A user-collaborative similar material recommendation method

Abstract

The present invention relates to a user-collaborative similar material recommendation method, which belongs to the technical field of material information service methods. The technical solution of the present invention is: to establish a material classification hierarchy structure tree and a material data set, and the user selects or inputs the reference materials for similarity calculation and the data set participating in the similarity calculation; the materials to be compared in the data set are sorted according to the size of the overall similarity, and the user can control the number of similar material recommendations by setting a similarity threshold. The beneficial effects of the present invention are: it can automatically implement similar material recommendation or material selection matching process according to the user's material selection requirements, and can realize similarity calculation and matching of different data types such as interval values, precise values, and enumeration values, thereby improving the scientific nature of material selection recommendation, having a wide range of applications, strong flexibility, and reliable recommendation results.

Description

Technical Field

[0001] The present invention relates to a user-coordinated similar material recommendation method, belonging to the technical field of material information service methods. Background Art

[0002] Traditional user material recommendation and unknown material identification processes are subject to randomness and subjectivity, often limited by personal experience, resulting in low efficiency and poor results. Furthermore, existing similarity calculation methods for material recommendation are mostly based on exact value similarity, raising concerns about the proper handling of material data containing interval values in material databases. Therefore, establishing a rational, comprehensive, quantitative evaluation method for material selection and recommendation is crucial for achieving scientific material selection. Summary of the Invention

[0003] The purpose of the present invention is to provide a user-collaborative similar material recommendation method, which can automatically realize the similar material recommendation or material selection matching process according to the user's material selection requirements, and can realize the similarity calculation and matching of different data types such as interval values, exact values, and enumeration values, thereby improving the scientific nature of material selection recommendation, having a wide range of applications, strong flexibility, and reliable recommendation results, and effectively solving the above-mentioned problems existing in the background technology.

[0004] The technical solution of the present invention is: a user-collaborative similar material recommendation method, comprising the following steps:

[0005] (1) Establish a hierarchical structure tree for material classification, with the following options: material type, material category, material subcategory, and material brand. Maintain the basic public information of material data in the corresponding hierarchy. The field type in each hierarchy is an enumeration value.

[0006] (2) Establish a material data set and maintain the data of chemical composition, material properties and key characteristic parameters at the bottom of the material classification hierarchy structure. The data types of chemical composition and material performance attribute parameters are precise values and interval values, and the data type of key characteristic parameters is enumeration value;

[0007] (3) The user selects or inputs the reference material for similarity calculation and the data set involved in the similarity calculation, including the material classification hierarchy structure, chemical composition, material performance parameters and key characteristic parameters; the chemical composition includes the chemical elements C, Si, Mn, P and S, but is not limited to the above elements; the material performance parameters include yield strength, tensile strength and elongation after fracture, but are not limited to the above parameters; the key characteristic parameters include purpose and material state, but are not limited to the above attributes;

[0008] (4) Calculate the similarity Sc of the material classification hierarchy structure tree. The calculation formula is as follows:

[0009] Sc(C1,C2)=∑ω i *Sbi +ω t Sa(C1,C2)

[0010] ∑ωi+ωt=1

[0011] Where ωi is the weight of the attribute value similarity of each level of the structure tree, ωt is the weight of the concept similarity of the structure tree, Sa is the concept similarity of the structure tree level, and Sbi is the attribute value similarity of the i-th level of the structure tree;

[0012] (5) Calculate the type similarity of chemical composition, material properties and key characteristic parameters in the dataset;

[0013] The chemical composition type similarity calculation Sd1 is calculated as follows:

[0014] Sd1=(C1∩C2) / (C1∪C2)

[0015] Where C1∩C2 is the number of chemical elements or parameters that are matched between the reference material C1 and the reference material C2 in the data set and that participate in the similarity calculation, and C1∪C2 is the total number of chemical elements or parameters that participate in the matching;

[0016] The type similarity calculation Sd2 of material properties is calculated as follows:

[0017] Sd2=(C1∩C2) / (C1∪C2)

[0018] Among them, C1∩C2 is the number of matching material performance parameters between the reference material C1 and the reference material C2 in the data set that participate in the similarity calculation, and C1∪C2 is the total number of performance parameters participating in the matching;

[0019] The type similarity calculation Sd3 of the key feature parameters is calculated as follows:

[0020] Sd3=(C1∩C2) / (C1∪C2)

[0021] Among them, C1∩C2 is the number of key feature parameters that match between the reference material C1 and the reference material C2 in the data set and participate in the similarity calculation, and C1∪C2 is the total number of key feature parameters participating in the matching.

[0022] (6) Formulate processing rules for data interval values and default values and perform data pre-processing;

[0023] The optional processing method for chemical element interval values is to treat interval values ≥ or > A as the endpoint value A, treat interval values ≤ or < A as [0, A], and treat (A, B) or (A, B] or [A, B) as [A, B]. The optional processing method for default values is to fill them with 0. The chemical composition dataset processed according to the above rules is standardized / normalized.

[0024] The optional method for handling material performance parameter interval values is to treat interval values ≥ or > A and ≤ or < A as endpoint value A, and to treat (A, B) or (A, B] or [A, B) as [A, B]. The optional method for handling default values is to stipulate that if the attribute index value of the compared material B in the attribute set is empty, the attribute values of both parties are equivalent. The material performance data set processed according to the above rules is standardized / normalized.

[0025] The optional default value handling method for key feature parameters is: if the parameter value X of the reference material A is empty, XA is agreed to be "none". If the reference material B in the attribute set has an empty value, the attribute values of both parties are agreed to be equivalent. If the parameter value is "other", it can be matched with "other" or an empty keyword;

[0026] (7) Calculate the similarity of chemical composition, material properties and key characteristic parameter attribute values respectively to obtain a similarity matrix;

[0027] (8) Calculate the combined weights of the parameters j in the similarity matrix of chemical composition, material properties and key characteristic parameters respectively; calculate the attribute value similarity of the considered attribute weights of the chemical composition, material properties and key characteristic parameters of each compared material in the data set respectively;

[0028] (9) Calculate the similarity of comprehensive material properties;

[0029] Calculate the chemical composition similarity Schem: Schem = Sd1*Se1;

[0030] Calculate the material performance attribute similarity Sprop: Sprop = Sd2*Se2;

[0031] Calculate the key feature parameter attribute similarity Schar: Schar = Sd3*Se3.

[0032] Calculate the similarity of comprehensive material properties: Ssub=ω1*Schem+ω2*Sprop+ω3*Schar, where ω1, ω2, and ω3 are the weights of chemical composition, material properties, and key characteristic properties, respectively.

[0033] ω2+ω3=1;

[0034] (10) Calculate the overall similarity of materials, the overall similarity of materials St = similarity of material classification hierarchy structure Sc * similarity of material comprehensive attributes Ssub;

[0035] (11) Sort the materials in the dataset according to their overall similarity St value. Users can control the number of similar materials recommended by setting a similarity threshold.

[0036] In step (4), the calculation method of the structure tree hierarchical concept similarity Sa may be a distance-based concept or semantic similarity calculation formula, including but not limited to the following optional formula:

[0037] Where C1 and C2 are the two values at the bottom of the material classification hierarchy tree, and optionally the material brand, which are represented in len(C1, C2) and represent the shortest path length between C1 and C2 in the hierarchy tree, and Depth represents the maximum depth value of the hierarchy tree.

[0038] Among them, N1 and N2 represent the shortest paths between the lowest-level concepts such as material grades C1 and C2 and the nearest common parent node, respectively, and H represents the number of levels from the nearest common parent node C to the root node;

[0039] Where des(C1, C2) represents the total hierarchical size of the structure tree, and the numerator represents the distance from the common parent node to the root node, which describes the common characteristics of C1 and C2.

[0040] The calculation method of the attribute value similarity Sbi at the i-th level of the structure tree is as follows:

[0041]

[0042] In step (6), the calculation method for standardization / normalization of the chemical composition and material property data sets is not limited to the following method:

[0043] A chemical composition or performance parameter value in the data set has an interval value or an exact value. The kth material instance is an exact value, denoted as Xk, and the jth material instance is an interval value, denoted as [Aj, Bj]. Then

[0044] minA=min{min{A1,A2,…,Aj,…},min{…Xk…}}

[0045] maxB=max{max{B1, B2,…,Bj,…},max{…Xk…}}

[0046] The interval value is standardized: Aj'=(Aj-minA) / (maxB-minA)

[0047] Bj'=(Bj-minA) / (maxB-minA)

[0048] The exact value is normalized: Xk'=(Xk-minAi) / (maxB-minA).

[0049] In step (7), the similarity calculation of attribute values is not limited to the following method:

[0050] Calculation of similarity between interval values and exact values:

[0051]

[0052] Similarity calculation between exact values: Sij(A1,A2)=1-|A1-A2|

[0053] Similarity calculation of enumeration values:

[0054]

[0055] In step (8), the combined weight calculation includes subjective weight calculation and objective weight calculation, and adopts the multiplication combined weighting method, which is as follows:

[0056]

[0057] Among them, ω 1 is the subjective weight, ω 2 is the objective weight;

[0058] The subjective weight calculation method may be an expert survey method or a hierarchical analysis method, but is not limited to the above methods; the objective weight calculation method may be a similarity deviation method or an entropy method, but is not limited to the above methods. The similarity deviation information method selected is as follows:

[0059]

[0060]

[0061] In step (8), the calculation method of the chemical component attribute value similarity Se1 includes but is not limited to the following method:

[0062] or

[0063] The calculation methods of material performance attribute value similarity Se2 include but are not limited to the following methods: or

[0064] The calculation methods of key feature attribute value similarity Se3 include but are not limited to the following methods:

[0065] or

[0066] The beneficial effects of the present invention are: it can automatically realize the similar material recommendation or material selection matching process according to the user's material selection requirements, can realize the similarity calculation and matching of different data types such as interval values, exact values, and enumeration values, improve the scientific nature of material selection recommendation, has a wide range of applications, strong flexibility, and reliable recommendation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 is a flow chart of the present invention;

[0068] Figure 2 This is a flow chart of chemical composition similarity calculation according to an embodiment of the present invention;

[0069] Figure 3 Flowchart of material property similarity calculation according to an embodiment of the present invention;

[0070] Figure 4 Flowchart of key feature parameter similarity calculation according to an embodiment of the present invention;

[0071] Figure 5 Attached diagram is a diagram showing similarity calculation of a material classification hierarchy structure tree according to an embodiment of the present invention;

[0072] Figure 6 Attached figure is the calculation of chemical composition similarity of materials in the embodiment of the present invention;

[0073] Figure 7 Attached diagram is a diagram showing the calculation of material performance similarity in an embodiment of the present invention;

[0074] Figure 8 Attached diagram is a diagram showing the calculation of similarity of key characteristic parameters of materials in an embodiment of the present invention;

[0075] Figure 9 Attached is the result of overall material similarity calculation and similarity recommendation sorting in an embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solutions and advantages of the invention implementation cases clearer, the technical solutions in the invention implementation cases will be clearly and completely described below in conjunction with the drawings in the implementation cases. Obviously, the implementation cases described are only a small part of the implementation cases of the present invention, rather than all the implementation cases. Based on the implementation cases in the present invention, all other implementation cases obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0077] A user-collaborative similar material recommendation method includes the following steps:

[0078] (1) Establish a hierarchical structure tree for material classification, with the following options: material type, material category, material subcategory, and material brand. Maintain the basic public information of material data in the corresponding hierarchy. The field type in each hierarchy is an enumeration value.

[0079] (2) Establish a material data set and maintain the data of chemical composition, material properties and key characteristic parameters at the bottom of the material classification hierarchy structure. The data types of chemical composition and material performance attribute parameters are precise values and interval values, and the data type of key characteristic parameters is enumeration value;

[0080] (3) The user selects or inputs the reference material for similarity calculation and the data set involved in the similarity calculation, including the material classification hierarchy structure, chemical composition, material performance parameters and key characteristic parameters; the chemical composition includes the chemical elements C, Si, Mn, P and S; the material performance parameters include yield strength, tensile strength and elongation after fracture; the key characteristic parameters include application and material state;

[0081] (4) Calculate the similarity Sc of the material classification hierarchy structure tree. The calculation formula is as follows:

[0082] Sc(C1,C2)=∑ω i *Sb i +ω t Sa(C1,C2)

[0083] ∑ωi+ωt=1

[0084] Where ωi is the weight of the attribute value similarity of each level of the structure tree, ωt is the weight of the concept similarity of the structure tree, Sa is the concept similarity of the structure tree level, and Sbi is the attribute value similarity of the i-th level of the structure tree;

[0085] (5) Calculate the type similarity of chemical composition, material properties and key characteristic parameters in the dataset;

[0086] The chemical composition type similarity calculation Sd1 is calculated as follows:

[0087] Sd1=(C1∩C2) / (C1∪C2)

[0088] Where C1∩C2 is the number of chemical elements or parameters that are matched between the reference material C1 and the reference material C2 in the data set and that participate in the similarity calculation, and C1∪C2 is the total number of chemical elements or parameters that participate in the matching;

[0089] The type similarity calculation Sd2 of material properties is calculated as follows:

[0090] Sd2=(C1∩C2) / (C1∪C2)

[0091] Among them, C1∩C2 is the number of matching material performance parameters between the reference material C1 and the reference material C2 in the data set that participate in the similarity calculation, and C1∪C2 is the total number of performance parameters participating in the matching;

[0092] The type similarity calculation Sd3 of the key feature parameters is calculated as follows:

[0093] Sd3=(C1∩C2) / (C1∪C2)

[0094] Among them, C1∩C2 is the number of key feature parameters that match between the reference material C1 and the reference material C2 in the data set and participate in the similarity calculation, and C1∪C2 is the total number of key feature parameters participating in the matching.

[0095] (6) Formulate processing rules for data interval values and default values and perform data pre-processing;

[0096] The optional processing method for chemical element interval values is to treat interval values ≥ or > A as the endpoint value A, treat interval values ≤ or < A as [0, A], and treat (A, B) or (A, B] or [A, B) as [A, B]. The optional processing method for default values is to fill them with 0. The chemical composition dataset processed according to the above rules is standardized / normalized.

[0097] The optional method for handling material performance parameter interval values is to treat interval values ≥ or > A and ≤ or < A as endpoint value A, and to treat (A, B) or (A, B] or [A, B) as [A, B]. The optional method for handling default values is to stipulate that if the attribute index value of the compared material B in the attribute set is empty, the attribute values of both parties are equivalent. The material performance data set processed according to the above rules is standardized / normalized.

[0098] The optional default value handling method for key feature parameters is: if the parameter value X of the reference material A is empty, XA is agreed to be "none". If the reference material B in the attribute set has an empty value, the attribute values of both parties are agreed to be equivalent. If the parameter value is "other", it can be matched with "other" or an empty keyword;

[0099] (7) Calculate the similarity of chemical composition, material properties and key characteristic parameter attribute values respectively to obtain a similarity matrix;

[0100] (8) Calculate the combined weights of the parameters j in the similarity matrix of chemical composition, material properties and key characteristic parameters respectively; calculate the similarity of the chemical composition, material properties and key characteristic parameter attribute values of each compared material in the data set respectively;

[0101] (9) Calculate the similarity of comprehensive material properties;

[0102] Calculate the chemical composition similarity Schem: Schem = Sd1*Se1;

[0103] Calculate the material performance attribute similarity Sprop: Sprop = Sd2*Se2;

[0104] Calculate the key feature parameter attribute similarity Schar: Schar = Sd3*Se3.

[0105] Calculate the similarity of comprehensive material properties: Ssub = ω1*Schem+ω2*Sprop+ω3*Schar, where ω1, ω2, and ω3 are the weights of chemical composition, material properties, and key characteristic attributes, respectively, and ω1+ω2+ω3=1;

[0106] (10) Calculate the overall similarity of materials, the overall similarity of materials St = similarity of material classification hierarchy structure Sc * similarity of material comprehensive attributes Ssub;

[0107] (11) Sort the materials in the dataset according to their overall similarity St value. Users can control the number of similar materials recommended by setting a similarity threshold.

[0108] In step (4), the calculation method of the structure tree hierarchical concept similarity Sa may be a distance-based concept or semantic similarity calculation formula, including but not limited to the following optional formula:

[0109] Where C1 and C2 are the two values at the bottom of the material classification hierarchy tree, and optionally the material brand, which are represented in len(C1, C2) and represent the shortest path length between C1 and C2 in the hierarchy tree, and Depth represents the maximum depth value of the hierarchy tree.

[0110] Among them, N1 and N2 represent the shortest paths between the lowest-level concepts such as material grades C1 and C2 and the nearest common parent node, respectively, and H represents the number of levels from the nearest common parent node C to the root node;

[0111] Where des(C1, C2) represents the total hierarchical size of the structure tree, and the numerator represents the distance from the common parent node to the root node, which describes the common characteristics of C1 and C2.

[0112] The calculation method of the attribute value similarity Sbi at the i-th level of the structure tree is as follows:

[0113]

[0114] In step (6), the calculation method for standardization / normalization of the chemical composition and material property data sets is not limited to the following method:

[0115] A chemical composition or performance parameter value in the data set has an interval value or an exact value. The kth material instance is an exact value, denoted as Xk, and the jth material instance is an interval value, denoted as [Aj, Bj]. Then

[0116] minA=min{min{A1,A2,…,Aj,…},min{…Xk…}}

[0117] maxB=max{max{B1, B2,…,Bj,…},max{…Xk…}}

[0118] The interval value is standardized: Aj'=(Aj-minA) / (maxB-minA)

[0119] Bj'=(Bj-minA) / (maxB-minA)

[0120] The exact value is normalized: Xk'=(Xk-minAi) / (maxB-minA).

[0121] In step (7), the similarity calculation of attribute values is not limited to the following method:

[0122] Calculation of similarity between interval values and exact values:

[0123]

[0124] Similarity calculation between exact values: Sij(A1,A2)=1-|A1-A2|

[0125] Similarity calculation of enumeration values:

[0126]

[0127] In step (8), the combined weight calculation includes subjective weight calculation and objective weight calculation, and adopts the multiplication combined weighting method, which is as follows:

[0128]

[0129] Among them, ω 1 is the subjective weight, ω 2 is the objective weight;

[0130] The subjective weight calculation method may be an expert survey method or a hierarchical analysis method, but is not limited to the above methods; the objective weight calculation method may be a similarity deviation method or an entropy method, but is not limited to the above methods. The similarity deviation information method selected is as follows:

[0131]

[0132]

[0133] In step (8), the calculation method of the chemical component attribute value similarity Se1 includes but is not limited to the following method:

[0134] or

[0135] The calculation methods of material performance attribute value similarity Se2 include but are not limited to the following methods: or

[0136] The calculation methods of key feature attribute value similarity Se3 include but are not limited to the following methods:

[0137] or

[0138] Example:

[0139] The present invention provides a user-collaborative similar material recommendation method, which specifically includes the following steps:

[0140] (1) Establish a material classification hierarchy structure, with five levels from high to low: material type, variety, product type, material category, and material brand. Maintain the basic public information of material data in the corresponding level, specifically as follows: Figure 5 The material classification hierarchy structure tree is characterized in that the field type in each level is an enumeration value.

[0141] (2) Establishing a material data set: Maintain material parameter attributes and data such as chemical composition, mechanical properties, and key characteristic parameters at the material grade level in the material classification structure tree. The data types for the elements in the chemical composition and material performance attribute parameters can be either exact values or interval values. The data type for the key characteristic attributes is enumerated values.

[0142] (3) Selection Figure 5-Figure 9 ID1 in the dataset is used as the reference material, and ID2 to ID7 are used as the reference materials for similarity recommendation in the dataset. Figure 6 The chemical elements C, Si, and S shown participate in the similarity calculation; select Figure 7 The material performance parameters shown, such as yield strength Rp0.2, tensile strength Rm, and elongation after fracture A, are included in the similarity calculation; select Figure 8The key characteristic parameters of the materials shown, such as material status, application, coating form, and topcoat type, are involved in the similarity calculation.

[0143] (4) Check the box that the material classification hierarchy structure tree participates in the similarity calculation, and then the material classification hierarchy structure tree similarity calculation Sc is performed. The calculation formula is as follows:

[0144] Sc(C1,C2)=∑ω i *Sb i +ω t Sa(C1,C2)

[0145] ∑ωi+ωt=1

[0146] Where ωi is the weight of the attribute value similarity at each level of the structure tree, ωt is the weight of the concept similarity of the structure tree, Sa is the concept similarity of the structure tree level, and Sbi is the attribute value similarity of the i-th level of the structure tree.

[0147] (5) Calculate the type similarity Sd1 of the chemical component attributes in the data set. The calculation formula is as follows:

[0148] Sd1=(C1∩C2) / (C1∪C2)

[0149] Wherein, C1∩C2 is the number of chemical elements or parameters that match between the reference material C1 and the reference material C2 in the data set and participate in the similarity calculation, and C1∪C2 is the total number of chemical elements or parameters participating in the matching.

[0150] The type similarity calculation Sd2 of the material performance attributes in the data set is performed, and the calculation formula is as follows:

[0151] Sd2=(C1∩C2) / (C1∪C2)

[0152] Among them, C1∩C2 is the number of matching material performance parameters between the reference material C1 and the reference material C2 in the data set that participate in the similarity calculation, and C1∪C2 is the total number of performance parameters participating in the matching.

[0153] The type similarity calculation Sd3 of the key feature parameters in the data set is performed, and the calculation formula is as follows:

[0154] Sd3=(C1∩C2) / (C1∪C2)

[0155] Among them, C1∩C2 is the number of key feature parameters that match between the reference material C1 and the reference material C2 in the data set and participate in the similarity calculation, and C1∪C2 is the total number of key feature parameters participating in the matching.

[0156] (6) Pre-process the chemical composition dataset involved in similarity calculation and formulate processing rules for chemical composition element interval values and default values. The processing method for chemical composition element interval values is to treat interval values ≥ or > A as the endpoint value A, treat interval values ≤ or < A as [0, A], and treat (A, B) or (A, B] or [A, B) as [A, B]. The default value processing method is to fill with 0. The chemical composition dataset processed according to the above rules is standardized / normalized.

[0157] Pre-process the material property datasets involved in the similarity calculation and develop rules for handling interval and default values for material property parameters. For interval values, treat interval values ≥ or > A and ≤ or < A as endpoint A, and treat (A, B) or (A, B] or [A, B) as [A, B]. For default value handling, if an attribute indicator value for the compared material B in the attribute set is null, the attribute values for both materials are considered equivalent. After processing according to these rules, the material property datasets are standardized / normalized.

[0158] Perform data pre-processing on the key feature parameter datasets involved in similarity calculations and develop default value handling rules for key feature parameters. The default value handling method is: if the parameter value X of reference material A is empty, then XA = "None". If the attribute set of reference material B contains an empty value, then the attribute values of both parties are agreed to be equivalent. If the parameter value is "Other", it can be matched with "Other" or an empty keyword.

[0159] (7) Perform similarity calculation on the chemical component attribute values of the chemical component data set involved in the similarity calculation to obtain the chemical component similarity matrix S1.

[0160] The similarity calculation of the material performance attribute values is performed on the material performance data set involved in the similarity calculation to obtain the material performance similarity matrix S2.

[0161] The similarity calculation of the key characteristic parameter attribute values is performed on the key characteristic parameter data set involved in the similarity calculation to obtain the material performance similarity matrix S3.

[0162] (8) Calculate the combined weights of the parameters C, Si, and S in the chemical composition similarity matrix S1, including subjective and objective weight calculations, to obtain the weight ωj of each parameter. Perform weighted chemical composition attribute value similarity calculations to obtain the chemical composition attribute value similarity Se1 of each referenced material in the data set:

[0163]

[0164] The combined weights of the parameters yield strength Rp0.2, tensile strength Rm, and elongation after fracture A in the material performance similarity matrix S2 are calculated, including subjective and objective weight calculations, to obtain the weight ωj of each parameter. The similarity calculation of the chemical composition attribute values taking into account the weights is performed to obtain the material performance attribute value similarity Se2 of each compared material in the data set:

[0165]

[0166] The combined weights of the parameters material state, application, coating form, and topcoat type in the key characteristic parameter similarity matrix S3 are calculated, including subjective weight calculation and objective weight calculation, to obtain the weight ωj of each parameter. The similarity calculation of the chemical composition attribute value considering the weight is performed to obtain the key characteristic parameter attribute value similarity Se3 of each compared material in the data set:

[0167]

[0168] (9) Calculate the chemical composition attribute similarity Schem: Schem = Sd1*Se1, such as Figure 6 ;

[0169] Calculate the material performance attribute similarity Sprop: Sprop = Sd2*Se2, such as Figure 7 ;

[0170] Calculate the key feature parameter attribute similarity Schar: Schar = Sd3*Se3, such as Figure 8 .

[0171] Calculate the similarity of comprehensive material properties: Ssub = ω1*Schem+ω2*Sprop+ω3*Schar, where ω1, ω2, and ω3 are the weights of chemical composition, material properties, and key characteristic attributes, respectively, and ω1+ω2+ω3=1.

[0172] (10) Calculate the overall similarity of materials:

[0173] Material overall similarity St = material classification hierarchy structure similarity Sc * material comprehensive attribute similarity Ssub

[0174] (11) Sort by the overall similarity St value of the compared materials in the data set, see Appendix Figure 9 ,Optionally set similarity threshold to control the number of similar materials recommended.

[0175] Furthermore, the classification structure tree hierarchical similarity calculation method Sa described in step (4) uses the distance-based concept similarity calculation formula:

[0176] Where C1 and C2 are the two values at the bottom of the material classification hierarchy tree, and the material brand is optional. Len(C1, C2) represents the shortest path length of C1 and C2 in the hierarchy tree, and Depth represents the maximum depth value of the hierarchy tree.

[0177] Furthermore, the attribute value similarity Sb of each layer of the classification structure tree in step (4) is i The calculation method is as follows:

[0178]

[0179] Furthermore, the calculation method for standardization / normalization of the chemical composition and material property data set described in step (6) is not limited to the following method:

[0180] A chemical composition or performance parameter value in the data set has an interval value or an exact value. The kth material instance is an exact value, denoted as Xk, and the jth material instance is an interval value, denoted as [Aj, Bj]. Then

[0181] minA=min{min{A1,A2,…,Aj,…},min{…Xk…}}

[0182] maxB=max{max{B1, B2,…,Bj,…},max{…Xk…}}

[0183] The interval value is standardized: Aj'=(Aj-minA) / (maxB-minA)

[0184] Bj'=(Bj-minA) / (maxB-minA)

[0185] Standardize the exact value: Xk'=(Xk-minAi) / (maxB-minA)

[0186] Furthermore, the similarity of the attribute values in step (7) is determined by the following method:

[0187] Calculation of similarity between interval values and exact values:

[0188]

[0189] Similarity calculation between exact values: Sij(A1,A2)=1-|A1-A2|

[0190] Similarity calculation of enumeration values:

[0191]

[0192] Furthermore, the combined weight calculation method described in step (8) is as follows:

[0193] The combination weight calculation includes subjective weight calculation and objective weight calculation, using the multiplication combination weighting method, as follows:

[0194]

[0195] Among them, ω 1 is the subjective weight, ω 2 is the objective weight.

[0196] The subjective weight calculation method uses the expert survey method. The objective weight calculation method uses the similarity deviation method:

[0197]

[0198] ωj≥0.

Claims

1. A user-collaborative similar material recommendation method, characterized in that The following steps are involved: (1) Establish a material classification hierarchy structure, select material type, material category, material subcategory and material brand, and maintain the basic public information of material data in the corresponding hierarchy. The field type in each hierarchy is an enumeration value; (2) Establish a material data set and maintain the data of chemical composition, material properties and key characteristic parameters at the bottom of the material classification hierarchy structure tree. The data types entered for chemical composition and material performance attribute parameters include exact values and interval values, and the data type entered for key characteristic parameters is enumeration value; (3) The user selects or inputs the reference material for similarity calculation and the data set involved in the similarity calculation, including the material classification hierarchy structure, chemical composition, material performance parameters and key characteristic parameters; the chemical composition includes the chemical elements C, Si, Mn, P and S, the material performance parameters include tensile strength, yield strength, elongation parameters, and the key characteristic parameters include specifications, applications, and material state parameters; (4) Calculate the similarity Sc of the material classification hierarchy structure. The similarity calculation formula between any two objects C1 and C2 is as follows: ， Where ωi is the weight of the attribute value similarity of each level of the structure tree, ωt is the weight of the concept similarity of the structure tree, Sa is the concept similarity of the structure tree level, and Sbi is the attribute value similarity of the i-th level of the structure tree; (5) Calculate the type similarity of chemical composition, material properties and key characteristic parameters in the dataset; The chemical composition type similarity calculation Sd1 is calculated as follows: Sd1=(C1∩C2) / (C1∪C2) Where C1∩C2 is the number of chemical elements or parameters that are matched between the reference material C1 and the reference material C2 in the data set and that participate in the similarity calculation, and C1∪C2 is the total number of chemical elements or parameters that participate in the matching; The type similarity calculation Sd2 of material properties is calculated as follows: Sd2=(C1∩C2) / (C1∪C2) Among them, C1∩C2 is the number of matching material performance parameters between the reference material C1 and the reference material C2 in the data set that participate in the similarity calculation, and C1∪C2 is the total number of performance parameters participating in the matching; The type similarity calculation Sd3 of the key feature parameters is calculated as follows: Sd3=(C1∩C2) / (C1∪C2) Wherein, C1∩C2 is the number of key feature parameters that match between the reference material C1 and the reference material C2 in the data set and participate in the similarity calculation, and C1∪C2 is the total number of key feature parameters participating in the matching; (6) Formulate processing rules for data interval values and default values, and perform data pre-processing; The method for handling chemical element interval values is to treat interval values ≥ or > A as the endpoint value A, treat interval values ≤ or < A as [0, A], and treat (A, B) or (A, B ] or [A, B) as [A, B]. The default value is filled with 0. The chemical composition dataset processed according to the above rules is normalized. The method for handling interval values of material performance parameters is to treat interval values ≥ or > A and ≤ or < A as the endpoint value A, and to treat (A, B) or (A, B] or [A, B) as [A, B]. The default value handling method is to stipulate that if the attribute index value of the compared material B in the attribute set is empty, the attribute values of both parties are equivalent. The material performance data set processed according to the above rules is standardized / normalized. The default value handling method for key feature parameters is: if the parameter value X of reference material A is empty, XA = "None" is agreed upon. If the reference material B in the attribute set has an empty value, the attribute values of both parties are agreed to be equivalent. If the parameter value is "Other", it matches "Other" or an empty keyword. (7) Calculate the chemical composition similarity matrix S1, material property similarity matrix S2 and key characteristic parameter attribute value similarity matrix S3; (8) The combined weight calculations are performed on the parameters in the chemical composition similarity matrix S1, the material performance similarity matrix S2, and the key characteristic parameter similarity matrix S3 to obtain the chemical composition attribute value similarity Se1, the material performance attribute value similarity Se2, and the key characteristic parameter attribute value similarity Se3: (9) Calculate the similarity of comprehensive material properties; Calculate the chemical composition similarity Schem: Schem=Sd1*Se1; Calculate the material performance attribute similarity Sprop: Sprop=Sd2*Se2; Calculate the key feature parameter attribute similarity Schar: Schar=Sd3*Se3; Calculate the similarity of comprehensive material properties: Ssub=ω1*Schem+ω2*Sprop+ω3*Schar, where ω1, ω2, and ω3 are the weights of chemical composition, material properties, and key characteristic attributes, respectively, and ω1+ω2+ω3=1; (10) Calculate the overall similarity of materials: the overall similarity of materials St = the similarity of the material classification hierarchy structure Sc * the similarity of the material comprehensive attributes Ssub; (11) The materials in the dataset are sorted according to their overall similarity St values. Users can control the number of similar materials recommended by setting a similarity threshold.

2. The user-collaborative similar material recommendation method according to claim 1, characterized in that: In step (4), the calculation method of the structure tree level concept similarity Sa is selected from the distance-based concept or semantic similarity calculation formula, including the following formula: , where C1 and C2 are the two values at the bottom of the material classification hierarchy tree. The material grade is selected. Len(C1, C2) represents the shortest path length between C1 and C2 in the hierarchy tree, and Depth represents the maximum depth value of the hierarchy tree. , where N1 and N2 represent the shortest paths between C1, C2 and the nearest common node concept word, respectively, and H represents the number of levels from the nearest common parent node C to the root node; , where des(C1,C2) represents the total hierarchical size of the structure tree, and the numerator represents the distance from the common parent node to the root node, which describes the commonality of C1 and C2; The calculation method of the attribute value similarity Sbi at the i-th level of the structure tree is as follows: 。 3. The user-collaborative similar material recommendation method according to claim 1, characterized in that: In step (6), the calculation method for standardization / normalization of chemical composition and material properties data set is as follows: A chemical composition or performance parameter value in the data set has both interval values and exact values, where the kth material instance is an exact value, denoted as Xk, and the jth material instance is an interval value, denoted as [Aj, Bj]. minA=min{min{A1, A2,…,Aj,…},min{…Xk…}} maxB=max{max{B1, B2,…,Bj,…}, max{…Xk…}} The interval value is standardized: Aj'=(Aj- minA) / (maxB- minA) Bj'=(Bj- minA) / (maxB- minA) The exact value is normalized: Xk'=(Xk- minAi) / (maxB- minA).

4. The user-collaborative similar material recommendation method according to claim 1, characterized in that: In step (7), the similarity calculation method of the attribute values is as follows: Calculate the similarity between interval values and exact values: ， Similarity calculation between exact values: , Similarity calculation of enumeration values: .

5. The user-collaborative similar material recommendation method according to claim 1, characterized in that: In step (8), the combined weight calculation includes subjective weight calculation and objective weight calculation, and adopts the multiplication combined weighting method, which is as follows: ， in, is the subjective weight, is the objective weight; The subjective weight calculation method uses the expert survey method or the hierarchical analysis method; the objective weight calculation method uses the similarity deviation method or the entropy method. The similarity deviation information method used is as follows: ， ， 。 6. The user-collaborative similar material recommendation method according to claim 1, characterized in that: In step (8), the chemical composition attribute value similarity Se1 is calculated as follows: or ; The calculation method of material performance attribute value similarity Se2 is as follows: or ; The calculation method of key feature attribute value similarity Se3 is as follows: or .

Images