Knowledge graph-based method for determining composition and process parameters of a strip

Abstract

The present application relates to the field of material processing manufacturing and intelligent technology, and provides a strip steel composition and process parameter determination method based on a knowledge graph, the method comprising: S1: collecting strip steel data and performing data preprocessing; S2: constructing a strip steel ontology model; S3: obtaining strip steel triple data using a mapping language and constructing a strip steel knowledge graph; S4: performing similarity calculation, screening the highest similarity instance, and obtaining a strip steel composition and process parameter that meets customer demand. The present application constructs a strip steel process knowledge graph, on the one hand, realizes the sedimentation of strip steel process knowledge, improves the knowledge reuse rate, and reduces the redundancy of knowledge storage; on the other hand, finds the best strip steel composition and process window for the personalized needs of customers, reduces production cost and quality risk, and improves enterprise production and research and development efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of material processing manufacturing and intelligent technology, and particularly relates to a method for determining strip steel composition and process parameters based on a knowledge graph. BACKGROUND

[0002] Steel is a basic material in modern industry and is widely used in mechanical manufacturing, architectural design, transportation and other fields.

[0003] The selection of strip steel composition and process parameters has a direct impact on product quality, performance and service life, and is a crucial link in the entire production process. The traditional recommendation method of strip steel composition and process parameters mainly relies on manual experience and trial and error, which is low in production efficiency and easy to cause unstable product quality. In the process of strip steel production, there are many manufacturing processes and knowledge levels, from steelmaking, continuous casting to hot rolling, cold rolling process knowledge is very complex, and meanwhile, various process parameters influence each other, resulting in low efficiency of composition and process design, poor knowledge reuse rate and knowledge redundancy and other problems.

[0004] Therefore, constructing a strip steel process knowledge base system to realize modeling and reuse of strip steel process knowledge can realize automatic processing and reasoning of various process parameter information and material composition. According to the personalized needs of customers, the best strip steel composition and process window is found, the production cost and quality risk are reduced, and it is of great significance to improve the strip steel knowledge reuse rate and improve the efficiency and stability of enterprise production. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art and provide a method for determining strip steel composition and process parameters based on a knowledge graph. By constructing a strip steel process knowledge graph, on the one hand, the multi-level structure of strip steel process knowledge is stored in the knowledge graph to realize the sedimentation of strip steel process knowledge, improve the knowledge reuse rate and reduce the redundancy of knowledge storage, and on the other hand, based on the constructed knowledge graph, an intelligent recommendation algorithm is used to find the best strip steel composition and process window for the personalized needs of customers.

[0006] To achieve the above purpose, the present application provides the following technical scheme:

[0007] A method for determining strip steel composition and process parameters based on a knowledge graph, comprising:

[0008] S1: Collecting strip steel data and performing data preprocessing;

[0009] S2: Constructing a strip steel ontology model according to the strip steel data of step S1;

[0010] S3: Using the strip steel ontology model of step S2 to obtain strip steel triple data using a mapping language and constructing a strip steel knowledge graph;

[0011] S4: For a given customer demand, similarity calculation is performed according to the strip steel knowledge graph of step S3, one or several strip steel instances with the highest similarity are screened, and the strip steel composition and process parameters meeting the customer demand are determined.

[0012] According to any possible implementation manner described above, further provided is an implementation manner, in step S1, the strip steel data includes data of steel grade, composition, process, structure and mechanical property of the steel material; and the data preprocessing includes removing repeated data, removing abnormal data and replacing missing data of the strip steel data.

[0013] According to any possible implementation manner described above, further provided is an implementation manner, in step S1, the source of the strip steel data includes a steel enterprise production business database, a strip steel production process manual, a strip steel national standard and expert knowledge.

[0014] According to any possible implementation manner described above, further provided is an implementation manner, in step S2, the strip steel ontology model includes a class, a hierarchical structure and attribute constraints of the strip steel.

[0015] According to any possible implementation manner described above, further provided is an implementation manner, the class of the strip steel includes steel grade, chemical composition, structure, production process and mechanical property;

[0016] The hierarchical structure is: the strip steel grade includes low-alloy high-strength steel, dual-phase steel, interstitial-atom-free steel, martensitic steel, phase transformation induced plasticity steel, complex phase steel, quenched and tempered steel and bake hardening steel; the chemical composition includes C element, Si element, Mn element, P element and S element; the structure includes austenite, ferrite, cementite, pearlite, martensite, ledeburite and bainite; the production process includes hot rolling process, pickling process, cold rolling process, continuous annealing process and galvanizing process; and the mechanical property includes yield strength, tensile strength, elongation after fracture, n value, r value and bake hardening value.

[0017] According to any possible implementation manner described above, further provided is an implementation manner, the attribute constraints specifically include:

[0018] A pair of attribute constraints “strip steel applicable chemical composition” is set, wherein the definition domain of the “strip steel applicable chemical composition” is strip steel, and the value domain is chemical composition;

[0019] A pair of attribute constraints “strip steel applicable production process” is set, wherein the definition domain of the “strip steel applicable production process” is strip steel, and the value domain is production process;

[0020] A pair of attribute constraints “strip steel applicable structure” is set, wherein the definition domain of the “strip steel applicable structure” is strip steel, and the value domain is structure.

[0021] A pair of attribute constraints "strip applicable mechanical properties" are set, wherein the definition domain of the "strip applicable mechanical properties" is the strip, and the value domain is the mechanical properties;

[0022] A pair of attribute constraints "mechanical properties applicable chemical composition" are set, wherein the definition domain of the "mechanical properties applicable chemical composition" is the mechanical properties, and the value domain is the chemical composition;

[0023] A pair of attribute constraints "microstructure applicable production process" are set, wherein the definition domain of the "microstructure applicable production process" is the microstructure, and the value domain is the production process.

[0024] According to any possible implementation manner described above, further provided is an implementation manner, and step S3 specifically comprises:

[0025] S3.1 in a database language, the data preprocessed in step S1 is stored in a relational database according to the strip grade, composition, process, microstructure and mechanical property data information in step S2;

[0026] S3.2 in a mapping language, the strip grade, composition, process, microstructure and mechanical property data information in the relational database is mapped according to the strip ontology model in step S2, to obtain strip triple data;

[0027] S3.3 the strip triple data is stored in a graph database and knowledge visualization is performed, to obtain a strip knowledge graph.

[0028] According to any possible implementation manner described above, further provided is an implementation manner, and the strip triple data set comprises two forms of <entity, relationship, entity> and <entity, attribute, attribute value>.

[0029] According to any possible implementation manner described above, further provided is an implementation manner, in step S4, a similarity algorithm is used to calculate the similarity between the customer demand input value and all nodes (with the same customer demand attribute) in the strip knowledge graph, and the nodes are sorted according to the similarity, and one or more instances with the highest similarity to the customer demand input value are selected as the determination result of the strip composition and process parameters.

[0030] According to any possible implementation manner described above, further provided is an implementation manner, and the relationship between the similarity calculation in step S4 and the construction of the strip knowledge graph in steps S1-S3 is that, based on the attributes and relationships between different entities of the constructed strip knowledge graph, the nodes with the same customer demand attribute are accessed as the instance object data of the similarity calculation through graph search traversal.

[0031] According to any possible implementation manner described above, further provided is an implementation manner, and step S4 specifically comprises:

[0032] S4.1 selecting attributes of customer demand as the basis for similarity calculation according to attribute information in the steel grade, composition, production process and mechanical properties of the strip steel;

[0033] S4.2 performing dimensionless processing on the attribute values of the strip steel data;

[0034] S4.3 performing similarity calculation on attribute values of different numerical types respectively;

[0035] S4.4 performing weight calculation on each different attribute according to variance fluctuation;

[0036] S4.5 calculating a comprehensive weight similarity, and selecting one or several instances with the highest similarity according to the comprehensive weight similarity as the strip steel composition and process parameters meeting the customer demand.

[0037] The beneficial effects of the present application at least include:

[0038] By constructing the strip steel process knowledge graph, on the one hand, the multi-level structure of the strip steel process knowledge is stored in the knowledge graph, the strip steel process knowledge is precipitated, the knowledge reuse rate is improved, and the redundancy of knowledge storage is reduced; on the other hand, based on the constructed knowledge graph, an intelligent recommendation algorithm is used to find the best strip steel composition and process window for the personalized demand of customers, reduce the production cost and quality risk, and improve the production and research and development efficiency of enterprises. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 Fig. 1 shows a whole flowchart of a method for determining strip steel composition and process parameters based on a knowledge graph according to an embodiment of the present application.

[0040] Figure 2 Fig. 2 shows a schematic diagram of a strip steel hierarchical relationship in the embodiment.

[0041] Figure 3 Fig. 3 shows a schematic diagram of a strip steel knowledge graph in the embodiment.

[0042] Figure 4 Fig. 4 shows a specific flowchart of strip steel composition and process parameter determination in the embodiment. DETAILED DESCRIPTION

[0043] Hereinafter, specific embodiments of the present application will be described in detail with reference to specific drawings. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered in isolation, and they can be combined with each other to achieve better technical effects.

[0044] As Figure 1As shown, the embodiment of the application is a method for determining strip composition and process parameters based on a knowledge graph, comprising the following steps:

[0045] S1: Collect strip data and perform data preprocessing;

[0046] Collect data such as steel grade, composition, process, structure, and mechanical properties of the steel material of the strip. The sources of the strip data include steel enterprise production business databases, strip production process manuals, strip national standards, and expert knowledge, etc.

[0047] Clean and preprocess the collected strip data to ensure the accuracy and integrity of the data. This includes:

[0048] (1) De-duplicate the strip data to ensure that there is no duplicate data in the strip data set.

[0049] (2) Use the box plot method to read the outliers in the steel grade, composition, process, structure, and mechanical properties in the strip data set. The interquartile range IQR represents the interval between the lower quartile Q1 and the upper quartile Q3 (IQR = Q3-Q1). Values between the upper limit line Q3+1.5xIQR and the lower limit line Q1-1.5xIQR are normal values, and the rest are outliers.

[0050] (3) Replace the outliers and missing values with the experience values determined based on historical production data and expert experience to obtain the preprocessed data set.

[0051] S2: Construct a strip ontology model based on the strip data of step S1;

[0052] (1) Define the classes of the strip and their hierarchical structure in the strip ontology model based on the characteristic information of the strip steel grade, composition, process, structure, and mechanical properties. The classes of the strip and their hierarchical structure are as follows:

[0053] The strip steel grade includes but is not limited to low-alloy high-strength steel, dual-phase steel, interstitial-free steel, martensitic steel, transformation-induced plasticity steel, complex phase steel, quenched and tempered steel, and bake-hardening steel.

[0054] The chemical composition includes but is not limited to C element, Si element, Mn element, P element, and S element.

[0055] The structure includes but is not limited to austenite, ferrite, cementite, pearlite, martensite, ledeburite, and bainite.

[0056] The production process includes but is not limited to hot rolling process, pickling process, cold rolling process, continuous annealing process, and galvanizing process.

[0057] Mechanical properties include, but are not limited to: yield strength, tensile strength, elongation at break, n-value, r-value, bake hardening value.

[0058] (3) According to the characteristics of the steel grade, composition, process, structure and mechanical properties of the strip steel, the attributes and attribute constraints of the class of the strip steel are defined. The attributes and attribute constraints of the class of the strip steel are as follows:

[0059] A pair of attribute constraints "strip steel applicable chemical composition" is set, wherein the definition domain of "strip steel applicable chemical composition" is strip steel, the value domain is chemical composition, and the applicable chemical composition is set for each strip steel grade separately. For example, the applicable chemical composition of dual-phase steel CR340 / 590DP includes C element, Si element, Mn element, P element, S element, etc.

[0060] A pair of attribute constraints "strip steel applicable production process" is set, wherein the definition domain of "strip steel applicable production process" is strip steel, the value domain is production process, and the applicable production process is set for each strip steel grade separately. For example, the applicable production process of dual-phase steel CR340 / 590DP includes cold rolling process and continuous annealing process.

[0061] A pair of attribute constraints "strip steel applicable structure" is set, wherein the definition domain of "strip steel applicable structure" is strip steel, the value domain is structure, and the applicable structure is set for each strip steel grade separately. For example, the applicable structure of dual-phase steel CR340 / 590DP is martensite, austenite or bainite and ferrite two-phase structure.

[0062] A pair of attribute constraints "strip steel applicable mechanical properties" is set, wherein the definition domain of "strip steel applicable mechanical properties" is strip steel, the value domain is mechanical properties, and the applicable structure is set for each strip steel grade separately. For example, the applicable mechanical properties of dual-phase steel CR340 / 590DP are: yield strength 340-460 MPa, tensile strength not less than 590 MPa, and elongation at break not less than 18%.

[0063] A pair of attribute constraints "mechanical properties applicable chemical composition" is set, wherein the definition domain of "mechanical properties applicable chemical composition" is mechanical properties, and the value domain is chemical composition. For example: when the C content is below 0.8%, the C content is positively correlated with yield strength and tensile strength, and negatively correlated with elongation at break; when the C content is above 1%, the C content is negatively correlated with yield strength and tensile strength.

[0064] A pair of attribute constraints "structure applicable production process" is set, wherein the definition domain of "structure applicable production process" is structure, and the value domain is production process. For example: in order to obtain the martensite and ferrite structure of dual-phase steel CR340 / 590DP, the austenite dual-phase method or the critical zone dual-phase method in the continuous annealing process is adopted.

[0065] After the design of the strip steel ontology model is completed, the Protégé software is applied to realize the documentation of the strip steel ontology model. Figure 2 A specific strip steel ontology model is shown in the following.

[0066] S3: Using the strip steel ontology model of step S2, obtaining strip steel triple data using a mapping language, and constructing a strip steel knowledge graph;

[0067] (1) Using the SQL database language, the preprocessed data in step S1 is stored in a relational database according to the strip steel grade, composition, process, organization, and mechanical property data information in step S2.

[0068] Specifically, a database table is established for each class of strip steel grade, composition, process, organization, and mechanical property, and a unique ID is set to mark each entity (specific strip steel number). Then, the classes with attributes are connected by foreign keys.

[0069] (2) Using the D2RQ mapping language, the strip steel grade, composition, process, organization, and mechanical property data information of the relational database is mapped according to the strip steel ontology model of step S2, and the mapping file after mapping is modified to obtain strip steel triple data in RDF form. The strip steel triple data set includes two forms: <entity, relationship, entity> and <entity, attribute, attribute value>; wherein the entity includes strip steel model, chemical composition, production process, mechanical property, and attribute value, and the relationship and attribute are attribute constraints; in the strip steel knowledge graph, each entity is represented as a node, and each relationship and attribute is represented as an edge.

[0070] (3) Using the neosemantics plug-in in the Neo4j graph database, importing the strip steel triple data in RDF form obtained after mapping into the Neo4j graph database, storing and visualizing the strip steel triple data, and obtaining the strip steel knowledge graph.

[0071] A specific strip steel knowledge graph is shown in the following. Figure 3

[0072] S4: For a given customer demand, similarity calculation is performed according to the strip steel knowledge graph of step S3, one or several instances with the highest similarity are selected, and the strip steel composition and process parameters that meet the customer demand are determined.

[0073] Using the similarity algorithm, the similarity between the customer demand input value and the node is calculated, and the similarity is sorted according to the size, and one or several strip steel instances with the highest similarity to the customer demand input value are selected as the determination result of the strip steel composition and process parameters. ​

[0074] The relationship between step S4 similarity calculation and steps S1-S3 of constructing the strip steel knowledge graph is that, based on the attributes and relationships between different entities of the constructed strip steel knowledge graph, the nodes with the same attributes as the customer demand are accessed as instance object data for similarity calculation through graph search traversal.

[0075] Specifically, the similarity calculation method for determining (recommending) the strip steel composition and process parameters is as follows:

[0076] (1) According to the attribute information in the strip steel grade, chemical composition, production process and mechanical properties, the attributes of customer demand are selected as the basis for similarity calculation. Which includes:

[0077] The strip steel grade includes the strip steel grade mark.

[0078] The chemical composition includes C content, Si content, Mn content, P content, S content, etc.

[0079] Production process: hot rolling process, pickling process, cold rolling process, continuous annealing process, galvanizing process parameters. For example: the parameters in the hot rolling process include discharge temperature, finishing temperature, coiling temperature, hot rolling thickness, hot rolling width.

[0080] Mechanical properties include yield strength, tensile strength, elongation after fracture, n value, r value, bake hardening value.

[0081] (2) In order to make the similarity calculation have commensurability, the attribute values are dimensionless processed, so that the attribute values are in the range of [0, 1].

[0082] (3) For the attribute values in a certain range, the similarity between the interval values is calculated.

[0083] According to the data of the strip steel knowledge graph, the to-be-recommended object (customer demand input) and the instance object (knowledge graph node) are calculated using the Euclidean distance similarity. Define the attribute interval value [a1, a2] of the to-be-recommended object, and the attribute interval value [b1, b2] of the instance object, then the similarity between the two is:

[0084]

[0085] Wherein, r is 2.

[0086] For numerical types with certain attribute values, the attribute values are directly calculated for similarity.

[0087] Define the attribute value a1 of the to-be-recommended object and the attribute value b1 of the instance object, then the similarity between the two is:

[0088]

[0089] Define cr C = {c1, c2, c3, c4, …, c n} is an instance set, c r and the jth attribute similarity of c i is denoted as s ij , then the attribute similarity of the object to be recommended c r and all instances in the instance set C constitutes a similarity matrix S.

[0090]

[0091] (4) Different attributes have different degrees of influence on similarity. Based on the calculation of weights according to the variance of each column of the similarity matrix, the variance of each column is used as the weight, which can represent the fluctuation degree of the column, so as to evaluate the importance of the column. The greater the variance, the greater the weight, which means that the fluctuation degree of the data in the column is greater, and it is relatively more important.

[0092] Define the attribute weight W = {w1, w2, w3, w4, …, w m} of each column of the similarity matrix S, then the variance of each column of the similarity matrix S is:

[0093]

[0094] Where s ij represents the element in the ith row and jth column of the similarity matrix, n represents the number of rows of the similarity matrix, and m represents the number of columns of the similarity matrix.

[0095] The column attribute weight is calculated as:

[0096]

[0097] (5) The obtained weight is used for similarity weighted calculation, and a similarity threshold is set to select one or several instances with the highest similarity as the determination result of the steel composition and process parameters.

[0098] Specifically, the comprehensive weight similarity is:

[0099]

[0100] Embodiment

[0101] The input mechanical property attribute is used as the basis for similarity calculation. For example, a galvanized steel with (yield strength, tensile strength, elongation after fracture) = (250, 360, 30%) is obtained, and 15 data traversed in the knowledge graph are used as instance object data for similarity calculation. The data is part of the composition data, process data and performance data of the material, as shown in the following table:

[0102]

[0103]

[0104] The selected mechanical property data is dimensionless, and the following data table is obtained:

[0105]

[0106]

[0107] Further, the similarity matrix S is obtained as follows:

[0108]

[0109] Further, the variance of each column is obtained according to the similarity matrix S, and the weight of each item is calculated.

[0110] w i =[0.40 0.40 0.20]

[0111] Further, the final similarity calculated by the comprehensive weight is:

[0112] [0.67 0.58 0.74 0.82 0.13 0.96 0.66 0.06 0.27 0.64 0.73 0.70 0.79 0.11 0.64] T .

[0113] Further, the similarity threshold is set to 0.90, and the recommended steel grade is DX51D according to the similarity calculation, and the comprehensive weight similarity is 0.96, and the recommended part of the composition and process parameters are: (C, Si, Mn, P, S, opening rolling temperature, finishing temperature) = (0.037, 0.028, 0.140, 0.012, 0.006, 1204, 883).

[0114] The specific process is shown in Figure 4 .

[0115] Although several embodiments of the present application have been given in the present text, those skilled in the art should understand that the embodiments in the present text can be changed without departing from the spirit of the present application. The above embodiments are only exemplary, and the embodiments in the present text should not be used as a limitation of the scope of the present application.

Claims

1. A method for determining strip composition and process parameters based on a knowledge graph, characterized in that, The method comprises: S1: collecting strip steel data and performing data preprocessing; S2: constructing a strip steel ontology model according to the strip steel data of step S1; S3: using the strip steel ontology model of step S2, obtaining strip steel triple data using a mapping language, and constructing a strip steel knowledge graph; S4: for a given customer demand, calculating similarity according to the strip steel knowledge graph of step S3, filtering one or several strip steel instances with the highest similarity, and obtaining strip steel components and process parameters that meet the customer demand; Step S4 specifically comprises: S4.1 selecting attributes of customer demand as the basis for similarity calculation according to attribute information in strip steel grade, composition, production process and mechanical properties; S4.2 performing dimensionless processing on the attribute values of the strip steel data; S4.3 performing similarity calculation on attribute values of different numerical types respectively; S4.4 calculating weights for different attributes according to variance fluctuations; The attribute weight W = {w1, w2, w3, w4,..., wn} defining each column of the similarity matrix S is determined by the following equation: m The variance of each column of the similarity matrix S is determined by the following equation: ； wherein s ij denotes the element in the i-th row and j-th column of the similarity matrix, n denotes the number of rows of the similarity matrix, and m denotes the number of columns of the similarity matrix. The column attribute weight is calculated as: ； S4.5 calculating a comprehensive weight similarity, and selecting one or several instances with the highest comprehensive weight similarity as strip steel components and process parameters that meet the customer demand.

2. The knowledge graph based determination of strip composition and process parameters method as claimed in claim 1, wherein, In step S1, the strip steel data includes data of the grade, composition, process, structure and mechanical properties of the strip steel; and the data preprocessing includes removing duplicate data, removing abnormal data and replacing missing data from the strip steel data.

3. The knowledge graph based strip composition and process parameter determination method of claim 1, wherein, In step S1, the sources of the strip steel data include a steel enterprise production business database, a strip steel production process manual, strip steel national standards and expert knowledge.

4. The knowledge graph based strip composition and process parameter determination method of claim 1, wherein, In step S2, the strip steel ontology model includes the class, hierarchical structure and attribute constraints of the strip steel.

5. The method for determining strip steel components and process parameters based on a knowledge graph according to claim 4, wherein the class of the strip steel includes the grade, chemical composition, structure, production process and mechanical properties of the strip steel; the hierarchical structure is that the strip steel grade includes low-alloy high-strength steel, dual-phase steel, interstitial-free steel, martensitic steel, transformation-induced plasticity steel, complex-phase steel, quenched and tempered steel and bake-hardening steel; the chemical composition includes C element, Si element, Mn element and P element; the structure includes austenite, ferrite, cementite, pearlite, martensite, ledeburite and bainite; the production process includes hot rolling process, pickling process, cold rolling process, continuous annealing process and galvanizing process; and the mechanical properties include yield strength, tensile strength, elongation after fracture, n value, r value and bake-hardening value. the attribute constraints specifically include: a pair of attribute constraints "strip steel applicable chemical composition" are set, wherein the domain of "strip steel applicable chemical composition" is strip steel, and the value domain is chemical composition; 6. The knowledge graph based strip composition and process parameter determination method of claim 5, wherein, a pair of attribute constraints "strip steel applicable production process" are set, wherein the domain of "strip steel applicable production process" is strip steel, and the value domain is production process; a pair of attribute constraints "strip steel applicable structure" are set, wherein the domain of "strip steel applicable structure" is strip steel, and the value domain is structure; a pair of attribute constraints "strip steel applicable mechanical properties" are set, wherein the domain of "strip steel applicable mechanical properties" is strip steel, and the value domain is mechanical properties; ​ ​ A pair of property constraints "mechanical performance applicable chemical composition" is set, wherein the definition domain of the "mechanical performance applicable chemical composition" is the mechanical performance, and the value domain is the chemical composition; A pair of property constraints "microstructure applicable production process" is set, wherein the definition domain of the "microstructure applicable production process" is the microstructure, and the value domain is the production process.

7. The knowledge graph based strip composition and process parameter determination method of claim 1, wherein, Step S3 specifically comprises: S3.1 storing the data preprocessed in step S1 in a relational database according to the strip steel grade, composition, process, microstructure and mechanical property data information in step S2 by using a database language; S3.2 mapping the strip steel grade, composition, process, microstructure and mechanical property data information in the relational database according to the strip steel ontology model in step S2 by using a mapping language to obtain strip steel triple data; S3.3 storing the strip steel triple data by using a graph database and performing knowledge visualization to obtain a strip steel knowledge graph.

8. The knowledge graph based strip composition and process parameter determination method of claim 7, wherein, The strip steel triple data set comprises two forms of <entity, relationship, entity> and <entity, attribute, attribute value>; wherein the entity comprises a strip steel model, a chemical composition, a production process, a mechanical property and an attribute value, and the relationship is a property constraint; in the strip steel knowledge graph, each entity is represented as a node, and each relationship is represented as an edge.

9. The knowledge graph based strip composition and process parameter determination method of claim 8, wherein, In step S4, a similarity algorithm is used to calculate the similarity between the customer demand input value and all nodes in the strip steel knowledge graph, and the similarity is sorted according to the size, and one or several strip steel instances with the highest similarity to the customer demand input value are selected as the determination result of the strip steel composition and process parameters.

Images