Smelting Industry Knowledge Graph Management Method and System

Abstract

The present invention discloses a management method and system for a smelting industry knowledge graph, which relates to the technical field of the smelting industry. It analyzes historical smelting records, extracts several smelting condition analysis data groups, including smelting process performance factors and condition interpretation data, uses semantic analysis technology to extract condition keywords and interpretation keywords from the data groups, determines the semantic relationship between condition parameters and interpretation keywords; constructs a knowledge graph index structure based on the condition keywords and parameters, associates the interpretation data, and classifies the index structure and branches according to the equivalence of the interpretation keywords; analyzes the expandable range of condition parameters in the index branches of the same category to determine fuzzy features; The present invention effectively solves the problems of unstructured smelting data, lack of knowledge association, and insufficient management of dynamic fuzziness, improves the practicability and adaptability of the knowledge graph, and supports the optimization of the smelting process and intelligent decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of smelting industry, and particularly to a method and system for managing a knowledge graph of the smelting industry. Background Art

[0002] As an important part of the modern industrial system, the production process of the smelting industry involves complex physical and chemical reactions and multivariable control. A large number of historical records generated (such as workload data, process parameters, condition descriptions, etc.) contain rich empirical knowledge. However, traditional smelting data management methods mainly rely on manual records or simple database storage, and the data presents unstructured or semi-structured characteristics, making it difficult to be directly used for systematic analysis and knowledge mining. At the same time, there is a lack of an effective association mechanism between various performance factors involved in the smelting process (such as temperature, pressure, output fluctuations) and condition interpretations (such as equipment failures, process optimization suggestions), resulting in the difficulty of fully utilizing the hidden laws and experiences. In addition, the smelting industry has significant dynamic characteristics, and the workload and process parameters often fluctuate over time, showing a certain degree of ambiguity and uncertainty. Existing technologies often lack scientific quantification methods and extensible knowledge representation frameworks when dealing with these dynamic changes. This makes it difficult for smelting enterprises to achieve efficient process optimization and intelligent decision-making support based on historical data. Summary of the Invention

[0003] The purpose of the present invention is to provide a management method and system capable of generating a knowledge graph of the smelting industry.

[0004] The present invention discloses a method for managing a knowledge graph of the smelting industry, including:

[0005] Step S100: Analyze historical smelting records to determine a number of smelting condition analysis data groups, where each smelting condition analysis data group includes a number of smelting process performance factors and condition interpretation data;

[0006] Step S200: Use semantic analysis technology to extract condition keywords and interpretation keywords from the smelting condition analysis data groups, determine the condition parameters corresponding to the condition keywords, and determine the semantic pointing relationship of the interpretation keywords;

[0007] Step S300: Based on the condition keywords and corresponding condition parameters of the smelting condition analysis data groups, construct an index structure of the industrial knowledge graph, associate the index structure with the corresponding condition interpretation data, classify the index structure based on the equivalence of the interpretation keywords of the smelting condition analysis data groups, and classify the index branches of the index structure;

[0008] Step S400: Determine the expandable range of each condition parameter of the index branches of the same category, and then determine the fuzzy characteristics of the index branches.

[0009] In some embodiments disclosed by the present invention, the method for determining a number of smelting condition analysis data groups includes:

[0010] Step S101: Determine the change of smelting workload in the time dimension in the historical smelting records, construct a number of smelting workload curves, classify the similarity of each smelting workload curve, and obtain a number of smelting work curve groups;

[0011] Step S102: Based on each smelting work curve group, truncate and classify the corresponding historical smelting records to obtain a number of historical smelting record section groups. Perform semantic analysis on the historical smelting record sections in each historical smelting record section group, mark the smelting process performance description sections and the condition interpretation description sections respectively, and associate the smelting process performance description sections belonging to the same description type, which is recorded as the smelting process performance factor, and record the condition interpretation description section as the condition interpretation data.

[0012] In some embodiments disclosed by the present invention, the method for classifying the similarity of each smelting work curve includes:

[0013] Step S1011: Set an upper workload comparison line and a lower workload comparison line for each smelting workload curve. If the smelting workload curve is higher than the upper workload comparison line, intercept the corresponding curve segment, which is recorded as the upper curve segment. If the smelting workload curve is lower than the lower workload comparison line, intercept the corresponding curve segment, which is recorded as the lower curve segment. Determine the length of the upper curve segment of each upper curve segment and the length of the lower curve segment of each lower curve segment, and determine the interval between the upper curve segments and the interval between the lower curve segments, and determine the curve peak value of the upper curve segment and the lowest curve value of the lower curve segment;

[0014] Step S1012: Construct a curve segment parameter group sequence according to the appearance order of the upper curve segments and the lower curve segments. The curve segment parameter group is divided into an upper curve segment parameter group and a lower curve segment parameter group. The upper curve segment parameter group includes the length of the upper curve segment, the interval between the upper curve segments, and the curve peak value. The lower curve segment parameter group includes the length of the lower curve segment, the interval between the lower curve segments, and the lowest curve value;

[0015] Step S1013: Compare the curve segment parameter group sequences with each other. Each time of comparison, randomly displace the curve segment parameter group sequences and determine the initial similarity degree between them. Select the highest initial similarity degree as the reference similarity degree between the curve segment parameter group sequences;

[0016] Step S1014: Classify the smelting work curves corresponding to the curve segment parameter group sequences with a reference similarity degree greater than or equal to the preset value.

[0017] In some embodiments disclosed by the present invention, the method for determining the initial similarity degree includes:

[0018] Step S10131: Align the sequence of curve segment parameter groups, compare the difference features between each curve parameter group in the sequence. The difference features include the upper curve segment length difference, the lower curve segment length difference, the upper curve segment interval difference, the lower curve segment interval difference, the curve peak difference, and the curve lowest value difference. Based on the continuous performance of the difference features of all curve parameter groups, determine the initial similarity degree.

[0019] In some embodiments disclosed by the present invention, the method for constructing the index structure of the industrial knowledge graph based on the situation keywords and corresponding situation parameters of the smelting condition analysis data group includes:

[0020] Step S301: Compare the determined situation keywords with a preset word order template to determine the connection order relationship between the situation keywords, and use the situation keywords as index relationship nodes for relationship connection;

[0021] Step S302: Set index relationship nodes for the synonyms of the situation keywords, and perform parallel expansion connection with the index relationship nodes of the corresponding situation keywords to form a synonymous relationship node cluster, and connect the synonymous relationship node cluster with other sequential index relationship nodes;

[0022] Step S303: Configure the situation parameters on the corresponding index relationship nodes or synonymous relationship node clusters to form the index structure of the industrial knowledge graph.

[0023] In some embodiments disclosed by the present invention, the method for determining the equivalence of the interpretation keywords of the smelting condition analysis data group includes:

[0024] Step S304: Count the total number of keywords of all interpretation keywords between the smelting condition analysis data groups, determine the corresponding keyword quantities of the corresponding interpretation keywords, and calculate the corresponding keyword ratio between the corresponding keyword quantity and the total number of keywords;

[0025] Step S305: Judge whether the semantic references of the corresponding interpretation keywords between the smelting condition analysis data groups are equivalent, and determine the equivalent semantic reference ratio of the equivalent semantic reference quantity to all semantic reference quantities;

[0026] Step S306: Based on the corresponding keyword ratio and the equivalent semantic reference ratio of the smelting condition analysis data group, determine the equivalence degree of the interpretation keywords between the smelting condition analysis data groups. If the equivalence degree is greater than or equal to the preset value, it is determined that there is an equivalent relationship between the interpretation keywords of the smelting analysis data groups.

[0027] In some embodiments disclosed by the present invention, the method for determining the expandable range of each condition parameter of the index branches of the same category includes:

[0028] Step S401: Construct a condition parameter reference axis for the condition parameter, map the condition parameters of the same type to the condition parameter reference axis in the form of parameter mapping points, determine the proportion of parameter mapping points near each parameter mapping point, and based on the proportion of parameter mapping points near each parameter mapping point, determine the expandable range of the condition parameter corresponding to the parameter mapping point, where the proportion of parameter mapping points is the proportion of the number of parameter mapping points within the range near the parameter mapping point to the total number of parameter mapping points on the condition parameter reference axis.

[0029] In some embodiments disclosed by the present invention, the method for determining the expandable range of the condition parameter corresponding to the parameter mapping point includes:

[0030] Step S4011: Set several levels of range sections near the parameter mapping point, each range section is set with a comparison mapping point proportion, compare the proportion of parameter mapping points with the comparison mapping point proportion successively. If the proportion of parameter mapping points is greater than or equal to the comparison mapping point proportion, then compare the size relationship between the proportion of parameter mapping points in the next range section and the comparison mapping point proportion until the proportion of parameter mapping points is less than the comparison mapping point proportion, and then determine the expandable range of the parameter mapping point as the range section at this time.

[0031] In some embodiments disclosed by the present invention, a smelting industrial knowledge graph management system is also disclosed, including:

[0032] The first module is used to analyze historical smelting records to determine several smelting condition analysis data groups, and the smelting condition analysis data groups include several smelting process performance factors and condition interpretation data;

[0033] The second module is used to use semantic analysis technology to extract condition keywords and interpretation keywords from the smelting condition analysis data groups, determine the condition parameters corresponding to the condition keywords, and determine the semantic pointing relationship of the interpretation keywords;

[0034] The third module is used to construct an index structure of the industrial knowledge graph based on the condition keywords and corresponding condition parameters of the smelting condition analysis data groups, associate the index structure with the corresponding condition interpretation data, classify the index structure based on the equivalence of the interpretation keywords of the smelting condition analysis data groups, and classify the index branches of the index structure;

[0035] The fourth module is used to determine the expandable range of each condition parameter of the index branches of the same category, and further determine the fuzzy features of the index branches.

[0036] The present invention discloses a management method and system for a knowledge graph in the smelting industry, which relates to the technical field of smelting industry. By analyzing historical smelting records, a number of smelting condition analysis data groups are extracted, including smelting process performance factors and condition interpretation data. Using semantic analysis technology, condition keywords and interpretation keywords are extracted from the data groups, and the semantic relationships between condition parameters and interpretation keywords are determined; Based on the condition keywords and parameters, a knowledge graph index structure is constructed, the interpretation data is associated, and the index structure and branches are classified according to the equivalence of the interpretation keywords; Analyze the expandable range of condition parameters in the same category of index branches to determine fuzzy features; The present invention effectively solves the problems of unstructured smelting data, lack of knowledge association, and insufficient management of dynamic fuzziness, improves the practicability and adaptability of the knowledge graph, and supports the optimization of the smelting process and intelligent decision-making.

[0037] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Brief Description of the Drawings

[0038] Figure 1 It is a method step diagram of the management method of the knowledge graph in the smelting industry disclosed in the embodiment of the present invention. Detailed Embodiments

[0039] The technical solution of the present invention will be further described below through the accompanying drawings and embodiments.

[0040] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and should not be construed as limiting the protection scope of the present invention. Those skilled in the art can make some non-essential improvements and adjustments based on the content of the present invention described below. In the present invention, unless otherwise clearly specified and limited, the technical terms used in the present invention should have the ordinary meaning understood by those skilled in the art of the present invention.

[0041] Embodiment:

[0042] The present invention discloses a management method for a knowledge graph in the smelting industry. Refer to Figure 1 , including:

[0043] Step S100, analyze historical smelting records to determine a number of smelting condition analysis data groups, and the smelting condition analysis data groups include a number of smelting process performance factors and condition interpretation data.

[0044] In the smelting industry, historical records usually exist in the form of logs, reports, or equipment monitoring data. These data are often unstructured or semi-structured, containing a large amount of text descriptions, numerical records, and time series information. The goal of step S100 is to transform these scattered data into a structured "smelting condition analysis data set" through systematic analysis means, providing basic materials for subsequent knowledge graph construction. Its core principle is based on data mining and information extraction technologies. First, it is necessary to preprocess the historical records, such as removing noise data and unifying the format, and then identify key information through statistical analysis or pattern recognition methods. Specifically, the analysis process will focus on two main outputs: one is the "smelting process performance factors", which are key indicators reflecting the state of the smelting process extracted from the records, and may include quantitative data such as temperature, pressure, and output rate, or qualitative descriptions such as "slag accumulation" and "equipment vibration"; the other is the "condition interpretation data", which is the explanatory information for the performance factors, such as diagnostic conclusions like "high temperature leads to a decrease in furnace efficiency" or "abnormal pressure is due to pipeline blockage". The technical means to achieve this step may include entity recognition in natural language processing (NLP) to extract factors from text, and time series analysis to mine trends and outliers from numerical data. In addition, in order to ensure the effectiveness of the data set, screening criteria may need to be set, such as data integrity or representativeness, so as to divide the massive records into several representative subsets. The logic of this division is similar to feature engineering in machine learning, abstracting the original data into more easily processed information units through dimensionality reduction and clustering methods. The principle of the whole process is to extract structured knowledge fragments from complex industrial data, laying a foundation for subsequent semantic analysis and graph construction, and ultimately realizing the transformation from "data" to "information".

[0045] Step S200: Using semantic analysis technology, extract condition keywords and interpret keywords from the smelting condition analysis data set, determine the condition parameters corresponding to the condition keywords, and determine the semantic pointing relationship of the interpret keywords.

[0046] Step S200 is the core link for in-depth semantic parsing of the smelting condition analysis data group generated in S100. Its principle is based on natural language processing (NLP) and semantic network construction. This process aims to further extract key semantic elements from structured data and establish the associations between them, providing a semantic foundation for the construction of the knowledge graph. First, semantic analysis techniques (such as word frequency statistics, word segmentation, and dependency syntax analysis) are used to process the "smelting process performance factors" and "condition interpretation data" in the data group, extracting two types of keywords: one is the "condition keywords", such as "high temperature", "blockage", "molten liquid flow rate", which directly reflect the core states or problems in the smelting process, and through context analysis, the corresponding "condition parameters" can be further determined, such as the specific temperature value associated with "high temperature" (e.g., 1350 °C) or the pressure difference corresponding to "blockage" (e.g., 0.5 MPa); the other is the "interpretation keywords", such as "reason for overheating", "efficiency decline", which are extracted from the condition interpretation data and are used to describe the reasons or results behind the states. After extracting the keywords, the technical focus shifts to the determination of semantic relationships, which requires the use of word vector models (such as Word2Vec) or knowledge graph embedding techniques (such as TransE) to analyze the semantic pointing relationships between the interpretation keywords and the condition keywords, such as "reason for overheating" pointing to "equipment aging" or "efficiency decline" pointing to "high temperature". The principle of this process is to transform discrete text information into a computable structured relationship network through semantic similarity and syntactic dependence. In addition, to improve accuracy, it may be necessary to combine domain dictionaries or expert knowledge to correct and supplement the extraction results. The implementation of the entire step relies on information extraction and relationship extraction techniques in NLP, aiming to make the implicit knowledge in the data group explicit, providing semantic support for the subsequent graph index structure, and ultimately realizing the transition from "information" to "knowledge".

[0047] In step S300, based on the condition keywords and the corresponding condition parameters of the smelting condition analysis data group, an index structure of the industrial knowledge graph is constructed, and the index structure is associated with the corresponding condition interpretation data. The index structure is classified based on the equivalence of the interpretation keywords of the smelting condition analysis data group, and the index branches of the index structure are classified.

[0048] Step S300 is the core link in constructing the knowledge graph of the smelting industry. Its principle is based on graph theory and semantic clustering techniques in knowledge representation, aiming to organize the semantic elements extracted in S200 into a hierarchical and queryable knowledge system. Specifically, first, an index structure of the graph is constructed based on the "condition keywords" and their corresponding "condition parameters". In this process, the keywords are regarded as nodes, and the parameters are regarded as attributes of the nodes. The nodes are connected into a network through semantic relationships (such as causality, chronological order). For example, "high temperature (1350°C)" is connected to "efficiency decline" to form a basic relationship graph. At the same time, the index structure will be associated with the corresponding "condition interpretation data", such as "the reason for the efficiency decline caused by high temperature is equipment aging". These interpretation data are attached to the nodes or edges as supplementary information to enhance the interpretability of the graph. After the construction is completed, the optimization and classification of the graph become crucial. The index structure is sorted out using the "equivalence of interpretation keywords", which requires judging the meaning overlap between keywords through semantic similarity analysis (such as cosine similarity). For example, "overheat" and "high temperature" may be recognized as having the same meaning, so the relevant nodes are clustered into one category. The principle of this classification is similar to clustering algorithms (such as K-means), which improves the simplicity of the graph by reducing redundancy. In addition, further classification is carried out on the "index branches" (i.e., sub-graph structures in the graph), which may be based on process stages (such as smelting, refining) or problem types (such as equipment failures, process anomalies) to form a hierarchical index system. The implementation of the entire step depends on the storage and query capabilities of a graph database (such as Neo4j) and classification techniques in semantic analysis to ensure that the graph can not only reflect the complex relationships in the smelting process but also be convenient for quick retrieval and expansion, ultimately achieving the leap from "knowledge" to "system".

[0049] In step S400, the expandable range of each condition parameter of the index branches of the same category is determined, and then the fuzzy features of the index branches are determined.

[0050] Step S400 is a key link for the dynamic and uncertainty management of the knowledge graph. Its principle integrates statistical analysis and fuzzy logic, aiming to enhance the adaptability of the graph to the complexity of the smelting process. In S300, the index branches of the graph have been classified by category, and each branch contains several condition parameters (such as temperature, pressure value). However, these parameters are often not fixed in the actual industrial scenario, but show a certain fluctuation range over time or conditions. Therefore, in this step, each condition parameter in the same category branch is first analyzed by statistical methods. For example, the distribution of "temperature" in historical data is fitted to determine its "expandable range", such as 1350°C ± 50°C, which reflects the possible change boundary of the parameter. Probability distribution models (such as normal distribution) or interval estimation techniques may be used to ensure that the range has statistical significance. Next, "fuzzy features" are further extracted based on these ranges, which is achieved through fuzzy mathematics. For example, temperature is divided into fuzzy states such as "normal", "slightly high", "too high", etc., and the possibility of each state is quantified by a membership function (such as "the membership degree of 1350°C belonging to too high is 0.8"). The principle of this fuzzification is to recognize the uncertainty in the smelting process and convert it into computable features, so that the graph can not only describe static facts but also reflect dynamic trends. The implementation of the entire step may combine data visualization (such as drawing parameter distribution maps) and fuzzy inference systems, and finally generate fuzzy feature descriptions for each index branch. For example, "the fuzzy feature of the high-temperature branch is that the probability of slightly high temperature is 80%". Through this process, the graph evolves from a static knowledge network into a dynamic decision support tool, significantly enhancing its flexibility and practicality in actual applications.

[0051] In some embodiments disclosed in the present invention, the method for determining several groups of smelting condition analysis data includes:

[0052] Step S101, determining the change of the smelting workload in the time dimension in the historical smelting records, constructing several smelting workload curves, and classifying the similarity of each smelting workload curve to obtain several groups of smelting work curves.

[0053] The purpose of step S101 is to extract the workload change characteristics in the time dimension from historical smelting records and classify them through similarity analysis, providing a data basis for subsequent condition analysis. Its principle is based on time series analysis and clustering techniques, aiming to mine regular patterns from complex industrial data. Specifically, first, the key indicator of "smelting workload" needs to be extracted from historical records. The workload may be the output (such as tons per hour), energy consumption (such as power consumption), or equipment operation load (such as rotational speed). These data are usually recorded in the form of timestamps, forming a continuous or discrete time series. The process of determining workload changes involves data cleaning (such as filling missing values, smoothing noise) and feature extraction (such as calculating trends, periodicity), and then these time series are transformed into visual "smelting workload curves", where each curve represents the dynamic change of workload over a period of time.

[0054] In step S102, based on each group of smelting work curves, the corresponding historical smelting records are truncated and classified to obtain several groups of historical smelting record segments. Semantic analysis is performed on the historical smelting record segments in each group of historical smelting record segments. The smelting process performance description segments and the condition interpretation description segments are respectively marked, and the smelting process performance description segments belonging to the same description type are associated and recorded as smelting process performance factors, and the condition interpretation description segments are recorded as condition interpretation data.

[0055] Step S102 further processes the smelting working curve group generated in S101. Its goal is to segment the historical records according to curve characteristics and extract key information, ultimately forming a structured "smelting condition analysis data group". Its principle combines time series segmentation and natural language processing (NLP) techniques, aiming to mine the semantic content of process performance and interpretation from the data. Specifically, first, based on the characteristics of each smelting working curve group (such as the time span or change trend of the curve), the corresponding historical smelting records are truncated by time period. For example, a curve showing "rising workload - stable" may correspond to the records of a certain day, and this section of the record is intercepted as an independent section. In this way, the original records are divided into several "historical smelting record section groups", each section group corresponding to a working curve group, retaining the relevance in the time dimension. Next, semantic analysis is performed on the record sections in each section group. This process relies on NLP techniques (such as named entity recognition, syntactic analysis) to identify and mark two types of text fragments: one is the "smelting process performance description section", such as "temperature rises to 1400°C" or "slag accumulation is obvious", which reflect the process state; the other is the "condition interpretation description section", such as "high temperature leads to reduced efficiency" or "accumulation is due to insufficient cooling", which provides explanations or reasons for the state. To structure this information, it is necessary to associate the performance description sections of the same type. For example, all descriptions mentioning "temperature" are grouped together and denoted as a "smelting process performance factor" (such as "high temperature factor"), while the interpretation description sections are directly marked as "condition interpretation data" (such as "reason for overheating").

[0056] In some embodiments disclosed in the present invention, the method for classifying the similarity of each smelting working curve includes:

[0057] Step S1011, set an upper workload comparison line and a lower workload comparison line for each smelting workload curve. If the smelting workload curve is higher than the upper workload comparison line, then intercept the corresponding curve segment and denote it as the upper curve segment. If the smelting workload curve is lower than the lower workload comparison line, then intercept the corresponding curve segment and denote it as the lower curve segment. Determine the length of each upper curve segment and the length of each lower curve segment, and determine the interval between upper curve segments and the interval between lower curve segments, and determine the curve peak value of the upper curve segment and the lowest curve value of the lower curve segment.

[0058] Step S1012, construct a curve segment parameter group sequence according to the order in which the upper curve segments and lower curve segments appear. The curve segment parameter group is divided into an upper curve segment parameter group and a lower curve segment parameter group. The upper curve segment parameter group includes the length of the upper curve segment, the interval between upper curve segments, and the curve peak value. The lower curve segment parameter group includes the length of the lower curve segment, the interval between lower curve segments, and the lowest curve value.

[0059] Step S1013: Compare the sequence of curve segment parameter groups with each other. Each time when comparing, randomly displace the sequence of curve segment parameter groups, and determine the initial similarity degree between them. Select the highest initial similarity degree as the reference similarity degree between the sequences of curve segment parameter groups.

[0060] Step S1014: Classify the smelting working curves corresponding to the sequences of curve segment parameter groups whose reference similarity degree is greater than or equal to the preset value.

[0061] In some embodiments disclosed by the present invention, the method for determining the initial similarity degree includes:

[0062] Step S10131: Align the sequence of curve segment parameter groups, and compare the difference features between each curve parameter group in the sequence. The difference features include the difference amount of the upper curve segment length, the difference amount of the lower curve segment length, the difference amount of the upper curve segment interval, the difference amount of the lower curve segment interval, the difference amount of the curve peak value, and the difference amount of the curve lowest value. Based on the continuous performance of the difference features of all curve parameter groups, determine the initial similarity degree.

[0063] Among them, the expression for calculating the initial similarity degree is:

[0064] .

[0065] Among them, S is the initial similarity degree, is a continuous performance determination function, which is used to determine whether the difference features between the i-th curve segment parameter group and the adjacent curve segment parameter groups in the sequence of curve segment parameter groups are equivalent. The determination method includes calculating the sub-similarity degree between the i-th curve segment parameter groups. If the sub-similarity degree is greater than or equal to the preset value, it is determined that the i-th curve segment parameter groups are similar. If the (i - 1)-th curve segment parameter groups are similar, or the (i + 1)-th curve segment parameter groups are similar, then output 1, otherwise output 0. L is the continuous performance influence adjustment coefficient, and C is the continuous performance influence adjustment constant;

[0066] Among them, the expression for calculating the sub-similarity degree between the i-th curve segment parameter groups is:

[0067] .

[0068] Among them, is the sub-similarity degree between the i-th curve segment parameter groups, is the preset maximum curve segment length difference amount, is the curve segment length difference amount, is the preset maximum curve segment interval difference amount, is the curve segment interval difference amount, is the preset maximum curve peak difference or the maximum curve minimum value difference, is the curve peak difference or the curve minimum value difference, is the influence adjustment coefficient of the curve segment length difference, is the influence adjustment coefficient of the curve segment interval difference, and b is the influence adjustment constant of the difference.

[0069] In some embodiments disclosed by the present invention, the method for constructing the index structure of the industrial knowledge graph based on the condition keywords and the corresponding condition parameters of the smelting condition analysis data group includes:

[0070] Step S301: Compare the determined condition keywords with the preset word order template to determine the connection order relationship between the condition keywords, and use the condition keywords as the index relationship nodes for relationship connection.

[0071] Step S302: Set index relationship nodes for the synonyms of the condition keywords, and perform parallel expansion connection with the index relationship nodes of the corresponding condition keywords to form a synonym relationship node cluster, and connect the synonym relationship node cluster with other index relationship nodes in sequence.

[0072] Step S303: Configure the condition parameters to the corresponding index relationship nodes or synonym relationship node clusters to form the index structure of the industrial knowledge graph.

[0073] In some embodiments disclosed by the present invention, the method for determining the equivalence of the interpretation keywords of the smelting condition analysis data group includes:

[0074] Step S304: Count the total number of keywords of all the interpretation keywords between the smelting condition analysis data groups, determine the corresponding keyword quantity of the corresponding interpretation keywords, and calculate the corresponding keyword ratio between the corresponding keyword quantity and the total number of keywords.

[0075] Step S305: Judge whether the semantic pointers of the corresponding interpretation keywords between the smelting condition analysis data groups are equivalent, and determine the equivalent semantic pointer ratio of the equivalent semantic pointer quantity to all the semantic pointer quantities.

[0076] Step S306: Based on the corresponding keyword ratio and the equivalent semantic pointer ratio of the smelting condition analysis data group, determine the equivalence degree of the interpretation keywords between the smelting condition analysis data groups. If the equivalence degree is greater than or equal to the preset value, it is determined that there is an equivalent relationship between the interpretation keywords of the smelting analysis data groups.

[0077] Among them, the expression for calculating the equivalence degree is:

[0078] .

[0079] Among them, D is the degree of equivalence, is the corresponding keyword ratio, is the equivalent semantic pointing ratio, is the equivalent weight coefficient of the corresponding keyword ratio, is the equivalent weight coefficient of the equivalent semantic pointing ratio.

[0080] In some embodiments disclosed by the present invention, the method for determining the expandable range of each condition parameter of the index branches of the same category includes:

[0081] Step S401: Construct a condition parameter reference axis for the condition parameter, map the condition parameters of the same type to the condition parameter reference axis in the form of parameter mapping points, determine the parameter mapping point ratio near each parameter mapping point, and based on the parameter mapping point ratio near each parameter mapping point, determine the expandable range of the condition parameter corresponding to the parameter mapping point, where the parameter mapping point ratio is the ratio of the number of parameter mapping points within the range near the parameter mapping point to the total number of parameter mapping points on the condition parameter reference axis.

[0082] The purpose of step S401 is to determine the expandable range of the condition parameters (such as temperature, pressure) in the index branches of the same category, so as to quantify the dynamic changes and uncertainties of these parameters during the smelting process. Its principle integrates the ideas of statistical analysis, data visualization, and spatial distribution, providing a basis for subsequent fuzzy feature extraction. Specifically, first, a "condition parameter reference axis" is constructed for a certain type of condition parameter (such as "temperature"). This is an abstract numerical axis, similar to a one-dimensional coordinate system, with a range covering the minimum value to the maximum value that the parameter may appear (for example, the temperature axis may be from 0°C to 2000°C). The construction of this reference axis is based on historical data statistics, and the upper and lower limits may be determined by analyzing the distribution of this parameter in all relevant records. Next, the condition parameters of the same type (for example, multiple temperature values recorded in a certain category branch, such as 1350°C, 1370°C, 1340°C) are projected onto the reference axis in the form of "parameter mapping points", and each mapping point corresponds to a specific value, forming a set of distribution points. This process is similar to the drawing of a scatter plot in data visualization, aiming to convert discrete parameter values into a spatial representation form.

[0083] In some embodiments disclosed by the present invention, the method for determining the expandable range of the condition parameter corresponding to the parameter mapping point includes:

[0084] Step S4011: Set several levels of range sections near the parameter mapping point. Each range section is set with a comparison mapping point ratio. Compare the parameter mapping point ratio with the comparison mapping point ratio successively. If the parameter mapping point ratio is greater than or equal to the comparison mapping point ratio, then compare the size relationship between the parameter mapping point ratio of the next range section and the comparison mapping point ratio until the parameter mapping point ratio is less than the comparison mapping point ratio. Then, recognize the range section at this time as the expansion range of the parameter mapping point.

[0085] In some embodiments disclosed by the present invention, a smelting industrial knowledge graph management system is also disclosed, including:

[0086] A first module, configured to analyze historical smelting records to determine several smelting condition analysis data groups. The smelting condition analysis data groups include several smelting process performance factors and condition interpretation data;

[0087] A second module, configured to use semantic analysis technology to extract condition keywords and interpretation keywords from the smelting condition analysis data groups, determine the condition parameters corresponding to the condition keywords, and determine the semantic pointing relationship of the interpretation keywords;

[0088] A third module, configured to construct an index structure of the industrial knowledge graph based on the condition keywords and corresponding condition parameters of the smelting condition analysis data groups, associate the index structure with the corresponding condition interpretation data, classify the index structure based on the equivalence of the interpretation keywords of the smelting condition analysis data groups, and classify the index branches of the index structure;

[0089] A fourth module, configured to determine the expandable range of each condition parameter of the index branches of the same category, and then determine the fuzzy features of the index branches.

[0090] The present invention discloses a smelting industrial knowledge graph management method and system, relating to the technical field of smelting industry. Analyze historical smelting records, extract several smelting condition analysis data groups, including smelting process performance factors and condition interpretation data, adopt semantic analysis technology to extract condition keywords and interpretation keywords from the data groups, determine the condition parameters and the semantic relationship of the interpretation keywords; construct a knowledge graph index structure based on the condition keywords and parameters, associate the interpretation data, and classify the index structure and branches according to the equivalence of the interpretation keywords; analyze the expandable range of the condition parameters in the index branches of the same category to determine the fuzzy features; the present invention effectively solves the problems of unstructured smelting data, lack of knowledge association, and insufficient management of dynamic fuzziness, improves the practicability and adaptability of the knowledge graph, and supports smelting process optimization and intelligent decision-making.

[0091] Through the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented by hardware or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various implementation scenarios of the present invention.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A knowledge graph management method for the smelting industry, characterized in that: include: Step S100, analyzing historical smelting records to determine a number of smelting status analysis data groups, the smelting status analysis data groups including a number of smelting process performance factors and status interpretation data; Step S200, using semantic analysis technology, extracting condition keywords and interpretation keywords from the smelting condition analysis data group, determining condition parameters corresponding to the condition keywords, and determining the semantic pointing relationship of the interpretation keywords; Step S300, based on the condition keywords and corresponding condition parameters of the smelting condition analysis data group, construct an index structure of the industrial knowledge graph, associate the index structure with the corresponding condition interpretation data, classify the index structure based on the equality of the interpretation keywords of the smelting condition analysis data group, and classify the index branches of the index structure; Step S400, determining the expandable range of each status parameter of the same category of index branches, and then determining the fuzzy features of the index branches.

2. The method for managing knowledge graphs for smelting industry according to claim 1, characterized in that: Methods for determining a number of metallurgical condition analysis data sets include: Step S101, determining the change of smelting workload in the time dimension in the historical smelting records, and constructing a plurality of smelting workload curves, and classifying each smelting workload curve by similarity to obtain a plurality of smelting workload curve groups; Step S102, based on each smelting work curve group, the corresponding historical smelting records are truncated and classified to obtain several historical smelting record segment groups, and semantic analysis is performed on the historical smelting record segments in each historical smelting record segment group, and the smelting process performance description segments and the status interpretation description segments are marked respectively, and the smelting process performance description segments belonging to the same description type are associated and recorded as smelting process performance factors, and the status interpretation description segments are recorded as status interpretation data.

3. The smelting industry knowledge graph management method according to claim 2 is characterized in that: The methods for similarity classification of each smelting working curve include: Step S1011, an upper workload comparison line and a lower workload comparison line are set for each smelting workload curve. If the smelting workload curve is higher than the upper workload comparison line, the corresponding curve segment is intercepted and recorded as the upper curve segment. If the smelting workload curve is lower than the lower workload comparison line, the corresponding curve segment is intercepted and recorded as the lower curve segment. The upper curve segment length of each upper curve segment and the lower curve length of the lower curve segment are determined, and the upper curve segment interval between the upper curve segments and the lower curve segment interval between the lower curve segments are determined, and the curve peak value of the upper curve segment and the curve minimum value of the lower curve segment are determined; Step S1012, constructing a curve segment parameter group sequence according to the order in which the upper curve segment and the lower curve segment appear, wherein the curve segment parameter group is divided into an upper curve segment parameter group and a lower curve segment parameter group, wherein the upper curve segment parameter group includes an upper curve segment length, an upper curve segment interval, and a curve peak value, and the lower curve segment parameter group includes a lower curve segment length, a lower curve segment interval, and a curve minimum value; Step S1013, comparing the curve segment parameter group sequences with each other. During each comparison, the curve segment parameter group sequences are randomly displaced, and the initial similarity between them is determined. The highest initial similarity is selected as the reference similarity between the curve segment parameter group sequences. Step S1014, classifying the smelting working curves corresponding to the curve segment parameter group sequence with a reference similarity greater than or equal to a preset value.

4. The method for managing knowledge graphs for smelting industry according to claim 3, characterized in that: Methods for determining the initial degree of similarity include: Step S10131, align the curve segment parameter group sequence, compare the difference features between each curve parameter group in the sequence, the difference features include the difference in upper curve segment length, the difference in lower curve segment length, the difference in upper curve segment interval, the difference in lower curve segment interval, the difference in curve peak value and the difference in curve minimum value, and determine the initial similarity based on the continuous performance of the difference features of all curve parameter groups.

5. The method for managing knowledge graphs for smelting industry according to claim 1, characterized in that: Based on the condition keywords and corresponding condition parameters of the smelting condition analysis data set, the method for constructing the index structure of the industrial knowledge graph includes: Step S301, comparing the determined situation keywords with the preset word order template, determining the connection sequence relationship between the situation keywords, and using the situation keywords as index relationship nodes for relationship connection; Step S302, setting an index relationship node for the synonym of the condition keyword, and performing parallel expansion connection with the index relationship node corresponding to the condition keyword to form a synonym relationship node cluster, and connecting the synonym relationship node cluster with other sequential index relationship nodes; Step S303, configure the status parameters to the corresponding index relationship nodes or synonymous relationship node clusters to form the index structure of the industrial knowledge graph.

6. The method for managing knowledge graphs for smelting industry according to claim 1, characterized in that: Methods for determining equivalence of interpretation keywords for a metallurgical analysis data set include: Step S304, summing the total number of all interpretation keywords between the smelting condition analysis data groups, determining the corresponding number of keywords of the corresponding interpretation keywords, and calculating the corresponding keyword ratio between the number of corresponding keywords and the total number of keywords; Step S305, determining whether the semantic orientations of the corresponding interpretation keywords between the smelting condition analysis data groups are identical, and determining the ratio of the number of identical semantic orientations to the number of identical semantic orientations; Step S306, based on the corresponding keyword ratio and equivalent semantic pointing ratio of the smelting condition analysis data group, determine the degree of similarity of the interpretation keywords between the smelting condition analysis data groups. If the degree of similarity is greater than or equal to the preset value, it is determined that there is an equivalent relationship between the interpretation keywords between the smelting analysis data groups.

7. The method for managing knowledge graphs for smelting industry according to claim 1, characterized in that: The method for determining the expandable range of each status parameter of the same category of index branches includes: Step S401, construct a status parameter reference axis for the status parameter, and map the same type of status parameters to the status parameter reference axis in the form of parameter mapping points, determine the ratio of parameter mapping points near each parameter mapping point, and based on the ratio of parameter mapping points near each parameter mapping point, determine the status parameter extension range corresponding to the parameter mapping point, wherein the parameter mapping point ratio is the ratio of the number of parameter mapping points in the range near the parameter mapping point to the number of all parameter mapping points on the status parameter reference axis.

8. The method for managing knowledge graphs for smelting industry according to claim 7, characterized in that: The method for determining the extended range of the status parameter corresponding to the parameter mapping point includes: Step S4011, set several levels of range segments near the parameter mapping point, each range segment is set with a comparison mapping point ratio, and compare the parameter mapping point ratio with the comparison mapping point ratio in sequence. If the parameter mapping point ratio is greater than or equal to the comparison mapping point ratio, compare the size relationship between the parameter mapping point ratio and the comparison mapping point ratio of the next range segment, until the parameter mapping point ratio is less than the comparison mapping point ratio, then the range segment at this time is identified as the extended range of the parameter mapping point.

9. The knowledge graph management system for the smelting industry is characterized by: A method for managing a knowledge graph for a smelting industry according to any one of claims 1 to 8, comprising: The first module is used to analyze the historical smelting records and determine a number of smelting status analysis data groups, which include a number of smelting process performance factors and status interpretation data; The second module is used to extract condition keywords and interpretation keywords from the smelting condition analysis data group using semantic analysis technology, determine the condition parameters corresponding to the condition keywords, and determine the semantic pointing relationship of the interpretation keywords; The third module is used to construct an index structure of the industrial knowledge graph based on the condition keywords and corresponding condition parameters of the smelting condition analysis data group, associate the index structure with the corresponding condition interpretation data, classify the index structure based on the equality of the interpretation keywords of the smelting condition analysis data group, and classify the index branches of the index structure; The fourth module is used to determine the expandable range of each status parameter of the index branch of the same category, and then determine the fuzzy characteristics of the index branch.

Images