Bolt connection mechanical model automatic generation method, device and equipment

Abstract

The invention provides a bolt connection mechanical model automatic generation method, device and equipment, and relates to the technical field of mechanical design, the generation method comprises the following steps: constructing a mechanical generation model, the mechanical generation model is constructed based on an autoregression language model; obtaining a description text of a user, and converting the description text to obtain structured data; wherein the description text comprises functional requirements and critical dimension parameters of the mechanical model; based on the structured data, performing a model generation process through a mechanical generation model to obtain three-dimensional mechanical model data of bolt connection; wherein the model generation process comprises a physical perception rollback operation. According to the method, the mechanical generation model based on the autoregressive language model is constructed, and the analysis of the user description text containing the requirements of mechanical model function requirements, key size parameters and the like is combined, so that the automatic generation of the bolt connection three-dimensional mechanical model is realized, and a user does not need to have professional 3D modeling skills; and the technical threshold of mechanical model design is greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of mechanical design technology, and more specifically, to a method, apparatus, and equipment for automatically generating bolted connection mechanical models. Background Technology

[0002] With the deep integration of computer 3D modeling technology and the field of intelligent manufacturing, the demand for rapidly generating practically applicable bolted connection mechanical models in industrial design, machining, and other scenarios is becoming increasingly urgent. Currently, mechanical model design largely relies on professional designers manually drawing using tools such as CAD. This not only requires operators to have solid 3D modeling skills, mechanical structure knowledge, and bolted connection design experience, but also requires a significant amount of time to adjust part dimensions, connection methods, and physical stability step by step. This results in high design thresholds and long cycles, making it difficult to meet the rapid design needs of small and medium-sized manufacturing enterprises or non-professional designers.

[0003] Meanwhile, while Large Language Models (LLMs) have made significant progress in the fields of natural language understanding and content generation, they have obvious limitations in mechanical model generation scenarios. On the one hand, existing LLM-based generation technologies mostly focus on unstructured content such as text and images, making it difficult to accurately convert users' textual descriptions (such as functional requirements and dimensional parameters) into three-dimensional mechanical models that include bolted connections. On the other hand, even if some technologies can generate simple geometric models, they lack consideration for the physical stability of mechanical models. The generated models often have problems such as unreasonable bolted connections and the inability of the overall structure to achieve static equilibrium, making it impossible to achieve actual assembly and use and difficult to adapt to the engineering requirements of mechanical design. Summary of the Invention

[0004] The problem solved by this invention is one or more of the aforementioned related technical problems.

[0005] To address the above problems, this invention provides a method, apparatus, and equipment for automatically generating mechanical models of bolted connections.

[0006] In a first aspect, the present invention provides a method for automatically generating a mechanical model of a bolted connection, comprising: A mechanical generation model is constructed, which is based on an autoregressive language model. Obtain the descriptive text and transform it into structured data; wherein, the descriptive text includes the functional requirements and key dimensional parameters of the mechanical model; Based on the structured data, a model generation process is performed using the mechanical generation model to obtain three-dimensional mechanical model data of the bolt connection; wherein, the model generation process includes a physical sensing rollback operation.

[0007] Optionally, the process of generating a model based on the structured data and using the mechanical generation model to obtain the corresponding three-dimensional mechanical model data includes: A preliminary perception model is obtained based on the structured data; The preliminary perception model is segmented step by step, and the segmentation module obtained at each step is subject to constraint mechanism judgment to obtain the judgment result. The three-dimensional mechanical model data is obtained based on the judgment result.

[0008] Optionally, the step of progressively segmenting the initial perception model and performing constraint mechanism judgment on each segmentation module obtained from the segmentation process to obtain a judgment result, and obtaining the three-dimensional mechanical model data based on the judgment result, includes: Based on the geometric structure of the preliminary perception model, segmentation processing is performed to obtain the current segmentation module and the remaining segmentation modules. The current segmentation module corresponds to one or a group of associated mechanical parts. The constraint mechanism is used to determine the current segmentation module; If the current segmentation module fails to pass any constraint mechanism verification, a subsequent physical perception rollback operation is triggered. If the current segmentation module passes all constraint mechanisms, the remaining segmentation modules are segmented to obtain a temporary segmentation model. The temporary segmentation model is then used as the current segmentation module to return to the execution constraint mechanism for judgment, until the judgment of the segmentation module is obtained after each step of the segmentation process is completed. The three-dimensional mechanical model data is determined based on all the segmentation modules that have passed all constraint mechanisms.

[0009] Optionally, the constraint mechanism includes part routine verification, part geometric stability verification, part machining stability verification, connection rationality verification, and physical stability verification.

[0010] Optionally, the specific process of the physical sense rollback operation includes: During model generation, if the constraint mechanism verification result of the current segmentation module fails, rollback to the state of the previous segmentation module that passed all constraint verifications. Based on the adjusted segmentation logic, the segmentation process and constraint mechanism judgment are re-executed until the newly generated current segmentation module passes all verifications, and then the subsequent model generation process continues.

[0011] Optionally, the method for automatically generating the mechanical model of the bolt connection further includes: During the model generation process, if the similarity coefficient between the assembly of multiple segmentation modules connected by bolts and a single part in the preset mechanical model library is greater than or equal to the corresponding preset threshold; The corresponding multiple segmentation modules are merged, and the merged module is subjected to a constraint mechanism for judgment.

[0012] Optionally, the output format of the three-dimensional mechanical model data includes at least one of rendered images, 3D visualization files, 2D projected engineering drawings, and 2D sectional engineering drawings.

[0013] Secondly, the present invention provides an automatic generation device for bolt connection mechanical models, comprising: A construction unit is used to construct a mechanical generative model, which is based on an autoregressive language model. An acquisition unit is used to acquire the user's description text and convert the description text into structured data; wherein, the description text includes the functional requirements and key dimensional parameters of the mechanical model; The generation unit is used to perform a model generation process based on the structured data and the mechanical generation model to obtain three-dimensional mechanical model data of the bolt connection; wherein the model generation process includes a physical sensing rollback operation.

[0014] Thirdly, the present invention provides an automatic bolt connection mechanical model generation device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the automatic bolt connection mechanical model generation method as described in the first aspect.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the automatic generation method for bolt connection mechanical models as described in the first aspect.

[0016] The beneficial effects of the automatic generation method, apparatus, and equipment for bolt connection mechanical models of the present invention are: By constructing a mechanical generation model based on an autoregressive language model and combining it with the parsing of user-described text containing functional requirements and key dimensional parameters of the mechanical model, the automated generation of 3D mechanical models for bolted connections is achieved. This eliminates the need for users to possess professional 3D modeling skills, significantly lowering the technical threshold for mechanical model design and allowing non-professionals to easily participate in the design process, effectively expanding the application scenarios of the technology. Simultaneously, the physical awareness rollback operation integrated into the model generation process can promptly correct unstable or unreasonable design steps during generation, ensuring that the final output 3D mechanical model conforms to physical laws and practical application requirements. This avoids the problems of difficult model assembly and poor physical stability in traditional 3D generation methods, guaranteeing the model's practicality and reliability. Furthermore, the fully automated process from text input to model output eliminates the tedious steps of manually drawing and adjusting the model step by step, significantly shortening the design cycle and greatly improving the design efficiency of bolted connection mechanical models. It is particularly suitable for scenarios with high efficiency requirements, such as rapid design and iterative solutions, providing efficient, convenient, and reliable technical support for mechanical design work. Attached Figure Description

[0017] Figure 1 This is one of the flowcharts illustrating an automatic generation method for a bolt connection mechanical model according to an embodiment of the present invention; Figure 2 This is a second schematic flowchart of an automatic generation method for a bolt connection mechanical model according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an automatic generation device for bolt connection mechanical models according to an embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0019] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0020] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0021] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0022] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0023] While traditional 3D model generation methods can produce diverse geometries, they are generally not optimized for the core mechanical design scenario of bolted connections. They cannot efficiently respond to connection requirements in user text descriptions, nor do they establish effective design error correction mechanisms. If the generated model has physical instability or connection defects, it requires manual disassembly and adjustment, further extending the design cycle. This fails to meet the technical demands of modern manufacturing for efficient, reliable, and low-barrier mechanical model design. Against this backdrop, there is an urgent need for a technical solution that can automatically generate physically stable and practically applicable bolted connection mechanical models based on user text descriptions, in order to overcome the bottlenecks of existing technologies.

[0024] To address the problems existing in the aforementioned related technologies, embodiments of the present invention provide a method, apparatus, and equipment for automatically generating mechanical models of bolted connections.

[0025] like Figure 1 As shown in the figure, an embodiment of the present invention provides a method for automatically generating a bolt connection mechanical model, comprising: Step S100: Construct a mechanical generation model, which is based on an autoregressive language model.

[0026] Specifically, a mechanical generative model based on an autoregressive language model is constructed. An autoregressive language model is a sequence prediction model commonly used in natural language processing, but in this application it is specifically adapted for the field of mechanical design. The model learns from a large amount of historical data on bolted mechanical models, including 3D geometric information, functional features, and design rules. The model employs a Transformer architecture or other autoregressive network (LLM) to progressively output the data structure of the 3D mechanical model (such as point clouds, meshes, or parametric features) based on the input sequence. During training, the model learns how to map textual descriptions to mechanical components and ensures that the generated design conforms to engineering standards and constraints.

[0027] A mechanical generative model based on an autoregressive language model can automatically learn and generate high-quality bolt connection models from rich design data. This approach significantly improves the accuracy and consistency of model generation, reducing errors and repetitive work in manual design. The sequential generation characteristic of the autoregressive model allows it to handle the step-by-step construction of complex mechanical structures, ensuring that each component is generated rationally based on its context, thereby improving design efficiency and reducing development costs. Furthermore, the model's scalability enables it to adapt to various mechanical design scenarios, providing a solid foundation for automated design.

[0028] Step S200: Obtain the description text and convert it into structured data; the description text includes the functional requirements and key dimensional parameters of the mechanical model.

[0029] Specifically, descriptive text is obtained from the user end. This text includes the functional requirements of the mechanical model (a mechanical model is a virtual model of a mechanical device (or mechanical component) containing bolted connections. Its core is a mechanical structure that achieves a specific function through the combination of parts and bolted connections, such as "shelf with horizontal partitions" or "retractable camera tripod". Such models contain multiple independent mechanical parts (such as the uprights and partitions of the shelf, the legs and connectors of the tripod), as well as the bolted connection relationships between the parts. They also have clear functional attributes (such as carrying items and supporting equipment) and are digital mechanical design results that can be used for subsequent processing, assembly or simulation), key dimensional parameters (such as the height of the bookshelf, the maximum and minimum height of the tripod, etc.), bolted connection requirements and load-bearing capacity requirements, etc. Users can provide this information through a natural language interface (such as text input or speech conversion). Then, natural language processing technology is used to parse the text, extract key features and parameters, and convert it into structured data as input for the mechanical generation model.

[0030] By acquiring users' descriptive text, a user-friendly and efficient design input method is achieved, enabling non-professional users to easily express their needs and generate professional-grade mechanical models. This lowers the barrier to entry, eliminating the need for users to master complex CAD software or design knowledge, thereby broadening the application scope of automated design tools. Simultaneously, the precise input of key dimensional parameters ensures that the generated model meets practical application requirements in terms of size and function, reducing the need for later modifications. This customized generation process improves design flexibility and user satisfaction, accelerating the product development cycle.

[0031] Step S300: Based on the structured data, a model generation process is performed using the mechanical generation model to obtain three-dimensional mechanical model data of the bolt connection; wherein, the model generation process includes a physical sensing rollback operation.

[0032] Specifically, using the mechanical generation model constructed in step S100, and based on the structured data obtained in step S200, the model generation process is executed, ultimately outputting three-dimensional mechanical model data of the bolted connection (such as an STL file or a parametric model). The generation process includes a physically-aware rollback operation: after each generation step, the model checks in real time whether the generated components conform to physical constraints (such as strength, stability, manufacturability, and assemblability). If a constraint violation is detected, the model automatically rolls back to the previous generation step and adjusts the generation strategy (such as modifying the geometry or replacing components). This rollback mechanism is based on predefined physical rules or simulation feedback to ensure that the final model is physically feasible and optimized.

[0033] The model generation process, combined with physics-aware rollback operations, significantly enhances the reliability and practicality of the generated models. By verifying physical constraints in real time, it automatically avoids generating impractical or high-risk designs, thus ensuring the performance and safety of bolted connection models in real-world applications. This mechanism reduces design iterations and post-testing costs, making the generation process more efficient and robust. Furthermore, the automatic generation of 3D model data saves considerable manual design time, supports rapid prototyping and optimization, and is suitable for various engineering scenarios, such as smart manufacturing and digital factories.

[0034] In this embodiment, by constructing a mechanical generation model based on an autoregressive language model and combining it with the parsing of user-described text containing functional requirements and key dimensional parameters of the mechanical model, the automated generation of a 3D mechanical model of bolted connections is achieved. This eliminates the need for users to possess professional 3D modeling skills, significantly lowering the technical threshold for mechanical model design and allowing non-professionals to easily participate in the design process, effectively expanding the application scenarios of the technology. Simultaneously, the physical perception rollback operation integrated into the model generation process can promptly correct unstable or unreasonable design steps during generation, ensuring that the final output 3D mechanical model conforms to physical laws and practical application requirements. This avoids the problems of difficult model assembly and poor physical stability in traditional 3D generation methods, guaranteeing the practicality and reliability of the model. Furthermore, the fully automated process from text input to model output eliminates the tedious steps of manually drawing and adjusting the model step by step, significantly shortening the design cycle and greatly improving the design efficiency of bolted connection mechanical models. It is particularly suitable for scenarios with high efficiency requirements, such as rapid design and iterative solutions, providing efficient, convenient, and reliable technical support for mechanical design work.

[0035] Optionally, the process of generating a model based on the structured data and using the mechanical generation model to obtain the corresponding three-dimensional mechanical model data includes: A preliminary perception model is obtained based on the structured data; The preliminary perception model is segmented step by step, and the segmentation module obtained at each step is subject to constraint mechanism judgment to obtain the judgment result. The three-dimensional mechanical model data is obtained based on the judgment result.

[0036] Optionally, the step of progressively segmenting the initial perception model and performing constraint mechanism judgment on each segmentation module obtained from the segmentation process to obtain a judgment result, and obtaining the three-dimensional mechanical model data based on the judgment result, includes: Based on the geometric structure of the preliminary perception model, segmentation processing is performed to obtain the current segmentation module and the remaining segmentation modules. The current segmentation module corresponds to one or a group of associated mechanical parts. The constraint mechanism is used to determine the current segmentation module; If the current segmentation module fails to pass any constraint mechanism verification, a subsequent physical perception rollback operation is triggered. If the current segmentation module passes all constraint mechanisms, the remaining segmentation modules are segmented to obtain a temporary segmentation model. The temporary segmentation model is then used as the current segmentation module to return to the execution constraint mechanism for judgment, until the judgment of the segmentation module is obtained after each step of the segmentation process is completed. The three-dimensional mechanical model data is determined based on all the segmentation modules that have passed all constraint mechanisms.

[0037] Optionally, the constraint mechanism includes part routine verification, part geometric stability verification, part machining stability verification, connection rationality verification, and physical stability verification.

[0038] Specifically, firstly, the mechanical generation model (built based on an autoregressive language model) generates a preliminary perceptual model from structured data, including the overall structure, core components, and connection relationships (e.g., the user's description text includes a tabletop that can support 5kg, equipped with 3 supporting legs, each 60cm high, connected to the tabletop by bolts; the preliminary perceptual model presents the overall framework of a circular tabletop + 3 cylindrical legs + bolt connection points between the tabletop and the legs). Secondly, the preliminary perceptual model is gradually segmented into independent segmented modules, and a constraint mechanism is applied to each module. Finally, modules that meet the requirements are selected based on the judgment results and combined to form three-dimensional mechanical model data.

[0039] The step-by-step segmentation process and constraint mechanism clearly define the iterative logic of segmentation-judgment-loop: Segmentation rules: Based on the geometric structure of the initial perception model (such as the connection boundary between the desktop and the support leg), functional partitioning (load-bearing components and support components), and bolt connection logic (the combination of parts where the bolt connection points are located), the first segmentation obtains the current segmentation module (such as the circular desktop in the example above) and the remaining segmentation modules (3 cylindrical legs and bolt connection structure). The current segmentation module corresponds to one or a group of associated parts (a single desktop is a single part, and if the desktop contains reinforcing ribs, it is a group of associated parts). Judgment trigger: The constraint mechanism is only executed on the current segmentation module, and the remaining modules are not processed in advance to ensure the verification accuracy of each module; If the current module passes all constraint verifications (such as the desktop's regularity and geometric stability meeting the standards), the remaining segmentation modules are segmented again (e.g., the first cylindrical leg is segmented as the new current module), and the judgment process is repeated; if the current module fails any verification (e.g., the desktop thickness is too thin, causing physical stability to fail), the physical perception rollback operation is immediately triggered. Termination condition: After all remaining segmentation modules have been segmented and verified, combine all verified modules according to their original connection relationships to obtain the final three-dimensional mechanical model data.

[0040] The scope of the constraint mechanism's judgment is clearly defined, covering five core dimensions: part conformity verification (determining whether the module conforms to common mechanical part shapes), part geometric stability verification (determining whether the module's geometric structure is regular), part machining stability verification (determining whether the module is machinable), connection rationality verification (determining whether the bolted connections between the module and other parts are compliant), and physical stability verification (determining whether the module meets static equilibrium and load-bearing requirements). These five types of verification form a comprehensive control from "part's own properties" to "connection relationships between parts" to "overall physical performance," ensuring that each segmented module conforms to mechanical design common sense and is adaptable to actual application scenarios.

[0041] In some embodiments, the quantification methods for determining the five core dimensions by the constraint mechanism include: 1. Part Conventional Verification (Determining if the module conforms to common mechanical part shapes): By comparing the similarity between the segmentation module and the preset common parts dataset (preset mechanical model library) (including 3D models and feature parameters of common mechanical parts such as shafts, plates, and brackets), it is determined whether the module belongs to a mature and easily processed conventional part shape, avoiding the generation of irregular and difficult-to-use parts.

[0042] Its quantification methods include: Feature extraction: Extract geometric classification features (such as whether it is a plate-type part), topological features (genus, Euler feature number), and key size ratios (such as the length-width-thickness ratio of a plate-type part) from the segmentation module. Similarity calculation: The cosine similarity algorithm is used to compare the extracted features with the feature vectors of each part in the common parts dataset, and the regularity coefficient is calculated (the value range is 0-1). Judgment criteria: If the normality coefficient is greater than the first preset threshold (e.g., 0.95), the judgment is passed; if it is less than or equal to the first preset threshold, the judgment is failed.

[0043] For example: Take a round workbench (50cm in diameter, 2cm thick) as an example: Extracted features: Plate-like parts, genus 0 (no holes), Euler feature number 2, aspect ratio 25:25:1; Comparison dataset: Compare with the feature vector of a circular support plate (dataset number PL-028, diameter 40-60cm, thickness 1.5-3cm, length-width-thickness ratio 20-30:20-30:1); Calculation result: regularity coefficient 0.98 > 0.95, which passes the verification.

[0044] 2. Part geometric stability verification (determining whether the module's geometric structure is regular and not easily deformed): Starting from geometric properties such as symmetry and size ratio, determine whether there are structural weak points in the segmented module (such as slender rods or ultra-thin plates) to ensure that it is not easily broken or deformed due to its geometric shape during subsequent assembly and use.

[0045] Its quantification methods include: Core indicator calculation: Rotational symmetry score: Calculate the degree of overlap of the module after rotating it 180° / 90° around the central axis (overlap = area of ​​overlapping region / total area of ​​module), score = overlap × 100 (value range 0-100); Mirror symmetry score: The degree of overlap of the module after mirroring along the central plane is calculated. Score = degree of overlap × 100 (value range 0-100). Length-to-diameter ratio / length-width-to-height ratio: For rod-type parts, calculate the length-to-diameter ratio (length / diameter); for plate-type parts, calculate the length-width-to-thickness ratio (maximum side length / thickness). A ratio ≤ 15 gets 100 points, 15 < ratio ≤ 20 gets 80 points, and a ratio > 20 gets 50 points. Comprehensive coefficient: Geometric stability coefficient = (rotational symmetry score + mirror symmetry score + proportional score) / 300 × 1 (value range 0-1); Judgment criteria: If the geometric stability coefficient is greater than the second preset threshold (e.g., 0.8), the judgment is passed; if it is less than or equal to the second preset threshold, the judgment is failed.

[0046] For example: Take the workbench support leg (originally designed with a diameter of 2cm and a height of 60cm) as an example: Rotational symmetry score: 100% coincidence when rotated 180° around the central axis, score 100; Mirror symmetry score: 100% coincidence along the central plane, score 100; Aspect ratio: 60cm / 2cm = 30 > 20, ratio score 50; The comprehensive coefficient is approximately 0.83 (100+100+50) / 300. When the second preset threshold is greater than 0.83, it indicates that the geometric stability verification of the part has failed.

[0047] However, for simple mechanical models like workbench legs, if the geometric stability verification of parts consistently fails during subsequent re-segmentation, it can be determined that the parameters in the user's description text are unreasonable or incorrect. Therefore, the subsequent generation process may require some correction. For example, a scale > 20 might score 50 points, resulting in (100+100+50) / 300 = 0.83. However, an actual length-to-diameter ratio of 30 indicates a severely slender structure, requiring adjustment of the judgment logic: when the scale > 20, the scale score is calculated as (30 - scale) × 5 (e.g., a scale of 30 scores 0 points). After correction, the scale score is 0, and the comprehensive coefficient = (100+100+0) / 300 ≈ 0.67 < 0.8, failing verification. After adjustment, the support leg diameter is 3cm, height is 60cm, length-to-diameter ratio is 20, scale score is 80, and the comprehensive coefficient = (100+100+80) / 300 ≈ 0.93 > 0.8, passing verification.

[0048] 3. Verification of the stability of part machining (determining whether the module can be manufactured by conventional machining or specific methods): Based on geometric topology analysis, determine whether the segmented module can be achieved by conventional cutting machining such as milling and turning. If additive manufacturing such as 3D printing is required, it needs to be marked separately and meet the structural requirements of additive manufacturing (such as no overhang structure) to avoid the disconnect between design and manufacturing.

[0049] Its quantification methods include: processing feasibility analysis; routine processing inspection: determining whether there are areas that cannot be reached by the cutting tool (such as closed cavities, slender internal circular holes with a depth > 5 times the diameter). If there are no such areas, 100 points are awarded; if there is one such area, 50 points are awarded; if there are two or more such areas, 0 points are awarded. Dimensional accuracy check: Whether the deviation of key dimensions (such as bolt hole diameter) is within the normal machining tolerance range (±0.05mm), 100 points are awarded if it meets the standard, and 50 points are awarded if it does not. Special annotation adaptation: If the module is labeled "3D printing" and has no overhang structure (overhang angle ≤45° no support is needed, >45° requires reserved support position), it will directly get 100 points; Comprehensive coefficient: Machining stability coefficient = (Conventional machining score + dimensional accuracy score) / 200 × 1 (value range 0-1), the coefficient for 3D printed standard parts is directly taken as 1; Judgment criteria: If the mechanical processing stability coefficient is greater than the third preset threshold (e.g., 0.8), or if the 3D printing is marked and meets the structural requirements, the judgment is passed; otherwise, it is not passed.

[0050] For example: Take a workbench reinforcement block with an internal blind hole (10cm long, 5cm wide, 3cm high, with an internal blind hole 1cm in diameter and 4cm deep) as an example: Routine machining inspection: Blind hole depth 4cm < 5×1cm = 5cm, the tool can reach it, 100 points; Dimensional accuracy check: Blind hole diameter tolerance ±0.03mm, meets conventional machining requirements, score 100 points; Overall coefficient: (100+100) / 200=1>0.8, verified; If the blind hole depth is 6cm > 5 × 1cm = 5cm, the routine processing inspection will score 0 points, and the comprehensive coefficient will be (0 + 100) / 200 = 0.5 < 0.8, thus failing the verification; if it is marked as 3D printed and the blind hole has no overhang structure, it will score 100 points and pass the verification.

[0051] 4. Verification of connection rationality (determining whether the bolted connection between the module and other parts is compliant and not prone to failure): Based on the core parameters of the bolted connection, determine whether the connection structure can achieve reliable fastening and avoid loosening or damage to parts due to insufficient connection surface or unreasonable bolt hole spacing.

[0052] Its quantification method includes: calculation of core parameters: area of ​​connection surface: effective contact area of ​​bolt connection surface (must be ≥ minimum bearing area corresponding to the nominal diameter of the bolt, such as M8 bolt must be ≥ 20mm) 2 If the answer is yes, you get 100 points; if it is no, you get 50 points. Bolt hole spacing: The center distance between adjacent bolt holes (must be ≥2 times the bolt diameter, such as ≥16mm for M8 bolts). If it meets the requirements, 100 points will be awarded; if it does not, 50 points will be awarded. Loading and unloading space allowance: The distance between the bolt head / nut and the surrounding structure (must be ≥ bolt wrench space, such as M8 bolts must be ≥ 10mm), if it meets the requirements, 100 points are awarded, if it does not meet the requirements, 50 points are awarded; Comprehensive coefficient: Connection rationality coefficient = (Connection surface area score + Bolt hole spacing score + Loading and unloading space margin score) / 300 × 1 (Value range 0-1); Judgment criteria: If the connection rationality coefficient is greater than the fourth preset threshold (e.g., 0.8), the judgment is passed; if it is less than or equal to the fourth preset threshold (e.g., 0.8), the judgment is failed.

[0053] For example: taking the connection between the workbench tabletop and the support leg using M8 bolts as an example: Contact area between the tabletop and the leg: 30mm 2 ≥20mm 2 100 points; Bolt hole spacing: The center distance between adjacent bolt holes is ≥16mm, and 20mm is required to get 100 points; Loading and unloading space allowance: The distance between the bolt head and the edge of the tabletop is 12mm or more than 10mm, which scores 100 points; Overall coefficient: (100+100+100) / 300=1>0.8, verification passed; Similarly, during the connection rationality verification process, if the geometric stability verification of the parts fails when re-segmenting or when only this segmentation method is available, it can be determined that the parameters in the user's description text are unreasonable or incorrect. In this case, the subsequent generation process may need to make certain corrections: If the bolt hole spacing is 12mm < 16mm, the bolt hole spacing score is 50, and the comprehensive coefficient = (100 + 50 + 100) / 300 ≈ 0.83 > 0.8 (critical pass); if the spacing is 10mm, the score is 50, and the comprehensive coefficient = (100 + 50 + 100) / 300 ≈ 0.83. The correction is as follows: a spacing < 1.5 times the bolt diameter (12mm) gets 0 points, a spacing of 10mm gets 0 points, and the comprehensive coefficient = (100 + 0 + 100) / 300 ≈ 0.67 < 0.8, thus failing the verification.

[0054] 5. Physical stability verification (determining whether the module meets static equilibrium and load-bearing risk): Based on the principles of mechanics, determine whether the segmented module can maintain static equilibrium under its own weight and the load of related parts, so as to avoid overturning or structural damage caused by excessive bending moment or center of gravity shift.

[0055] Its quantification methods include: core mechanical calculations; gravity load: based on the default density of the module material (steel 7.85 g / cm³). 3 Aluminum 2.7g / cm 3 ) and calculate weight G from volume; Associated load: The load F transmitted by the associated parts (e.g., the workbench table needs to bear 5kg, i.e., F=50N). Bending moment calculation: Taking the bolt connection point as the fulcrum, calculate the total bending moment M = (G × L1 + F × L2), where L1 is the distance from the module's center of gravity to the fulcrum, and L2 is the distance from the point of application of the associated load to the fulcrum; the bending moment must be ≤ the allowable bending moment of the material (e.g., the allowable bending moment of Q235 steel M8 bolt is 10 N·m). If it meets the requirement, 100 points will be awarded; if it does not, 50 points will be awarded. Center of gravity offset: The center of gravity of the module must fall within the support surface (e.g., the center of gravity of the desktop must fall within the triangular support surface formed by the three legs). If it meets the requirements, 100 points will be awarded; otherwise, 0 points will be awarded. Comprehensive coefficient: Physical stability coefficient = (bending moment score + center of gravity offset score) / 200 × 1 (value range 0-1); Judgment criteria: If the physical stability coefficient is greater than the fifth preset threshold (0.8), the judgment is passed; if it is less than or equal to the fifth preset threshold (0.8), the judgment is not passed.

[0056] For example: Take the workbench tabletop (steel material, volume = π × 25) as an example. 2 ×2=3925cm 3 (Weight G≈30.8N) Associated load: F=50N (supporting a 5kg camera), with the point of application at the center of the table; Bending moment calculation: The support point is the bolt connection point between the tabletop and the leg (15cm from the center), M=(30.8×15 + 50×15)=1212N·mm=1.212N·m≤10N·m, bending moment score 100; Center of gravity shift: The center of gravity of the desktop is in the center and falls within the triangular support surface formed by the three legs. The score for center of gravity shift is 100. Overall coefficient: (100+100) / 200=1>0.8, verified; If the diameter of the tabletop is reduced to 30cm, the distance between the supporting legs remains unchanged, and the center of gravity shifts off the supporting surface, the score for center of gravity shift is 0, and the comprehensive coefficient = (100+0) / 200 = 0.5 < 0.8, thus failing the verification.

[0057] It should be noted that for simple mechanical models, if the constraint mechanism fails to pass verification after re-performing subsequent segmentation or after only simple segmentation, it can be determined that the parameters in the user's description text are unreasonable or incorrect. Therefore, the subsequent generation process may need to make certain corrections, that is, outputting 3D mechanical model data along with its parameter descriptions (which do not conform to the parameter descriptions in the description text, including the reasons for modification, etc.).

[0058] The construction process of the mechanical generative model includes: We selected mechanical models containing bolted connections (covering various common mechanical types such as workbenches, shelves, and supports) from industrial-grade 3D model libraries (such as GrabCAD and TraceParts) and mechanical design standard databases. Each model was ensured to consist of multiple detachable mechanical parts connected by standard bolts (such as M4-M20 bolts). Simultaneously, we collected detailed text descriptions of the corresponding models, including functional requirements (such as a fixed shelf supporting 10kg), key dimensions (such as a shelf height of 30cm), component composition (such as 4 uprights + 3 shelves), and bolt connection requirements (such as uprights and shelves connected by M8 bolts). This data formed training data for model-text pairing. The initial generative model (autoregressive language model such as GPT-4, LLaMA 3, etc.) is trained and optimized using training data, and the optimized initial generative model is used as the final mechanical generative model.

[0059] Optionally, the specific process of the physical sense rollback operation includes: During model generation, if the constraint mechanism verification result of the current segmentation module fails, rollback to the state of the previous segmentation module that passed all constraint verifications. Based on the adjusted segmentation logic, the segmentation process and constraint mechanism judgment are re-executed until the newly generated current segmentation module passes all verifications, and then the subsequent model generation process continues.

[0060] Specifically, the implementation of the physical perception rollback operation clarifies the rollback-adjustment-retry error correction process: When the current segmentation module fails constraint verification (e.g., the aspect ratio of the first cylindrical leg in the example above is too large, resulting in substandard geometric stability), the remaining modules are not further segmented. Instead, the process rolls back to the previous state where all verifications were passed (i.e., the circular tabletop has passed verification). Then, the segmentation logic is adjusted based on the rollback state. This adjustment includes optimizing the segmentation boundaries (e.g., dividing the original single slender leg into two short legs plus an intermediate connector) and adjusting the part dimensions (e.g., increasing the leg diameter from 2cm to 3cm). After adjustment, the segmentation process (creating new short-leg modules) and constraint checks are re-executed until the new module passes verification, at which point the subsequent segmentation process continues. This step solves the problem of timely error correction during the generation process, avoiding the accumulation of invalid designs.

[0061] In some specific embodiments, taking "generating a small workbench capable of supporting 5kg (tabletop diameter 50cm, 3 support legs 60cm high, legs connected to the tabletop by bolts)" as an example, the process of generating the three-dimensional mechanical model data includes: After the user input description text is converted into structured data, a preliminary perception model is obtained by analyzing the mechanical generation model, which includes a circular desktop, three cylindrical legs, and four bolt connections between the desktop and the legs. It is determined that three-dimensional data needs to be generated through stepwise segmentation and constraint judgment. Segmentation: Based on geometric boundaries, the initial perception model is segmented into the current segmentation module "circular desktop (2cm thick)" and the remaining segmentation module "3 cylindrical legs (2cm in diameter, 60cm in height) + bolt connection structure"; Judgment: The five constraint verifications performed on the "circular tabletop" (routineness: 98% similarity to commonly used circular tabletops, coefficient > 0.95; geometric stability: high symmetry score, reasonable aspect ratio, coefficient > 0.8; machining stability: can be machined, coefficient > 0.8; connection rationality: the reserved bolt hole positions are compliant, coefficient > 0.8; physical stability: 2cm thickness can bear 5kg, coefficient > 0.8) all passed the verification; Loop: Divide the remaining modules into the current module "first cylindrical leg (diameter 2cm, height 60cm)" and proceed to the next round of judgment; Constraint verification was performed on the "first cylindrical leg". In the geometric stability verification, the aspect ratio (60cm / 2cm=30) was too large, and the coefficient = 0.6 < 0.8, so the verification failed. The system was rolled back to the state of "circular tabletop has passed verification". The segmentation logic was adjusted to "segment the original 60cm long leg into two 30cm short legs + 1 connecting flange", and the current module "first short leg (diameter 2.5cm, height 30cm)" was obtained by re-segmenting. The verification was repeated: the aspect ratio (30cm / 2.5cm=12) of the newly segmented "first short leg" was reasonable, and the geometric stability coefficient was 0.85, thus passing all verifications; the subsequent segments "connecting flange", "second short leg", "other two legs" and "bolted connection structure" were also verified, and then combined to obtain the complete three-dimensional mechanical model data of the workbench.

[0062] By clearly defining the segmentation-judgment-rollback process, two core benefits are achieved: First, it significantly improves the accuracy and reliability of 3D mechanical model data generation. Through the iterative logic of "segmenting and verifying one by one," it avoids the shortcomings of traditional batch generation, such as "difficulty in locating problems after overall errors." Combined with five types of full-dimensional constraint verification and timely physical perception rollback, it ensures that each segmentation module meets the engineering requirements of mechanical design. The final generated model does not have problems such as unreasonable bolt connections, unprocessable parts, or physical instability, and can be directly used for subsequent assembly and manufacturing. Second, it significantly reduces the cost of manual intervention in the model generation process. The entire process, from segmentation rules and verification standards to rollback adjustments, is executed automatically by the system. There is no need for manual disassembly of the model or correction of errors. Even if design defects occur during the generation process, they can be quickly corrected through the rollback mechanism, avoiding the time wasted on "starting over" in traditional design. While ensuring model quality, it further shortens the design cycle. It is especially suitable for mechanical model generation scenarios with multiple parts and complex bolt connection relationships, providing technical support for efficient design.

[0063] Optionally, the method for automatically generating the mechanical model of the bolt connection further includes: During the model generation process, if the similarity coefficient between the assembly of multiple segmentation modules connected by bolts and a single part in the preset mechanical model library is greater than or equal to the corresponding preset threshold; The corresponding multiple segmentation modules are merged, and the merged module is subjected to a constraint mechanism for judgment.

[0064] Specifically, this process is the optimization and error correction stage in the generation of the bolt connection mechanical model. It relies on the real-time detection and verification mechanism in the model generation process and proceeds in an orderly manner in three steps: detection triggering, merging execution, and secondary verification. The specific operation details are as follows: Triggering conditions: Similarity detection and threshold determination: Detection timing: It is carried out simultaneously with the "stepwise segmentation-constraint judgment" process. After the constraint verification of each segmentation module is completed, the combination of multiple related segmentation modules and bolted connection structures that have passed the verification is automatically detected (e.g., after the separate verification of the two short legs of the workbench and the intermediate connector is completed, the combination of the three connected by bolts is immediately detected). Comparison object: The preset mechanical model library is the "bolted connection mechanical model training dataset" used when building mechanical generative models. The library stores the 3D models, feature parameters (geometric classification features, topological features, bolt hole features, etc.) and corresponding text descriptions of various single mechanical parts, covering common integrated parts such as shafts, rods, and brackets. Similarity calculation: Using the same feature extraction and cosine similarity algorithm as "part routine verification", the overall geometric features (such as the total length, total diameter, and overall topology of the assembly), functional features (such as load-bearing parts and connection interface positions), and the equivalent structure after bolt connection simplification (the core shape of the assembly after removing redundant bolts) of the assembly are extracted and compared with the feature vector of a single part in the preset mechanical model library to generate a "similarity coefficient" (value range 0-1). Threshold trigger: A preset threshold is set, for example, to 0.9 (based on extensive mechanical design case verification, this threshold can effectively distinguish between integrable assemblies and multi-part structures that need to be retained independently). When the similarity coefficient is ≥0.9, the assembly is determined to have merging value, and the merging operation is triggered; if the similarity coefficient is <0.9, the assembly is determined to be a "necessary multi-part structure" (such as a gear set with complex transmission relationships), the merging is not performed, and the subsequent module segmentation continues.

[0065] Merging and Execution: Structural Simplification and Function Retention: First, analyze the bolt connection logic of the assembly to identify "redundant bolt connections" (i.e., connections that do not affect the overall structural stability and functional realization of the assembly after removal, such as the bolt connections between the two short legs and the middle connecting parts, which can be replaced by integrated molding after merging, without the need to retain the bolts). At the same time, mark the "core functional structure" of the assembly (such as the load-bearing section of the short legs and the interface end connected to the desktop) and "key dimensional parameters" (such as the total length of the assembly and the dimensions of the connection surfaces at both ends). Using a geometric fusion algorithm, the 3D models of multiple segmented modules are reconstructed as a whole according to their original relative positions. Redundant bolt connection structures are removed, while core functional structures and key dimensions are retained. For example, the combination of the workbench with "two short legs + middle connector" is merged and reconstructed into "a single complete support leg". The total length (60cm), the dimensions of the connecting surfaces at both ends (diameter 3cm), and the position of the bolt holes at the top of the original combination are retained, while the bolt holes of the middle connector and the short legs are removed. After merging, the feature parameters of the new module are automatically generated (such as the geometric classification feature of the integrated support leg being "rod type" and the topological feature being "genus 0"), and updated to the parameter library generated by the model, replacing the independent parameters of the original multiple segmented modules, providing a basis for subsequent verification and assembly.

[0066] Secondary Verification: Compliance Assurance of the Merged Module: After the merging operation is completed, the merged new module must undergo full-dimensional verification according to the aforementioned "constraint mechanism." This includes: routine part verification (judging the similarity coefficient between the integrated support leg and the "complete support leg" in the library, which must be >0.95), geometric stability verification (calculating the aspect ratio and symmetry score of the integrated support leg, which must be >0.8), machining stability verification (analyzing whether the integrated support leg can be machined, which must be >0.8), connection rationality verification (checking the bolt connection parameters between the integrated support leg and the tabletop, which must be >0.8), and physical stability verification (calculating the load-bearing bending moment and center of gravity position of the integrated support leg, which must be >0.8). Only when the merged module passes all constraint verifications will it be included in the 3D mechanical model data; if it fails (e.g., the aspect ratio of the merged module is too large, resulting in substandard geometric stability), the merging operation will be canceled, reverting to the original state of multiple segmented modules, and the original process will continue.

[0067] This merging process significantly improves the practicality and engineering adaptability of the mechanical model by real-time detection of redundant connection structures and integrated optimization. On the one hand, by merging the complex combination of "multi-module + redundant bolts" into a single integrated part, the number of bolt connections in the model is effectively reduced, avoiding the cumbersome assembly, increased costs, and potential connection failure risks (such as bolt loosening and stress concentration) caused by too many bolts. At the same time, the model structure is closer to the "integrated part" form in actual mechanical design, reducing the difficulty of subsequent processing and manufacturing (such as reducing the number of parts and assembly steps). On the other hand, the secondary constraint verification mechanism after merging ensures that the merging operation does not introduce new design defects (such as geometric instability and unmanufacturability). It retains the core functions of the original combination and improves the overall performance of the modules through structural optimization. The final generated 3D mechanical model not only meets the user's text description requirements but also has the engineering attributes of "less redundancy, high integration, and easy manufacturing," further shortening the conversion cycle from model design to actual application and adapting to the demand for efficient and reliable mechanical design in intelligent manufacturing scenarios.

[0068] Optionally, the output format of the three-dimensional mechanical model data includes at least one of rendered images, 3D visualization files, 2D projected engineering drawings, and 2D sectional engineering drawings.

[0069] Specifically, the multi-format output process of 3D mechanical model data relies on the system's built-in "format conversion and rendering engine." After verifying and merging all segmentation modules, it follows a four-step process of "data extraction - format conversion - quality verification - file output" to ensure that different format files not only match the original model data but also meet the requirements of subsequent application scenarios. Specific details are as follows: First, extract complete core information from the final generated 3D mechanical model data, including: the model's geometric topology data (vertex, edge, and face coordinates and connection relationships of all parts), bolt connection parameters (bolt type, hole coordinates, and connection surface dimensions), dimensional annotation information (part length, width, height, tolerance range, and fit clearance), material properties (default or user-specified material density and mechanical performance parameters), and functional annotations (load-bearing parts and assembly reference surfaces). This information will serve as the basic data source for format conversion, ensuring that each output format can completely reproduce the design details of the model.

[0070] Format conversion: Adapting to different application scenarios: Image rendering output: Based on the extracted geometric topology data and material properties, high-fidelity visualization images are generated using ray tracing rendering algorithms. The system allows users to select rendering perspectives (such as front view, side view, and perspective view), lighting effects (such as natural light and industrial lighting), and background styles (such as solid color background and workshop scene background). The default image resolution is set to 300 DPI (which can be adjusted to 600 DPI to meet printing requirements). Supported formats include JPG, PNG, and TIFF, making it suitable for use in scenarios such as presentations and illustrations in design documents.

[0071] 3D visualization file output: Primarily outputs common industry standard formats, including .stp (STEP format, compatible with cross-CAD software such as AutoCAD and SolidWorks), .igs (IGES format, compatible with older versions of CAD software), and .obj (suitable for 3D visualization software such as Blender). The conversion preserves the model's complete geometry and bolt connection relationships, and embeds dimension annotations and material information, facilitating subsequent model modifications, assembly simulations, or model checks before 3D printing by designers.

[0072] 2D Projection Engineering Drawing Output: In accordance with the national standard for mechanical drawing (GB / T 4458), the 3D model is automatically projected into a front view, top view, and left view (and a right view and bottom view are generated when necessary). Dimensions (such as the total length of the part, bolt hole diameter and spacing), tolerances (such as fit tolerance H7 / g6), technical requirements (such as surface roughness Ra3.2), and a title block (including model name, scale, and material) are automatically added. The format supports .dwg (AutoCAD native format) and .dxf (general 2D format), which can be directly used for workshop processing drawing handover.

[0073] 2D sectional engineering drawing output: Based on the complexity of the model, the system automatically selects the cutting location (such as bolted connections or internal cavities of parts) to generate full, half, or partial sectional views. These views clearly show the internal structure of the parts (such as bolt hole depth and internal step dimensions) and assembly relationships (such as the fit clearance between bolts and holes). The cutting symbols and section line styles conform to mechanical drawing standards, and the format is consistent with 2D projection engineering drawings. Figure 1 This allows for the inclusion of internal details that cannot be shown in the projection drawings, thus meeting the drawing requirements for precision machining scenarios.

[0074] After each file format is generated, an automatic quality check is performed: for rendered images, color reproduction and detail clarity are checked (e.g., whether bolt holes are clearly visible); for 3D visualization files, the integrity of geometric data is checked (e.g., whether there are missing faces or coordinate offsets); for 2D engineering drawings, the accuracy of dimension annotations is checked (e.g., whether the annotation values ​​are consistent with the actual dimensions of the 3D model), the compliance of the view scale (e.g., whether it conforms to standard scales such as 1:1 or 1:2), and the compliance of drafting standards (e.g., whether the direction of section lines is correct). If the check fails, the system automatically backtracks to the format conversion step for reprocessing until the file meets the requirements.

[0075] After verification, the system provides a variety of output options: users can choose to download locally (the file is packaged as a ZIP archive), store in the cloud (linked to the user account, supporting retrieval at any time), or send it directly to a specified email address. At the same time, a file version number (including generation time and model version) is generated for easy version management and traceability.

[0076] This multi-format output process significantly enhances the practicality and application flexibility of bolted connection mechanical models, greatly facilitating subsequent design implementation and multi-scenario applications. On the one hand, by covering multiple formats such as rendered images, 3D visualization files, and 2D engineering drawings, it can simultaneously meet the needs of different roles and stages. Designers can optimize models using 3D files, production personnel can directly use 2D engineering drawings for processing, and marketing or reporting personnel can use high-fidelity rendered images to showcase solutions without the need for additional third-party software for format conversion, greatly reducing the time cost and operational barriers of cross-stage collaboration. On the other hand, all output formats strictly adhere to industry standards (such as GB / T 4458 for mechanical drawings and STEP standard for 3D files) and undergo automatic quality verification to ensure the accuracy and compliance of file information. This avoids design rework or processing errors caused by format incompatibility or incorrect dimensioning, further ensuring the smoothness of the process from model design to actual manufacturing, effectively connecting the design and production ends, and improving the efficiency and reliability of the overall mechanical design work.

[0077] In some specific embodiments, such as Figure 2 As shown, the steps of the automatic generation method for the bolt connection mechanical model include: 1. Generating the original shape model using natural language description: The system receives a natural language description of the mechanical model input by the user (such as "a desktop that can support 5kg + 3 supporting legs"). Based on the mechanical generative model (autoregressive language model), the system parses the text requirements and generates an original shape model containing the overall structure, core parts and bolt connection relationships, which serves as the basic framework for subsequent processing.

[0078] 2. Divide the parts according to the general shape and process them sequentially: Using the "general mechanical part form" as a reference, the original shape model is decomposed into independent segmented modules (such as first segmenting the "desktop" and then segmenting the "support leg"). Following the logic of "processing each module one by one", subsequent verification steps are performed on each segmented module in sequence.

[0079] 3. Calculation and judgment of routine coefficients: Extract the geometric features of the current segmentation module, compare them with common part datasets, calculate the "regularity coefficient," and determine if it is greater than 0.95. If "yes", proceed to the next verification step; If "No", trigger "rollback and re-segmentation", adjust the segmentation logic of the current module (such as optimizing the shape), and then re-execute the verification of this step.

[0080] 4. Calculation and determination of geometric stability coefficient: Calculate the rotational / mirror symmetry score, aspect ratio, and other indicators of the current module to obtain the "geometric stability coefficient," and determine if it is greater than 0.8. If "yes", proceed to the next verification step; If "No", trigger "Rollback and Re-segmentation", adjust module size / shape and re-verify.

[0081] 5. Calculation and judgment of machining stability coefficient: Analyze the machinability of the current module (e.g., whether there are uncuttable structures), calculate the "machining stability coefficient," and determine if it is greater than 0.8. If "yes", proceed to the next verification step; If "No", trigger "rollback and re-segmentation", adjust the module structure (such as removing closed cavities) and then re-verify.

[0082] 6. Calculation and judgment of connection rationality coefficient: For the bolted connections between the current module and other parts, calculate the "connection rationality coefficient" for parameters such as the area of ​​the connection surface and the bolt hole spacing, and determine whether it is greater than 0.8. If "yes", proceed to the next verification step; If "No", trigger "Rollback and Re-segmentation", adjust bolt connection parameters and re-verify.

[0083] 7. Calculation and determination of physical stability coefficient: By combining the module's own weight and associated loads to calculate the bending moment and center of gravity position, a "physical stability coefficient" is obtained, and it is determined whether it is greater than 0.8. If "yes", proceed to the next step; If "No", trigger "rollback and re-segmentation", adjust module size / support structure and re-verify.

[0084] 8. Components are properly identified and recorded: After the current module passes all 5 verifications, it is determined to be a "reasonable part" and its parameters (dimensions, connection relationships, etc.) are written into the database of the 3D mechanical model.

[0085] 9. Other parts are present: Determine if there are any unprocessed segmentation modules in the original shape model: If "yes", proceed to the "next" step and repeat the verification process of steps 3-8 for the next segmentation module; If "No", proceed to the final output stage.

[0086] 10. Complete the design output: After all the segmentation modules have been verified and written into the database, the system integrates all part parameters, generates a complete bolt connection mechanical model, and outputs rendered images, 3D files, 2D engineering drawings, and other formats as required, thus completing the design process.

[0087] like Figure 3 As shown in the figure, an automatic generation device for bolt connection mechanical models provided in this embodiment of the invention includes: A construction unit is used to construct a mechanical generative model, which is based on an autoregressive language model. An acquisition unit is used to acquire the user's description text and convert the description text into structured data; wherein, the description text includes the functional requirements and key dimensional parameters of the mechanical model; The generation unit is used to perform a model generation process based on the structured data and the mechanical generation model to obtain three-dimensional mechanical model data of the bolt connection; wherein the model generation process includes a physical sensing rollback operation.

[0088] An embodiment of the present invention provides an automatic bolt connection mechanical model generation device, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to implement the automatic bolt connection mechanical model generation method as described above when the computer program is executed.

[0089] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the method for automatically generating a bolt connection mechanical model as described above.

[0090] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for automatically generating a mechanical model of a bolted connection, characterized in that, include: A mechanical generation model is constructed, which is based on an autoregressive language model. Obtain the descriptive text and transform it into structured data; wherein, the descriptive text includes the functional requirements and key dimensional parameters of the mechanical model; Based on the structured data, a model generation process is performed using the mechanical generation model to obtain three-dimensional mechanical model data of the bolt connection; wherein, the model generation process includes a physical sensing rollback operation.

2. The method for automatically generating a bolt connection mechanical model according to claim 1, characterized in that, The process of generating a model based on the structured data and using the mechanical generation model to obtain the corresponding three-dimensional mechanical model data includes: A preliminary perception model is obtained based on the structured data; The preliminary perception model is segmented step by step, and the segmentation module obtained at each step is subject to constraint mechanism judgment to obtain the judgment result. The three-dimensional mechanical model data is obtained based on the judgment result.

3. The method for automatically generating a bolt connection mechanical model according to claim 2, characterized in that, The process involves progressively segmenting the initial perception model, and performing constraint mechanism judgments on each segmentation module to obtain judgment results. Based on these judgment results, the three-dimensional mechanical model data is obtained, including: Based on the geometric structure of the preliminary perception model, segmentation processing is performed to obtain the current segmentation module and the remaining segmentation modules. The current segmentation module corresponds to one or a group of associated mechanical parts. The constraint mechanism is used to determine the current segmentation module; If the current segmentation module fails to pass any constraint mechanism verification, a subsequent physical perception rollback operation is triggered. If the current segmentation module passes all constraint mechanisms, the remaining segmentation modules are segmented to obtain a temporary segmentation model. The temporary segmentation model is then used as the current segmentation module to return to the execution constraint mechanism for judgment, until the judgment of the segmentation module is obtained after each step of the segmentation process is completed. The three-dimensional mechanical model data is determined based on all the segmentation modules that have passed all constraint mechanisms.

4. The method for automatically generating a bolt connection mechanical model according to claim 3, characterized in that, The constraint mechanism includes part routine verification, part geometric stability verification, part machining stability verification, connection rationality verification, and physical stability verification.

5. The method for automatically generating a bolt connection mechanical model according to claim 3, characterized in that, The specific process of the physical sensing rollback operation includes: During model generation, if the constraint mechanism verification result of the current segmentation module fails, rollback to the state of the previous segmentation module that passed all constraint verifications. Based on the adjusted segmentation logic, the segmentation process and constraint mechanism judgment are re-executed until the newly generated current segmentation module passes all verifications, and then the subsequent model generation process continues.

6. The method for automatically generating a bolt connection mechanical model according to claim 3, characterized in that, The automatic generation method for bolt connection mechanical models also includes: During the model generation process, if the similarity coefficient between the assembly of multiple segmentation modules connected by bolts and a single part in the preset mechanical model library is greater than or equal to the corresponding preset threshold; The corresponding multiple segmentation modules are merged, and the merged module is subjected to a constraint mechanism for judgment.

7. The method for automatically generating a bolt connection mechanical model according to claim 1, characterized in that, The output format of the three-dimensional mechanical model data includes at least one of rendered images, 3D visualization files, 2D projected engineering drawings, and 2D sectional engineering drawings.

8. An automatic generation device for bolt connection mechanical models, characterized in that, include: A construction unit is used to construct a mechanical generative model, which is based on an autoregressive language model. An acquisition unit is used to acquire the user's description text and convert the description text into structured data; wherein, the description text includes the functional requirements and key dimensional parameters of the mechanical model; The generation unit is used to perform a model generation process based on the structured data and the mechanical generation model to obtain three-dimensional mechanical model data of the bolt connection; wherein the model generation process includes a physical sensing rollback operation.

9. An automatic model generation device for bolted connections, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the automatic generation method for bolt connection mechanical models as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the automatic generation method for bolt connection mechanical models as described in any one of claims 1 to 7.

Images