A modular configuration method for complex products based on knowledge graph

Abstract

The present invention discloses a method for modular configuration of complex products based on a knowledge graph. Automatically generate a product family main structure based on historical product bills of materials; construct a configuration knowledge graph; automatically generate a configuration generation method from demand to material; and automatically configure products based on the configuration generation method from demand to material and the configuration knowledge graph. The present invention can quickly and intelligently formulate a product bill of materials that meets the personalized needs of users, can effectively solve the problem of difficulty in establishing the main structure and configuration of the product family, and improves product configuration efficiency. The intelligent modular configuration method provides support for mass personalized customization of complex products.

Description

Technical Field

[0001] The present invention relates to a product modular configuration method in the field of intelligent manufacturing, and in particular to a complex product modular configuration method based on a knowledge graph. Background Art

[0002] In the manufacturing industry, modular configuration technology for complex products is an important part of achieving mass customization. Mass customization is a production method that uses the cost and delivery time of mass production to manufacture personalized customized products, and is the goal pursued by Industry 4.0. Complex products usually refer to products with high R&D costs, intensive technology, complex structures, and a high degree of customization. With the increasing demand for low cost and short delivery time of personalized customized products, as well as the continuous development of intelligent manufacturing, the intelligence and explainability of modular configuration of products will help improve the efficiency of customized services in the manufacturing industry, shorten product design and production cycles, reduce product and material inventory, reduce product costs, and improve user satisfaction with customized products.

[0003] Product modular configuration refers to the rapid configuration of products that meet user needs from predefined or historical product series and materials. The modular configuration of complex products is a process of knowledge inheritance and reuse, and this knowledge is scattered and implicit in the bill of materials of a large number of instance products. In order to quickly configure the products required by users, it is necessary to establish a product modular configuration design platform before the modular configuration of products. The most important content is the establishment of the main structure and configuration method of the product family. In the main structure of the product family, users can select modules according to the configuration generation method, and can add personalized customization modules to obtain personalized customized products. Most of the materials in this customized product are general materials, with large batches and low costs. However, the customization cost of traditional manufacturing is high, and the customization production efficiency is low, which cannot meet people's actual needs. This will bring a lot of workload, and the analysis of multiple people will bring some subjective problems, such as different technicians, or even the same technician at different times, will give different configuration methods and terms, resulting in long time to establish the configuration system, poor consistency of configuration methods and names, etc., which cannot adapt to the diverse explainable and rapid configuration needs of users, and also make it difficult for the configuration system to form a closed loop of continuous optimization. From the existing inventions, all the current inventions involve highly professional domain knowledge, requiring a large amount of expert prior knowledge, and the complexity of configuration solution reduces the user's participation. For example, CN110114769A implements a system and method for semantic comparison product configuration model; CN1479904A designs a method for configuring products with a directed acyclic graph, through which the internal association of components can be tracked through a preferred embodiment (virtual tabulation), and intelligent search can calculate the internal association between components; CN106130305A discloses an electronic product equipped with a longitudinal vibration motor; CN111165002A discloses a parameter configuration method and related products for improving the flexibility of PT-RS configuration BWP; CN111038419A discloses a configurable vehicle power socket system.

[0004] In view of this, it is necessary to improve the existing product configuration methods and propose a set of big data-driven and explainable modular configuration methods for complex products, so as to realize the intelligent establishment and optimization of the main structure and configuration method of the product family and improve the efficiency of product modularization. Summary of the invention

[0005] In order to solve the technical problems existing in the background technology, the present invention aims to propose a modular configuration method for complex products based on knowledge graph.

[0006] The method of the present invention automatically generates the main structure of the product family based on the historical product material list, adopts the knowledge graph as the knowledge representation method to construct the configuration knowledge graph, automatically generates the configuration generation method from demand to material according to the mapping relationship between demand and material, and automatically configures the product according to the configuration generation method from demand to material and the configuration knowledge graph.

[0007] The technical solution adopted by the present invention to solve the technical problem is:

[0008] 1. A modular configuration method for complex products based on knowledge graph:

[0009] The specific process flow chart is as follows Figure 1 As shown, the method mainly includes:

[0010] Step 1: Generate the main structure of the product family;

[0011] Step 2: Build a configuration knowledge graph;

[0012] Step 3: Generate configuration generation method from demand to materials;

[0013] Step 4: Automatically configure the product based on the configuration generation method from demand to material and the configuration knowledge graph.

[0014] The S1 is specifically:

[0015] Step 1.1: Create a material classification template based on the key materials in the known historical product bill of materials. The material classification template contains the categories of key materials and the known relative values ​​of the materials.

[0016] The product bill of materials includes key materials and non-key materials. All materials have categories and relative values. The relative values ​​measure the importance of the materials in manufacturing.

[0017] The historical product bill of materials includes products of the same type but different models. Modules are composed of different materials according to the function and structure of the product.

[0018] Step 1.2: Extract key materials at each level in the historical product bill of materials according to the material classification template, and generate a weighted fuzzy matrix based on the quantity and relative value of key materials;

[0019] Step 1.3: For the weighted fuzzy matrix obtained at each level in the product bill of materials, cluster the products using a clustering algorithm. The clustering result obtained at one level is used as a product family division scheme. The weighted fuzzy matrices at each level are clustered separately to generate multiple product family division schemes.

[0020] Step 1.4: According to the product family division scheme, the materials of the product family main structure in the product family division scheme are divided into basic modules, mandatory modules, and optional modules to support the modular configuration of complex products based on the product family main structure;

[0021] A basic module means that the material corresponding to the module is selected by default during selection. A required module means that at least one material corresponding to the module of this category must be selected during configuration. An optional module means that the material corresponding to the module of this category can be selected or not selected during configuration.

[0022] Step 1.5: For each main structure of a product family in the product family division scheme, establish the universality, reuse times, and independence parameters of the main structure of the product family, and conduct a comprehensive analysis and evaluation of each product family division scheme in combination with the demand type to obtain the confidence of various product family division schemes;

[0023] Step 1.6: Select a product family division scheme based on comprehensive analysis and processing, and select the product family division scheme with the highest confidence as the optimal product family division scheme, so that the differences in the structure of the product bill of materials within the product family are as small as possible (material commonality and reuse times are as large as possible, and material independence is as small as possible), and the differences in the structure of the product bill of materials between product families are as large as possible (material commonality and reuse times are as small as possible, and material independence is as large as possible).

[0024] According to the product family division plan, all products in the historical product bill of materials are divided into various product families, and the bill of materials of each product under each product family are integrated to form the main structure bill of materials of the product family.

[0025] The step 1.5 is specifically as follows:

[0026] Obtain the evaluation index parameters such as universality, reuse times, independence, etc. of each product family main structure under each product family division scheme generated in step 1.3:

[0027] The product family universality mty is calculated as follows:

[0028]

[0029] Among them, JMS represents the material M of the product family j The total number of times used by all products TMS represents the total number of times all materials in the product family are used by all products.

[0030] The product family reuse times mfy are calculated as follows:

[0031]

[0032] Among them, FM j Indicates the number of times material j is reused in the product family.

[0033] Independence is calculated based on the design characteristics of the material:

[0034] Processing to obtain each material M j In each design characteristic T s The weight value is used to establish the design characteristic table of the material;

[0035] Table (5) Material design characteristics table

[0036] Products\Materials <![CDATA[T1]]> <![CDATA[T2]]> <![CDATA[T3]]> … <![CDATA[T k ]]> <![CDATA[M1]]> <![CDATA[A 11 ]]> <![CDATA[A 12 ]]> <![CDATA[A 13 ]]> … <![CDATA[A 1k ]]> <![CDATA[M2]]> <![CDATA[A 21 ]]> <![CDATA[A 22 ]]> <![CDATA[A 23 ]]> … <![CDATA[A 2k ]]> <![CDATA[M3]]> <![CDATA[A 31 ]]> <![CDATA[A 32 ]]> <![CDATA[A 33 ]]> … <![CDATA[A 3k ]]> … … … … … … <![CDATA[M m ]]> <![CDATA[A n1 ]]> <![CDATA[A n2 ]]> <![CDATA[A n3 ]]> … <![CDATA[A mk ]]>

[0037] In the design characteristics table, A js Indicates material M j In the design characteristic T s The weight value of .

[0038] Each weight value A js The design features are assigned values ​​according to their importance as follows:

[0039] For material M j , with k design characteristics {T1,T2,...,T k}, and if the importance of each design feature is ranked as T1>T2>…>T k , then the material M j The weight value A of the design characteristic T1 j1 Assign a value of 1, and then calculate material M according to the following formula j In the design characteristic T s The weight value A js :

[0040]

[0041] Among them, k represents material M j The number of design characteristics possessed, r represents the material M j Design characteristics of T s The importance ranking of material M j Any design feature that is not present is assigned a weight of 0.

[0042] Then calculate the independence between two materials according to the following formula:

[0043]

[0044] Among them, d ab Indicates material M a and material M b The independence between them, k is the number of design characteristics, A as Indicates material M a The weight value of the sth design feature;

[0045] Finally, according to the independence between two materials, the following formula is used to obtain the average value of the independence of all materials used in the product family as the independence of the product family:

[0046]

[0047] Among them, m represents the total number of materials in the product family;

[0048] Then, the confidence of each product family division scheme is obtained by weighted calculation based on the universality, number of reuses, and independence of each product family division scheme. The specific formula is as follows:

[0049]

[0050] Among them, EVA q represents the confidence of the product family partitioning scheme q, w mty represents the weight value of universality, represents the average value of the universality of all product families under the product family division scheme q, w mfy Indicates the weight value of the number of reuses, represents the average reuse times of all product families under product family partitioning scheme q, w mdl represents the weight value of universality, It represents the average independence of all product families under product family partitioning scheme q.

[0051] The step 2 is specifically as follows:

[0052] Step 2.1: Classify all materials under the bill of materials of all historical products and the bill of materials of all product family master structures according to the material coding method;

[0053] Step 2.4: Generate the first knowledge graph from all historical product bills of materials as the historical product knowledge graph, and generate the second knowledge graph from all product family main structure bills of materials as the product family main structure knowledge graph;

[0054] Specifically, it is divided into the following steps

[0055] 1) Extract the characteristics of the material, obtain the material category based on the material coding method, and build a transaction characteristic table based on the material characteristics and categories;

[0056] 2) Extract all materials in the bill of materials and map them to nodes in the knowledge graph. Use the material code as the main feature of the node to distinguish different materials. Other features of the material are used as attributes of the node in the configuration knowledge graph.

[0057] 3) Extract the hierarchical relationship between materials and map it to the relationship between nodes in the configuration knowledge graph.

[0058] Step 2.5: Merge the constructed historical product knowledge graph and the product family main structure knowledge graph to obtain the configuration knowledge graph.

[0059] The step three is specifically as follows:

[0060] Step 3.1: Use natural language processing tools to extract product function feature keywords in the requirements input by users, associate the product function feature keywords in the requirements with the keywords in the material characteristics, and thus construct a requirement-material semantic association table:

[0061] Table (7) Example of semantic association table between requirements and materials

[0062]

[0063] Step 3.2:

[0064] The association probability indicates the probability of having an association relationship between the two at the time of configuration.

[0065] 1) Establish the following requirements - material configuration generation method:

[0066] The probability of associating materials with requirements is set by the frequency of keywords in the materials, as shown in the following formula:

[0067]

[0068] in, Displays the material and requirement characteristic keyword R of category a r The probability of association between Indicates that it contains the demand characteristic keyword R r And the material category is the jth material of a, Indicates that it contains the demand characteristic keyword R r And the material category is the jth material of b, Indicates all the keywords containing the required characteristics R r The quantity of materials;

[0069] 2) Establish the following material-material configuration generation method:

[0070] Calculation and demand characteristics Keywords R r The probability of association between related materials is as follows:

[0071]

[0072] in, It represents the probability of association between materials of material category a and materials of material category b. represents the αth historical product bill of materials, express Keyword R in demand characteristics r Appears in materials of material category a, Indicated in The quantity of materials of material category b under the condition of Indicates the statistical summation of all historical product bills of materials. Indicates that all historical product bills of materials contain the demand characteristic keyword R r And the quantity of materials with material category b;

[0073] Step 3.3: Set the association probability threshold, and generate a third knowledge graph as the configuration generation method knowledge graph from the demand-material generation method and material-material generation method that are greater than the association probability threshold, so that candidate materials that meet the demand can be automatically selected according to the configuration generation method knowledge graph during configuration.

[0074] 2. A product configuration system based on knowledge graph, the system comprising:

[0075] Product family master structure generation unit, used to automatically generate product family master structure based on product historical bill of materials;

[0076] A first knowledge graph generation unit, used to create a configuration knowledge graph based on a product history bill of materials and a product family main structure bill of materials;

[0077] A second knowledge graph generation unit is used to extract user demand knowledge and obtain a configuration generation method knowledge graph based on the configuration generation method from the demand to the material;

[0078] A product configuration unit is used to generate corresponding complex product modular configuration results based on the configuration knowledge graph and the configuration generation method knowledge graph.

[0079] 3. A computer program of a browser / server architecture, which implements the method described above when the computer program is requested to be executed.

[0080] The present invention utilizes the historical product bill of materials to automatically obtain the main structure of the product family, and uses a structured knowledge graph to represent materials and the relationships between materials. Then, based on the configuration generation method from demand to materials, the materials are reasonably configured and combined, thereby quickly and intelligently formulating a product bill of materials that meets the personalized needs of users. This can effectively solve the problem of difficulty in establishing the main structure and configuration of the product family, and reduce the workload of product modularization.

[0081] Advantages and significance of the present invention:

[0082] Compared with the prior art, the method proposed in the present invention is a data-driven intelligent method. Through the construction of a knowledge graph, it improves the reusability of knowledge, realizes configuration explainability, and reduces the subjectivity in the modular configuration of complex products, thereby improving the efficiency and quality of modularization of complex products, reducing the workload of modularization, and improving the automatic configuration efficiency and configuration accuracy of complex products.

[0083] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations of the present invention.

[0085] In the figure:

[0086] Figure 1 The figure is an overall flow chart of the configuration method of the present invention.

[0087] Figure 2 A configuration system framework diagram constructed for the present invention. DETAILED DESCRIPTION

[0088] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0089] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0090] The embodiments of the present invention and their implementation are as follows:

[0091] Step 1: Automatically generate the main structure of the product family

[0092] Step 1.1: Take the main parts involved in the R&D, design, production, manufacturing, and procurement and supply of self-made and purchased parts of the product as key materials. According to the key materials in the known historical product bill of materials, establish a material classification template. The material classification template contains the categories and relative values ​​of key materials;

[0093] A product bill of materials is made up of materials.

[0094] Self-made parts refer to parts manufactured by ourselves, and purchased parts refer to parts purchased from outside.

[0095] The material classification template includes key materials of different categories and the relative value of each key material.

[0096] Relative value represents the relative importance of a material in the product life cycle and is a relatively subjective known value.

[0097] Step 1.2: Extract key materials at each level in the historical product bill of materials according to the material classification template, and generate a weighted fuzzy matrix based on the quantity and relative value of key materials;

[0098] In the specific implementation, the materials at each level are traversed, and a tree structure is established for the materials at each level. If a material has no child node in the next existing level, the material is copied as its child node in the next existing level.

[0099] The historical product bill of materials includes products of the same type but different models. For example, the P30 mobile phone and the Mate30 smart phone are all in the same historical product bill of materials, and the same model of products only appears once in the product bill of materials.

[0100] The levels in the key bill of materials are consistent with the levels in the historical product bill of materials, and the levels are set according to the manufacturing requirements of the product.

[0101] The key materials at each level in the bill of materials are processed as follows:

[0102] First, establish the product-material quantity matrix, as shown in Table (1):

[0103] Table (1) Product-Material Quantity Matrix

[0104]

[0105]

[0106] In the product-material quantity matrix, C ij Indicates product P i Used material M j the number of

[0107] Then, each column in the product-material quantity matrix is ​​normalized by a linear function according to the following formula:

[0108]

[0109] Among them, N ij is the quantity C ij The normalized result of θ is a non-zero parameter, and its specific value is an arbitrary small value, so that the normalized interval is (0, 1); min(C j ) indicates that for material M j The following is a single product P i The minimum number used, max(C j ) indicates that for material Mj The following is a single product P i The maximum number used;

[0110] After normalizing each column in the product-material quantity matrix, the product-material fuzzy matrix is ​​obtained, as shown in Table (2):

[0111] Table (2) Product-material fuzzy matrix

[0112] Products\Materials <![CDATA[M1]]> <![CDATA[M2]]> <![CDATA[M3]]> … <![CDATA[M m ]]> <![CDATA[P1]]> <![CDATA[N 11 ]]> <![CDATA[N 12 ]]> <![CDATA[N 13 ]]> … <![CDATA[N 1m ]]> <![CDATA[P2]]> <![CDATA[N 21 ]]> <![CDATA[N 22 ]]> <![CDATA[N 23 ]]> … <![CDATA[N 2m ]]> <![CDATA[P3]]> <![CDATA[N 31 ]]> <![CDATA[N 32 ]]> <![CDATA[N 33 ]]> … <![CDATA[N 3m ]]> … … … … … … <![CDATA[P n ]]> <![CDATA[N n1 ]]> <![CDATA[N n2 ]]> <![CDATA[N n3 ]]> … <![CDATA[N nm ]]>

[0113] Then use the relative value of the materials to assign weights to the normalized results according to the following formula:

[0114] W ij =N ij *V j

[0115] Among them, W ij For product P i Use material M j The importance of V j Indicates material M j the relative value of

[0116] After assigning weights, the product-material weighted fuzzy matrix is ​​obtained, as shown in Table (3):

[0117] Table (3) Product-material weighted fuzzy matrix

[0118] Products\Materials <![CDATA[M1]]> <![CDATA[M2]]> <![CDATA[M3]]> … <![CDATA[M m ]]> <![CDATA[P1]]> <![CDATA[W 11 ]]> <![CDATA[W 12 ]]> <![CDATA[W 13 ]]> … <![CDATA[W 1m ]]> <![CDATA[P2]]> <![CDATA[W 21 ]]> <![CDATA[W 22 ]]> <![CDATA[W 23 ]]> … <![CDATA[W 2m ]]> <![CDATA[P3]]> <![CDATA[W 31 ]]> <![CDATA[W 32 ]]> <![CDATA[W 33 ]]> … <![CDATA[W 3m ]]> … … … … … … <![CDATA[P n ]]> <![CDATA[W n1 ]]> <![CDATA[W n2 ]]> <![CDATA[W n3 ]]> … <![CDATA[W nm ]]>

[0119] Different product-material weighted fuzzy matrices are obtained corresponding to the key material lists at different levels.

[0120] Step 1.3: Using the K-means++ clustering algorithm, each product-material weighted fuzzy matrix generated in step 1.2 is used to cluster the products, divide the product families, and obtain various product family division schemes.

[0121] Among them, the steps of K-means++ clustering algorithm are as follows:

[0122] ① Randomly select a product from the historical product bill of materials as the initial cluster center;

[0123] ② Calculate the distance between each product sample and the nearest cluster center, expressed as:

[0124]

[0125]

[0126] Among them, D ic Indicates product Pi The product P with the cluster center i The distance between Indicates product P i Using the importance vector of all materials, The product representing cluster center c adopts the importance vector of all materials;

[0127] ③ Calculate the probability of each product other than the current cluster center being selected as the next cluster center, expressed as:

[0128]

[0129] The cumulative probability interval of each product is calculated by the above formula, and then the system generates a random number, and the product corresponding to the random number falling in the cumulative probability interval is the next cluster center.

[0130] ④ Repeat steps 2 and 3 until k cluster centers are selected;

[0131] ⑤Calculate P for each product i The distances to the k cluster centers are calculated respectively, and the clusters are divided into the classes corresponding to the cluster centers with the smallest distances;

[0132] ⑥ Recalculate the cluster center of each class, expressed as:

[0133]

[0134] In the formula, new xc represents the new cluster center c, s represents the number of products contained in cluster c;

[0135] ⑦ Repeat steps 5 and 6 until the cluster centers no longer change. At this point, clustering ends and k product families are generated. n is the total number of products.

[0136] Step 1.4: According to the product family division plan, materials are divided into three types of modules: basic modules, mandatory modules, and optional modules to support the modular configuration of complex products based on the main structure of the product family.

[0137] Basic modules: materials used by every product in the product family.

[0138] Required module: Select at least one material of the same type in the product family except the basic module to form the material of the product.

[0139] Optional modules: Select or not select materials that make up the product from the same type of materials in the product family except the basic module.

[0140] In the product family division scheme, each product P iUse each material M j The number of binary processing is performed, and a product material 0 / 1 matrix is ​​established for each material of the same type in each product family in the product family division scheme, as shown in Table (4):

[0141] Table (4) Product-Material 0 / 1 Matrix for a Product Family

[0142] Products\Materials <![CDATA[M1]]> <![CDATA[M2]]> <![CDATA[M3]]> … <![CDATA[M m ]]> <![CDATA[P1]]> <![CDATA[K 11 ]]> <![CDATA[K 12 ]]> <![CDATA[K 13 ]]> … <![CDATA[K 1m ]]> <![CDATA[P2]]> <![CDATA[K 21 ]]> <![CDATA[K 22 ]]> <![CDATA[K 23 ]]> … <![CDATA[K 2m ]]> <![CDATA[P3]]> <![CDATA[K 31 ]]> <![CDATA[K 32 ]]> <![CDATA[K 33 ]]> … <![CDATA[K 3m ]]> … … … … … … <![CDATA[P n ]]> <![CDATA[K n1 ]]> <![CDATA[K n2 ]]> <![CDATA[K n3 ]]> … <![CDATA[K nm ]]>

[0143] In the product material 0 / 1 matrix, K ij Indicates product P i Is material M used? j The state quantity K ij The value is 0 or 1, K ij =0 means product P i Unused material M j , K ij =1 indicates product P i Used material M j .

[0144] Different types of materials establish different product material 0 / 1 matrices. For each material in each product material 0 / 1 matrix, calculate the material M j The number of times the status is used by all products MP j :

[0145]

[0146] Then make a judgment:

[0147] If material M j MP j If it is equal to i, it means material M j Used by every product, that is, material M j is the basic module; otherwise, material M j Not a basic module;

[0148] If material M j MP j If it is less than i, then further calculate the product P i Used and materials M j PM status count of other materials of the same category that are not basic modules i :

[0149]

[0150] If product P i Status times PM i If it is greater than 0, it means that the material M jOther materials of the same category that are not basic modules are mandatory modules for the main structure of the product family;

[0151] If product P i Status times PM i If it is 0, it means that material M j Other materials of the same category that are not basic modules are optional modules for the main structure of the product family.

[0152] Step 1.5: Establish the following universality, reuse times, and independence parameters to obtain the universality, reuse times, independence and other evaluation index parameters of each product family main structure under each product family division scheme generated in step 1.3:

[0153] The product family universality mty is calculated as follows:

[0154]

[0155] Among them, JMS represents the material M of the product family j The total number of times used by all products, that is, the material M in the product-material quantity matrix j All C in the same column ij Cumulative result: TMS represents the total number of times all materials in the product family are used by all products, that is, the cumulative result of JMS of all materials.

[0156] The product family reuse times mfy are calculated as follows:

[0157]

[0158] Among them, FM j Represents the number of times material j is reused in a product family, which is the sum of the number of times material j is used by all products in the product family minus one.

[0159] Independence is calculated based on the design characteristics of the material:

[0160] Processing to obtain each material M j In each design characteristic T s The weight value is used to establish the material design characteristic table, as shown in the following table;

[0161] Table (5) Material design characteristics table

[0162]

[0163]

[0164] In the design characteristics table, A js Indicates material M j In the design characteristic T s The weight value of .

[0165] Each weight value A js The design features are assigned values ​​according to their importance as follows:

[0166] For material M j , with k design characteristics {T1,T2,...,T k}, and if the importance of each design feature is ranked as T1>T2>…>T k , then the material M j The weight value A of the design characteristic T1 j1 Assign a value of 1, and then calculate material M according to the following formula j In the design characteristic T s The weight value A js :

[0167]

[0168] Among them, k represents material M j The number of design characteristics possessed, r represents the material M j Design characteristics of T s The importance ranking of material M j Any design feature that is not present is assigned a weight of 0.

[0169] There are m materials and k design characteristics in total. All materials are assigned design characteristic weights in this way, and the material design characteristic table shown in Table (5) is obtained.

[0170] Then calculate the independence between two materials according to the following formula:

[0171]

[0172] Among them, d ab Indicates material M a and material M b The independence between them, k is the number of design characteristics, A as Indicates material M a The weight value of the sth design feature;

[0173] Finally, according to the independence between two materials, the following formula is used to obtain the average value of the independence of all materials used in the product family as the independence of the product family:

[0174]

[0175] Where m represents the total number of materials in the product family.

[0176] Then, according to the universality, number of reuses, and independence of each product family division scheme, a weighted calculation is performed to obtain the confidence of each product family division scheme. According to the confidence, a division scheme that meets the requirements is selected to generate the main structure material list of the product family. The specific formula is as follows:

[0177]

[0178] Among them, EVA q represents the confidence of the product family partitioning scheme q, w mty represents the weight value of universality, represents the average value of the universality of all product families under the product family division scheme q, w mfy Indicates the weight value of the number of reuses, represents the average reuse times of all product families under product family partitioning scheme q, w mdl represents the weight value of universality, It represents the average independence of all product families under product family partitioning scheme q.

[0179] Each product family division scheme includes multiple product families, and the weighted average of the product family evaluations is calculated as the confidence of the division scheme.

[0180] Step 1.6: Select the product family division scheme with the highest confidence as the optimal product family division scheme, and use the product family generated by the optimal product family division scheme as the product family main structure, which is the optimal main structure; merge the bill of materials of each product in each product family main structure as the bill of materials of the product family main structure of the product family.

[0181] Step 2: Build a configuration knowledge graph

[0182] Step 2.1: Analyze the historical product bill of materials and the main structure bill of materials of all product families, and classify all materials under the historical product bill of materials and the main structure bill of materials of all product families according to the material coding method;

[0183] Step 2.2: Extract material information based on templates and natural language processing methods, generate material characteristics, obtain material categories based on material coding methods, and build a property table based on material attributes and categories;

[0184] Step 2.3: Extract all materials in the bill of materials and map them to nodes in the knowledge graph. The material code is used as the main feature of the node to distinguish different materials, and each node contains all the material features of the corresponding material.

[0185] Step 2.4: Extract the hierarchical relationship between materials from the historical product bill of materials and the main structure bill of materials of all product families, and map them into the relationship between nodes in the configuration knowledge graph;

[0186] Step 2.4: Based on the nodes and relationships obtained in steps 2.2, 2.3 and 2.4, a first knowledge graph is generated from all historical product bills of materials as the historical product knowledge graph, and a second knowledge graph is generated from all product family main structure bills of materials as the product family main structure knowledge graph;

[0187] Step 2.5: Fuse the constructed historical product knowledge graph with the product family main structure knowledge graph to obtain the configuration knowledge graph to support the automatic configuration of complex products;

[0188] The specific implementation can re-analyze the hierarchy of historical product bills of materials and the main structure bills of materials of all product families, extract the hierarchical relationship between materials, and make the hierarchical relationship between nodes in the knowledge graph consistent with the hierarchical relationship between materials.

[0189] Extract other information of materials based on templates and natural language processing methods, generate material attributes, identify material categories based on material coding methods, and build a property table based on material attributes and categories;

[0190] Step 3: Automatically generate configuration generation method from demand to materials

[0191] Step 3.1: Classify the user's needs and use the Jiagu natural language processing tool to extract the product function feature keywords in the user's input needs, map the corresponding demand scenarios to the product's functions and features, and associate the product function feature keywords in the needs with the keywords in the material characteristics, thereby constructing a demand-material semantic association table, as shown in Table (7):

[0192] Table (7) Example of semantic association table between requirements and materials

[0193]

[0194] Step 3.2:

[0195] 1) Establish the following requirements - material configuration generation method:

[0196] The probability of associating materials with requirements is set by the frequency of keywords in the materials, as shown in the following formula:

[0197]

[0198] in, Displays materials of material category a and required characteristic keyword R r The probability of association between Indicates that it contains the demand characteristic keyword R r And the material category is the jth material of a, Indicates that it contains the demand characteristic keyword R r And the material category is the jth material of b, Indicates all the keywords containing the required characteristics R r The quantity of materials;

[0199] 2) Establish the following material-material configuration generation method:

[0200] Calculation and demand characteristics Keywords R r The probability of association between related materials is as follows:

[0201]

[0202] in, It represents the probability of association between materials of material category a and materials of material category b. represents the αth historical product bill of materials, express Keyword R in demand characteristics r Appears in materials of material category a, Indicated in The quantity of materials of material category b under the condition of Indicates the statistical summation of all historical product bills of materials. Indicates that all historical product bills of materials contain the demand characteristic keyword R r And the quantity of materials with material category b;

[0203] In this way, based on the product function feature keywords and material keywords, all material nodes in the configuration knowledge graph are traversed to generate and complete the configuration.

[0204] Step 3.3: Set an association probability threshold, and generate a third knowledge graph as the configuration generation method knowledge graph from the demand-material generation method and material-material generation method that are greater than the association probability threshold, so that candidate materials that meet the demand can be automatically selected according to the configuration generation method knowledge graph during configuration;

[0205] Step 4: Automatic configuration of complex products

[0206] Step 4.1: Use natural language processing tools to extract the demand keywords entered by the user, and select the automatic configuration method for complex products based on the configuration generation method from demand to materials;

[0207] Automatic configuration methods include complex product modular configuration based on historical products, complex product modular configuration based on product family main structure, and complex product modular configuration based on thing characteristic table.

[0208] Step 4.2: Select a complex product modular configuration method based on historical products, match user needs with materials through the configuration generation method knowledge graph, select a historical product containing the material, and for each material in the selected historical product bill of materials, calculate the graph structure similarity between the material and other materials of the same category that meet the needs based on the resource allocation algorithm. The same type of materials with high graph structure similarity are used as candidate materials for the material, thereby obtaining a historical product bill of materials that meets user needs. The resource allocation algorithm calculates the association between nodes based on the shared neighbors of the nodes, as shown in the following formula:

[0209]

[0210] Where N(u) represents the set of neighboring nodes of material node u. When the value approaches 0, it means that the correlation between the two materials is the smallest. Conversely, the greater the correlation, the greater the substitutability between the two materials.

[0211] Step 4.3: If no historical product meets the user's needs, a complex product modular configuration method based on the main structure of the product family is selected, and the user's needs are matched with the materials through the configuration generation method knowledge graph. A main structure of a product family that contains the material is selected. For the selected main structure of the product family, the materials of each level of the main structure of the product family are automatically selected according to the configuration generation method related to the user's needs in the configuration generation method knowledge graph. The basic modules are automatically selected. For any category of mandatory modules, one or more materials that meet the user's needs are automatically selected according to the cost and delivery time of the materials. For any category of optional modules, the user needs to select them according to their needs, so as to obtain a product family main structure material list that meets the user's needs;

[0212] Step 4.4: If any material in the main structure of the product family still cannot meet the user's needs, then select the variant configuration method based on the thing characteristic table. For the materials in the candidate historical products or product family main structure that do not meet the needs, query all candidate materials of the same category according to the material's thing characteristic table. The user selects one or more candidate materials according to the needs and adds them to the candidate historical product material list or the product family main structure material list. If there is still no material in the thing characteristic table that meets the needs, start the new material design process and add a material that meets the needs to the configuration knowledge graph as a new node in the configuration knowledge graph. Then, add the material as a dedicated module to the candidate historical product material list or the product family main structure material list, and finally complete the automatic configuration of complex products.

[0213] By performing the above steps, the modular configuration problem of complex products can be solved.

[0214] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0215] Those skilled in the art will appreciate that all or part of the steps in the above-mentioned embodiments can be accomplished by a computer program of a browser / server architecture, which is stored in a distributed server and executes all or part of the steps of the methods described in the various embodiments of the present application on the client. The client includes: computers, mobile devices, embedded systems, and other media that can store program codes.

[0216] It should also be noted that the embodiments of the present invention are not limited to the specific details in the above-mentioned embodiments. Within the technical concept of the embodiments of the present invention, the technical solutions of the embodiments of the present invention can be subjected to a variety of simple modifications, and these simple modifications all belong to the protection scope of the embodiments of the present invention. At the same time, the various specific technical features described in the above-mentioned specific embodiments can be combined in any suitable manner without contradiction, and as long as they do not violate the concept of the embodiments of the present invention, they should also be regarded as the contents disclosed by the embodiments of the present invention.

Claims

1. A modular configuration method for complex products based on knowledge graph, characterized by: The methods mainly include: Step 1: Generate the main structure of the product family; Step 2: Build a configuration knowledge graph; Step 3: Generate configuration generation method from demand to materials; Step 4: Automatically configure the product based on the configuration generation method from demand to material and the configuration knowledge graph; The specific step 4 is: Automatic configuration of complex products Step 4.1: Use natural language processing tools to extract the demand keywords entered by the user, and select the automatic configuration method for complex products based on the configuration generation method from demand to materials; Automatic configuration methods include complex product modular configuration based on historical products, complex product modular configuration based on product family main structure, and complex product modular configuration based on thing characteristic table. Step 4.2: Select a complex product modular configuration method based on historical products, match user needs with materials through the configuration generation method knowledge graph, select a historical product containing the material, and for each material in the selected historical product bill of materials, calculate the graph structure similarity between the material and other materials of the same category that meet the needs based on the resource allocation algorithm. The same type of materials with high graph structure similarity are used as candidate materials for the material, thereby obtaining a historical product bill of materials that meets user needs. The resource allocation algorithm calculates the association between nodes based on the shared neighbors of the nodes, as shown in the following formula: Where N(u) represents the set of neighboring nodes of material node u. When the value approaches 0, it means that the correlation between the two materials is the smallest. Conversely, the greater the correlation, the greater the substitutability between the two materials. Step 4.3: If no historical product meets the user's needs, a complex product modular configuration method based on the main structure of the product family is selected, and the user's needs are matched with the materials through the configuration generation method knowledge graph. A main structure of a product family that contains the material is selected. For the selected main structure of the product family, the materials of each level of the main structure of the product family are automatically selected according to the configuration generation method related to the user's needs in the configuration generation method knowledge graph. The basic modules are automatically selected. For any category of mandatory modules, one or more materials that meet the user's needs are automatically selected according to the cost and delivery time of the materials. For any category of optional modules, the user needs to select them according to their needs, so as to obtain a product family main structure material list that meets the user's needs; Step 4.4: If any material in the main structure of the product family still cannot meet the user's needs, then select the variant configuration method based on the thing characteristic table. For the materials in the candidate historical products or product family main structure that do not meet the needs, query all candidate materials of the same category according to the material's thing characteristic table. The user selects one or more candidate materials according to the needs and adds them to the candidate historical product material list or the product family main structure material list. If there is still no material in the thing characteristic table that meets the needs, start the new material design process and add a material that meets the needs to the configuration knowledge graph as a new node in the configuration knowledge graph. Then, add the material as a dedicated module to the candidate historical product material list or the product family main structure material list, and finally complete the automatic configuration of complex products.

2. According to claim 1, a complex product modular configuration method based on knowledge graph is characterized in that: The step 1 is specifically as follows: Step 1.1: Create a material classification template based on the key materials in the historical product bill of materials. The material classification template contains the categories of key materials and the relative value of the materials; Step 1.2: Extract key materials at each level in the historical product bill of materials according to the material classification template, and generate a weighted fuzzy matrix based on the quantity and relative value of key materials; Step 1.3: For the weighted fuzzy matrix obtained at each level in the product bill of materials, cluster the products using a clustering algorithm. The clustering result obtained at one level is used as a product family division scheme. The weighted fuzzy matrices at each level are clustered separately to generate multiple product family division schemes. Step 1.4: According to the product family division plan, divide the materials of the main structure of the product family in the product family division plan into basic modules, mandatory modules, and optional modules; Step 1.5: For each main structure of a product family in the product family division scheme, establish the universality, reuse times, and independence parameters of the main structure of the product family, and conduct a comprehensive analysis and evaluation of each product family division scheme in combination with the demand type to obtain the confidence of various product family division schemes; Step 1.6: Select the product family division scheme with the highest confidence as the optimal product family division scheme. Divide all products in the historical product bill of materials into product families according to the product family division scheme. The bill of materials of each product under each product family is integrated to form the main structure bill of materials of the product family.

3. The method for modular configuration of complex products based on knowledge graph according to claim 2, characterized in that: The step 1.5 is specifically as follows: Obtain the evaluation index parameters such as universality, reuse times, independence, etc. of each product family main structure under each product family division scheme generated in step 1.3: The product family universality mty is calculated as follows: Among them, JMS represents the material M of the product family j The total number of times a material is used by all products. TMS represents the total number of times all materials in a product family are used by all products. The product family reuse times mfy are calculated as follows: Among them, FM j Indicates the number of times material j is reused in the product family; Independence is calculated based on the design characteristics of the material: Processing to obtain each material M j In each design characteristic T s The weight value is used to establish the design characteristic table of the material; Table (5) Material design characteristics table In the design characteristics table, A js Indicates material M j In the design characteristic T s The weight value of Each weight value A js The design features are assigned values ​​according to their importance as follows: For material M j , with k design characteristics {T1,T2,...,T k }, and if the importance of each design feature is ranked as T1>T2>…>T k , then the material M j The weight value A of the design characteristic T1 j1 Assign a value of 1, and then calculate material M according to the following formula j In the design characteristic T s The weight value A js : Among them, k represents material M j The number of design characteristics possessed, r represents the material M j Design characteristics of T s The importance ranking of material M j Any design feature that is not present is assigned a weight of 0; Then calculate the independence between two materials according to the following formula: Among them, d ab Indicates material M a and material M b The independence between them, k is the number of design characteristics, A as Indicates material M a The weight value of the sth design feature; A bs Indicates material M b The weight value of the sth design feature; Finally, according to the independence between two materials, the following formula is used to obtain the average value of the independence of all materials used in the product family as the independence of the product family: Among them, m represents the total number of materials in the product family; Then, the confidence of each product family division scheme is obtained by weighted calculation based on the universality, number of reuses, and independence of each product family division scheme. The specific formula is as follows: Among them, EVA q represents the confidence of the product family partitioning scheme q, w mty represents the weight value of universality, represents the average value of the universality of all product families under the product family division scheme q, w mfy Indicates the weight value of the number of reuses, represents the average reuse times of all product families under product family partitioning scheme q, w mdl represents the weight value of universality, It represents the average independence of all product families under product family partitioning scheme q.

4. The method for modular configuration of complex products based on knowledge graph according to claim 1, characterized in that: The step 2 is specifically as follows: Step 2.1: Classify all materials under the bill of materials of all historical products and the bill of materials of all product family master structures according to the material coding method; Step 2.4: Generate the first knowledge graph from all historical product bills of materials as the historical product knowledge graph, and generate the second knowledge graph from all product family main structure bills of materials as the product family main structure knowledge graph; Step 2.5: Merge the constructed historical product knowledge graph and the product family main structure knowledge graph to obtain the configuration knowledge graph.

5. The method for modular configuration of complex products based on knowledge graph according to claim 1, characterized in that: The step three is specifically as follows: Step 3.1: Use natural language processing tools to extract product function feature keywords in the requirements input by users, associate the product function feature keywords in the requirements with the keywords in the material characteristics, and thus construct a requirement-material semantic association table: Table (7) Example of semantic association table between requirements and materials Step 3.2: 1) Establish the following requirements - material configuration generation method: Set the probability of associating materials with requirements by the frequency of keywords in the materials: in, Displays the material and requirement characteristic keyword R of category a r The probability of association between Indicates that it contains the demand characteristic keyword R r And the material category is the jth material of a, Indicates that it contains the demand characteristic keyword R r And the material category is the jth material of b, Indicates all the keywords containing the required characteristics R r The quantity of materials; 2) Establish the following material-material configuration generation method: Calculation and demand characteristics Keywords R r The probability of association between related materials: in, It represents the probability of association between materials of material category a and materials of material category b. represents the αth historical product bill of materials, express Keyword R in demand characteristics r Appears in materials of material category a, Indicated in The quantity of materials of material category b under the condition of Indicates the statistical summation of all historical product bills of materials. Indicates that all historical product bills of materials contain the demand characteristic keyword R r And the quantity of materials with material category b; Step 3.3: Set the association probability threshold, and generate a third knowledge graph as the configuration generation method knowledge graph from the demand-material generation method and material-material generation method that are greater than the association probability threshold, so that candidate materials that meet the demand can be automatically selected according to the configuration generation method knowledge graph during configuration.

6. A product configuration system based on knowledge graph, characterized in that: The system comprises: Product family master structure generation unit, used to automatically generate product family master structure based on product historical bill of materials; A first knowledge graph generation unit, used to create a configuration knowledge graph based on a product history bill of materials and a product family main structure bill of materials; A second knowledge graph generation unit is used to extract user demand knowledge and obtain a configuration generation method knowledge graph based on the configuration generation method from the demand to the material; A product configuration unit, used to generate a corresponding complex product modular configuration result based on the configuration knowledge graph and the configuration generation method knowledge graph; The way in which the product configuration unit generates the corresponding complex product modular configuration result is as follows: Step 4.1: Use natural language processing tools to extract the demand keywords entered by the user, and select the automatic configuration method for complex products based on the configuration generation method from demand to materials; Automatic configuration methods include complex product modular configuration based on historical products, complex product modular configuration based on product family main structure, and complex product modular configuration based on thing characteristic table. Step 4.2: Select a complex product modular configuration method based on historical products, match user needs with materials through the configuration generation method knowledge graph, select a historical product containing the material, and for each material in the selected historical product bill of materials, calculate the graph structure similarity between the material and other materials of the same category that meet the needs based on the resource allocation algorithm. The same type of materials with high graph structure similarity are used as candidate materials for the material, thereby obtaining a historical product bill of materials that meets user needs. The resource allocation algorithm calculates the association between nodes based on the shared neighbors of the nodes, as shown in the following formula: Where N(u) represents the set of neighboring nodes of material node u. When the value approaches 0, it means that the correlation between the two materials is the smallest. Conversely, the greater the correlation, the greater the substitutability between the two materials. Step 4.3: If no historical product meets the user's needs, a complex product modular configuration method based on the main structure of the product family is selected, and the user's needs are matched with the materials through the configuration generation method knowledge graph. A main structure of a product family that contains the material is selected. For the selected main structure of the product family, the materials of each level of the main structure of the product family are automatically selected according to the configuration generation method related to the user's needs in the configuration generation method knowledge graph. The basic modules are automatically selected. For any category of mandatory modules, one or more materials that meet the user's needs are automatically selected according to the cost and delivery time of the materials. For any category of optional modules, the user needs to select them according to their needs, so as to obtain a product family main structure material list that meets the user's needs; Step 4.4: If any material in the main structure of the product family still cannot meet the user's needs, then select the variant configuration method based on the thing characteristic table. For the materials in the candidate historical products or product family main structure that do not meet the needs, query all candidate materials of the same category according to the material's thing characteristic table. The user selects one or more candidate materials according to the needs and adds them to the candidate historical product material list or the product family main structure material list. If there is still no material in the thing characteristic table that meets the needs, start the new material design process and add a material that meets the needs to the configuration knowledge graph as a new node in the configuration knowledge graph. Then, add the material as a dedicated module to the candidate historical product material list or the product family main structure material list, and finally complete the automatic configuration of complex products.

Images