with the first historical ore blending structure, and is a data point that records specific data or results of the sintering process under the ore blending structure.

[0096] In this embodiment, the midpoint data refers to one or more representative data points selected within the sintering process parameter data interval, which are used to quantitatively analyze the impact of these parameters on the sintering target parameters.

[0097] In this embodiment, the sintering quantitative impact knowledge graph is constructed by quantitatively analyzing the impact of sintering process parameters on sintering target parameters to describe and predict the relationship between various parameters in the sintering process.

[0098] In this embodiment, the real-time production status refers to the production status of the current sintering process, including various process parameters and operation results.

[0099] In this embodiment, the historical production status set is a set of production status recorded for each sintering in multiple sintering processes in the past.

[0100] In this embodiment, the second comparison is to compare the real-time production status with each historical production status in the historical production status set to find a historical production status similar to the real-time production status.

[0101] In this embodiment, the first historical production state is the historical production state in the historical production state set that has the highest degree of consistency with the real-time production state.

[0102] In this embodiment, the second historical data point is associated with the first historical production state, and is a data point that records specific data or results of the sintering process under the production state.

[0103] In this embodiment, the first influence weight and the second influence weight are determined based on the historical experience of the sintering trend influence knowledge graph and the sintering quantification influence knowledge graph.

[0104] In this embodiment, the sintering quantitative impact knowledge graph is a knowledge graph that further quantifies the influence of each parameter on the sintering target parameter based on the sintering trend impact knowledge graph and combines the specific sintering target parameter prediction results. It can help people understand the influence of each parameter on the sintering process more accurately and make targeted production adjustments and optimizations.

[0105] The beneficial effect of the above technical solution is: by constructing a sintering trend knowledge graph and a sintering quantitative influence knowledge graph, the graph display of the target sintering process can be comprehensively determined, which can improve the efficiency and accuracy of the knowledge graph and realize the fusion of mechanism data.

[0106] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a sintering dynamic knowledge graph, characterized in that: include: Step 1: Collect the full chain data of the real-time sintering process, and classify the parameters to obtain the sintering target parameters and sintering process parameters; Step 2: Analyze the trend relationship between sintering target parameters and sintering process parameters based on the preset data analysis method, so as to establish a sintering trend influence knowledge map; Step 3: Classify the sintering process parameters based on the target ore blending cycle, and match the corresponding machine learning algorithm based on each classification result to build a sintering target parameter prediction model for the sintering process; Step 4: Obtain the sintering trend influence knowledge graph corresponding to the first historical data point with the highest degree of consistency with the real-time ore blending structure, and at the same time obtain the second historical data point with the highest degree of consistency with the current production status based on the sintering target parameter prediction model, obtain the sintering quantitative influence knowledge graph, and determine the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore blending structure; Among them, based on matching the corresponding machine learning algorithm with each classification result, a sintering target parameter prediction model for the sintering process is constructed, including: Extract the machine learning algorithm with the highest matching degree with the target sintering process from the machine learning algorithms corresponding to each classification result as the optimal algorithm; The historical sintering process parameters and historical sintering target parameters of the target sintering process are extracted from the sintering database, and the initial sintering target parameter prediction model is obtained by combining the optimal algorithm, so as to perform model verification and obtain the sintering target parameter prediction model of the target sintering process.

2. A method for constructing a sintering dynamic knowledge graph according to claim 1, characterized in that: Collect the whole chain data of the real-time sintering process, classify the parameters, and obtain the sintering target parameters and sintering process parameters, including: Step 11: collecting the whole chain data of the real-time sintering process, and processing the whole chain data to obtain the first processed data; Step 12: Obtain the data type of the first processed data, and classify the first processed data according to the data type and sintering production experience to obtain sintering target parameters and sintering process parameters.

3. A method for constructing a sintering dynamic knowledge graph according to claim 2, characterized in that: Based on the preset data analysis method, the trend relationship between the sintering target parameters and the sintering process parameters is analyzed to establish a sintering trend impact knowledge map, including: Step 21: extracting a data analysis method matching the sintering target parameters and the sintering process parameters from a statistical analysis database based on the sintering characteristics of the ironmaking process to obtain a first data analysis method; Step 22: Analyze the parameter correlation between the sintering target parameters and the sintering process parameters based on the first data analysis method to obtain a first analysis result; Step 23: Obtain the data type of each data in the first processed data as a node of the sintering trend dynamic impact knowledge graph, and establish the sintering trend impact knowledge graph in combination with the first analysis result.

4. A method for constructing a sintering dynamic knowledge graph according to claim 3, characterized in that: The sintering process parameters are classified based on the target ore blending cycle, and the corresponding machine learning algorithm is matched based on each classification result to build a sintering target parameter prediction model for the sintering process, including: Step 31: obtaining an ore blending cycle of a target sintering process, and classifying sintering process parameters and sintering target parameters based on the ore blending cycle to obtain a first cycle parameter set; Step 32: extracting a corresponding machine learning algorithm from an algorithm database based on parameter characteristics in each first cycle parameter subset in the first cycle parameter set, obtaining a first machine learning algorithm corresponding to each first cycle parameter subset, thereby obtaining a first machine learning algorithm set for the target sintering process; Step 33: Compare and analyze each first machine learning algorithm in the first machine learning algorithm set to determine the first machine learning algorithm with the highest matching degree with the target sintering process as the optimal algorithm, thereby obtaining a sintering target parameter prediction model for the target sintering process.

5. A method for constructing a sintering dynamic knowledge graph according to claim 4, characterized in that: Comparing and analyzing each first machine learning algorithm in the first machine learning algorithm set, thereby determining the first machine learning algorithm with the highest matching degree with the target sintering process as the optimal algorithm, including: Step 331: different ore blending structures are distinguished based on sintering production experience, and the first cycle parameters corresponding to each ore blending cycle are obtained, so as to obtain the trend influence between the sintering target parameters and the sintering process parameters determined by different machine learning algorithms in different ore blending cycles, and obtain a trend influence set; Step 332: determining the trend influence result corresponding to each sintering process parameter based on the trend influence set; If the trend influence of the target sintering process parameters and the corresponding sintering target parameters are consistent with the mechanism knowledge graph, the mechanism data matching index is increased by 1; If the trend influence of the target sintering process parameters and the corresponding sintering target parameters are inconsistent with the mechanism knowledge graph, the mechanism data matching index is reduced by 1, and at the same time, the method with a smaller mechanism data matching index is deleted; Step 333: determining a corresponding historical mechanism knowledge graph based on historical sintering process parameters and historical sintering target parameters of the target sintering process; Step 334: Determine the mechanism knowledge graph of the target sintering process based on the mean of the historical mechanism knowledge graph, so as to obtain the preset group of sintering process parameters that have the greatest impact on the sintering target parameters, and determine the optimal algorithm of the target sintering process based on the relative ranking result of each sintering process parameter corresponding to the first machine learning algorithm.

6. A method for constructing a sintering dynamic knowledge graph according to claim 5, characterized in that: Determining an optimal algorithm for a target sintering process based on a relative ranking result of each sintering process parameter corresponding to the first machine learning algorithm includes: Obtaining the sequence number corresponding to the relative ranking result of each first machine learning algorithm in the first machine learning algorithm set of the preset group of sintering process parameters that have the greatest impact on the sintering target parameters, to obtain a first sequence number set; Each serial number in the first serial number set is added together, and combined with the corresponding mechanism data matching index to obtain a corresponding mechanism data matching coefficient; The first machine learning algorithm with the smallest mechanism data matching coefficient is taken as the optimal algorithm.

7. The method for constructing a sintering dynamic knowledge graph according to claim 1, characterized in that: The historical sintering process parameters and historical sintering target parameters of the target sintering process are extracted from the sintering database, and the initial sintering target parameter prediction model is obtained by combining the optimal algorithm, so as to perform model verification and obtain the sintering target parameter prediction model of the target sintering process, including: Extracting historical sintering process parameters and historical sintering target parameters of a target sintering process from a sintering database, and randomly extracting historical sintering process parameters and corresponding historical sintering target parameters as a first training set of the target sintering process, and taking the remaining historical sintering process parameters and corresponding historical sintering target parameters in the historical sintering process as a first validation set; The first training set is trained based on the optimal algorithm to obtain an initial sintering target parameter prediction model; The initial sintering target parameter prediction model is verified based on the first verification set to obtain the sintering target parameter prediction model.

8. A method for constructing a sintering dynamic knowledge graph according to claim 7, characterized in that: The sintering trend influence knowledge graph corresponding to the first historical data point with the highest degree of consistency with the real-time ore blending structure is obtained, and the second historical data point with the highest degree of consistency with the current production status is obtained based on the sintering target parameter prediction model to obtain the sintering quantitative influence knowledge graph, thereby comprehensively determining the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore blending structure, including: Step 41: obtaining a real-time ore blending structure of a real-time sintering process and a set of historical ore blending structures of historical sintering processes, and performing a first comparison between the real-time ore blending structure and each historical ore blending structure in the set of historical ore blending structures; Step 42: extracting the historical ore matching structure with the highest degree of consistency with the real-time ore matching structure from the historical ore matching structure set based on the first comparison result as the first historical ore matching structure, and obtaining the first historical data point corresponding to the first historical ore matching, thereby obtaining the sintering trend influence knowledge graph corresponding to the first historical data point; Step 43: predicting the quantitative influence of the midpoint data corresponding to each sintering process parameter on the sintering target parameter based on the sintering target parameter prediction model, thereby establishing a sintering quantitative influence knowledge graph based on the quantitative influence of the midpoint data corresponding to the sintering process parameter data interval on the sintering target parameter as a relationship; Step 44: acquiring the real-time production status of the real-time sintering process and the historical production status set of the historical sintering process based on the sintering quantitative impact knowledge graph, and performing a second comparison between the real-time production status and each historical production status in the historical production status set; Step 45: extracting the historical production state with the highest degree of consistency with the real-time production state from the historical production state set based on the second comparison result as the first historical production state, and obtaining a second historical data point corresponding to the first historical production state; Step 46: obtaining a sintering trend influence knowledge graph corresponding to the first historical data point, and at the same time, obtaining a sintering quantification influence knowledge graph corresponding to the second historical data point; Step 47: Obtain the first influence weight of the sintering trend influence knowledge graph on the construction of the sintering dynamic knowledge graph, and at the same time obtain the second influence weight of the sintering quantitative influence knowledge graph on the construction of the sintering dynamic knowledge graph, combine the first influence weight with the sintering trend influence knowledge graph, and integrate the second influence weight with the sintering quantitative influence knowledge graph, so as to realize the integration of the sintering trend influence knowledge graph and the sintering quantitative influence knowledge graph as the sintering trend and quantitative influence knowledge graph results corresponding to the real-time ore matching structure.

Images