A part machining parameter automatic generation method based on a knowledge graph and a graph neural network
By using knowledge graph and graph neural network-based methods, part processing parameters are automatically generated, solving the problem that traditional process parameters rely on human experience, and improving processing efficiency and adaptability to new part models.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENYANG INST OF AUTOMATION - CHINESE ACAD OF SCI
- Filing Date
- 2024-12-18
- Publication Date
- 2026-06-19
AI Technical Summary
Traditional machining process parameters rely on human experience, resulting in low processing efficiency, difficulty in achieving full automation, and poor adaptability to new parts models.
A knowledge graph and graph neural network-based approach is adopted. By converting part machining files into triple files, a process knowledge graph is constructed, and RED-GNN is used for training to generate part machining parameters.
It enables the automated generation of part machining parameters, improving design efficiency and adaptability to new part models, and reducing debugging time.
Smart Images
Figure CN122241891A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of automatic control technology, specifically a method for automatically generating part processing parameters based on knowledge graphs and graph neural networks. Background Technology
[0002] In modern manufacturing, automation and intelligence are key to improving efficiency and competitiveness. Generating machining process parameters is one of the core steps in achieving automated machining. Traditionally, the formulation of machining process parameters has relied heavily on engineers' experience and trial and error, making it difficult to derive directly from theoretical models and exhibiting high complexity and flexibility. This reliance on manual process parameter formulation has severely hampered the development of fully automated machining.
[0003] Traditional machining production lines typically only process specific or a few types of parts. When new products need to be manufactured or existing products require design changes, the process department needs to revise the machining parameters based on experience and adjust the production line and equipment configuration. This process usually takes several months to debug, resulting in poor adaptability of the production line to new parts.
[0004] To achieve fully automated machining design, there is an urgent need for a method that can automatically generate machining process parameters to provide a precise parameter list for machining process programming. Summary of the Invention
[0005] To address the aforementioned problems, the present invention aims to design an automatic part processing parameter generation method based on knowledge graphs and graph neural networks, in order to solve the problem of automatic generation of processing parameters during product adjustments.
[0006] The technical solution adopted by the present invention to achieve the above objectives is as follows:
[0007] A method for automatically generating part machining parameters based on knowledge graphs and graph neural networks includes the following steps:
[0008] 1) Extract knowledge from the part machining documents and extract the process information into a ternary file;
[0009] 2) Construct a knowledge graph of part process information;
[0010] 3) Construct a parts processing dataset based on the triples stored in the knowledge graph;
[0011] 4) Train the graph neural network using the parts machining dataset;
[0012] 5) Input the new part into the trained graph neural network to obtain the machining parameter information of the part.
[0013] Step 1) specifically refers to:
[0014] Extract the process information from the existing part machining file into a ternary file, represented as (part name, process category, process content).
[0015] Step 2) specifically refers to:
[0016] gCloud is used to convert assembly process structured documents into process knowledge graphs.
[0017] The process knowledge graph is divided into a schema layer and a data layer, wherein...
[0018] The pattern layer defines the structural patterns of the part body and its relationships;
[0019] The data layer extracts the processing data of existing parts in the form of a pattern layer and stores it in gCloud as a triple.
[0020] The pattern layer includes 12 ontology types and 10 relationships. The 12 ontology types include part name, part material, part type, machining features, machining method, machining process, accuracy, roughness, machine tool, machining parameters, cutting tool, fixture, and cutting fluid. The 10 relationships include the associations and interactions between ontology types in the part process knowledge graph.
[0021] Step 3) specifically refers to:
[0022] Call the gCloud API interface to read all the triples stored in gCloud as a total triple file, shuffle the total triple file, and divide it into training set, validation set and test set according to the ratio.
[0023] The graph neural network is RED-GNN.
[0024] An automatic part machining parameter generation system based on knowledge graphs and graph neural networks includes:
[0025] The knowledge extraction module is used to extract knowledge from the part processing files and extract the process information into triplet files.
[0026] The knowledge graph construction module is used to build a knowledge graph of part process information.
[0027] The Parts Processing Dataset Construction Module is used to construct parts processing datasets based on triples stored in the knowledge graph.
[0028] The network training module is used to train the graph neural network using the parts machining dataset;
[0029] The machining parameter generation module is used to input a new part into a trained graph neural network and obtain the machining parameter information of that part.
[0030] An automatic part machining parameter generation device based on knowledge graphs and graph neural networks includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the automatic part machining parameter generation method based on knowledge graphs and graph neural networks when the computer program is executed.
[0031] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for automatically generating part processing parameters based on knowledge graphs and graph neural networks.
[0032] The present invention has the following beneficial effects and advantages:
[0033] This invention proposes an automatic part machining parameter generation method based on knowledge graphs and graph neural networks. The method involves extracting knowledge from existing part machining files, converting process information into triplet files, constructing a knowledge graph of the part's process information using gCloud, building a part machining dataset based on this knowledge graph, and then training the dataset using a graph neural network. When a part changes, the known part parameters are input into the graph neural network, which then outputs the remaining unknown machining parameters. This algorithm provides machining parameter references for the process department after part changes, improving design efficiency. Attached Figure Description
[0034] Figure 1 This is a schematic diagram of the algorithm for the automatic generation method of part machining parameters provided in an embodiment of the present invention.
[0035] Figure 2 This is a schematic diagram of the component knowledge graph pattern layer provided in an embodiment of the present invention.
[0036] Figure 3 This is a schematic diagram illustrating the reasoning of machining parameters for shaft parts provided in an embodiment of the present invention. Detailed Implementation
[0037] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.
[0038] An automatic generation method for assembly process steps based on knowledge graphs and graph neural networks includes a knowledge graph construction layer, a graph neural network training layer, and a processing parameter generation layer.
[0039] The solution first extracts knowledge from existing part machining files, converting process information into triplet files. A knowledge graph of the part's process information is then constructed using gCloud. Based on this knowledge graph, a part machining dataset is built and trained using a graph neural network. When a part changes, the known part parameters are input into the graph neural network, which then outputs the remaining unknown machining parameters. This invention provides an automatic part machining parameter generation method based on knowledge graphs and graph neural networks. Figure 1 As shown, follow these steps:
[0040] Step 1: Extract knowledge from existing part machining files and extract process information into ternary files;
[0041] Step 2: Construct a knowledge graph of part manufacturing process information;
[0042] Step 2 describes a method for constructing a process knowledge graph by using gCloud to convert the historical machining files of the part into a process knowledge graph. Specifically:
[0043] The process knowledge graph is divided into a schema layer and a data layer.
[0044] The schema layer includes 12 ontology types and 10 relation types, such as Figure 2 As shown. These 12 ontologies cover part name, part material, part type, machining features, machining method, machining process, accuracy, roughness, machine tool, machining parameters, cutting tool, fixture, and cutting fluid, while the 10 relationships cover the associations and interactions between ontologies in the part process knowledge graph.
[0045] The data layer extracts the processing data of existing parts in the form of a schema layer and stores it in gCloud as a triple.
[0046] Step 3: Construct a parts processing dataset based on the triples stored in gCloud;
[0047] The method for constructing the part machining dataset mentioned in step 3 is as follows:
[0048] Call the gCloud API to create a master triplet file containing all triples stored in gCloud. Read and shuffle the master triplet file, dividing it into training, validation, and test sets in a 7:2:1 ratio.
[0049] Step 4: Train using a graph neural network;
[0050] The graph neural network mentioned in step 4 is specifically RED-GNN, such as... Figure 3As shown, RED-GNN (RelationalDigraph-based Graph Neural Network) is a graph neural network model for knowledge graph reasoning. By introducing a relation-directed graph (r-digraph) structure, it combines the advantages of paths and subgraphs, providing an efficient and interpretable method for knowledge graph reasoning.
[0051] Step 5: Input the new part into the graph neural network and output the machining parameter information of this part.
[0052] Example 1:
[0053] Example 1 illustrates the beneficial effects of the present invention by automatically generating machining parameters based on shaft parts.
[0054] S1. Extract knowledge from existing part machining files and extract process information into ternary files. ： For example: (shaft, part material, cast iron).
[0055] S2. Use gCloud to construct a knowledge graph of part process information.
[0056] S3. Construct a parts processing dataset based on the triples stored in gCloud.
[0057] S4. Train the model using a graph neural network. Save the model after training.
[0058] S5. After training, save the model, then load the shaft part to be processed to automatically generate the processing parameters of the shaft part.
Claims
1. A method for automatically generating part machining parameters based on knowledge graphs and graph neural networks, characterized in that, Includes the following steps: 1) Extract knowledge from the part machining documents and extract the process information into a ternary file; 2) Construct a knowledge graph of part process information; 3) Construct a parts processing dataset based on the triples stored in the knowledge graph; 4) Train the graph neural network using the parts machining dataset; 5) Input the new part into the trained graph neural network to obtain the machining parameter information of the part.
2. The method for automatically generating part machining parameters based on knowledge graphs and graph neural networks according to claim 1, characterized in that, Step 1) specifically refers to: Extract the process information from the existing part machining file into a ternary file, represented as (part name, process category, process content).
3. The method for automatically generating part machining parameters based on knowledge graphs and graph neural networks according to claim 1, characterized in that, Step 2) specifically refers to: gCloud is used to convert assembly process structured documents into process knowledge graphs.
4. The method for automatically generating part machining parameters based on knowledge graphs and graph neural networks according to claim 3, characterized in that, The process knowledge graph is divided into a pattern layer and a data layer. The pattern layer defines the structural pattern of the part body and its relationships. The data layer extracts the processing data of existing parts in the form of a pattern layer and stores it in gCloud as a triple.
5. The method for automatically generating part machining parameters based on knowledge graphs and graph neural networks according to claim 4, characterized in that, The pattern layer includes 12 ontology types and 10 relationships. The 12 ontology types include part name, part material, part type, machining features, machining method, machining process, accuracy, roughness, machine tool, machining parameters, cutting tool, fixture, and cutting fluid. The 10 relationships include the associations and interactions between ontology types in the part process knowledge graph.
6. The method for automatically generating part machining parameters based on knowledge graphs and graph neural networks according to claim 1, characterized in that, Step 3) specifically refers to: Call the gCloud API interface to read all the triples stored in gCloud as a total triple file, shuffle the total triple file, and divide it into training set, validation set and test set according to the ratio.
7. The method for automatically generating part machining parameters based on knowledge graphs and graph neural networks according to claim 1, characterized in that, The graph neural network is RED-GNN.
8. An automatic part machining parameter generation system based on knowledge graphs and graph neural networks, characterized in that, include: The knowledge extraction module is used to extract knowledge from the part processing files and extract the process information into triplet files. The knowledge graph construction module is used to build a knowledge graph of part process information. The Parts Processing Dataset Construction Module is used to construct parts processing datasets based on triples stored in the knowledge graph. The network training module is used to train the graph neural network using the parts machining dataset; The machining parameter generation module is used to input a new part into a trained graph neural network and obtain the machining parameter information of that part.
9. An automatic part machining parameter generation device based on knowledge graphs and graph neural networks, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement, when executing the computer program, a method for automatically generating part processing parameters based on knowledge graphs and graph neural networks as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the automatic generation method for part processing parameters based on knowledge graphs and graph neural networks as described in any one of claims 1-7.