Welding fixture design method and system based on large language model

Abstract

The invention provides a welding fixture design method and system based on a large language model, and the method comprises the steps: carrying out the clustering and grouping of a plurality of historical welding fixture assemblies, obtaining a plurality of welding fixture groups, and building a parameterized template; traversing the initial welding fixture assembly to obtain parameters of the initial welding fixture assembly; welding fixture design requirements are input into the large language model to obtain a matched parameterized template; analyzing the parameters of the initial welding fixture assembly through a large language model, obtaining a parameter modification instruction in combination with the matched parameterized template, adjusting the initial welding fixture assembly until the parameters of the initial welding fixture assembly meet optimization requirements, obtaining parameters of a target welding fixture assembly, and generating the target welding fixture assembly. And completing the welding fixture design. According to the method, welding fixture design experience parameterization is achieved, parameter semantics are extracted through a large language model to correspondingly optimize and adjust welding fixture assembly parameters, then welding fixture design meeting engineering requirements is generated, and the welding fixture design reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of automated design of vehicle body fixtures, and in particular to a welding fixture design method and system based on a large language model. Background Technology

[0002] Welding fixtures are used on automobile production assembly lines to position, constrain, and support welded workpieces during the welding process. They come in various structural types and are generally customized non-standard parts. With the accumulation of historical CAD model data by enterprises, there are a large number of reusable design resources. How to reliably adapt these design resources to various complex engineering needs has become a technical problem that needs to be studied.

[0003] Currently, existing technologies rely on engineers' experience and manual modeling. For the same welding workpiece model under different engineering needs, this involves repeatedly building and adjusting parameters in 3D modeling software using parameters or sketches, followed by interference checks and process verification. However, this method is cumbersome and susceptible to human subjective factors, resulting in poor reliability of welding fixture design. Furthermore, existing technologies employ automated paths such as rule scripts and template-driven approaches. When facing various engineering requirements, this method typically requires extensive manual maintenance of rules and templates, relying on human experience. Under complex working conditions, it is also easily affected by human subjective factors, leading to poor reliability of welding fixture design. Therefore, it is evident that existing technologies struggle to meet complex engineering requirements under complex working conditions, resulting in poor reliability of welding fixture design. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes a welding fixture design method and system based on a large language model. This method parameterizes welding fixture design experience and extracts parameter semantics through a large language model to optimize and adjust the parameters of the welding fixture assembly accordingly, thereby generating a welding fixture design that meets engineering requirements and improving the reliability of the welding fixture design.

[0005] To achieve the above objectives, embodiments of the present invention provide a welding fixture design method based on a large language model, comprising: clustering and grouping several historical welding fixture assemblies based on pre-acquired data to obtain several welding fixture groups and establishing parameterized templates based on these welding fixture groups; traversing the pre-acquired initial welding fixture assemblies to obtain initial welding fixture assembly parameters; based on pre-acquired welding fixture design requirements, inputting the welding fixture design requirements into a preset large language model to match the corresponding parameterized template to obtain a matching parameterized template; inputting the initial welding fixture assembly parameters and the matching parameterized template into the preset large language model, parsing the initial welding fixture assembly parameters through the preset large language model, and obtaining parameter modification instructions based on the matching parameterized template; adjusting the initial welding fixture assembly based on the parameter modification instructions until the initial welding fixture assembly parameters meet preset optimization requirements to obtain target welding fixture assembly parameters; and generating a target welding fixture assembly based on the target welding fixture assembly parameters to complete the welding fixture design.

[0006] This invention proposes a welding fixture design method based on a large language model. It constructs a parametric template to solidify variable design experience, extracts initial welding fixture assembly parameters understandable by the pre-defined large language model from the initial welding fixture assembly, and obtains a matching parametric template based on the welding fixture design requirements. Then, it inputs the initial welding fixture assembly parameters and the matching parametric template into the pre-defined large language model to replace manual reasoning, generating parameter modification instructions to adjust the initial welding fixture assembly. An iterative optimization method is used to obtain the target welding fixture assembly parameters and generate the target welding fixture assembly, ultimately completing the welding fixture design. Therefore, by constructing a parametric template library and extracting structured parameters from the assembly, using a pre-defined large language model for semantic parsing and parameter reasoning, and then optimizing and adjusting the welding fixture assembly parameters to generate a welding fixture design that meets engineering requirements, the reliability of the welding fixture design is improved.

[0007] Furthermore, based on several pre-acquired historical welding fixture assemblies, these assemblies are clustered and grouped to obtain several welding fixture groups. A parameterized template is then established based on these welding fixture groups. This process includes: acquiring several historical welding fixture assemblies from a pre-set historical design library; extracting historical semantic information corresponding to these historical welding fixture assemblies; performing semantic clustering on the historical semantic information corresponding to each historical welding fixture assembly according to pre-set clustering rules, and grouping historical welding fixture assemblies with the same historical semantic information into a corresponding welding fixture group, thus obtaining several welding fixture groups; and selecting several target parameters that satisfy pre-set parameter dimensions within each welding fixture group, and setting constraints on each target parameter to obtain the parameterized template corresponding to each welding fixture group.

[0008] In the above scheme, several historical welding fixture assemblies are obtained from the historical design library, and their corresponding historical semantic information is clustered and grouped according to preset clustering rules. Within each welding fixture group, several target parameters satisfying preset parameter dimensions are selected, and parameterized templates for each welding fixture group are obtained by setting constraint relationships. This transforms scattered engineering experience into structured and parameterizable design units. Thus, welding fixture design experience is parameterized, and by setting parameter dimensions and constraint relationships, subsequent reasoning in the large language model can focus on key and effective parameters, effectively improving the reliability of welding fixture design.

[0009] Furthermore, the initial welding fixture assembly is traversed to obtain its parameters, including: traversing the initial welding fixture assembly through a preset CAD script interface to extract the spatial relationship parameters, functional semantic parameters, and dependency parameters corresponding to each initial welding fixture assembly; and converting the spatial relationship parameters, functional semantic parameters, and dependency parameters into a preset structured text format based on a preset parameter structuring standard to obtain the initial welding fixture assembly parameters.

[0010] In the above scheme, a CAD script interface is used to extract key semantics such as spatial relationships, functional semantics, and dependencies of the initial welding fixture assembly. These spatial relationship parameters, functional semantic parameters, and dependency parameters are then converted into a pre-defined structured text format using a pre-defined parameter structured standard. This yields the initial welding fixture assembly parameters, enabling the subsequent large language model to parse the corresponding key semantic information, obtain the fixture's structure and design intent, and provide accurate and rich input information for subsequent intelligent reasoning and parameter optimization by the large language model. Therefore, converting complex semantic information into a format that the large language model can parse helps improve the reliability of welding fixture design.

[0011] Furthermore, based on the pre-acquired welding fixture design requirements, the welding fixture design requirements are input into a preset large language model to match the corresponding parametric template, resulting in a matching parametric template. This includes: based on the pre-acquired geometric feature information, welding process information, and fixture installation environment information, converting the geometric feature information, welding process information, and fixture installation environment information into target text format to obtain the welding fixture design requirements; inputting the welding fixture design requirements into a preset large language model to match a parametric template that meets the preset design requirements, resulting in a matching parametric template.

[0012] In the above scheme, complex multi-dimensional information such as geometric feature information, welding process information, and fixture installation environment information are converted into a unified target text format, so that the subsequent large language model can parse the corresponding requirement information. Then, the welding fixture design requirements are input into the preset large language model. By utilizing the powerful semantic understanding and matching capabilities of the large language model, the initial design scheme that best fits the current working conditions is automatically selected from the template library. This avoids initial design deviations caused by improper subjective selection and helps to improve the reliability of welding fixture design.

[0013] Furthermore, the initial welding fixture assembly parameters and the matching parameterized template are input into a preset large language model. The preset large language model parses the initial welding fixture assembly parameters and generates parameter modification instructions based on the matching parameterized template. This includes: inputting the initial welding fixture assembly parameters and the matching parameterized template into the preset large language model; parsing the functional relationships of the initial welding fixture assembly through the preset large language model to obtain a parameter dependency graph; generating parameter modification suggestions based on the matching parameterized template and the parameter dependency graph; and converting the parameter modification suggestions into an instruction script format to obtain parameter modification instructions.

[0014] In the above scheme, the initial welding fixture assembly parameters and the matching parameterized template are input into a preset large language model. The preset large language model parses the structured parameters to construct a parameter dependency graph reflecting the functional relationships between components. Then, it combines the parameterized template with the parameters to generate corresponding parameter modification suggestions. Finally, the parameter modification suggestions are converted into an executable instruction script format to obtain parameter modification instructions. Thus, by parsing the initial welding fixture assembly to obtain the parameter dependency graph and analyzing the differences between the parameter dependency graph and the matching parameterized template to generate parameter modification suggestions, the parameter modification thinking is closer to human reasoning but is not affected by subjective factors, which helps to improve the reliability of welding fixture design.

[0015] Furthermore, the initial welding fixture assembly is adjusted based on parameter modification instructions until its parameters meet preset optimization requirements, thereby obtaining the target welding fixture assembly parameters. This includes: adjusting the initial welding fixture assembly based on parameter modification instructions to obtain an optimized welding fixture assembly; performing interference and stability tests on the optimized welding fixture assembly sequentially to obtain collision and stability reports; inputting the collision and stability reports into a preset large model, evaluating the initial welding fixture assembly parameters based on a pre-built loss function, and generating evaluation feedback results; if the evaluation feedback results do not meet the preset optimization requirements, the initial welding fixture assembly is adjusted until the evaluation feedback results meet the preset optimization requirements, thereby obtaining the target welding fixture assembly parameters. In the steps of inputting collision reports and stability reports into a preset large model, evaluating the parameters of the initial welding fixture assembly based on a pre-built loss function, and generating evaluation feedback results, the pre-construction process of the loss function includes: obtaining initial points based on the initial welding fixture assembly; obtaining target points based on a matching parameterized template; calculating the distance error between the initial points and the target points to obtain the clamping point error term; and setting corresponding weight coefficients for the preset collision penalty term, the preset clamping stability index, and the clamping point error term to obtain the loss function.

[0016] In the above scheme, after each adjustment of the initial welding fixture assembly, interference tests and stability tests are sequentially performed on the optimized welding fixture assembly. The physical verification results, such as collision reports and stability reports, are fed back to the optimization loop. A loss function is constructed for quantitative evaluation. The construction of the loss function integrates the clamping point error term, the collision penalty term, and the clamping stability index, and assigns adjustable weight coefficients to them. This transforms the complex engineering quality judgment into a calculable scalar value. The construction of the loss function enables the large language model to clearly and efficiently evaluate the comprehensive score of each design iteration and guide the parameter adjustment strategy based on the direction of the loss value change. Finally, the initial welding fixture assembly is continuously iterated and optimized until the evaluation feedback results meet the preset optimization requirements, and the parameters of the target welding fixture assembly are obtained. This allows for the subsequent generation of a reliable target welding fixture assembly to complete the welding fixture design, effectively improving the reliability of the welding fixture design.

[0017] This invention also provides a welding fixture design system based on a large language model, including: a parametric template construction module, a parameter acquisition module, a matching parametric template acquisition module, a parameter modification instruction generation module, a parameter optimization module, and a welding fixture design module; the parametric template construction module is used to cluster and group several historical welding fixture assemblies based on several pre-acquired historical welding fixture assemblies to obtain several welding fixture groups and establish parametric templates based on several welding fixture groups; the parameter acquisition module is used to traverse the pre-acquired initial welding fixture assemblies to obtain the initial welding fixture assembly parameters; the matching parametric template acquisition module is used to, based on the pre-acquired welding fixture design requirements, and... The welding fixture design requirements are input into a preset large language model and matched with a corresponding parametric template to obtain the matched parametric template. A parameter modification instruction generation module is used to input the initial welding fixture assembly parameters and the matched parametric template into the preset large language model. The preset large language model parses the initial welding fixture assembly parameters and generates parameter modification instructions based on the matched parametric template. A parameter optimization module is used to adjust the initial welding fixture assembly based on the parameter modification instructions until the initial welding fixture assembly parameters meet preset optimization requirements, obtaining the target welding fixture assembly parameters. A welding fixture design module is used to generate the target welding fixture assembly based on the target welding fixture assembly parameters to complete the welding fixture design.

[0018] This invention proposes a welding fixture design system based on a large language model. It constructs parameterized templates to solidify varying design experience, extracts initial welding fixture assembly parameters understandable by the pre-defined large language model from the initial welding fixture assembly, and matches a parameterized template based on the welding fixture design requirements. Then, it inputs the initial welding fixture assembly parameters and the matching parameterized template into the pre-defined large language model to replace manual reasoning, generating parameter modification instructions to adjust the initial welding fixture assembly. An iterative optimization method is used to obtain the target welding fixture assembly parameters and generate the target welding fixture assembly, ultimately completing the welding fixture design. Therefore, by constructing a parameterized template library and extracting structured parameters from the assembly, using a pre-defined large language model for semantic parsing and parameter reasoning, and then optimizing and adjusting the welding fixture assembly parameters to generate a welding fixture design that meets engineering requirements, the reliability of the welding fixture design is improved.

[0019] Furthermore, the parameterized template construction module is used to cluster and group several historical welding fixture assemblies based on pre-acquired historical welding fixture assemblies to obtain several welding fixture groups and to build parameterized templates based on these welding fixture groups. This module includes: a historical assembly acquisition unit, a historical semantic information acquisition unit, a welding fixture group construction unit, and a template construction unit. The historical assembly acquisition unit is used to acquire several historical welding fixture assemblies from a preset historical design library. The historical semantic information acquisition unit is used to extract the historical semantic information corresponding to several historical welding fixture assemblies. The welding fixture group construction unit is used to perform semantic clustering on the historical semantic information corresponding to each historical welding fixture assembly according to preset clustering rules, and to group historical welding fixture assemblies with the same historical semantic information into a corresponding welding fixture group, thus obtaining several welding fixture groups. The template construction unit is used to select several target parameters that satisfy preset parameter dimensions in each welding fixture group, and to set constraint relationships for each target parameter, thus obtaining the parameterized template corresponding to each welding fixture group.

[0020] Furthermore, the parameter acquisition module is used to traverse the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters, including: a parameter extraction unit and a text format conversion unit; the parameter extraction unit is used to traverse the pre-acquired initial welding fixture assembly through a preset CAD script interface to extract the spatial relationship parameters, functional semantic parameters, and dependency relationship parameters corresponding to each initial welding fixture assembly; the text format conversion unit is used to convert the spatial relationship parameters, functional semantic parameters, and dependency relationship parameters into a preset structured text format based on a preset parameter structuring standard to obtain the initial welding fixture assembly parameters. Attached Figure Description

[0021] Figure 1 A flowchart illustrating the steps of a welding fixture design method based on a large language model, provided for a certain embodiment of the present invention; Figure 2 A comparative schematic diagram of the initial welding fixture assembly and the target welding fixture assembly of a welding fixture design method based on a large language model provided in a certain embodiment of the present invention; Figure 3 A schematic diagram of a fixture parameterization template for a welding fixture design method based on a large language model, provided in a certain embodiment of the present invention. Figure 1 ; Figure 4 A schematic diagram of a fixture parameterization template for a welding fixture design method based on a large language model, provided in a certain embodiment of the present invention. Figure 2 ; Figure 5 This is a schematic diagram of the module structure of a welding fixture design system based on a large language model, provided for one embodiment of the present invention. Detailed Implementation

[0022] 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.

[0023] Example 1 See Figure 1 , Figure 1 This is a flowchart illustrating the steps of a welding fixture design method based on a large language model, provided in one embodiment of the present invention. Figure 1 As shown in the figure, this embodiment of the invention proposes a welding fixture design method based on a large language model, including steps 101 to 106, each step of which is as follows: Step 101: Based on the pre-acquired historical welding fixture assemblies, the historical welding fixture assemblies are clustered and grouped to obtain several welding fixture groups, and a parameterized template is established based on the several welding fixture groups. Step 102: Traverse the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters; Step 103: Based on the pre-acquired welding fixture design requirements, input the welding fixture design requirements into the preset large language model to match the corresponding parameterized template to obtain the matching parameterized template; Step 104: Input the initial welding fixture assembly parameters and the matching parameterized template into the preset large language model, parse the initial welding fixture assembly parameters through the preset large language model, and obtain parameter modification instructions based on the matching parameterized template; Step 105: Adjust the initial welding fixture assembly based on the parameter modification command until the parameters of the initial welding fixture assembly meet the preset optimization requirements, and obtain the target welding fixture assembly parameters. Step 106: Generate the target welding fixture assembly based on the target welding fixture assembly parameters to complete the welding fixture design.

[0024] One possible implementation involves retrieving an initial welding fixture assembly from a historical design library. In this embodiment, the initial welding fixture assembly is interpreted as a set of CAD assemblies for welding fixtures. Then, based on the clamping position, clamping form, and assembly structural features of the initial welding fixture assembly, semantic clustering and grouping are performed on the CAD assemblies. A parameterized template is then established for each group, containing a minimum set of parameters driving changes in the key geometric features of that group of fixtures, and defining the constraint relationships between each parameter. Next, the initial fixture assembly is traversed through the script interface of the CAD software, extracting information such as component names, assembly hierarchy relationships, bounding boxes, assembly constraints, and functional annotations. The extracted information is then organized into structured JSON (JavaScript Object Notation Schema) according to a predefined JSON Schema. The initial welding fixture assembly parameters are obtained by converting the spatial relationships, functional semantics, and parameter dependencies of the initial fixture assembly into a machine-readable text format using a Notation file. In this embodiment, the initial welding fixture assembly parameters are represented as a structured JSON file. The structured JSON file and the user-inputted design requirements are then input into a preset large language model. In this embodiment, the preset large language model is represented as an LLM-Agent (Large Language Model). The LLM-Agent (or similar agent) defines the design requirements, including welding workpiece information, target clamping point coordinates, and process constraints. Based on these requirements, the LLM-Agent matches the corresponding parameterized template to obtain a matching parameterized template. Then, the LLM-Agent parses the component functions and spatial relationships in the JSON file, combines the design requirements with the matching parameterized template, and generates parameter modification instructions that drive the CAD software. Based on these instructions, the initial welding fixture assembly is adjusted, and interference detection is performed on the adjusted assembly to generate a collision report. The error between the clamping point and the target clamping point of the adjusted assembly is calculated, and its clamping stability is evaluated, generating an evaluation result. A multi-objective loss function is then constructed by combining a collision penalty term, a stability index term, and a clamping point error term. The collision report and evaluation result are fed back to the LLM-Agent. The LLM-Agent, based on the constructed multi-objective loss function and historical optimization records, determines whether the current design meets the convergence condition. If not, a new round of parameter adjustment suggestions is generated, and the process returns to step 104 for iteration until the convergence condition is met or the iteration limit is reached, obtaining the target welding fixture assembly parameters. Finally, see Figure 2 , Figure 2 A comparative schematic diagram of the initial welding fixture assembly and the target welding fixture assembly provided for a welding fixture design method based on a large language model according to a certain embodiment of the present invention; as shown. Figure 2 As shown, the target welding fixture assembly that has passed verification is generated and exported. The export format is generally STEP, CATPart, or Parasolid, and the standard name of the target welding fixture assembly is automatically generated. For example, the comparison results of the initial welding fixture assembly parameter Current_params and the target welding fixture assembly parameter params are shown in Table 1: Table 1 Comparison of Initial Welding Fixture Assembly Parameters and Target Welding Fixture Assembly Parameters The comparison results show that, depending on the different working conditions, the required target welding fixture assembly can be generated simply by adjusting the corresponding parameters.

[0025] After the optimization iteration is completed, the final parameters, iteration process data, simulation verification results and manual correction records of this design task are written into the knowledge base. This knowledge base can be used to expand the parameterized template library and / or fine-tune the training of LLM-Agent in the future. The parameterized template library includes several parameterized templates.

[0026] This invention proposes a welding fixture design method based on a large language model. It constructs a parametric template to solidify variable design experience, extracts initial welding fixture assembly parameters understandable by the pre-defined large language model from the initial welding fixture assembly, and obtains a matching parametric template based on the welding fixture design requirements. Then, it inputs the initial welding fixture assembly parameters and the matching parametric template into the pre-defined large language model to replace manual reasoning, generating parameter modification instructions to adjust the initial welding fixture assembly. An iterative optimization method is used to obtain the target welding fixture assembly parameters and generate the target welding fixture assembly, ultimately completing the welding fixture design. Therefore, by constructing a parametric template library and extracting structured parameters from the assembly, using a pre-defined large language model for semantic parsing and parameter reasoning, and then optimizing and adjusting the welding fixture assembly parameters to generate a welding fixture design that meets engineering requirements, the reliability of the welding fixture design is improved.

[0027] A preferred embodiment involves clustering several historical welding fixture assemblies based on pre-acquired data to obtain several welding fixture groups, and establishing parameterized templates based on these welding fixture groups. This includes: acquiring several historical welding fixture assemblies from a pre-set historical design library; extracting historical semantic information corresponding to the several historical welding fixture assemblies; performing semantic clustering on the historical semantic information corresponding to each historical welding fixture assembly according to pre-set clustering rules, and grouping historical welding fixture assemblies with the same historical semantic information category as a corresponding welding fixture group, thus obtaining several welding fixture groups; and selecting several target parameters that satisfy pre-set parameter dimensions within each welding fixture group, and setting constraint relationships for each target parameter, thereby obtaining the parameterized template corresponding to each welding fixture group.

[0028] One preferred implementation method is described in [reference]. Figure 3 and Figure 4 , Figure 3 A schematic diagram of a fixture parameterization template for a welding fixture design method based on a large language model, provided in a certain embodiment of the present invention. Figure 1 ; Figure 4 A schematic diagram of a fixture parameterization template for a welding fixture design method based on a large language model, provided in a certain embodiment of the present invention. Figure 2 ;like Figure 3 and Figure 4As shown, the initial welding fixture assembly is obtained from the historical design library. In this embodiment, the initial welding fixture assembly is interpreted as a set of CAD assemblies of welding fixtures. Then, based on the clamping position, clamping form, and assembly structure features of the initial welding fixture assembly, semantic clustering and grouping are performed on the CAD assemblies. Then, a parameterized template is established for each group. The parameterized template contains the minimum set of parameters that drive the changes in the key geometric features of the fixture in that group, and defines the constraint relationship of each parameter. Specifically, in this embodiment, the preset clustering rule is set to perform semantic clustering and grouping based on the clamping position, clamping form, and assembly features. According to the function of the fixture, it can be divided into four categories: positioning fixture, support fixture, clamping fixture, and comprehensive fixture. According to the clamping form, it can be divided into top-pressure type, side-pressure type, rotary type, and combination type. The clamping positions include the edge of the outer panel and the position of the reinforcing beam, and the clamping forms include top-pressing, side-pressing, and three-point support. Assembly features include locating pins and base shapes. Then, a category of historical welded fixture assemblies is grouped together. Within each group, several target parameters that satisfy preset parameter dimensions are selected, and corresponding parametric templates are established. The data dimension of the parametric template refers to the minimum set of parameters that can drive key geometric changes. The preset parameter dimension is characterized by containing the minimum set of parameters that can drive key geometric changes. Generally, the minimum set of parameters that can drive key geometric changes includes, but is not limited to, base length, locating pin height, clamp arm length, clamping spacing, and pressure head angle. The constraint relationships defined in the template can generally be set as minimum clearance and maximum stroke. Thus, parametric templates corresponding to each group are obtained. For example, in CATIA software, parametric formulas can be used to define fixture feature dimensions and assembly constraints, giving the template flexible geometric reconfigurability. Figure 3 and Figure 4 In this context, Clamp_i_vertical is the parameter for the vertical direction of clamp i, where Clamp is the clamp and vertical represents the vertical direction, i∈[1,n]; Clamp_j_horizon is the parameter for the horizontal direction of clamp j, where Clamp is the clamp and horizontal represents the horizontal direction, j∈[1,n]; L_Plate_vertical is the parameter for the horizontal direction of clamp 1, representing the vertical length range of the L-shaped fixing seat (variable), where L_Plate represents the L-shaped fixing plate / seat and vertical represents the vertical direction; Support_height is the control parameter for the height of the base, where Support represents support and height represents height.

[0029] In the above scheme, several historical welding fixture assemblies are obtained from the historical design library, and their corresponding historical semantic information is clustered and grouped according to preset clustering rules. Within each welding fixture group, several target parameters satisfying preset parameter dimensions are selected, and parameterized templates for each welding fixture group are obtained by setting constraint relationships. This transforms scattered engineering experience into structured and parameterizable design units. Thus, welding fixture design experience is parameterized, and by setting parameter dimensions and constraint relationships, subsequent reasoning in the large language model can focus on key and effective parameters, effectively improving the reliability of welding fixture design.

[0030] A preferred approach involves traversing the pre-acquired initial welding fixture assemblies to obtain initial welding fixture assembly parameters, including: traversing the pre-acquired initial welding fixture assemblies through a preset CAD script interface, extracting spatial relationship parameters, functional semantic parameters, and dependency relationship parameters corresponding to each initial welding fixture assembly; and converting the spatial relationship parameters, functional semantic parameters, and dependency relationship parameters into a preset structured text format based on a preset parameter structured standard to obtain the initial welding fixture assembly parameters.

[0031] One preferred implementation involves traversing the initial fixture assembly through the script interface of CAD software, extracting information such as part names, assembly hierarchy relationships, spatial bounding boxes, assembly constraints, and functional annotations. The extracted information is then organized into a structured JSON (JavaScript Object Notation) file according to a predefined JSON Schema (JavaScript Object Notation Schema), transforming the spatial relationships, functional semantics, and parameter dependencies of the initial fixture assembly into a machine-readable text format, thus obtaining the parameters of the initial welding fixture assembly. In this embodiment, the script interface of the CAD software can generally be CATIA, UG, Siemens NX, Parasolid API, or Python scripts. The extracted spatial relationship parameters, functional semantic parameters, and dependency parameters corresponding to each initial welding fixture assembly generally include part names, assembly hierarchy relationships (Parent-Child Tree), constraints, bounding boxes (BBoxes), functional annotations, and parameter descriptions. Constraints can generally be Mate, Align, or Axis. Then, the corresponding spatial relationship parameters, functional semantic parameters, and dependency relationship parameters are converted into a preset structured text format. In this embodiment, the preset structured text format can be JSON format. An example illustrating the extraction results organized as a standard JSON structure file is as follows: "clamp_007_a7175".1":{"description":"007 fixture template, commonly used for clamping the upper window frame of a car door. The structure is as follows: an I-shaped base is mounted on the workpiece platform; an L-shaped plate is mounted on the upper end of the I-shaped base; a cylinder is mounted on the vertical end of the L-shaped plate; a B-shaped bottom beam is mounted on the horizontal end; a chuck is located on both the vertical and horizontal ends of the B-shaped bottom beam; a top beam extends from the upper interface of the cylinder, and two chucks are also mounted on the top beam. The four chucks clamp the workpiece.","clamping_method":{"description":"Clamps the upper window frame of the car door, with four contact points with the window frame: a support below the window frame, a support on the outer side of the window frame, a clamping point on the upper side of the window frame, and a support on the inner side of the window frame.","clamping_head": ["BHAFD6-1B18CQ01L_1", "BHAFD6-1B18CQ02L_1", "BHAFD6-1B18CQ03L_1", "BHAFD6-1B18CQ04L_1"]}, "parts": {"BHAFD6-1B18CQ00L-PUR_1": {"description":"Cylinder assembly, standard parts, upper interface for mounting the top beam"}, "BHAFD6-1B18CQ00L-STD_1": {"description":"I-shaped base and L-shaped plate assembly"}, "parts": {"ZCSLB080_001-L BLOCK 080_2": {"description":"L-shaped plate"}, "Part1.1": {"description":"I-shaped base"}}}},"BHAFD6-1B18CQ01L_1": {"description":"Clamp 1, installed on the horizontal section of the top beam"},"BHAFD6-1B18CQ02L_1": {"description":"Clamp 2, installed on the vertical section of the top beam"},"BHAFD6-1B18CQ03L_1": {"description":"Clamp 3, installed on the horizontal section of the bottom beam"},"BHAFD6-1B18CQ04L_1": {"description":"Clamp 4, installed on the vertical section of the bottom beam"},"BHAFD6-1B18CQ05L_1":{"description":"Top beam, extending from the top interface of the cylinder, first extends to a horizontal section, clamp 1 is installed on the horizontal section, the end of the horizontal section is a downward vertical section, clamp 2 is installed"},"BHAFD6-1B18CQ06L_1": {"description":"The bottom beam is installed on the L-shaped plate, with clamp 3 installed on the horizontal section and clamp 4 installed on the vertical section."}}

[0032] In the above scheme, a CAD script interface is used to extract key semantics such as spatial relationships, functional semantics, and dependencies of the initial welding fixture assembly. These spatial relationship parameters, functional semantic parameters, and dependency parameters are then converted into a pre-defined structured text format using a pre-defined parameter structured standard. This yields the initial welding fixture assembly parameters, enabling the subsequent large language model to parse the corresponding key semantic information, obtain the fixture's structure and design intent, and provide accurate and rich input information for subsequent intelligent reasoning and parameter optimization by the large language model. Therefore, converting complex semantic information into a format that the large language model can parse helps improve the reliability of welding fixture design.

[0033] A preferred approach involves, based on pre-acquired welding fixture design requirements, inputting these requirements into a preset large language model to match a corresponding parametric template, thereby obtaining a matching parametric template. This includes: converting pre-acquired geometric feature information, welding process information, and fixture installation environment information into a target text format to obtain the welding fixture design requirements; inputting these requirements into a preset large language model and matching a parametric template that meets the preset design requirements to obtain the matching parametric template.

[0034] In one preferred implementation, a structured JSON file and user-inputted design requirements are input into a preset large language model. In this embodiment, the preset large language model is represented as an LLM-Agent (Large Language Model Agent). The design requirements include welding workpiece information, target clamping point coordinates, and process constraints. Based on these design requirements, the LLM-Agent can be used to match the corresponding parameterized template to obtain the matching parameterized template. Specifically, taking a car door as an example, based on the pre-acquired geometric feature information, welding process information, and fixture installation environment information, the geometric feature information, welding process information, and fixture installation environment information are converted into target text format to obtain the welding fixture design requirements. Among them, the geometric feature information includes outer plates, inner plates, and reinforcing ribs, etc.; the welding process information includes weld point coordinates, welding torch reachability, installation posture, and clearance constraints, etc.; and the fixture installation environment information includes tooling position, process reference surface, and operating space limitations, etc. The target text format can be natural language or structured format. Then, the welding fixture design requirements are input into the LLM-Agent, and a parametric template that meets the preset design requirements is matched from the parametric template library to obtain the matched parametric template. The preset design requirements are adjustable in real time according to the working conditions. For example, when the weld point distribution is located in the upper edge area of ​​the workpiece, the model preferentially selects the top-pressure template group and calls the corresponding parameter set to complete the design initialization.

[0035] In the above scheme, complex multi-dimensional information such as geometric feature information, welding process information, and fixture installation environment information are converted into a unified target text format, so that the subsequent large language model can parse the corresponding requirement information. Then, the welding fixture design requirements are input into the preset large language model. By utilizing the powerful semantic understanding and matching capabilities of the large language model, the initial design scheme that best fits the current working conditions is automatically selected from the template library. This avoids initial design deviations caused by improper subjective selection and helps to improve the reliability of welding fixture design.

[0036] A preferred embodiment involves inputting initial welding fixture assembly parameters and a matching parameterized template into a preset large language model, parsing the initial welding fixture assembly parameters through the preset large language model, and obtaining parameter modification instructions based on the matching parameterized template. This includes: inputting the initial welding fixture assembly parameters and the matching parameterized template into the preset large language model; parsing the functional relationships of the initial welding fixture assembly through the preset large language model to obtain a parameter dependency graph; generating parameter modification suggestions based on the matching parameterized template and the parameter dependency graph; and converting the parameter modification suggestions into an instruction script format to obtain parameter modification instructions.

[0037] In one preferred implementation, the LLM-Agent parses the component functions and spatial relationships in a JSON file, combines design requirements with corresponding matching parametric templates, and generates parameter modification instructions that can drive CAD software. Specifically, the LLM-Agent parses the functional relationships of the initial welding fixture assembly; for example, the locating pin aligns with the workpiece hole, and the pressure head corresponds to the force direction. Based on the functional relationships of the initial welding fixture assembly, a parameter dependency graph is constructed to identify the key parameters that play a dominant role in the position and orientation of the clamping point. Then, parameter suggestions are given based on the matching parametric template, such as increasing the height of LocPin1 by 5mm or shortening the chuck arm length by 10mm. The LLM-Agent also generates parameter modification instructions in a corresponding instruction script format based on the parameter modification suggestions. Generally, the parameter modification instructions can be a sequence of CAD API calls to drive the subsequent modeling process.

[0038] In the above scheme, the initial welding fixture assembly parameters and the matching parameterized template are input into a preset large language model. The preset large language model parses the structured parameters to construct a parameter dependency graph reflecting the functional relationships between components. Then, it combines the parameterized template with the parameters to generate corresponding parameter modification suggestions. Finally, the parameter modification suggestions are converted into an executable instruction script format to obtain parameter modification instructions. Thus, by parsing the initial welding fixture assembly to obtain the parameter dependency graph and analyzing the differences between the parameter dependency graph and the matching parameterized template to generate parameter modification suggestions, the parameter modification thinking is closer to human reasoning but is not affected by subjective factors, which helps to improve the reliability of welding fixture design.

[0039] A preferred embodiment involves adjusting the initial welding fixture assembly based on parameter modification instructions until the parameters of the initial welding fixture assembly meet preset optimization requirements, thereby obtaining the target welding fixture assembly parameters. This includes: adjusting the initial welding fixture assembly based on parameter modification instructions to obtain an optimized welding fixture assembly; sequentially performing interference and stability tests on the optimized welding fixture assembly to obtain collision reports and stability reports; inputting the collision and stability reports into a preset large model, evaluating the parameters of the initial welding fixture assembly based on a pre-built loss function, and generating evaluation feedback results; if the evaluation feedback results do not meet the preset optimization requirements, adjusting the initial welding fixture assembly until the evaluation feedback results meet the preset optimization requirements, thereby obtaining the target welding fixture assembly parameters. In the steps of inputting collision reports and stability reports into a preset large model, evaluating the parameters of the initial welding fixture assembly based on a pre-built loss function, and generating evaluation feedback results, the pre-construction process of the loss function includes: obtaining initial points based on the initial welding fixture assembly; obtaining target points based on a matching parameterized template; calculating the distance error between the initial points and the target points to obtain the clamping point error term; and setting corresponding weight coefficients for the preset collision penalty term, the preset clamping stability index, and the clamping point error term to obtain the loss function.

[0040] One preferred implementation involves adjusting the initial welding fixture assembly according to the parameter modification instruction, performing interference detection on the adjusted initial welding fixture assembly to generate a collision report, calculating the error between the clamping point and the target clamping point of the adjusted initial welding fixture assembly, evaluating its clamping stability, and generating an evaluation result. Then, a multi-objective loss function is constructed by combining the collision penalty term, the stability index term, and the clamping point error term. The collision report and evaluation result are fed back to the LLM-Agent. The LLM-Agent determines whether the current design meets the convergence condition based on the constructed multi-objective loss function and historical optimization records. If not, a new round of parameter adjustment suggestions is generated, and the process returns to step 104 for iteration until the convergence condition is met or the iteration limit is reached, thus obtaining the target welding fixture assembly parameters. Specifically, upon receiving a parameter modification instruction, the initial welding fixture assembly is adjusted, and a template is run to generate new 3D geometry and update the assembly. Each time an optimized initial welding fixture assembly is obtained, interference detection is performed. If a collision is detected, the system feeds back the list of conflicting components and bounding box (BBox) information to the LLM-Agent, automatically adjusting relevant parameters and reconstructing the model. Generally, interference detection can be based on the CAD's own detection or by calling a physics simulation engine such as PyBullet. Then, stability indices such as the error between the clamping point and the target point, force closure determination, or simple mechanics are calculated. Typically, stability indices also include the condition number of the force closure matrix or the consistency index of the clamping force direction. Finally, collision reports and stability reports are fed back to the LLM-Agent, which uses a pre-built loss function to determine whether the current design meets the requirements. In this embodiment, the loss function is characterized as follows: ; In the formula, clamping point With the target point The distance error, i.e., the clamping point error term, For collision penalties, For clamping stability indicators, , , The weighting coefficients are also proposed in this embodiment of the invention. A parameter search strategy is proposed to gradually adjust key parameters under safety boundary constraints; when two consecutive rounds of improvement are insufficient, an alternative strategy is switched or the solution is rolled back to the nearest optimal solution; if necessary, human-machine collaboration is triggered for manual intervention to accelerate convergence.

[0041] LLM-Agent evaluates whether the design has been optimized based on feedback. If it fails to meet the requirements, it returns to step 104 to proceed to the next round of parameter reasoning. When the number of collisions is 0 and the minimum safety gap is reached, the iteration stops.

[0042] It is worth mentioning that in this embodiment, when the LLM-Agent fails to converge within a specified number of iterations or when the collision situation is complex, such as multiple interferences and conflicting adjustment directions, the human-machine collaboration mode is triggered. Specifically, the current geometry snapshot, parameter change history, collision report, and LLM suggestions are displayed to the engineer through a visual interface. The engineer can directly modify parameters or confirm the LLM suggestions in the interface. After confirmation, the system will continue to iterate or write the suggestions into the knowledge base as manual rules. The engineer's confirmed design corrections and manufacturing feedback are structured and added to the knowledge base. The engineer's confirmed design corrections and manufacturing feedback include assembly difficulties, fixture life, and on-site modifications. This data can be used to expand the parameterized template set and update the rule engine, such as updating the upper and lower bounds of parameters. The LLM-Agent can be further trained or fine-tuned through fine-tuning or a small number of examples.

[0043] In the above scheme, after each adjustment of the initial welding fixture assembly, interference tests and stability tests are sequentially performed on the optimized welding fixture assembly. The physical verification results, such as collision reports and stability reports, are fed back to the optimization loop. A loss function is constructed for quantitative evaluation. The construction of the loss function integrates the clamping point error term, the collision penalty term, and the clamping stability index, and assigns adjustable weight coefficients to them. This transforms the complex engineering quality judgment into a calculable scalar value. The construction of the loss function enables the large language model to clearly and efficiently evaluate the comprehensive score of each design iteration and guide the parameter adjustment strategy based on the direction of the loss value change. Finally, the initial welding fixture assembly is continuously iterated and optimized until the evaluation feedback results meet the preset optimization requirements, and the parameters of the target welding fixture assembly are obtained. This allows for the subsequent generation of a reliable target welding fixture assembly to complete the welding fixture design, effectively improving the reliability of the welding fixture design.

[0044] Example 2 See Figure 5 , Figure 5 This is a schematic diagram of the module structure of a welding fixture design system based on a large language model, provided in one embodiment of the present invention. Figure 5As shown, this embodiment of the invention proposes a welding fixture design system based on a large language model, including: This embodiment of the invention also provides a welding fixture design system based on a large language model, including: a parameterized template construction module 201, used to cluster and group several historical welding fixture assemblies based on pre-acquired several historical welding fixture assemblies to obtain several welding fixture groups and establish parameterized templates based on several welding fixture groups; a parameter acquisition module 202, used to traverse pre-acquired initial welding fixture assemblies to obtain initial welding fixture assembly parameters; and a matching parameterized template acquisition module 203, used to input welding fixture design requirements based on pre-acquired welding fixture design requirements into a pre-acquired template. A parameterized template is matched to the large language model to obtain the matching parameterized template; the parameter modification instruction generation module 204 is used to input the initial welding fixture assembly parameters and the matching parameterized template into the preset large language model, parse the initial welding fixture assembly parameters through the preset large language model, and obtain parameter modification instructions based on the matching parameterized template; the parameter optimization module 205 is used to adjust the initial welding fixture assembly based on the parameter modification instructions until the initial welding fixture assembly parameters meet the preset optimization requirements to obtain the target welding fixture assembly parameters; the welding fixture design module 206 is used to generate the target welding fixture assembly based on the target welding fixture assembly parameters to complete the welding fixture design.

[0045] This invention proposes a welding fixture design system based on a large language model. It constructs parameterized templates to solidify varying design experience, extracts initial welding fixture assembly parameters understandable by the pre-defined large language model from the initial welding fixture assembly, and matches a parameterized template based on the welding fixture design requirements. Then, it inputs the initial welding fixture assembly parameters and the matching parameterized template into the pre-defined large language model to replace manual reasoning, generating parameter modification instructions to adjust the initial welding fixture assembly. An iterative optimization method is used to obtain the target welding fixture assembly parameters and generate the target welding fixture assembly, ultimately completing the welding fixture design. Therefore, by constructing a parameterized template library and extracting structured parameters from the assembly, using a pre-defined large language model for semantic parsing and parameter reasoning, and then optimizing and adjusting the welding fixture assembly parameters to generate a welding fixture design that meets engineering requirements, the reliability of the welding fixture design is improved.

[0046] Furthermore, the parameterized template construction module 201 is used to cluster and group several historical welding fixture assemblies based on several pre-acquired historical welding fixture assemblies to obtain several welding fixture groups and to establish parameterized templates based on several welding fixture groups. This includes: a historical assembly acquisition unit 301, used to acquire several historical welding fixture assemblies from a preset historical design library; a historical semantic information acquisition unit 302, used to extract historical semantic information corresponding to several historical welding fixture assemblies; a welding fixture group construction unit 303, used to perform semantic clustering on the historical semantic information corresponding to each historical welding fixture assembly according to preset clustering rules, and to group historical welding fixture assemblies with historical semantic information belonging to the same category as a corresponding welding fixture group, thus obtaining several welding fixture groups; and a template construction unit 304, used to select several target parameters that satisfy preset parameter dimensions in each welding fixture group, and to set constraint relationships for each target parameter, thus obtaining the parameterized template corresponding to each welding fixture group.

[0047] Furthermore, the parameter acquisition module 202 is used to traverse the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters, including: a parameter extraction unit 401, used to traverse the pre-acquired initial welding fixture assembly through a preset CAD script interface to extract the spatial relationship parameters, functional semantic parameters, and dependency relationship parameters corresponding to each initial welding fixture assembly; and a text format conversion unit 402, used to convert the spatial relationship parameters, functional semantic parameters, and dependency relationship parameters into a preset structured text format based on a preset parameter structuring standard to obtain the initial welding fixture assembly parameters.

[0048] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

[0049] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the described specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0050] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

Claims

1. A welding fixture design method based on a large language model, characterized in that, include: Based on several pre-acquired historical welding fixture assemblies, the several historical welding fixture assemblies are clustered and grouped to obtain several welding fixture groups, and a parameterized template is established based on the several welding fixture groups. Traverse the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters; Based on the pre-acquired welding fixture design requirements, the welding fixture design requirements are input into a preset large language model to match the corresponding parameterized template, and the matching parameterized template is obtained. The initial welding fixture assembly parameters and the matching parameterized template are input into the preset large language model. The initial welding fixture assembly parameters are parsed by the preset large language model, and parameter modification instructions are obtained based on the matching parameterized template. Based on the parameter modification instructions, the initial welding fixture assembly is adjusted until the parameters of the initial welding fixture assembly meet the preset optimization requirements, thereby obtaining the target welding fixture assembly parameters; The target welding fixture assembly is generated based on the parameters of the target welding fixture assembly to complete the welding fixture design.

2. The welding fixture design method based on a large language model as described in claim 1, characterized in that, The process involves clustering and grouping several historical welding fixture assemblies based on pre-acquired data to obtain several welding fixture groups, and establishing parameterized templates based on these welding fixture groups, including: Obtain several historical welding fixture assemblies from the preset historical design library; Extract historical semantic information corresponding to several of the historical welding fixture assemblies; According to the preset clustering rules, the historical semantic information corresponding to each historical welding fixture assembly is semantically clustered, and each historical welding fixture assembly whose historical semantic information belongs to the same category is taken as a corresponding welding fixture group, thus obtaining several welding fixture groups. In each welding fixture group, several target parameters that satisfy the preset parameter dimensions are selected, and constraint relationships are set for each target parameter to obtain the parameterized template corresponding to each welding fixture group.

3. The welding fixture design method based on a large language model as described in claim 1, characterized in that, The process of traversing the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters includes: By traversing the pre-acquired initial welding fixture assemblies through the preset CAD script interface, the spatial relationship parameters, functional semantic parameters and dependency relationship parameters corresponding to each initial welding fixture assembly are extracted. Based on a preset parameter structured standard, the spatial relationship parameters, the functional semantic parameters, and the dependency relationship parameters are converted into a preset structured text format to obtain the initial welding fixture assembly parameters.

4. The welding fixture design method based on a large language model as described in claim 3, characterized in that, Based on the pre-acquired welding fixture design requirements, and by inputting these requirements into a preset large language model to match the corresponding parameterized template, a matching parameterized template is obtained, including: Based on the pre-acquired geometric feature information, welding process information, and fixture installation environment information, the geometric feature information, welding process information, and fixture installation environment information are converted into target text format to obtain the welding fixture design requirements; The design requirements of the welding fixture are input into a preset large language model, and the parameterized template that meets the preset design requirements is matched to obtain the matching parameterized template.

5. The welding fixture design method based on a large language model as described in claim 4, characterized in that, The initial welding fixture assembly parameters and the matching parameterized template are input into the preset large language model. The preset large language model parses the initial welding fixture assembly parameters and obtains parameter modification instructions based on the matching parameterized template, including: The initial welding fixture assembly parameters and the matching parameterized template are input into a preset large language model. The functional relationships of the initial welding fixture assembly are analyzed through the preset large language model to obtain a parameter dependency graph. Based on the matching parameterization template and the parameter dependency graph, parameter modification suggestions are generated; The parameter modification suggestions are converted into an instruction script format to obtain parameter modification instructions.

6. The welding fixture design method based on a large language model as described in claim 1, characterized in that, Based on the parameter modification instructions, the initial welding fixture assembly is adjusted until the parameters of the initial welding fixture assembly meet the preset optimization requirements, thereby obtaining the target welding fixture assembly parameters, including: The initial welding fixture assembly is adjusted based on the parameter modification instructions to obtain an optimized welding fixture assembly; Interference tests and stability tests were performed on the optimized welding fixture assembly in sequence to obtain collision reports and stability reports; The collision report and the stability report are input into a preset large model, and the parameters of the initial welding fixture assembly are evaluated based on a pre-built loss function to generate evaluation feedback results. If the evaluation feedback result does not meet the preset optimization requirements, the initial welding fixture assembly is adjusted until the evaluation feedback result meets the preset optimization requirements, and the target welding fixture assembly parameters are obtained.

7. The welding fixture design method based on a large language model as described in claim 6, characterized in that, In the step of inputting the collision report and the stability report into a preset large model, evaluating the parameters of the initial welding fixture assembly based on a pre-built loss function, and generating evaluation feedback results, the pre-building process of the loss function includes: The initial position is obtained based on the initial welding fixture assembly; The target location is obtained based on the matching parameterized template; Calculate the distance error between the initial point and the target point to obtain the clamping point error term; By setting corresponding weight coefficients for the preset collision penalty term, the preset clamping stability index, and the clamping point error term, a loss function is obtained.

8. A welding fixture design system based on a large language model, characterized in that, Performing the welding fixture design method based on a large language model as described in any one of claims 1-7, comprising: The module includes a parameterized template construction module, a parameter acquisition module, a matching parameterized template acquisition module, a parameter modification instruction generation module, a parameter optimization module, and a welding fixture design module. The parameterized template construction module is used to cluster and group several historical welding fixture assemblies based on several pre-acquired historical welding fixture assemblies to obtain several welding fixture groups and to establish a parameterized template based on several welding fixture groups. The parameter acquisition module is used to traverse the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters. The matching parameterized template acquisition module is used to obtain the matching parameterized template by inputting the pre-acquired welding fixture design requirements into a preset large language model and matching the corresponding parameterized template. The parameter modification instruction generation module is used to input the initial welding fixture assembly parameters and the matching parameterized template into the preset large language model, parse the initial welding fixture assembly parameters through the preset large language model, and obtain parameter modification instructions based on the matching parameterized template. The parameter optimization module is used to adjust the initial welding fixture assembly based on the parameter modification instruction until the parameters of the initial welding fixture assembly meet the preset optimization requirements, thereby obtaining the target welding fixture assembly parameters. The welding fixture design module is used to generate a target welding fixture assembly based on the parameters of the target welding fixture assembly, so as to complete the welding fixture design.

9. The welding fixture design system based on a large language model as described in claim 8, characterized in that, The parameterized template construction module is used to cluster and group several historical welding fixture assemblies based on pre-acquired data to obtain several welding fixture groups, and to establish parameterized templates based on these welding fixture groups, including: Historical assembly acquisition unit, historical semantic information acquisition unit, welding fixture assembly construction unit, and template construction unit; The historical assembly acquisition unit is used to acquire several historical welding fixture assemblies from a preset historical design library; The historical semantic information acquisition unit is used to extract historical semantic information corresponding to several of the historical welding fixture assemblies. The welding fixture group construction unit is used to perform semantic clustering on the historical semantic information corresponding to each historical welding fixture assembly according to a preset clustering rule, and to take each historical welding fixture assembly with the same historical semantic information as a corresponding welding fixture group, thereby obtaining several welding fixture groups. The template construction unit is used to select several target parameters that satisfy the preset parameter dimensions in each welding fixture group, and set constraint relationships for each target parameter to obtain the parameterized template corresponding to each welding fixture group.

10. The welding fixture design system based on a large language model as described in claim 8, characterized in that, The parameter acquisition module is used to traverse the pre-acquired initial welding fixture assembly to obtain the initial welding fixture assembly parameters, including: Parameter extraction unit and text format conversion unit; The parameter extraction unit is used to traverse the pre-acquired initial welding fixture assembly through a preset CAD script interface and extract the spatial relationship parameters, functional semantic parameters and dependency relationship parameters corresponding to each initial welding fixture assembly. The text format conversion unit is used to convert the spatial relationship parameters, the functional semantic parameters, and the dependency relationship parameters into a preset structured text format based on a preset parameter structured standard, so as to obtain the initial welding fixture assembly parameters.

Images