A knowledge graph-based method and system for generating cable processing technology

Abstract

This invention discloses a method and system for generating cable processing technology based on knowledge graphs, comprising the following steps: 1. Modeling cable processing technology knowledge, establishing a CPKG model, extracting process information from cable processing technology documents and 3D models, converting the relationships between entity subtypes and attributes into relationships between classes, and creating a cable processing process pattern layer. 2. Generating a structure graph and embedding it into the cable processing process, creating an ICPKG structure tree, constructing a distributed embedding reasoning model, and continuously completing the CPKG by calculating the subgraph similarity between ICPKG and CPKGD. 3. Producing a cable processing technology planning RDF / XML document for output. This invention aims to use knowledge graph technology to automatically generate cable processing technology flows, reduce repetitive calculations and waste of process knowledge in cable processing, thereby improving the automation and intelligence of the cable processing process.

Description

Technical Field

[0001] This invention belongs to the field of cable manufacturing technology, specifically relating to a method and system for generating cable processing technology based on knowledge graphs. Background Technology

[0002] Traditional cable processing planning is cumbersome, repetitive, time-consuming, and inefficient. It largely relies on manual design, which can easily limit product quality. Furthermore, the differing experience and design approaches of different process engineers make it difficult to establish consistent, standardized, and standardized process planning. These problems make cable manufacturing increasingly ill-suited to market demands and lagging behind modern production. While some intelligent generation solutions for cable process planning have been proposed in recent years, which can identify products through characteristic descriptions and thus determine their processes, this process still relies on process engineers calculating cable parameters. However, the diverse applications of cables place varying requirements on different parameters and processes, making calculations tedious, repetitive, and prone to errors. The limited utilization of existing cable information documents further slows down the generation of cable process flow documents, severely hindering the improvement of cable processing preparation efficiency. Therefore, improving the efficiency and standardization of cable processing planning, and achieving knowledge-based and intelligent cable processing planning, is urgently needed in cable manufacturing.

[0003] The development of artificial intelligence (AI) technology has brought new impetus to the manufacturing industry, promoting its transformation from automation to intelligence. This necessitates the development of intelligent process planning to meet the requirements of intelligent manufacturing. In intelligent process design, accurately representing and utilizing process knowledge is a crucial issue. This process knowledge originates from expert experience, process documents, and design documents. Therefore, without effective organization and management of process knowledge, it is impossible to establish an effective and accurate knowledge base. Knowledge graphs (KG), as a branch of AI, are effective tools for knowledge management. Process knowledge extraction, part drawing construction, and process route generation are important components of intelligent process design. A PKG is a domain-specific KG. PKG primarily manages process knowledge and executes manufacturing process planning. A general-purpose KG is a structured knowledge base for a general domain, containing a large amount of common-sense knowledge from the open world. The main difference between PKG and general-purpose KG lies in the data source and object orientation. Process knowledge is dynamic, highly interconnected, and describes process entities that change over time. Open-world knowledge is robust and, once established, may not change for a long period. PKG is also similar to general-purpose KG in many ways. First, many methods used in general knowledge bases can be migrated to PKG, such as knowledge extraction, entity alignment, and knowledge fusion. Second, some mature tools can be used with PKG, such as the Neo4j database.

[0004] Computer-aided process planning (CAPP) has been widely applied to process planning of various parts and products, bringing convenience to process engineers. However, it cannot reuse process knowledge and experience. PKG can perfectly solve this problem. PKG-based process design has the following advantages compared with traditional CAPP: (1) PKG can more easily store and process various forms of complex process knowledge, including structured process knowledge, semi-structured process knowledge, and structured process knowledge; (2) PKG can reveal the explicit and implicit relationships between process knowledge, connecting seemingly isolated process knowledge. The structure of PKG determines its ability to complete process knowledge; (3) PKG-based process generation is the globally optimal result obtained by combining all process knowledge in the graph. Compared with traditional CAPP, it has stronger versatility and intelligence; (4) PKG-based process planning is flexible and can be applied to large-scale dynamic manufacturing processes. In contrast, traditional CAPP is not suitable for large-scale dynamic manufacturing processes because it cannot truly adjust the process and balance resource load in real time; (5) PKG is a key technology for the transformation from traditional manufacturing to intelligent manufacturing. These advantages allow PKG to reuse process knowledge and automatically generate process plans, promoting the standardization and automation of processing technology planning. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a cable processing technology generation method and system based on knowledge graphs. The aim is to use knowledge graph technology to automatically generate cable processing procedures, reduce repetitive calculations and waste of process knowledge in cable processing, thereby improving the automation and intelligence of the cable processing process.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A knowledge graph-based method for generating cable processing technology includes the following steps:

[0008] Model the knowledge of cable processing technology, establish a CPKG model, and based on the CPKG model, convert the relationships between the subtypes and attributes of entities into the relationships between classes, and create a cable processing process pattern layer;

[0009] A structure diagram is generated and embedded into the cable processing mode layer, an ICPKG structure tree is created, and a distributed embedded inference model is constructed based on the ICPKG structure tree.

[0010] Based on the aforementioned distributed embedded inference model, a production cable processing technology planning RDF / XML document is generated.

[0011] Preferred methods for modeling cable processing technology knowledge and establishing a CPKG model include:

[0012] Extract the process information from the cable processing documentation and 3D model, and describe the CPKG model as follows:

[0013] CPKGM={P∪F∪G∪C∪T∪S}

[0014] Where P represents a process step, F represents cable processing characteristics, G represents cable assembly, C represents processing requirements, T represents tooling, and S represents semantic relationships.

[0015] Preferably, based on the CPKG model, the method for creating a cable processing process pattern layer by converting the relationships between entity subtypes and attributes into relationships between classes includes:

[0016] Preset semantic relationship S;

[0017] Based on semantic relation S, six classes are established: process, processing feature, component, requirement, tool and relationship, and subclasses are created through inheritance.

[0018] Based on the six classes established and the subclasses established through inheritance, a cable processing pattern layer is created.

[0019] The semantic relation S is described as follows:

[0020] S = {Has, Connect, Sequence, Parallel, Recycle} where Has represents inclusion, Connect represents the connection method of wires and components, Sequence represents the sequential relationship between process steps, Parallel represents the parallel relationship between process steps, and Recycle represents the cycle of a certain process step;

[0021] Based on semantic relations S, the methods for establishing six classes—process, processing features, components, requirements, tools, and relationships—include:

[0022] (F,P) represents the inclusion relationship between cable processing features and process steps, (P,G) represents the inclusion relationship between process steps and cable assemblies, (P,T) represents the inclusion relationship between process steps and tooling, (P,C) represents the inclusion relationship between process steps and processing requirements, (F,C) represents the inclusion relationship between cable processing features and technical requirements, and (G,F) represents the inclusion relationship between cable assemblies and processing features.

[0023] The subclasses established through inheritance include: wire and component classes inherit from component class; geometric features, tolerance dimensions, and material features classes inherit from machining features class; dimensional deviation, heat treatment requirements, and batching requirements classes inherit from requirements class; and process steps classes inherit from process class.

[0024] Preferably, the method for generating a structure diagram and embedding it into the cable processing pattern layer, creating an ICPKG structure tree, and constructing a distributed embedded inference model based on the ICPKG structure tree includes:

[0025] A cable assembly structure tree is constructed to describe the connection relationships between cable assemblies. The cable assembly structure tree is constructed as a triple and stored. The components are connected in the form of the cable assembly structure tree to construct an initial processing technology knowledge graph model (ICPKG). The cable assembly structure tree information includes the ID number of the component, the connected components and wires, and the semantic relationship "Connect".

[0026] Based on the initial machining process knowledge graph model ICPKG, the machining process documents and CAD models are scanned to generate a machining process knowledge graph database CPKGD;

[0027] A distributed embedded reasoning model is established based on the CPKGD knowledge graph database of manufacturing processes.

[0028] Preferred methods for establishing distributed embedded reasoning models based on the CPKGD (Chemical Process Knowledge Graph Database) include:

[0029] The SKGCN method was used for knowledge reasoning to extract multiple processing case graph data from the CPKGD processing technology knowledge graph database.

[0030] CompGCN is used to learn the aggregation information of each processing case graph data, generate the embedding vector corresponding to each processing technology sequence, and connect the embedding vectors corresponding to different processing technology situations.

[0031] Based on the connection-based embedding vectors, an evaluation function is designed using capsule networks to learn the implicit semantic associations between cable processing instances at the global level, thus establishing a distributed embedding inference model.

[0032] Preferably, the method of using CompGCN to learn the aggregated information of each processing case graph data and generate the embedding vector corresponding to each processing technology sequence includes:

[0033]

[0034] Among them, h o This represents the update representation of node o. It is the nearest neighbor set of the output edge of o. It is a specific parameter of a relation type, φ represents a compound operation, x s , z r Let S and R represent the initial features of node S and relation R, respectively.

[0035]

[0036] Among them, h r Indicates relational embedding. It is a learnable transformation matrix that projects all relations into the same embedded mapping space as the nodes. It describes learnable scalar weights for relations and specific bases. V represents a set of learnable basis vectors. b =∈V1,V2,......,V β Let β represent a set of learnable basis vectors, where β is the number of learnable basis vector sets.

[0037]

[0038] in, This represents node O obtained after k layers. It is a specific parameter of the k-th layer. Represents the node features of the k-th layer. This represents the relation embedding at the k-th level;

[0039]

[0040] in, This represents the representation of relation r after level k.

[0041] Preferably, a method for designing an evaluation function using capsule networks includes:

[0042] f(s, r, o)=||capsnet(g([v, r, u]*Ω))||

[0043] Where s, r, and o are the output triplet information, s and o represent nodes, and r represents relations; similarly, v, r, and u represent the input triplet information, v and u represent nodes, and r represents relations; Ω represents the weight matrix, and the g function represents the nonlinear transformation of the triplet after the weight matrix transformation; caspsnet represents the capsule neural network structure.

[0044] Preferably, the method for generating cable processing technology planning RDF / XML documents based on the distributed embedded inference model includes:

[0045] Store the nodes and edges in the CPKG model in XML format;

[0046] The edge triples are stored by traversing the "start and end node" information of each edge.

[0047] The attribute value information of each node is obtained through iterative querying.

[0048] This invention also provides a knowledge graph-based cable processing technology generation system, comprising: a pattern layer creation module, a model building module, and a generation module;

[0049] The pattern layer creation module is used to model cable processing technology knowledge, establish a CPKG model, and based on the CPKG model, convert the relationships between entity subtypes and attributes into relationships between classes, and create a cable processing process pattern layer.

[0050] The model building module is used to generate a structure diagram and embed it into the cable processing mode layer, create an ICPKG structure tree, and build a distributed embedded inference model based on the ICPKG structure tree.

[0051] The generation module is used to generate RDF / XML documents for cable processing planning based on the distributed embedded inference model.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] This invention discloses: 1. Modeling cable processing technology knowledge, establishing a CPKG model, extracting process information from cable processing process documents and 3D models, converting the relationships between entity subtypes and attributes into relationships between classes, and creating a cable processing process pattern layer. 2. Generating a structure graph and embedding it into the cable processing process, creating an IPKG structure tree, constructing a distributed embedded reasoning model, and continuously completing the CPKG by calculating the subgraph similarity between IPKG and CPKGD. 3. Outputting a cable processing process planning RDF / XML document. This invention aims to use knowledge graph technology to automatically generate cable processing flow charts, reduce repetitive calculations and waste of process knowledge in cable processing, thereby improving the automation and intelligence of the cable processing process. Attached Figure Description

[0054] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram illustrating the mapping relationship of some data between mode layers in an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram illustrating the construction method of the distributed embedded reasoning model according to an embodiment of the present invention;

[0057] Figure 3This is a schematic diagram illustrating a method for converting information within a structure tree into triples according to an embodiment of the present invention.

[0058] Figure 4 This is a schematic diagram of the cable processing mode layer according to an embodiment of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Example 1

[0062] This invention provides a method for generating cable processing technology based on knowledge graphs, comprising the following steps:

[0063] Model the knowledge of cable processing technology and establish a processing technology knowledge graph (CPKG) model. Based on the CPKG model, convert the relationships between the subtypes and attributes of entities into the relationships between classes and classes, and create a cable processing process pattern layer.

[0064] A structure diagram is generated and embedded into the cable processing mode layer, an ICPKG structure tree is created, and a distributed embedded inference model is constructed based on the ICPKG structure tree.

[0065] Based on the aforementioned distributed embedded inference model, a production cable processing technology planning RDF / XML document is generated.

[0066] In this embodiment, the method for modeling cable processing technology knowledge and establishing a CPKG model includes:

[0067] Extract the process information from cable manufacturing process documents and saved 3D models of a certain type of cable, and describe the CPKG model as follows:

[0068] CPKGM={P∪F∪G∪C∪T∪S}

[0069] Where P represents the process step, F represents the cable processing characteristics, G represents the cable assembly, C represents the processing requirements, T represents the tooling, and S represents the semantic relationship. F is 3D model information.

[0070] In this embodiment, based on the CPKG model, the method for creating a cable processing process pattern layer by converting the relationships between entity subtypes and attributes into relationships between classes includes:

[0071] Preset semantic relationship S;

[0072] Based on semantic relation S, six classes are established: process, processing feature, component, requirement, tool and relationship, and subclasses are created through inheritance.

[0073] Based on the six classes established and the subclasses established through inheritance, a cable processing pattern layer is created.

[0074] The semantic relation S is described as follows:

[0075] S = {Has, Connect, Sequence, Parallel, Recycle} where Has represents inclusion, Connect represents the connection method of wires and components, Sequence represents the sequential relationship between process steps, Parallel represents the parallel relationship between process steps, and Recycle represents the cycle of a certain process step;

[0076] Based on semantic relations S, the methods for establishing six classes—process, processing features, components, requirements, tools, and relationships—include:

[0077] Has represents an inclusion relationship, which can be described as:

[0078] S={(F,P),(P,G),(P,T),(P,C),(F,C),(G,F)}

[0079] In this description, (F,P) represents the inclusion relationship between cable processing features and process steps, (P,G) represents the inclusion relationship between process steps and cable assemblies, (P,T) represents the inclusion relationship between process steps and tooling, (P,C) represents the inclusion relationship between process steps and processing requirements, (F,C) represents the inclusion relationship between cable processing features and technical requirements, and (G,F) represents the inclusion relationship between cable assemblies and processing features.

[0080] The subclasses established through inheritance include: wire and component classes inheriting from the component class; geometric features, tolerance dimensions, and material features classes inheriting from the machining features class; dimensional deviations, heat treatment requirements, and batching requirements classes inheriting from the requirements class; and process steps classes inheriting from the process class. This transforms the relationship between two entities having subtypes or attributes into a specific and unique node relationship between ontology pattern classes.

[0081] For example: Figure 1This diagram illustrates the specific mapping from entity relationships to class-to-class relationships. As can be seen from the diagram, subtype or attribute relationships between two entities are transformed into node relationships between classes. For example, the subtype relationship between cable assemblies and cable components is mapped to a subclass relationship between a component class and a component subclass; the attribute relationship between cable components and processing features is mapped to a feature-based relationship between a component class and a feature class.

[0082] Finally, a schema layer is generated based on the relational mapping. To clearly describe the complex semantic information of the cable processing process, the cable processing schema layer (a semantic relational graph where various semantic relations connect the attributes of the six classes mentioned earlier) can be divided into a cable structure schema layer, a cable processing step schema layer, and a cable processing procedure schema layer, such as... Figure 4 As shown. The semantic relationships Has and Connect in the processing structure are mainly reflected in the cable processing structure mode layer. The semantic relationships in the processing sequence include Sequence, Parallel, and Recycle, which are mainly reflected in the cable processing step mode layer and the cable processing process mode layer.

[0083] For example: Figure 2 This diagram illustrates the construction method of a distributed embedded inference model.

[0084] In this embodiment, the method for generating a structure diagram and embedding it into the cable processing pattern layer, creating an ICPKG structure tree, and constructing a distributed embedded inference model based on the ICPKG structure tree includes:

[0085] Construct a cable assembly structure tree to describe the connection relationships between cable assemblies, and save the structure tree as a triple. The information in the structure tree includes the ID number of the component, the connected components and wires, and the semantic relationship "Connect". Connect the various component and wire entities in the form of this structure tree to construct an ICPKG (Initial Process Knowledge Graph) model.

[0086] Scan the processing technology documents and CAD models containing cable processing features (including cable processing features), define and generate CPKGD (processing technology knowledge graph database).

[0087] Constructing a distributed embedded inference model, the construction method is as follows: Figure 2As shown, the Sequence Knowledge Graph Convolutional Network (SKGCN) algorithm is used for knowledge reasoning. This involves extracting multiple processing case graph data from CPKGD, and using a Combinatorial Multi-Relational Graph Convolutional Network (CompGCN) to learn the aggregation information of each processing case graph data. This generates an embedding vector corresponding to each processing sequence, and concatenates the embedding vectors corresponding to different processing situations. Based on a capsule network, an evaluation function is designed to learn the implicit semantic associations between cable processing instances at the global level. The specific steps are as follows:

[0088] 1) Define CPKGD as in ε represents the set of nodes and the set of edges, respectively. This represents the d0-dimensional input feature of each node. A set representing relations. This indicates the initial relational characteristics.

[0089] 2) Define h o For node embedding, it is represented as follows:

[0090]

[0091] Among them, h o This represents the update representation of node o. It is the nearest neighbor set of the output edge of o. It is a specific parameter of a relation type. Indicates compound operation, x s , z r Let S and R represent the initial features of node S and relation R, respectively.

[0092] 3) Define relation embedding h r as follows:

[0093]

[0094] It is a learnable transformation matrix that projects all relations into the same embedded mapping space as the nodes. V describes the learnable scalar weights of relations and specific bases. b =∈V1,V2,......,V β Let represent a set of learnable basis vectors, where β is the number of learnable basis vector sets. Furthermore, h... r These relationships are also allowed to be used in the next SKGCN layer.

[0095] 4) Definition The node o obtained after layer k is represented as follows:

[0096]

[0097] It is a specific parameter of the k-th layer. Represents the node features of the k-th layer. This represents the relation embedding at the k-th level.

[0098] 5) Definition The mathematical representation of relation r after k layers is as follows:

[0099]

[0100] 6) The evaluation function based on capsule networks is defined as follows:

[0101] f(s, r, o)=||caosnet(g([v, r, u]*Ω))|| (5)

[0102] s, r, o are the output triplet information, where s and o represent nodes and r represents relations; similarly, v, r, u represent the input triplet information, where v and u represent nodes and r represents relations; Ω represents the weight matrix, and the g function represents the nonlinear transformation performed on the triplet after the weight matrix transformation; caspsnet represents the capsule neural network structure.

[0103] 7) Define the graph similarity method as follows:

[0104]

[0105] in, It is a subgraph of CPKG, which can be described as a multi-hop sequence S. i ={s, r1, e1...r i e i ...r l ,o}, l represents the number of jumps. Let h represent the embedding vector of nodes and edges from a multi-hop sequence with a center node, where h v and h r It can be obtained from equations (3) and (4).

[0106] 8) Define formula (7) to obtain the similar subgraphs corresponding to the subgraph. If the subgraph contains multiple multi-hop sequences, multiple corresponding similar subgraphs can be obtained through formula (7), as shown below:

[0107]

[0108] 9) Based on the calculations in the above steps, the knowledge structure is learned from CPKGD to obtain multiple initial processing schemes.

[0109] 10) Reuse formula (7) to calculate missing processing information, including processing steps, procedures, etc. The specific implementation steps are as follows: First, input an ICPKG containing basic nodes and edges, then complete a certain type of node, complete the edges involved in the node type, then determine whether all types of nodes have been completed, and finally, when all types of nodes and edges are completed, output the completed ICPKG.

[0110] In this embodiment, the nodes and edges in the CPKG are first stored in XML format. Different nodes have different attributes. When generating RDF / XML, they are divided into different nodes, including cable parts, leads, process steps, and processes. The attributes of edges include name, head node, and tail node.

[0111] Then, the edge triples are stored by traversing the "start and end node" information of each edge.

[0112] The attribute value information of each node can be obtained through iterative queries.

[0113] Application Example: Taking the processing of a certain type of cable as an example. First, based on the cable's step model and the scan sequence of the processing technology document, triples are created to obtain a processing technology knowledge database for the cable. This database stores the scanned entities and relationships as nodes and edges. Nodes should include cable components, processing features, technical requirements, process steps, and tooling. Edges (relationships) include Has, Connect, Sequence, Parallel, and Recycle. Next, a processing structure tree is created, and the structure tree information is stored as triples, as detailed below. Figure 3As shown. Input a processing structure tree, create an entity list listN and a relation listR. By traversing the structure tree, save nodes to listN, and save the relation between every two nodes to listR. Output the triple <component, connecting component, connecting wire>. Construct an ICPKG using the structure trees of each component and wire. Based on the distributed embedded inference model mentioned in step two, input the triples from the ICPKG. Combine this with the process modeling from step one, and update the model through incremental training. Then generate implicit processing relations and entities. This is because the ICPKG only contains three types of information: component and wire nodes, "connecting wire" edges between components and wires, and "connecting component" edges between components. Therefore, the similarity between subgraphs in ICPKG and CPKGD can be calculated by loading nodes for "components, wires" and edges for "connecting wires, connecting components". The subgraph with the highest similarity is selected from multiple subgraphs to complete the cable structure pattern layer, which is then used as the CPKG for the current generation. Similarity is calculated by loading multi-hop sequences between the current CPKG and subgraphs in CPKGD, including edges for "Has, Connect" and nodes for "cable wires, cable components, tooling, processing features". Similarly, the subgraph with the highest similarity is selected to complete the CPKG processing step pattern layer. Then, similarly, similarity is calculated by loading multi-hop sequences between the current CPKG and subgraphs in CPKGD, including nodes for "cable wires, cable components, tooling, processing features, process steps" and edges for "Has, Connect, Sequence, Parallel, Recycle". Finally, the subgraph with the highest similarity is selected again to obtain the cable processing solution. Finally, by completing the CPKG loading and the triples related to process steps and operations, the semantic information of "Process-parts, Process-leads, Process-Process, ProcessStep-parts, ProcessStep-leads, Process-ProcessStep" is obtained, thereby converting the processing sequence information into RDF / XML format output.

[0114] Example 2

[0115] This invention also provides a knowledge graph-based cable processing technology generation system, comprising: a pattern layer creation module, a model building module, and a generation module;

[0116] The pattern layer creation module is used to model cable processing technology knowledge, establish a CPKG model, and based on the CPKG model, convert the relationships between entity subtypes and attributes into relationships between classes, and create a cable processing process pattern layer.

[0117] The model building module is used to generate a structure diagram and embed it into the cable processing mode layer, create an ICPKG structure tree, and build a distributed embedded inference model based on the ICPKG structure tree.

[0118] The generation module is used to generate RDF / XML documents for cable processing planning based on the distributed embedded inference model.

[0119] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for generating cable processing technology based on knowledge graphs, characterized in that, Includes the following steps: Cable processing technology knowledge is modeled, and a processing technology knowledge graph (CPKG) model is established. Based on the CPKG model, the relationships between the subtypes and attributes of entities are transformed into relationships between classes, and a cable processing process pattern layer is created. A structure diagram is generated and embedded into the cable processing mode layer, an ICPKG structure tree is created, and a distributed embedded inference model is constructed based on the ICPKG structure tree. Based on the aforementioned distributed embedded inference model, a production cable processing technology planning RDF / XML document is generated. The method for generating a structure diagram and embedding it into the cable processing pattern layer, creating an ICPKG structure tree, and constructing a distributed embedded inference model based on the ICPKG structure tree includes: A cable assembly structure tree is constructed to describe the connection relationships between cable assemblies. The cable assembly structure tree is constructed as a triple and stored. The various components and wire entities are connected in the form of the cable assembly structure tree to construct an initial processing technology knowledge graph model ICPKG. The cable assembly structure tree information includes the ID number of the component, the connected components and wires, and the semantic relationship "Connect". Based on the initial machining process knowledge graph model ICPKG, the machining process documents and CAD models are scanned to generate a machining process knowledge graph database CPKGD; A distributed embedded reasoning model is established based on the CPKGD knowledge graph database of processing technology. Methods for establishing distributed embedded reasoning models based on the CPKGD (Chemical Process Knowledge Graph Database) include: The SKGCN method was used for knowledge reasoning to extract multiple processing case graph data from the CPKGD processing technology knowledge graph database. CompGCN is used to learn the aggregation information of each processing case graph data, generate the embedding vector corresponding to each processing technology sequence, and connect the embedding vectors corresponding to different processing technology situations. Based on the connection-based embedding vectors, an evaluation function is designed using capsule networks to learn the implicit association semantics between cable processing instances at the global level, and a distributed embedding inference model is established. The methods for using CompGCN to learn the aggregated information of each processing case graph data and generate the embedding vector corresponding to each processing technology sequence include: ； in, Represents a node The update represents the quantity. yes Output the nearest neighbor set of the edge. It is a specific parameter of a relation type, where φ represents a compound operation. Representing nodes respectively s and relationships r Initial characteristics; ； in, Indicates relational embedding. It is a learnable transformation matrix that projects all relations into the same embedded mapping space as the nodes. It describes learnable scalar weights for relations and specific bases. This represents a set of learnable basis vectors. This represents a set of learnable basis vectors. The number of sets of learnable basis vectors; ； in, express k Nodes obtained after layer , It is the first k A specific parameter of the layer, Indicates the first k Layer node characteristics, Indicates the first k Layer relationship embedding; ； in, express k Layer Relationship r The representation of; Methods for designing an evaluation function using capsule networks include: ； in, s , r , o It outputs the triplet information. s , o Represents a node. r Indicates a relationship; similarly, v , r , u This represents the input triplet information. v , u Represents a node. r Indicates a relationship; Represents the weight matrix. g The function represents a nonlinear transformation of the triples after the weight matrix transformation; caspsnet This represents the structure of a capsule neural network; Based on the aforementioned distributed embedded inference model, the method for producing RDF / XML documents for cable processing planning includes: Calculate the similarity between subgraphs in ICPKG and CPKGD, select the subgraph with the highest similarity from multiple subgraphs to complete the pattern layer of the cable structure, and use it as the current CPKG. Calculate the similarity by loading multi-hop sequences between the current CPKG and the subgraphs in CPKGD, select the subgraph with the highest similarity, obtain the cable processing scheme, and convert the processing sequence information into RDF / XML format output using the completed CPKG.

2. The cable processing technology generation method based on knowledge graph according to claim 1, characterized in that, Methods for modeling cable processing technology knowledge and establishing CPKG models include: Extract the process information from the cable processing documentation and 3D model, and describe the CPKG model as follows: ； in, P Indicates process steps, F Indicates cable processing characteristics, G Indicates cable assembly, C Indicates processing requirements. T Indicates tooling, S Indicates semantic relationships.

3. The cable processing technology generation method based on knowledge graphs according to claim 2, characterized in that, Based on the CPKG model, the method for creating a cable processing pattern layer by converting the relationships between entity subtypes and attributes into relationships between classes includes: Preset semantic relationships S ; Based on semantic relations S Establish six classes: process, processing characteristics, components, requirements, tools, and relationships, and create subclasses through inheritance. Based on the six classes established and the subclasses established through inheritance, a cable processing pattern layer is created. Among them, semantic relations S Described as: ； in, Has Indicates an inclusion relationship. Connect Indicates the connection method of wires and components. Sequence Indicates the sequential relationship between process steps. Parallel This indicates the parallel relationship between process steps. Recycle Indicates a cycle of a certain process step; Based on semantic relations S The methods for establishing six categories—process, processing features, components, requirements, tools, and relationships—include: ( F , P This indicates the inclusion relationship between cable processing characteristics and process steps. P , G This indicates the inclusion relationship between process steps and cable assemblies. P , T This indicates the inclusion relationship between process steps and tooling. P , C The inclusion relationship between process steps and processing requirements, F , C This indicates the inclusion relationship between cable processing characteristics and technical requirements. G , F This indicates the inclusion relationship between cable assemblies and processing features; The subclasses established through inheritance include: wire and component classes inherit from component class; geometric features, tolerance dimensions, and material features classes inherit from machining features class; dimensional deviation, heat treatment requirements, and batching requirements classes inherit from requirements class; and process steps classes inherit from process class.

4. The cable processing technology generation method based on knowledge graph according to claim 1, characterized in that, Based on the aforementioned distributed embedded inference model, the method for producing RDF / XML documents for cable processing planning includes: Store the nodes and edges in the CPKG model in XML format; The edge triples are stored by traversing the "start and end node" information of each edge. The attribute value information of each node is obtained through iterative querying.

5. A cable processing technology generation system based on knowledge graphs, the system being used to implement the method described in any one of claims 1-4, characterized in that, include: The schema layer includes a creation module, a model building module, and a generation module. The pattern layer creation module is used to model cable processing technology knowledge, establish a CPKG model, and based on the CPKG model, convert the relationships between entity subtypes and attributes into relationships between classes, and create a cable processing process pattern layer. The model building module is used to generate a structure diagram and embed it into the cable processing mode layer, create an ICPKG structure tree, and build a distributed embedded inference model based on the ICPKG structure tree. The generation module is used to generate RDF / XML documents for cable processing planning based on the distributed embedded inference model.

Images