Welding technological process three-dimensional expression and construction method based on domain knowledge graph

Abstract

The invention discloses a three-dimensional expression and construction method for a welding process flow based on a domain knowledge graph, which comprises the following steps of: 1, collecting multi-source heterogeneous welding process data, and carrying out cleaning, standardization and classified storage on the data; 2, on the basis of the preprocessed data, welding entities, relations and attributes are recognized, three-dimensional space attributes are associated, and a domain knowledge graph supporting spatial semantics is constructed through an ontology modeling tool and stored in a graph database; 3, expanding an entity-relation-entity triple structure, adding three-dimensional space attributes, and generating a reasonable triple chain; 4, welding requirements and space and process constraints are input, an initial technological process scheme is generated through rule engine reasoning, and technological parameters are globally optimized in combination with a multi-objective optimization algorithm; and 5, storing the optimization scheme into a graph database, evaluating the performance of the scheme through a quantitative index, and feeding back the scheme performance to the knowledge graph to realize closed-loop optimization.

Description

Technical Field

[0001] The present invention belongs to the field of welding process technology, and specifically provides a three-dimensional expression and construction method of welding process flow based on domain knowledge graph. Background Art

[0002] A knowledge graph is a structured semantic network whose core goal is to integrate dispersed knowledge into an interconnected network. It can store knowledge in a graph format, enabling the visual representation of discrete knowledge. A domain knowledge graph, on the other hand, is a knowledge graph modeled and applied to specific industries or domains. Building on conventional knowledge graphs, it further focuses on a specific domain, providing a more in-depth depiction of entities, concepts, attributes, and their semantic relationships within that domain.

[0003] The welding process encompasses various aspects, including the selection of welding materials, welding methods, and welding equipment. It is a core element of the manufacturing industry and is widely used in industrial manufacturing and assembly. However, as the manufacturing industry evolves towards intelligence and digitalization, traditional welding process management methods suffer from fragmented knowledge management, low intelligence levels, and insufficient three-dimensional visualization. In particular, existing research on knowledge graphs in the field of welding processes primarily focuses on constructing knowledge graphs corresponding to existing typical processes to retrieve and select relevant process element information. Few methods address the welding process flow and express the actual welding process in three dimensions using triples. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a three-dimensional expression and construction method of welding process flow based on domain knowledge graph, which can effectively improve the digital level of welding process flow, meet the dynamic welding production needs and improve the intuitiveness and operability of process management, and provide auxiliary domain knowledge graph related research for welding process flow reasoning construction and visual expression.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A three-dimensional expression and construction method for welding process flow based on domain knowledge graph includes the following steps:

[0007] Step 1: Welding process knowledge collection and preprocessing

[0008] Collect multi-source heterogeneous welding process data, clean, standardize and classify the data for storage;

[0009] Step 2: Construction of domain knowledge graph

[0010] Based on pre-processed data, the welding entities, relationships, and attributes are identified, and three-dimensional spatial attributes are associated. An ontology modeling tool is used to construct a domain knowledge graph that supports spatial semantics and store it in a graph database.

[0011] Step 3: Three-dimensional welding process expression based on triples

[0012] Expand the "entity-relationship-entity" triple structure, add three-dimensional spatial attributes, and generate reasonable triple chains;

[0013] Step 4: Intelligent Reasoning Construction of Welding Process

[0014] Input welding requirements and space and process constraints, use rule engine reasoning to generate the initial process flow plan, and combine multi-objective optimization algorithm to perform global optimization of process parameters;

[0015] Step 5: Welding process storage and evaluation

[0016] The optimization plan is stored in the graph database, and the performance of the plan is evaluated through quantitative indicators and fed back to the knowledge graph to achieve closed-loop optimization.

[0017] Furthermore, in step 1, the method steps for collecting and preprocessing welding process knowledge are as follows:

[0018] 11) Data classification: Classify data sources according to the welding process flow of welding preparation, processing, post-processing and quality inspection;

[0019] 12) Data extraction: extracting three-dimensional data and unstructured process parameters;

[0020] 13) Data preprocessing: including data cleaning and data standardization. The data cleaning includes data deduplication, missing data repair, and erroneous data correction; the data standardization is to a unified data format and data unit;

[0021] 14) Data storage: Store the pre-processed welding process data into the database.

[0022] Furthermore, in step 2, the method steps for constructing the domain knowledge graph are as follows:

[0023] 21) Knowledge extraction of welding process data: identifying key entities and relationships between entities from welding process data, and extracting attribute information of key entities;

[0024] 22) 3D attribute association: associate the relationship information related to 3D spatial attributes in the welding process; the 3D spatial attributes include welding equipment layout coordinates, welding path spatial sequence and robot workspace range;

[0025] 23) Constructing ontology model: defining the language model of welding process knowledge, starting from the process logic relationship and three-dimensional space constraint relationship, clarifying the types, relationships and attributes between entities;

[0026] 24) Knowledge storage: Use graph database to store the constructed domain knowledge graph.

[0027] Furthermore, in step 23), Protégé is used to construct an ontology model, and the core classes of the ontology model include welding equipment, welding materials, process parameters, welding paths, three-dimensional coordinates and equipment subclasses.

[0028] Furthermore, in step three, the method steps for expressing the three-dimensional welding process based on the triplet are as follows:

[0029] 31) Triple extension design: Expanding the "entity-relationship-entity" triple structure to add three-dimensional spatial attributes;

[0030] 32) 3D semantic mapping: Based on the 3D spatial attributes in the domain knowledge graph, create intermediate nodes for welding process instances, decompose multi-dimensional relationships into triple chains containing central semantics, and realize the mapping of 3D semantics in the domain knowledge graph;

[0031] 33) Knowledge graph association integration: The triples that extend and add three-dimensional spatial attributes are bound and associated with knowledge graph entities to form a semantic network that can be queried and indexed.

[0032] Furthermore, in step 4, the method steps for constructing the intelligent reasoning of the welding process are as follows:

[0033] 41) Determine input requirements and loading constraints: Determine input information including material type and welding equipment from actual welding production requirements, load space and process dual constraints to ensure non-interference in model layout and reasonable process flow;

[0034] 42) Generate a plan: Define a rule engine based on the welding experience specifications in actual production, and infer and generate an initialization process flow plan; the rule engine uses the Drools engine, and its rule file includes logical rules, business rules and calculation rules.

[0035] 43) Target optimization: Set optimization targets including production efficiency, production cost and production quality, and use multi-objective optimization algorithms to globally optimize process parameters.

[0036] Furthermore, in step 43), the multi-objective optimization algorithm adopts the NSGA-II algorithm, and its optimization objective function includes:

[0037] Production efficiency objective function:

[0038]

[0039] Where: L i is the length of the welding path of the i-th section; t i (x) is the welding time under the current decision variable configuration x = [I, V, S, T]; I is the welding current; V is the welding voltage; S is the welding speed; T is the path time interval;

[0040] Production cost objective function:

[0041]

[0042] Where: C energy and C materials are energy cost coefficient and material cost coefficient respectively;

[0043] Quality deviation objective function:

[0044]

[0045] in: Score the target; Q i (x) is the welding quality score under the decision variable configuration x = [I, V, S, T];

[0046] The constraints include: the upper limit of the welding production line's safe current, the lower limit of welding quality, and the lower limit of the welding robot's motion parameters.

[0047] Furthermore, in step 5, the method steps for storing and evaluating the welding process are as follows:

[0048] 51) Quantitative evaluation indicators: define multi-dimensional evaluation indicators and normalize the calculation results of different dimensions contained in each indicator;

[0049] 52) Scheme evaluation and feedback: Classify and evaluate production schemes based on scoring values, evaluate generated process flow schemes based on defined comprehensive evaluation indicators, and generate feedback recommendations;

[0050] 53) Storage and Update: The process flow plan after evaluation feedback is stored in the knowledge graph database, the configuration rule file is updated, and the parameters are optimized to achieve closed-loop management.

[0051] Furthermore, in step 51), the evaluation indicators include efficiency indicators, cost indicators and quality indicators;

[0052] The efficiency index is expressed as:

[0053] E=α1T t ′+α2U e ′+α3L r '

[0054] Where: E is the efficiency index; T t ′ represents the total welding time normalized to the value of [0,1] interval; U e ′ represents the value of equipment utilization efficiency normalized to the interval [0,1]; L r ′ represents the path length ratio normalized to the value of the interval [0,1]; α1, α2 and α3 are weight coefficients;

[0055] The cost indicator is expressed as:

[0056] C=β1C e ′+β2C m '

[0057] Among them: C is the cost index; C e ′ represents the energy consumption cost normalized to the value in the range [0,1]; C m ′ represents the value of material loss cost normalized to the interval [0,1]; β1 and β2 are weight coefficients;

[0058] The quality index is expressed as:

[0059] Q=γ1ΔQ′+γ2R d '

[0060] Where: Q is the quality index; ΔQ′ represents the value of the intensity deviation normalized to the interval [0,1]; R d ′ represents the surface defect rate normalized to the value in the interval [0,1]; γ1 and γ2 are weight coefficients.

[0061] Furthermore, in step 52), the comprehensive evaluation index is expressed as:

[0062] S=w E E+w C (1―C)+w Q (1-Q)

[0063] Among them: S is the comprehensive evaluation index; w E 、w C and w Q is the weight coefficient, and w E +w C +w Q =1.

[0064] The beneficial effects of the present invention are:

[0065] The three-dimensional expression and construction method of the welding process based on the domain knowledge graph of the present invention has the following technical effects:

[0066] (1) Three-dimensional expression of welding process flow: By expanding the triple structure and associating three-dimensional spatial attributes, the traditional two-dimensional process knowledge is upgraded to a three-dimensional semantic level, solving the accuracy issues of welding path planning and equipment interference detection;

[0067] (2) Intelligent decision-making closed-loop optimization: Combined with the rule engine, the initial process flow plan is quickly generated, and the multi-objective optimization algorithm is used to perform multi-objective optimization of welding process parameters to achieve collaborative optimization of production efficiency, cost, and quality, significantly shortening the process design cycle;

[0068] (3) Enhanced knowledge reusability: By feeding back quantitative evaluation indicators to the knowledge graph, the rule files are dynamically updated to form a closed loop of process knowledge precipitation-optimization-reuse, thereby improving the digital management level of the welding process.

[0069] In summary, the three-dimensional expression and construction method of the welding process flow based on the domain knowledge graph of the present invention can effectively improve the digital level of the welding process flow, meet the dynamic welding production needs and improve the intuitiveness and operability of process management, and provide auxiliary domain knowledge graph related research for the reasoning construction and visual expression of the welding process flow. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to make the purpose, technical solutions and beneficial effects of the present invention more clear, the present invention provides the following drawings for illustration:

[0071] Figure 1 This is a flowchart of the three-dimensional expression and construction method of the welding process based on the domain knowledge graph of the present invention;

[0072] Figure 2 This is the structural framework diagram of the Drools rule engine. DETAILED DESCRIPTION

[0073] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0074] Taking the visualization expression of the welding process flow of alloy steel structural parts in a welding production line of a certain production workshop as an example, the specific implementation method of the three-dimensional expression and construction method of the welding process flow based on the domain knowledge graph of the present invention is described in detail.

[0075] like Figure 1 As shown, the three-dimensional expression and construction method of the welding process flow based on the domain knowledge graph in this embodiment includes the following steps.

[0076] Step 1: Welding process knowledge collection and preprocessing

[0077] Collect multi-source heterogeneous welding process data, clean, standardize and classify the data for storage, and provide high-quality data support for the construction of domain knowledge graphs.

[0078] Specifically, in this embodiment, the method steps for collecting and preprocessing welding process knowledge are as follows.

[0079] 11) Data classification: Classify data sources according to the welding process flow of welding preparation, processing, post-processing and quality inspection. Specifically, welding process knowledge collection requires the collection of multi-source heterogeneous data of the entire welding process (including preparation, processing, post-processing and quality inspection), so the collected data sources can be classified according to each process stage. In the collection of welding preparation data sources, it is necessary to collect the material characteristic parameters of alloy steel structural parts and welded parts, the layout location information of the entire production workshop, and the CAD three-dimensional models of key equipment such as production, welding, and logistics. In the collection of welding processing data sources, it is necessary to collect the coordinate sequence of the welding path and welding process parameters such as current and speed. For data collection in the post-processing and quality inspection stages of welding, it is necessary to collect standardized paper information such as the welding process manual and product quality standards in the production workshop.

[0080] 12) Data Extraction: Extract 3D data and unstructured process parameters. Specifically, data extraction from the data source involves not only extracting unstructured process parameter information from welding process documents, but also calibrating and extracting the 3D model information of the production workshop using CAD tools.

[0081] 13) Data preprocessing: This includes data cleaning and standardization. Data cleaning involves deduplication, missing data correction, and erroneous data correction. Data standardization involves standardizing the data format and units. Specifically, preprocessing the collected data files involves deleting duplicate data, adding missing data, removing irrelevant data, and filtering and correcting erroneous information. The collected data formats and units are standardized and categorized by knowledge type, providing clear data support for the subsequent construction of domain knowledge graphs.

[0082] 14) Data storage: Store the pre-processed welding process data into the database.

[0083] Step 2: Construction of domain knowledge graph

[0084] Based on the preprocessed data, welding entities, relationships and attributes are identified, three-dimensional spatial attributes are associated, and a domain knowledge graph supporting spatial semantics is constructed using ontology modeling tools and stored in a graph database.

[0085] Specifically, the method and steps for constructing the domain knowledge graph are as follows.

[0086] 21) Knowledge Extraction from Welding Process Data: Identify key entities and relationships between them from the welding process data, and extract attribute information for these key entities. Specifically, first, identify key entities from the collected data that contribute to the knowledge of the welding process, such as "welding equipment," "welding materials," and "welding process parameters." Second, identify relationships between entities based on the logical relationships within the process, such as "welding equipment → use → welding materials," and extract attribute information for these key entities.

[0087] 22) Three-dimensional Attribute Association: Attribute association is performed on the relationship information involving three-dimensional spatial attributes in the welding process; the three-dimensional spatial attributes include the welding equipment layout coordinates, the welding path spatial sequence, and the robot workspace range. Specifically, the identified key entities need to be dimensionally classified. Entities that do not require three-dimensional attribute association, such as alloy steel material properties, pretreatment temperature, and welding method, are defined as two-dimensional entities. Entities that involve three-dimensional layers in the welding process are defined as three-dimensional entities. Three-dimensional coordinate information is extracted from the data source, and attribute association is performed on the three-dimensional entity relationship information, such as the equipment layout coordinates, the welding path sequence, and the welding robot workspace.

[0088] 23) Constructing an ontology model: This defines a language model for welding process knowledge. Based on process logic relationships and three-dimensional spatial constraints, the types, relationships, and attributes between entities are clarified. In this embodiment, Protégé is used to construct the ontology model. The core classes of the ontology model include welding equipment, welding materials, process parameters, welding paths, three-dimensional coordinates, and equipment subclasses. Table 1 shows the core classes of the ontology model constructed using Protégé, along with their main data attributes and main objects.

[0089] Table 1 Core classes of the ontology model and their main data attributes and main objects

[0090]

[0091] 24) Knowledge storage: Use a graph database (such as Neo4j) to store the constructed domain knowledge graph, supporting efficient query and reasoning.

[0092] Step 3: Three-dimensional welding process expression based on triples

[0093] The "entity-relationship-entity" triple structure is expanded to add three-dimensional spatial attributes to generate a reasonable triple chain. Specifically, the traditional triple is expanded on the basis of adding three-dimensional spatial attributes, and the three-dimensional welding process information is encoded and integrated into a reasonable triple form to achieve the precise expression of three-dimensional semantic knowledge.

[0094] In this embodiment, the method steps for expressing a three-dimensional welding process based on a triplet are as follows.

[0095] 31) Triple Extension Design: Expand the "entity-relationship-entity" triple structure by adding three-dimensional spatial attributes. Traditional triples simply contain "entity-relationship-entity," which is a static triple that lacks three-dimensional information and cannot express the spatial hierarchical information of the welding process. Therefore, this embodiment expands the spatial semantic triple design for the three-dimensional entity defined in step 22) based on the traditional triple design, adding three-dimensional spatial attributes to include the three-dimensional information of the welding process data. The example of the structured data generated by the extended triple is: Equipment ID: EQ001; Model: ARC200; Maximum Current: 200A; Coordinates: (100, 150, 20).

[0096] 32) Three-dimensional semantic mapping: Based on the three-dimensional spatial attributes in the domain knowledge graph, create intermediate nodes of welding process instances, decompose multi-dimensional relationships into triple chains containing central semantics, and realize the mapping of three-dimensional semantics in the domain knowledge graph.

[0097] 33) Knowledge graph association integration: The triples that extend and add three-dimensional spatial attributes are bound and associated with knowledge graph entities to form a semantic network that can be queried and indexed.

[0098] Step 4: Intelligent Reasoning Construction of Welding Process

[0099] Input welding requirements as well as space and process constraints, use rule engine reasoning to generate the initial process flow plan, and combine multi-objective optimization algorithm to perform global optimization of process parameters.

[0100] In this embodiment, the method steps for constructing the intelligent reasoning of the welding process are as follows:

[0101] 41) Determine input requirements and load constraints: Determine input information including material type and welding equipment from actual welding production requirements, and load spatial and process constraints to ensure that the model layout does not interfere and the process flow is reasonable. Specifically, input information such as material type and welding equipment is determined from actual welding production requirements. For example, in this example, the structural material is alloy steel and the welding equipment is a six-axis industrial general-purpose welding robot. The input requirement information can be refined based on the input end of the production task. Loading constraints mainly include two categories: spatial constraints and process constraints. Spatial constraints are loaded for three-dimensional process entities, and overlap is calculated by calculating the spatial position range to avoid interference. Specifically, in the welding process, this includes obstacle avoidance for the welding path and equipment accessibility. Process constraints are loaded by applying process constraint rules to process entities based on standardized data sources such as process knowledge manuals to ensure the process rationality of the inference-constructed process flow. Specifically, in the welding process, this includes matching welding materials and welding methods, setting thresholds for welding parameters, and ensuring a reasonable range of weld strength at the end of welding.

[0102] 42) Generate a plan: Define a rule engine based on actual welding experience and specifications, and generate an initial process plan by reasoning. In this embodiment, the rule engine uses the Drools engine, and its rule file includes logic rules, business rules, and calculation rules.

[0103] Specifically, a knowledge graph rule engine was defined based on welding experience specifications used in actual production. The rule file for the rule engine contains logical rules (OWL ontology classes and attribute constraints), business rules (welding process expert experience decisions), and calculation rules (key index value condition judgments such as parameter thresholds). Based on the complex scenarios and multi-semantic situations in actual welding production lines, the Drools rule engine was selected. The Drools rule file was imported into the engine and configured with a linked domain knowledge graph database. The engine then uses this inference to construct and generate an initial process flow plan.

[0104] 43) Target optimization: Set optimization targets including production efficiency, production cost and production quality, and use multi-objective optimization algorithms to globally optimize process parameters.

[0105] Specifically, in this embodiment, the multi-objective optimization algorithm adopts the NSGA-II algorithm, and the global optimization process for the key process parameters in the initial process flow plan is as follows.

[0106] (1) Define decision variables and optimization objective function:

[0107] Decision variables: welding current I (unit: A); welding voltage V (unit: V); welding speed S (unit: mm / s); robot welding path interval time T (unit: s).

[0108] The objective functions include production efficiency objective function, production cost objective function and quality deviation objective function.

[0109] Production efficiency objective function (maximize efficiency, minimize the negative sign):

[0110]

[0111] Where: L i is the length of the welding path of the i-th section; t i (x) is the welding time under the current decision variable configuration x = [I, V, S, T]; I is the welding current; V is the welding voltage; S is the welding speed; T is the path time interval.

[0112] Production cost objective function:

[0113]

[0114] Where: C energy and C materials are the energy consumption cost coefficient and material cost coefficient respectively.

[0115] Quality deviation objective function:

[0116]

[0117] in: Score the target; Q i (x) is the welding quality score under the decision variable configuration x = [I, V, S, T];

[0118] (2) Constraints and parameter settings

[0119] The value range of decision variables is: 50≤I≤600(A),20≤V≤35(V),1.0≤S≤10.0(mm / s),0.5≤T≤5.0(s).

[0120] The constraints include: the upper limit of the welding production line's safe current, the lower limit of welding quality, and the lower limit of the welding robot's motion parameters.

[0121] (3) Result processing and output

[0122] Based on the production focus, such as lowest cost or highest quality, a suitable solution is selected from the result set, the selected decision variable parameter x* is stored in the form of a Json file, and stored in the domain knowledge graph knowledge base to complete the optimization closed loop.

[0123] Step 5: Welding process storage and evaluation

[0124] The optimization plan is stored in the graph database, and the performance of the plan is evaluated through quantitative indicators and fed back to the knowledge graph to achieve closed-loop optimization.

[0125] In this embodiment, the method steps for storing and evaluating the welding process are as follows:

[0126] 51) Quantitative evaluation indicators: Define multi-dimensional evaluation indicators and normalize the calculation results of different dimensions contained in each indicator. Specifically, starting from the three main indicators of efficiency, cost and quality, the three indicators are further refined according to the production characteristics of the welding production line, and the main indicators are quantified with multiple sub-indicators. Among them, the efficiency indicator can be quantified by the total welding time, equipment utilization rate and path length ratio; the cost indicator can be quantified by energy consumption cost and material loss cost; the quality indicator can be quantified by strength deviation and surface defect rate. The calculation results of the sub-indicators with different dimensions contained in each main indicator are uniformly normalized to the interval [0,1] for the next step of evaluation feedback.

[0127] That is, in this embodiment, the evaluation indicators include efficiency indicators, cost indicators and quality indicators.

[0128] This embodiment divides the efficiency index into three sub-indicators: "total welding time", "equipment utilization rate" and "path length ratio", which comprehensively constitute the efficiency performance evaluation. Specifically, the efficiency index is expressed as:

[0129] E=α1T t ′+α2U e ′+α3L r '

[0130] Where: E is the efficiency index; T t ′ represents the total welding time normalized to the value of [0,1] interval; U e ′ represents the value of equipment utilization efficiency normalized to the interval [0,1]; L r ′ represents the path length ratio normalized to the value of the interval [0,1]; α1, α2 and α3 are weight coefficients.

[0131] The total welding time is:

[0132]

[0133] Where: T t Indicates the total welding time; n is the number of welding sections; t i is the welding time of the i-th segment, in seconds.

[0134] The equipment utilization rate is:

[0135]

[0136] Among them: Ue Indicates equipment utilization efficiency; T w is the actual welding working time; T a The total time the device is available, including standby and maintenance time.

[0137] The path length ratio is:

[0138]

[0139] Where: L r represents the path length ratio; L i is the length of the i-th path; L max is the total path length at maximum load in millimeters.

[0140] This embodiment divides the cost index into two sub-indicators: "energy consumption cost" and "material loss cost", which are used to comprehensively constitute the cost evaluation. Specifically, the cost index is expressed as:

[0141] C=β1C e ′+β2C m '

[0142] Among them: C is the cost index; C e ′ represents the energy consumption cost normalized to the value in the range [0,1]; C m ′ represents the material loss cost normalized to the value in the interval [0,1]; β1 and β2 are weight coefficients.

[0143] The energy cost is:

[0144] C e =T t ×P avg ×c e

[0145] Where: C e represents energy consumption cost; P avg Indicates the average power of welding equipment; c e It is the electricity price coefficient, and its unit is yuan per kilowatt-hour.

[0146] The cost of material loss is:

[0147]

[0148] Where: C m Indicates material loss cost; Δm j is the loss weight of the jth type of material; c mj The unit cost of the material is RMB per kilogram.

[0149] This embodiment divides the quality index into two sub-indicators: "strength deviation" and "surface defect rate", which comprehensively constitute the quality performance evaluation. Specifically, the quality index is expressed as:

[0150] Q=γ1ΔQ′+γ2R d '

[0151] Where: Q is the quality index; ΔQ′ represents the value of the intensity deviation normalized to the interval [0,1]; R d ′ represents the surface defect rate normalized to the value in the interval [0,1]; γ1 and γ2 are weight coefficients.

[0152] The intensity deviation is:

[0153]

[0154] Where: ΔQ represents the intensity deviation; represents the measured strength of the weld in section i; is the corresponding target intensity value in MPa.

[0155] The surface defect rate is:

[0156]

[0157] Where: R d Represents the surface defect rate; where N d is the number of defect points detected; N i is the total number of checkpoints.

[0158] 52) Scheme Evaluation and Feedback: The production scheme is evaluated and categorized based on the scoring values. The generated process flow scheme is evaluated and feedback results are generated based on the defined comprehensive evaluation indicators. The comprehensive score of the scheme is calculated based on the quantitative indicator values ​​calculated and solved by the weighted summation method, and the production scheme is evaluated and categorized based on the scoring values. The calculation results of the quantitative indicators can be fed back to the target optimization step 43) to feedback and redesign the optimization target weights. Specifically, in this embodiment, the comprehensive evaluation indicator is expressed as:

[0159] S=w E E+w C (1―C)+w Q (1-Q)

[0160] Among them: S is the comprehensive evaluation index; w E 、w C and W Q is the weight coefficient, and w E +w C +w Q =1.

[0161] Specifically, this embodiment takes the total minus normalized value of the cost and quality factors, and unifies the dimensions with the principle of "the larger the better".

[0162] 53) Storage and Update: The process flow plan after evaluation feedback is stored in the knowledge graph database, the configured rule file is updated, and the parameters are optimized to achieve closed-loop management. The process plan after evaluation feedback (including welding process parameters, welding methods, evaluation results, etc.) is stored in the graph database Neo4j. The configured rule file is then updated to add constraint rules and adjust the optimization target parameter weights to complete the knowledge graph update and completion. The stored process plan is also added with a version model and timestamp to support historical backtracking.

[0163] The above embodiments are merely preferred embodiments for the purpose of fully illustrating the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are within the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

Claims

1. A three-dimensional expression and construction method for welding process flow based on domain knowledge graph, characterized by: The steps include: Step 1: Welding process knowledge collection and preprocessing Collect multi-source heterogeneous welding process data, clean, standardize and classify the data for storage; Step 2: Construction of domain knowledge graph Based on pre-processed data, the welding entities, relationships, and attributes are identified, and three-dimensional spatial attributes are associated. An ontology modeling tool is used to construct a domain knowledge graph that supports spatial semantics and store it in a graph database. Step 3: Three-dimensional welding process expression based on triples Expand the "entity-relationship-entity" triple structure, add three-dimensional spatial attributes, and generate reasonable triple chains; Step 4: Intelligent Reasoning Construction of Welding Process Input welding requirements and space and process constraints, use rule engine reasoning to generate the initial process flow plan, and combine multi-objective optimization algorithm to perform global optimization of process parameters; Step 5: Welding process storage and evaluation The optimization plan is stored in the graph database, and the performance of the plan is evaluated through quantitative indicators and fed back to the knowledge graph to achieve closed-loop optimization.

2. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 1 is characterized in that: In step 1, the method steps for collecting and preprocessing welding process knowledge are as follows: 11) Data classification: Classify data sources according to the welding process flow of welding preparation, processing, post-processing and quality inspection; 12) Data extraction: extracting three-dimensional data and unstructured process parameters; 13) Data preprocessing: including data cleaning and data standardization. The data cleaning includes data deduplication, missing data repair, and erroneous data correction; the data standardization is to a unified data format and data unit; 14) Data storage: Store the pre-processed welding process data into the database.

3. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 1 is characterized in that: In step 2, the method steps for constructing the domain knowledge graph are as follows: 21) Knowledge extraction of welding process data: identifying key entities and relationships between entities from welding process data, and extracting attribute information of key entities; 22) 3D attribute association: associate the relationship information related to 3D spatial attributes in the welding process; the 3D spatial attributes include welding equipment layout coordinates, welding path spatial sequence and robot workspace range; 23) Constructing ontology model: defining the language model of welding process knowledge, starting from the process logic relationship and three-dimensional space constraint relationship, clarifying the types, relationships and attributes between entities; 24) Knowledge storage: Use graph database to store the constructed domain knowledge graph.

4. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 3 is characterized by: In the step 23), Protégé is used to construct an ontology model, and the core classes of the ontology model include welding equipment, welding materials, process parameters, welding paths, three-dimensional coordinates and equipment subclasses.

5. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 1 is characterized in that: In step 3, the method steps for expressing the three-dimensional welding process based on the triplet are as follows: 31) Triple extension design: Expanding the "entity-relationship-entity" triple structure to add three-dimensional spatial attributes; 32) 3D semantic mapping: Based on the 3D spatial attributes in the domain knowledge graph, create intermediate nodes for welding process instances, decompose multi-dimensional relationships into triple chains containing central semantics, and realize the mapping of 3D semantics in the domain knowledge graph; 33) Knowledge graph association integration: The triples that extend and add three-dimensional spatial attributes are bound and associated with knowledge graph entities to form a semantic network that can be queried and indexed.

6. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 1 is characterized in that: In step 4, the method steps for constructing the intelligent reasoning of the welding process are as follows: 41) Determine input requirements and loading constraints: Determine input information including material type and welding equipment from actual welding production requirements, load space and process dual constraints to ensure non-interference in model layout and reasonable process flow; 42) Generate a plan: Define a rule engine based on actual welding experience specifications in production, and generate an initial process flow plan by reasoning; the rule engine uses the Drools engine, and its rule file includes logical rules, business rules, and calculation rules; 43) Target optimization: Set optimization targets including production efficiency, production cost and production quality, and use multi-objective optimization algorithms to globally optimize process parameters.

7. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 6 is characterized in that: In step 43), the multi-objective optimization algorithm adopts the NSGA-II algorithm, and its optimization objective function includes: Production efficiency objective function: Where: L i is the length of the welding path of the i-th section; t i (x) is the welding time under the current decision variable configuration x = [I, V, S, T]; I is the welding current; V is the welding voltage; S is the welding speed; T is the path time interval; Production cost objective function: Where: C energy and C materials are energy cost coefficient and material cost coefficient respectively; Quality deviation objective function: in: Score the target; Q i (x) is the welding quality score under the decision variable configuration x = [I, V, S, T]; The constraints include: the upper limit of the welding production line's safe current, the lower limit of welding quality, and the lower limit of the welding robot's motion parameters.

8. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 1 is characterized in that: In step 5, the method steps for storing and evaluating the welding process are as follows: 51) Quantitative evaluation indicators: define multi-dimensional evaluation indicators and normalize the calculation results of different dimensions contained in each indicator; 52) Scheme evaluation and feedback: Classify and evaluate production schemes based on scoring values, evaluate generated process flow schemes based on defined comprehensive evaluation indicators, and generate feedback recommendations; 53) Storage and Update: The process flow plan after evaluation feedback is stored in the knowledge graph database, the configuration rule file is updated, and the parameters are optimized to achieve closed-loop management.

9. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 8 is characterized in that: In the step 51), the evaluation indicators include efficiency indicators, cost indicators and quality indicators; The efficiency index is expressed as: E=α1T t ′+α2U e ′+α3L r ′ Where: E is the efficiency index; T t ′ represents the total welding time normalized to the value of [0,1] interval; U e ′ represents the value of equipment utilization efficiency normalized to the interval [0,1]; L r ′ represents the path length ratio normalized to the value of the interval [0,1]; α1, α2 and α3 are weight coefficients; The cost indicator is expressed as: C=β1C e ′+β2C m ′ Among them: C is the cost index; C e ′ represents the energy consumption cost normalized to the value in the range [0,1]; C m ′ represents the value of material loss cost normalized to the interval [0,1]; β1 and β2 are weight coefficients; The quality index is expressed as: Q=γ1ΔQ′+γ2R d ′ Where: Q is the quality index; ΔQ′ represents the value of the intensity deviation normalized to the interval [0,1]; R d ′ represents the surface defect rate normalized to the value in the interval [0,1]; γ1 and γ2 are weight coefficients.

10. The method for three-dimensionally expressing and constructing a welding process flow based on a domain knowledge graph according to claim 9 is characterized in that: In step 52), the comprehensive evaluation index is expressed as: S=w E E+w C (1―C)+w Q (1―Q) Among them: S is the comprehensive evaluation index; w E 、w C and w Q is the weight coefficient, and w E +w C +w Q =1.

Images