An artificial intelligence-based tooling fixture design method, system and device

Abstract

The application provides a kind of tooling fixture design method, system and equipment based on artificial intelligence, relating to artificial intelligence technical field, the design method, by obtaining and fusing the three-dimensional model of the part to be processed, process information and manufacturing constraint, constructs the clamping state model and design semantic graph of unified representation clamping demand, and then based on the semantic graph, similar cases are intelligently retrieved from the historical case library and reusable constraint patterns are extracted, finally, the fixture design scheme with historical engineering experience and current constraints is generated combined with large model. This method converts discrete geometry, process and constraint information into structured and computable design semantics, enhances the engineering rationality of large model reasoning using historical case knowledge, significantly reduces the dependence of the design process on individual human experience, and significantly improves the automation level and knowledge reuse efficiency of tooling fixture design under the premise of ensuring stable positioning of the design scheme, reliable clamping and meeting engineering constraints such as space interference.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a tooling fixture design method, system, and device based on artificial intelligence. Background Technology

[0002] As a key process equipment in mechanical manufacturing systems for achieving accurate workpiece positioning, reliable clamping, and stable support, the design quality of tooling fixtures directly determines machining accuracy, production efficiency, and process costs. Traditional fixture design relies heavily on the personal experience and knowledge accumulation of designers, typically employing computer-aided design methods based on rule bases or instance reasoning. This approach suffers from inherent limitations such as long design cycles, repetitive work, and difficulties in effectively reusing and standardizing design results within an enterprise.

[0003] In recent years, with the development of artificial intelligence technology, automated design methods based on deep learning or large models have begun to be introduced into this field. However, most of these methods focus on generating conceptual solutions from text or simple rules, resulting in designs with low engineering credibility that are difficult to apply directly to actual production. Furthermore, the large number of historical fixture design cases accumulated by manufacturing enterprises over a long period, which have been validated through practice, are usually stored in unstructured or semi-structured forms. These cases are difficult to effectively transform into structured knowledge that can be retrieved and reasoned about by intelligent systems, leading to the waste and accumulation of valuable engineering experience.

[0004] Therefore, there is a lack of a tooling and fixture design method based on artificial intelligence reasoning capabilities in the existing technology, which can ensure the reliability of the engineering solution while realizing the automation, intelligence and continuous reuse of knowledge in the design process. Summary of the Invention

[0005] The problem solved by this invention is one or more of the aforementioned related technical problems.

[0006] To address the above problems, this invention provides a tooling fixture design method, system, and device based on artificial intelligence.

[0007] In a first aspect, the present invention provides a tooling fixture design method based on artificial intelligence, comprising:

[0008] Obtain the 3D model data, machining process information, and manufacturing constraints of the part to be processed;

[0009] Based on the three-dimensional model data and the processing technology information, a clamping state model is constructed to characterize the clamping requirements of the part to be processed during the processing.

[0010] The clamping state model is fused with the manufacturing constraints to obtain a design semantic graph;

[0011] Based on the design semantic graph, similar historical fixture cases are retrieved from the historical fixture case library, and reusable constraint patterns and configuration rules are extracted from the retrieved historical fixture cases to form a structured case context.

[0012] The design semantic graph and the case context are input into a preset large model to generate a tooling fixture design scheme for the part to be processed. The tooling fixture design scheme is a structured scheme that includes fixture structure, positioning scheme, clamping configuration and key parameters.

[0013] Optionally, the step of constructing a clamping state model based on the three-dimensional model data and the machining process information to characterize the clamping requirements of the part to be processed during the machining process includes:

[0014] Geometric analysis is performed on the three-dimensional model data to obtain geometric features, which include a positioning reference plane, a clamping area, and a key force-bearing area.

[0015] Based on the processing technology information, determine the process characteristic parameters and constraint requirements;

[0016] The geometric features, process feature parameters, and constraint requirements are modeled in a unified manner to obtain a structured clamping state model.

[0017] Optionally, fusing the clamping state model with the manufacturing constraints to obtain a design semantic graph includes:

[0018] Extract part structural feature data and process feature data from the clamping state model, and extract engineering constraint data from the manufacturing constraint data;

[0019] Based on the clamping logic information in the clamping state model and the engineering constraint information in the manufacturing constraints, the positioning relationship, clamping relationship and constraint relationship are determined.

[0020] The part structural feature data, the process feature data, and the engineering constraint data are used as semantic nodes, and the positioning relationship, the clamping relationship, and the constraint relationship are used as semantic edges.

[0021] Based on the semantic nodes and semantic edges, the design semantic graph is constructed.

[0022] Optionally, the step of retrieving similar historical fixture cases from the historical fixture case library based on the design semantic graph includes:

[0023] The design semantic graph is subjected to multimodal feature analysis to obtain geometric modal features, process modal features and engineering constraint modal features;

[0024] The geometric modal features, the process modal features, and the engineering constraint modal features are fused using a preset fusion method to obtain a joint feature vector;

[0025] The joint feature vector is matched with the corresponding feature vector of each case in the historical fixture case library for similarity, and one or more similar historical fixture cases are determined based on the similarity matching results.

[0026] Optionally, the geometric modal features include positioning reference plane, clamping area, hole features, and key stress area information; the process modal features include processing method, processing direction, number of clamping operations, and machine tool type information; the engineering constraint modal features include degree-of-freedom constraint set, allowable clamping force range, spatial interference restrictions, and interface constraint information.

[0027] Optionally, after inputting the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme, the process further includes:

[0028] Based on the preset inspection method, the engineering feasibility of the tooling fixture design scheme is verified, and the verification results are obtained.

[0029] If the verification result is unsuccessful, verification feedback information containing the specific constraint violation type is generated and fed back to the preset large model to guide the correction of the tooling fixture design scheme and to re-execute the step of verifying the engineering feasibility of the tooling fixture design scheme until the tooling fixture design scheme meets the engineering constraints.

[0030] Optionally, the preset inspection methods include freedom constraint integrity verification, clamping mechanics feasibility verification, and machining space interference verification.

[0031] Optionally, the manufacturing constraints include at least one of the following: machining accuracy requirements, clamping stiffness requirements, clamping time requirements, machine tool interface constraints, and fixture manufacturing cost constraints.

[0032] Secondly, the present invention provides an artificial intelligence-based tooling fixture design system, comprising:

[0033] The acquisition unit is used to acquire the 3D model data, machining process information, and manufacturing constraints of the part to be processed.

[0034] A construction unit is used to construct a clamping state model that characterizes the clamping requirements of the part to be processed during the processing, based on the three-dimensional model data and the processing technology information.

[0035] The fusion unit is used to fuse the clamping state model with the manufacturing constraints to obtain a design semantic graph;

[0036] The processing unit is used to retrieve similar historical fixture cases from the historical fixture case library based on the design semantic graph, and extract reusable constraint patterns and configuration rules from the retrieved historical fixture cases to form a structured case context.

[0037] The processing unit is also used to input the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme for the part to be processed. The tooling fixture design scheme is a structured scheme that includes fixture structure, positioning scheme, clamping configuration and key parameters.

[0038] Thirdly, the present invention provides an artificial intelligence-based tooling fixture design 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 artificial intelligence-based tooling fixture design method described in the first aspect.

[0039] The beneficial effects of the tooling fixture design method, system, and equipment based on artificial intelligence of the present invention are:

[0040] By acquiring and integrating the 3D model, process information, and manufacturing constraints of the parts to be processed, a unified clamping state model and design semantic graph representing clamping requirements are constructed. Based on this semantic graph, similar cases are intelligently retrieved from a historical case library, and reusable constraint patterns are extracted. Finally, a fixture design scheme that combines historical engineering experience with current constraints is generated by combining the large model. This method transforms discrete geometric, process, and constraint information into structured, computable design semantics. It leverages historical case knowledge to enhance the engineering rationality of the large model's reasoning, significantly reducing the reliance on individual human experience in the design process. While ensuring stable positioning, reliable clamping, and compliance with engineering constraints such as spatial interference, it greatly improves the automation level, knowledge reuse efficiency, and consistency of solution generation in tooling and fixture design. Attached Figure Description

[0041] Figure 1 This is one of the flowcharts illustrating an artificial intelligence-based tooling fixture design method according to an embodiment of the present invention;

[0042] Figure 2 This is a second flowchart illustrating an artificial intelligence-based tooling fixture design method according to an embodiment of the present invention.

[0043] Figure 3 This is a schematic diagram of a tooling fixture design system based on artificial intelligence, according to an embodiment of the present invention. Detailed Implementation

[0044] 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.

[0045] 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.

[0046] 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.

[0047] 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".

[0048] 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.

[0049] like Figure 1 As shown in the figure, an embodiment of the present invention provides an artificial intelligence-based tooling fixture design method, comprising:

[0050] Step S100: Obtain the three-dimensional model data, processing technology information and manufacturing constraints of the part to be processed.

[0051] Specifically, 3D model data refers to the digital geometric representation of the part to be processed, typically derived from files generated by a computer-aided design (CAD) system (such as STEP, IGES format, or a native format of a specific CAD software). This data precisely defines the part's shape, dimensions, topology, and all geometric features. For example, the 3D model data of an aluminum alloy flange part to be processed contains precise 3D information on all features such as the flange body, center hole, bolt holes, and mounting flanges.

[0052] Machining process information: This refers to the process planning data describing how the part will be manufactured. It specifies the type, sequence, and orientation of machining operations, as well as the equipment resources used. For example, for the flange mentioned above, the process information might specify: "On a vertical machining center, the upper surface will be machined using a face milling operation, with the spindle axis vertically downwards, and a machining allowance of 2mm." This information directly affects the support and constraint directions that the fixture needs to provide, as well as the machining areas that need to be avoided.

[0053] Manufacturing constraints refer to a series of engineering limitations and performance requirements that must be met when designing and applying fixtures. These conditions are usually derived from process specifications, machine tool parameters, quality standards, and cost considerations. For example, they may include: "The part must be fully constrained in all five degrees of freedom except rotation about the spindle during machining," "The clamping force must not exceed 500N to prevent part deformation," "The total height of the fixture must be less than 300mm to fit the machine tool's working chamber," and "The positioning error must be guaranteed to be within ±0.02mm," etc.

[0054] By systematically aggregating three types of heterogeneous information—part geometry, manufacturing processes, and engineering rules—it provides a structured and computable data foundation for the entire intelligent design process. It transforms the vague requirements that rely on manual interpretation and synthesis in traditional design into explicit input parameters that can be directly processed by computers. This ensures that all subsequent automated reasoning (such as state modeling, case retrieval, and solution generation) begins with a complete, accurate, and unambiguous problem definition, improving the objectivity, repeatability, and precise attainability of engineering goals in the design process.

[0055] Step S200: Based on the three-dimensional model data and the processing technology information, construct a clamping state model to characterize the clamping requirements of the part to be processed during the processing.

[0056] Specifically, the raw data (3D model and process information) obtained in step S100 is intelligently analyzed and integrated to construct a "clamping state model" that can explicitly and structurally represent all clamping technical requirements of the part under a specific processing procedure. This process is not simply data packaging, but a deep interpretation and engineering definition of the design intent.

[0057] For example, the three-dimensional model data is analyzed to determine geometric features such as positioning reference planes, clamping areas and key stress areas, and the machining process information is analyzed to determine the machining posture and direction, main stress analysis, constraint requirements, etc.

[0058] Ultimately, the geometric elements, process parameters, and constraint requirements derived from the above analysis are unified into a model, outputting a structured "clamping state model." This model, in a computer-understandable form, clearly indicates "where (geometric position)," "in what way (positioning / clamping)," "what is constrained (degrees of freedom)," and "what forces need to be resisted (force conditions)."

[0059] By automatically converting unstructured geometric models and textual process information into a structured "clamping state model" rich in engineering semantics, precise digital capture of the designer's clamping intent is achieved. This solves the problems of unclear reasoning direction and weak engineering foundation of solutions caused by fuzzy input information and missing semantics in traditional intelligent design methods. This model provides a unified engineering constraint basis for subsequent operations such as case matching, solution generation, and interference verification, ensuring that each step of the design process is based on consistent and clear constraints, thereby improving the reliability and engineering credibility of the design.

[0060] Step S300: The clamping state model is fused with the manufacturing constraints to obtain a design semantic graph.

[0061] Specifically, the "clamping state model" generated in step S200, which describes the clamping requirements of the part itself, is deeply integrated with the "manufacturing constraints" obtained in step S100, which describes external engineering limitations, to construct a unified, graph-structured "design semantic graph." This graph aims to comprehensively transform engineering problems into a computer-analyzable, searchable, and reasonable networked knowledge representation.

[0062] For example, the specific integration and construction process is as follows:

[0063] Specific part features (such as the main positioning bottom surface, side through holes, and top clamping surface) are extracted from the "clamping state model," and explicit restriction rules (such as positioning accuracy ≤ 0.05mm and prohibition of setting up structures within 30mm above the machining area) are extracted from the "manufacturing constraints." These discrete "features" and "rules" are abstracted into nodes in a semantic graph.

[0064] Further analyze the logical and engineering relationships between nodes and abstract them as edges connecting the nodes. Examples include positioning / clamping relationships, constraint relationships, and association relationships.

[0065] Each node and edge is assigned specific attribute values ​​to achieve parameterized description. For example, the node "Main Positioning Bottom Surface" has attributes including geometric position coordinates, surface normal vector, and flatness requirements. The edge "Clamping Relationship" connects the node "Top Clamping Surface" and the virtual node "Pressure Plate" and its attributes include the suggested clamping force range (200-400N), force direction, and contact type (surface contact).

[0066] Ultimately, all nodes, edges, and their attributes together constitute a complete design semantic graph. This design semantic graph organizes the relationships between various elements in the current fixture design task in a graph structure, including not only positioning surfaces and clamping points, but also clearly defining constraints, spatial relationships, and accuracy and interference limitations.

[0067] By constructing a "design semantic graph," a unified, standardized, and relational representation of multi-source, heterogeneous design information (geometric requirements, process logic, and engineering rules) is achieved. It integrates design constraints, originally scattered across different dimensions and forms, into a clearly structured, semantically rich graph model capable of performing complex relational operations directly on a computer. This design semantic graph can be used for subsequent historical case retrieval based on semantic similarity and provides a structured design context for large models, improving the retrieval accuracy problems caused by information isolation and semantic fragmentation in traditional methods.

[0068] Step S400: Based on the design semantic graph, retrieve similar historical fixture cases from the historical fixture case library, and extract reusable constraint patterns and configuration rules from the retrieved historical fixture cases to form a structured case context.

[0069] Specifically, based on the "design semantic graph" generated in step S300 that accurately describes the current design task, the system intelligently searches for and learns from past successful experiences in the established "historical fixture case library." This process is not a simple keyword matching, but rather a similarity retrieval and knowledge extraction based on deep semantic understanding. Each case in the historical fixture case library includes information on the fixture structure type, positioning method, clamping method, and actual application effect.

[0070] The specific execution process may include the following:

[0071] The multi-dimensional information (geometry, process, constraints) contained in the "design semantic graph" is jointly encoded and transformed into a comprehensive "feature vector." Simultaneously, each case in the historical case library is also pre-encoded as a feature vector in the same way. The best-matching case is found by calculating the semantic similarity between the current task's feature vector and the feature vectors of all historical cases.

[0072] For example, the current task is to design a milling fixture for an aluminum bracket part, whose semantic graph emphasizes "the need for lateral constraint and primary resistance to horizontal cutting forces". The system may retrieve a historical case of a fixture designed for milling a similar "cast iron bearing housing" because the semantic graphs of the two are highly similar in modal features such as "clamping posture (vertical)", "primary constraint direction (lateral)", and "critical force (horizontal force)", despite the different part geometries.

[0073] Instead of copying the entire design from one or a few of the most similar historical cases retrieved, the system intelligently analyzes and extracts universal and transferable "design patterns" and "rules".

[0074] Constraint pattern: refers to the principle arrangement of positioning and clamping abstracted from a specific structure. For example, from historical cases, a pattern such as "using one surface and two pins (one plane and two positioning pins) to achieve complete positioning of disc-shaped parts" may be extracted.

[0075] Configuration rules: These refer to the specific engineering parameters or component selection logic that correspond to the constraint mode. For example, a rule that corresponds to the above mode might be: "When the part hole diameter is H7 level, select a combination of fixed cylindrical pins and diamond pins of g6 level."

[0076] These extracted patterns and rules are organized into a structured "case context" document. This document clearly explains "what design patterns were used in similar situations (How), and what the underlying configuration logic was (Why)."

[0077] By leveraging deep semantic-based case retrieval and knowledge extraction, fragmented and implicit engineering experience embedded in historical data is transformed into explicit, structured design knowledge that can be directly accessed by computers. This approach represents a leap from "data retrieval" to "knowledge retrieval," enabling large models to not only be constrained by the current task but also to receive contextual guidance on "how similar problems were successfully solved" during subsequent reasoning. This improves the engineering rationality and maturity of generated solutions, reduces trial and error in the design process, and enhances design efficiency and the first-time success rate of solutions.

[0078] Step S500: Input the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme for the part to be processed. The tooling fixture design scheme is a structured scheme that includes fixture structure form, positioning scheme, clamping configuration and key parameters.

[0079] Specifically, the “design semantic graph” generated in step S300, which accurately describes the current design constraints, and the “structured case context” extracted in step S400, which provides guidance based on historical experience, are used together as input to drive a pre-set large model to perform engineering reasoning, and finally automatically generate a complete, detailed tooling fixture design scheme that can be directly used to guide subsequent detailed design and manufacturing.

[0080] The pre-defined large model is a domain-adapted, large-scale artificial intelligence model capable of processing mixed inputs of structured engineering semantics and natural language. Its core function is to receive and understand the engineering constraints represented by the design semantic graph and the historical experience carried by the structured case context, and then perform engineering reasoning to generate detailed design solutions that meet all constraints. It receives two key inputs.

[0081] Design semantic graphs, input in structured data formats (such as adjacency lists, attribute lists, or specific serialization formats), clearly define the "design space" and "hard constraints" of the current task for the larger model. They tell the model: parts are here, they need to be positioned this way, clamped that way, and these spatial and mechanical rules cannot be violated.

[0082] Structured case context: Input in the form of natural language or semi-structured text provides design inspiration for the model. It suggests to the model: historically, when solving similar problems, experts have adopted this structural layout, used those types of components, and followed the following configuration principles.

[0083] The preset large model can be selected from various basic model types, such as general large language models (LLMs), including the GPT series (such as GPT-4), LLaMA series, Claude series, etc.; multimodal large models, including models that support mixed image and text input, such as Gemini and GPT-4V; and domain-adapted or fine-tuned specialized models, including models that are pre-trained or supervised fine-tuned based on general LLMs (such as LLaMA and ChatGLM) using a large amount of professional corpus in the fields of mechanical design and fixture design (such as design manuals, patent documents, and historical case reports).

[0084] The large model first deeply understands the engineering meaning represented by all nodes and edges in the "design semantic graph".

[0085] Then, it adapts and integrates the historical patterns in the "case context" with the details of the current drawing. For example, if the case suggests "one surface, two pins", the model will calculate the precise installation position and selection specifications of the locating pins based on the specific reference surface dimensions and hole diameter of the current part.

[0086] Throughout the reasoning process, the model continuously verifies whether the generated content violates any constraints (such as interference or degrees of freedom) defined in the "design semantic graph," ensuring the engineering feasibility of the solution.

[0087] Finally, the model outputs a complete structured design scheme. For example, for the aforementioned task of milling grooves in an aluminum bracket, the output might be specified as follows: Overall fixture structure: Bent plate fixture. Positioning scheme: The bottom surface of the bracket is used as the main positioning surface (constraining three degrees of freedom), and complete positioning is achieved by using two process holes on the side combined with a combination of "one cylindrical pin and one diamond pin". Clamping configuration: Two adjustable spiral pressure plates are used to apply clamping force in the non-machined area at the top of the bracket. Key parameters: Positioning pin diameter Φ8h6, clamping bolt specification M10, pressure plate opening height must be ≥50mm.

[0088] By deeply integrating precise engineering constraints (design semantic graphs) with valuable prior knowledge (case context), this approach guides large-scale models to conduct intelligent design reasoning with clear objectives and boundaries. It successfully combines the general generative capabilities of large-scale models with engineering rules and historical experience specific to certain domains, resulting in design solutions that are not only innovative but also highly engineering-rational, manufacturable, and reliable. This overcomes the shortcomings of traditional automated design methods, such as vague and impractical solutions, and the uncontrollable and unreliable results generated by general-purpose large-scale models in specialized fields. It achieves a paradigm shift in tooling and fixture design from "human experience-driven" to "knowledge data and AI collaboratively driven," significantly improving the automation level, solution quality, and decision-making efficiency of the design process.

[0089] In this embodiment, by acquiring and fusing the 3D model, process information, and manufacturing constraints of the part to be processed, a unified clamping state model and design semantic graph representing clamping requirements are constructed. Then, based on this semantic graph, similar cases are intelligently retrieved from a historical case library, and reusable constraint patterns are extracted. Finally, a fixture design scheme that combines historical engineering experience with current constraints is generated by combining the large model. This method transforms discrete geometric, process, and constraint information into structured, computable design semantics, enhances the engineering rationality of large model reasoning by utilizing historical case knowledge, significantly reduces the reliance on individual human experience in the design process, and greatly improves the automation level, knowledge reuse efficiency, and consistency of solution generation in tooling fixture design while ensuring stable positioning, reliable clamping, and compliance with engineering constraints such as spatial interference.

[0090] Optionally, the step of constructing a clamping state model based on the three-dimensional model data and the machining process information to characterize the clamping requirements of the part to be processed during the machining process includes:

[0091] Geometric analysis is performed on the three-dimensional model data to obtain geometric features, which include a positioning reference plane, a clamping area, and a key force-bearing area.

[0092] Based on the processing technology information, determine the process characteristic parameters and constraint requirements;

[0093] The geometric features, process feature parameters, and constraint requirements are modeled in a unified manner to obtain a structured clamping state model.

[0094] Specifically, the goal of 3D model geometry analysis is to automatically identify geometric features that are crucial to fixture design from the digital model of the part.

[0095] Positioning reference surface identification: Based on rules such as area, flatness, and relationship with machined surfaces, all planes or regular curved surfaces of the analysis model are automatically filtered or confirmed interactively as the main positioning reference surfaces. For example, for a box-shaped part, its bottom surface is usually identified as the main positioning reference surface because it has the largest area and is perpendicular to most machined surfaces.

[0096] Clampable area identification: By analyzing the accessibility, stiffness (avoiding thin-walled areas), and whether it is outside the machining path, a safe area where clamping force can be applied is automatically defined. For example, the upper part of the two symmetrical side walls of the aforementioned housing part, which are not scheduled for machining operations, can be marked as a clampable area.

[0097] Critical stress area identification: Through finite element analysis pre-calculation or rule-based reasoning (such as identifying cantilever and thin-walled structures), areas that may experience significant deformation or vibration under a given processing load are predicted. For example, a thin-walled lug structure used for connection on a box may be identified as a critical stress area during the drilling process, indicating the need for auxiliary support.

[0098] Machining process information parsing and constraint requirement derivation: Transforming textual or structured machining process information into specific clamping parameters and constraint targets.

[0099] Determine process characteristic parameters: Analyze the operation card or process document to extract parameters that directly affect the fixture type. For example, from "Operation 10: Milling the top surface on a vertical machining center", extract the machining equipment type (vertical machining), machining direction (spindle axis, usually Z-axis), and operation content (milling a plane). From this, it can be inferred that the fixture needs to be adapted to the vertical worktable and provide positioning and clamping perpendicular to the Z-axis.

[0100] Analyze and determine constraint requirements (degrees of freedom set): This analysis is based on classic clamping theories such as the "six-point positioning principle". Based on the part's set posture on the machine tool (determined by machining process information) and process requirements, the system automatically analyzes the degrees of freedom that must be restricted (i.e., degree-of-freedom constraints).

[0101] For example, for a box-shaped part placed flat on a worktable, during the top surface milling process, it is necessary to prevent its movement in the Z direction (1 translational degree of freedom) and rotation about the X and Y axes (2 rotational degrees of freedom) to ensure the thickness dimension. Simultaneously, to prevent its movement in the horizontal plane, it is usually also necessary to constrain the translation in the X and Y directions (2 translational degrees of freedom). Therefore, the constraint requirements (degrees of freedom set) for this process can be clearly defined as: constraining Z-axis translation, X-axis translation, Y-axis translation, rotation about X, and rotation about Y, a total of 5 degrees of freedom. Rotation about the Z-axis may be allowed freely due to machining symmetry or may require additional constraints.

[0102] The discrete features and requirements identified above are integrated into a unified data structure. This can be achieved using an object-oriented data model or a domain-specific language (DSL). For example, a "clamping state model" object can be created, whose attributes include: a list of constrained degrees of freedom, a list of positioning features, a list of clampable regions, main force conditions, and prohibited interference regions, etc.

[0103] Geometric analysis can be performed using CAD kernel APIs (such as ACIS and Parasolid) for programmatic recognition, or by applying a deep learning-based geometric feature recognition model. Process analysis can be performed using natural language processing (NLP) techniques to parse text process documents, or by directly reading from a structured process database (such as an MES system).

[0104] Through the aforementioned analysis and modeling process, the part geometry and process intent, which originally relied on human experience for interpretation, are transformed into a precise, structured, and unambiguous "clamping state model." This model serves as the source of engineering requirements for all subsequent intelligent steps (such as semantic graph construction and case retrieval), ensuring the consistency of input in the design process. It enables the computer to understand the specific problems that clamping needs to solve (which degrees of freedom to constrain, where to apply forces, and which areas to avoid), thus providing a data foundation for generating engineeringly reasonable and feasible design solutions, significantly reducing design rework caused by ambiguous requirements or misunderstandings.

[0105] Optionally, fusing the clamping state model with the manufacturing constraints to obtain a design semantic graph includes:

[0106] Extract part structural feature data and process feature data from the clamping state model, and extract engineering constraint data from the manufacturing constraint data;

[0107] Based on the clamping logic information in the clamping state model and the engineering constraint information in the manufacturing constraints, the positioning relationship, clamping relationship and constraint relationship are determined.

[0108] The part structural feature data, the process feature data, and the engineering constraint data are used as semantic nodes, and the positioning relationship, the clamping relationship, and the constraint relationship are used as semantic edges.

[0109] Based on the semantic nodes and semantic edges, the design semantic graph is constructed.

[0110] Specifically, the "clamping state model" and "manufacturing constraints" are deeply integrated to construct a "design semantic graph" that uniformly represents design intent and engineering rules. This process aims to transform discrete technical parameters and abstract rules into a computable, searchable, and reasonable graphical knowledge network.

[0111] First, extract the core elements from the existing model to prepare "entity" materials for building the semantic network:

[0112] Extracting part structural feature data from the clamping state model: This refers to the solidified features and their parameters obtained from geometric analysis. For example, extracting the main locating surface {ID: F1, geometry: plane equation, normal vector: [0,0,1], roughness requirement: Ra1.6}.

[0113] Process characteristic data: refers to process parameters that affect fixture design and are derived from process information. For example, extracting machining conditions {machining direction: +Z, main cutting force direction: [1,0,0], force estimation: 300N}.

[0114] Extracting engineering constraint data from manufacturing constraints: These refer to explicit restrictive rules or thresholds. For example, extracting accuracy constraints {positioning error: ≤0.03mm}, spatial constraints {maximum fixture profile: X<400mm, Y<300mm, Z<250mm}, and mechanical constraints {maximum allowable clamping force: 1000N}.

[0115] Relationship Analysis and Determination: Based on engineering logic, analyze the interactions and constraints between the above entities, and define the types and attributes of "relationships".

[0116] The positioning / clamping relationship is determined based on the clamping logic:

[0117] Positioning Relationship: Analyze the constraint correspondence between part features and fixture positioning elements. For example, determine the positioning relationship {Type: Main Positioning, Applied to: Part Feature F1, Constraint Degrees of Freedom: [Tz, Rx, Ry], Ideal Contact Type: Surface Contact}.

[0118] Clamping Relationship: Analyze the force relationship between the clampable area of ​​the part and the clamping elements of the fixture. For example, determine the clamping relationship {Type: spiral pressure plate, Applied to: part feature (side wall area), Direction of action: [0,0,-1], Purpose: to maintain contact}.

[0119] Determining constraints based on engineering limitations: This involves translating manufacturing constraints into explicit "prohibitions" or "requirements" in the diagram. For example, determining spatial interference constraints {Main body: fixture body, Prohibited entry: toolpath envelope + safety zone, minimum distance: 5mm}, and stiffness constraints {Object: clamping point, maximum allowable deformation: 0.01mm}.

[0120] Entities are abstracted as "nodes", relationships are abstracted as "edges", attributes are assigned, and they are assembled into a graph.

[0121] Node creation and attribute assignment: The extracted "part structural feature data", "process feature data", and "engineering constraint data" are instantiated as semantic nodes respectively. For example, create a node Node_F1 (type: positioning datum plane) and add attributes {normal vector, dimension, accuracy} to it.

[0122] At the same time, necessary auxiliary nodes will be created, such as Node_Machine Table (Type: External Reference) and Node_Toolpath (Type: Process Geometry).

[0123] Edge creation and attribute assignment: Instantiate the defined "positioning relationship", "clamping relationship", and "constraint relationship" as semantic edges connecting related nodes. For example, create an edge Edge_P1 (type: is_located_on) between node Node_F1 and Node_Machine Tool Table, and add the attribute {list of degrees of freedom of the constraint, allowable error} to it.

[0124] Graph Construction and Output: Using the Structured Query Language (SMQ) of a graph database (such as Neo4j) or an in-memory graph data structure (such as the NetworkX library), all nodes and edges are assembled into an attributed graph with attributes and types. This graph is the design semantic graph, whose data structure clearly expresses "what entities (nodes)" and "how they are related (edges and attributes)," thus completely encapsulating the design semantics.

[0125] Semantic graph construction can be implemented in various ways:

[0126] Pattern definition: A "fixture design domain ontology" or pattern can be predefined to standardize node types, edge types and their set of attributes, ensuring semantic consistency and standardization.

[0127] Automated Build: Through a rules engine or script, the clamping state model (usually in JSON / XML format) and the list of manufacturing constraints are automatically converted into nodes and edges that conform to a predetermined pattern, thereby achieving automatic graph generation.

[0128] Through multi-source data extraction, relationship analysis, and graph structure construction, heterogeneous design information from different dimensions (geometry, process, rules) was successfully integrated and sublimated into a unified knowledge graph rich in machine-readable semantics—the design semantic graph. This graph structure not only preserves the accuracy of all original constraints but, more importantly, reveals the intrinsic connections between them, transforming design requirements from a list of parameters into an interconnected networked knowledge system. This enables subsequent accurate retrieval of historical cases based on semantic similarity rather than simple keyword matching, and provides deeply structured, logically rich contextual information for design reasoning in large models. It solves the problem of information silos and represents a crucial step in advancing design intelligence from the "data processing" level to the "knowledge understanding and application" level.

[0129] Optionally, the step of retrieving similar historical fixture cases from the historical fixture case library based on the design semantic graph includes:

[0130] The design semantic graph is subjected to multimodal feature analysis to obtain geometric modal features, process modal features and engineering constraint modal features;

[0131] The geometric modal features, the process modal features, and the engineering constraint modal features are fused using a preset fusion method to obtain a joint feature vector;

[0132] The joint feature vector is matched with the corresponding feature vector of each case in the historical fixture case library for similarity, and one or more similar historical fixture cases are determined based on the similarity matching results.

[0133] Optionally, the geometric modal features include positioning reference plane, clamping area, hole features, and key stress area information; the process modal features include processing method, processing direction, number of clamping operations, and machine tool type information; the engineering constraint modal features include degree-of-freedom constraint set, allowable clamping force range, spatial interference restrictions, and interface constraint information.

[0134] Specifically, multimodal feature parsing involves extracting core dimensions from the semantic graph. This step involves uniformly designing the semantic graph and parsing it into several independent feature dimensions for in-depth comparison.

[0135] Geometric modal feature analysis: Extracting nodes and attributes directly related to the geometry and structure of the part from the semantic graph. This includes identified positioning reference surfaces (such as a large plane), clamping areas (such as two thick bosses), hole features (such as a set of holes for bolts or pin holes for positioning), and critical stress areas (such as a slender cantilever).

[0136] During parsing, not only is the type of the feature recorded, but its key attributes are also extracted. For example, for a hole feature, it is not only labeled as a "hole," but its type (through hole / blind hole / threaded hole), approximate diameter, location, orientation, and other information are also extracted. This can be achieved by traversing all nodes of type "geometric feature" in the semantic graph and reading their attributes.

[0137] Process modal feature analysis: Extracting nodes and relationships related to the machining process from the semantic graph. These include machining method (e.g., milling, drilling), machining direction (the orientation of the spindle relative to the workpiece), number of clamping operations (how many times is this operation clamped), and type of machine tool used (e.g., vertical machining center, horizontal machining center).

[0138] For example: from the edge connecting the "Part" node and the "Machining Operation" node, read the "Machining Type" attribute as "Face Milling"; from the attributes of the "Machining Operation" node, read the "Machine Tool" as "Five-Axis Vertical Machining Center" and the "Clamping Index" as "1".

[0139] Engineering constraint modal feature analysis: Extracting boundary constraints imposed by manufacturing conditions from the semantic graph. This includes the set of degrees of freedom that must be constrained, for example, {Tx, Ty, Tz, Rx, Ry}, where: Tx: Translation along the X-axis. This can be understood as the part sliding in the left-right direction (assuming the X-axis is horizontal). Ty: Translation along the Y-axis. This can be understood as the part sliding in the front-back direction (assuming the Y-axis is horizontal). Tz: Translation along the Z-axis. This usually refers to the part moving up and down (vertically). Rx: Rotation about the X-axis. Imagine the part rotating (rolling back and forth) about an axis passing through it. Ry: Rotation about the Y-axis. Imagine the part rotating (rolling left and right) about an axis passing through it. Allowable clamping force ranges (e.g., [200N, 800N]), spatial interference constraints (e.g., "clamp height < 150mm"), and interface constraints (e.g., "using ISO standard T-slot bolts") are also considered. For example, extract from a node of type "Engineering Constraint". For instance, a "Spatial Constraint" node might have the attribute {Direction: Z, Constraint Type: Maximum, Value: 150, Unit: mm}.

[0140] The heterogeneous features of the different modalities mentioned above are fused into a machine-comparable comprehensive representation. The aim is to address the problem that "a plate-shaped part milled on a large plane on a vertical machining center" and "a block-shaped part milled on a side on a horizontal machining center" are geometrically different, but their clamping semantics (both require a large-area plane for positioning and resistance to unidirectional cutting forces) may be highly similar.

[0141] Preset fusion methods (selectable):

[0142] 1. Feature concatenation followed by encoding: The feature list (such as text description or attribute set) parsed from each modality is simply concatenated into a long string or vector, and then an encoder model (such as BERT or Sentence Transformer) is used to map it into a fixed-dimensional vector.

[0143] 2. Weighted fusion: Train an encoder independently for each modality, map each modality's features into a vector, and then assign weights according to the importance of the modality and perform weighted summation to obtain a joint feature vector.

[0144] 3. Joint Encoding Network Based on Attention Mechanism: A neural network is constructed whose input is the features of each modality. The network dynamically learns the contribution of different modality features to the final similarity judgment through an attention mechanism and directly outputs a joint feature vector. This is the most complex but usually the most effective approach.

[0145] Regardless of the method used, a joint feature vector (e.g., a 384-dimensional array of floating-point numbers) is ultimately generated for the current design semantic graph. Each case in the historical case library also undergoes the exact same process, with its corresponding joint feature vector pre-calculated and stored.

[0146] In vector space, the similarity between the joint feature vector of the current task and all case vectors in the historical case library is calculated. Commonly used methods include cosine similarity (which measures the proximity of vector directions) or Euclidean distance (which measures the spatial distance between vector points).

[0147] Similar cases can be selected:

[0148] Thresholding method: Set a similarity threshold (e.g., 0.85) and return all cases with a similarity higher than this value.

[0149] Top-K method: Returns the top K (e.g., K=3) cases with the highest similarity.

[0150] Hybrid method: Returns the top K cases with similarity higher than a threshold.

[0151] For example, after calculation, the joint feature vector of a current task of "milling the top surface of an aluminum alloy shell" may have a cosine similarity of 0.92 with the vector of a case in the historical database of "milling the plane of a composite material cover plate". Although the materials are different, the two are highly consistent in core clamping semantics such as "positioning with a large plane", "constraining the vertical degree of freedom" and "resisting the vertical cutting force". Therefore, the historical case was successfully retrieved.

[0152] By using multimodal feature parsing and fusion, the "design semantic graph" describing design requirements is transformed into a "joint feature vector" that comprehensively represents its core clamping semantics. This enables historical case retrieval based on deep engineering semantics rather than surface geometry. This method can "learn by analogy," intelligently discovering commonalities in clamping principles among different parts and processes, expanding the scope of historical experience reuse. It effectively solves the pain points of traditional retrieval methods that rely on a single dimension (such as only based on 3D model shape matching), resulting in incomplete retrieval and omission of effective cases. This allows subsequent design reasoning to be based on the most relevant and reliable engineering experience, significantly improving the knowledge utilization efficiency and solution generation quality of the intelligent design system.

[0153] Optionally, after inputting the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme, the process further includes:

[0154] Based on the preset inspection method, the engineering feasibility of the tooling fixture design scheme is verified, and the verification results are obtained.

[0155] If the verification result is unsuccessful, verification feedback information containing the specific constraint violation type is generated and fed back to the preset large model to guide the correction of the tooling fixture design scheme and to re-execute the step of verifying the engineering feasibility of the tooling fixture design scheme until the tooling fixture design scheme meets the engineering constraints.

[0156] Optionally, the preset inspection methods include freedom constraint integrity verification, clamping mechanics feasibility verification, and machining space interference verification.

[0157] Specifically, after generating a preliminary design scheme from a large model, how can the feasibility of the scheme be ensured through automated engineering verification, forming an intelligent closed-loop optimization process of "generation-verification-correction"? This process aims to simulate the design review process of senior engineers, verifying the physical feasibility of the scheme from multiple dimensions.

[0158] Engineering feasibility verification based on preset verification methods:

[0159] Freedom constraint integrity verification: Verify whether the arrangement of positioning elements in the generated scheme fully constrains all degrees of freedom defined in the "clamping state model" and necessary for machining, and whether there is over-positioning (repeated constraint of the same degree of freedom leads to interference or uncertainty).

[0160] Verification process and optional methods:

[0161] 1. Establish a constraint mapping matrix: Describe the directions of freedom (transitions Tx / Ty / Tz along the X / Y / Z axes, rotations Rx / Ry / Rz about the X / Y / Z axes) of each positioning element (such as the constraint spiral matrix in spiral theory) in the design scheme using a mathematical model. Here, Rz represents rotation about the Z-axis. Imagine the part rotating about a vertical axis passing through it (rotation in the horizontal plane).

[0162] 2. Perform rank and linear correlation analysis: Calculate the joint rank of the constraint matrices of all positioning elements. If the joint rank equals the number of degrees of freedom of the required constraints (e.g., 5), and the properties of each constraint line are independent, the verification passes. If the rank is insufficient, under-positioning exists, and the part may move; if linear correlation exists, over-positioning exists, which may lead to clamping deformation or interference.

[0163] For example, in a milling operation that requires constraints on 5 degrees of freedom, if the design only includes three support pins on the bottom surface (constraints Tz, Rx, Ry) and one side stop (constraint Tx), then the constraint on Ty is missing, and the system will determine that "the degree of freedom constraint is incomplete: the constraint on the Y-direction translational degree of freedom is missing".

[0164] Feasibility verification of clamping mechanics: Verify whether the clamping force configuration of the scheme is sufficient to resist the cutting force, centrifugal force, etc. generated during processing, to ensure that the workpiece does not shift or vibrate during processing, and that the clamping force itself does not cause unacceptable deformation or surface damage to the workpiece.

[0165] Verification process and optional methods:

[0166] 1. Force Modeling and Calculation: Estimate the direction and magnitude of the maximum cutting force / torque based on process characteristic data (cutting parameters, material). Calculate the minimum clamping force required to prevent workpiece slippage based on the friction coefficient.

[0167] 2. Clamping force verification: Check whether the theoretical clamping force provided by the clamping mechanism (such as spiral pressure plate, hydraulic cylinder) selected in the design scheme is greater than the required minimum clamping force and less than the maximum allowable clamping force specified in the "Manufacturing Constraints" (to prevent deformation).

[0168] For critical or easily deformable parts, a simplified finite element analysis (FEA) simulation can be initiated, applying clamping and cutting forces as loads to the three-dimensional model of the part, calculating stress and deformation, and comparing them with allowable values.

[0169] For example: The design for the aluminum alloy thin-walled part has four pneumatic clamping plates. After system calculation, the feedback is: "Clamping force feasibility warning: The estimated main cutting force requires a minimum clamping force of 1200N. The current clamping force limit is 900N, which poses a risk of slippage. It is recommended to switch to hydraulic clamping or increase the number of clamping points."

[0170] Machining space interference verification: Verify whether there are static or dynamic collisions between the fixture entities (base plate, column, pressure plate, etc.) in the design scheme and the tool path, machine tool moving parts (spindle head, tool magazine), and non-machining parts of the workpiece during the entire machining process.

[0171] Verification process and optional methods:

[0172] 1. 3D geometry construction and Boolean operations: Based on the design scheme, the 3D model library of fixture components is generated or called, and assembled with the workpiece and tool model (toolpath envelope generated according to the process) into an assembly.

[0173] 2. Interference Check: Static interference checks are performed using the collision detection algorithm of the CAD kernel. For dynamic interference, the machine tool movement is simulated to check for collisions at key points along the toolpath.

[0174] For example: After the system performs a dynamic simulation, it reports: "Spatial interference conflict: When the tool moves along the negative Y-axis to a stroke of -120mm, the support column at the upper left corner of the fixture collides with the tool holder, with a minimum interval of -3.5mm."

[0175] The verification process does not simply output "pass" or "fail," but generates structured, actionable feedback. For example: {"Constraint violation type": "Spatial interference", "Violation object": "Support column_01 and tool holder", "Violation description": "Dynamic collision, insufficient safety distance", "Suggested correction direction": "Move column_01 at least 15mm in the positive X-axis direction, or reduce its height"}.

[0176] When the verification result is unsuccessful, the verification feedback information, along with the original design semantic diagram and case context, is re-input into the large model. The large model's prompt will be updated, for example, by adding: "Based on the following engineering verification feedback, please revise the solution while maintaining the original design intent: 1. Resolve the collision problem between the column and the tool..." The large model then performs further reasoning under the constraints to generate the revised solution.

[0177] The new solution will undergo the same engineering verification process again until all verification items pass, at which point the solution will be considered a mature design solution that can be delivered in the end.

[0178] By introducing automated, multi-dimensional (kinematic, mechanical, and geometric) engineering feasibility verification and a closed-loop iterative correction based on precise feedback, this solution successfully integrates the "creativity" of generative artificial intelligence with the rigorous "regularity" of engineering physics. It endows the intelligent design system with the ability to self-verify and optimize, ensuring that the final output solution is not only conceptually novel but also possesses high engineering reliability, safety, and manufacturability. This process enhances the practicality and feasibility of intelligent design for complex tooling fixtures.

[0179] Optionally, the manufacturing constraints include at least one of the following: machining accuracy requirements, clamping stiffness requirements, clamping time requirements, machine tool interface constraints, and fixture manufacturing cost constraints.

[0180] Specifically, in the task of tooling and fixture design, in addition to part geometry and machining technology, various boundary conditions derived from the actual production environment and business objectives must be considered. These "manufacturing constraints" constitute the hard boundaries and optimization objectives of the design scheme, ensuring that the generated fixture is not only technically feasible, but also economical, efficient, and adaptable to the production system.

[0181] Machining accuracy requirements: These refer to the dimensional, shape, positional, or surface quality tolerances that the workpiece must meet during this process. These requirements directly determine the positioning accuracy and rigidity required by the fixture.

[0182] For example, flatness ≤ 0.05 mm, hole position tolerance Φ0.1 mm, and surface roughness Ra ≤ 1.6 μm. This means that the accuracy of the positioning elements of the fixture, the repeatability of the fixture, and the overall rigidity must be sufficient to ensure that the workpiece meets these specifications after machining.

[0183] Clamping stiffness requirement: This refers to the maximum allowable elastic deformation of the fixture-workpiece system under a given clamping force and machining load. This requirement is used to prevent machining errors or vibrations caused by system deformation.

[0184] For example, under maximum cutting force, the overall deformation of the workpiece at the tool contact point should be less than 0.02 mm. This constraint will affect the selection of the fixture base material, the arrangement of structural stiffeners, and the distribution of support points.

[0185] Clamping time requirement: This refers to the maximum allowable time to complete one clamping and unclamping operation of a workpiece, and is usually linked to the production cycle time. This requirement affects the complexity and degree of automation of fixture operations.

[0186] For example, single-piece clamping time ≤ 90 seconds. This may guide the system to prioritize quick-change clamping mechanisms (such as pneumatic or hydraulic), modular positioning elements, and optimize the ergonomics of clamping actions.

[0187] Machine tool interface constraints refer to the compatibility requirements between the fixture and the target machine tool in terms of physical connection, size, and control.

[0188] For example: Physical interface: The fixture base plate must match the T-slot specifications of the machine tool table (e.g., slot width 18mm, spacing 100mm). Dimensional constraints: The total height of the fixture (including the workpiece) must not exceed the minimum distance from the machine tool spindle nose to the table. Control interface: If a hydraulic fixture is used, the machine tool must have a reserved hydraulic interface and control signal points.

[0189] The fixture manufacturing cost constraint refers to the upper limit of the budget for resources such as funds, materials, and labor invested in manufacturing the fixture. This is a key constraint that links technical solutions with economic viability.

[0190] For example, the total manufacturing cost of the fixture (including materials, processing, and purchased parts) should be controlled within ¥8,000. This constraint will affect the trade-offs between the ratio of standard parts to customized parts, material grades (e.g., 45 steel vs. alloy steel), and the complexity of the processing technology during the solution generation stage.

[0191] In step S100 (Design Information Acquisition), these constraints can be entered through interactive forms, configuration files, or integration with a Product Lifecycle Management (PLM) / Enterprise Resource Planning (ERP) system. In subsequent steps, they will be incorporated into the design semantic graph as "engineering constraint data" and ultimately become the basis for engineering verification (such as stiffness verification and interference verification) and scheme optimization (such as cost estimation).

[0192] By explicitly incorporating and structuring the aforementioned multi-dimensional manufacturing constraints, this method tightly anchors the AI-driven design process to real-world engineering and production contexts. It not only answers the question "How to design a usable fixture?" but also "How to design an optimal or feasible fixture that is within a specific cost budget, adaptable to given machine tools, and meets production cycle and quality requirements?" This significantly improves the engineering practicality, economy, and feasibility of the generated solutions, transforming intelligent design from an idealized concept generation into a reliable tool that directly supports production decisions, effectively bridging the gap between innovative design and manufacturing practice.

[0193] In some specific embodiments, such as Figure 2 As shown, taking the aluminum alloy box-shaped part as an example, the specific implementation process of the artificial intelligence-based tooling and fixture design method includes:

[0194] Step S1: Obtaining Design Information

[0195] Obtain the 3D CAD model data (3D model data) of the aluminum alloy box part. Its material is 6061-T6, and its dimensions are approximately 220 mm × 160 mm × 90 mm. Simultaneously, obtain its machining process information: perform precision milling on its right outer wall on a vertical CNC milling machine to form an assembly reference surface, requiring a flatness of ≤0.02 mm after machining. Manufacturing constraints are also specified, such as machining accuracy requirements (flatness 0.02 mm) and equipment space limitations (vertical milling machine table dimensions).

[0196] Step S2: Clamping state modeling:

[0197] Geometric analysis of the 3D model of the part: identify its bottom surface (already machined) as the main positioning reference surface; identify the through-hole structure on the side of the part as potential auxiliary positioning features; mark the right outer wall to be machined as the machining area, and delineate it as a prohibited clamping area because it is a machining surface; at the same time, identify non-machined areas such as the top of the part as clampable areas.

[0198] Based on process information (vertical milling machine, finish milling of the right side), the system analysis determines that: the part is clamped with its bottom face down; the machining direction is horizontal (X or Y axis); and the main cutting force acts horizontally on the right side wall. Based on the "six-point positioning principle," the degrees of freedom that need to be constrained include: movement along the vertical direction (Z-axis), movement along the two horizontal directions (X and Y axes), and rotation about the vertical axis (Z-axis). Therefore, the system establishes a structured clamping state model, the core contents of which include: the set of constrained degrees of freedom (e.g., {Tz, Tx, Ty, Rz}), the corresponding positioning features (bottom surface, side hole), the main force direction (horizontal), and the prohibited clamping area (right side wall).

[0199] Step S3: Design and construct the semantic graph:

[0200] By integrating the clamping state model with manufacturing constraints, a design semantic graph is constructed. Specifically:

[0201] Entities such as "part bottom surface", "lateral through hole" and "top clamping area" are abstracted into semantic nodes and assigned attributes (such as position coordinates and normal vector).

[0202] The logical relationships such as "bottom surface constraint Z-axis movement", "side hole constraint X, Y-axis movement and Rz rotation", "top area can be used for clamping", and "right side wall clamping is prohibited" are abstracted into semantic edges of connecting nodes and assigned attributes (such as constraint type and direction).

[0203] Manufacturing constraints (such as flatness of 0.02 mm) are also added to the graph as constraint nodes or edge attributes.

[0204] Ultimately, a design semantic graph, represented by a graph data structure, is generated, which uniformly and formally expresses all design requirements and engineering constraints.

[0205] Step S4: Historical Case Retrieval and Knowledge Extraction:

[0206] The design semantic graph of the current task is subjected to multimodal feature parsing (geometry, process, constraints), and transformed into a joint feature vector through joint encoding. Subsequently, the similarity between this vector and all historical case vectors is calculated in the historical fixture case library. Based on similarity matching, the system retrieves historical successful cases that are highly similar in clamping semantics such as "box-type parts", "side wall milling", and "one-face-one-pin positioning". These similar historical successful cases are further analyzed to extract reusable positioning-clamping structure patterns (e.g., the general pattern of "using bottom surface main positioning + side pin auxiliary positioning + top pressure plate clamping") and configuration rules, forming structured case context information.

[0207] Step S5: Retrieval Enhanced Reasoning and Solution Generation:

[0208] The design semantic map of the current task and the extracted case context are input into a pre-defined large model. While understanding the specific constraints (such as bottom datum, side hole positions, and prohibited clamping areas), the large model also draws on successful patterns from historical cases (bottom positioning + side pin positioning + top clamping plate) for fusion reasoning. Finally, the large model generates a specific fixture design description, such as: "A three-point support structure is used to position the bottom surface of the part to constrain the Z-axis degree of freedom; a cylindrical locating pin is set at the identified side through-hole position to constrain X and Y-axis movement and rotation around the Z-axis; two spiral clamping plate mechanisms are arranged in the non-machined area on the top of the part to provide a stable vertical downward clamping force and completely avoid the right-side machining area."

[0209] Step S6: Project feasibility verification and iterative correction:

[0210] Perform automated engineering verification on the generated preliminary plan:

[0211] Freedom constraint integrity verification: Verify whether the combination of "three-point support + one cylindrical pin" completely and without over-positioning constrains the required degrees of freedom {Tz, Tx, Ty, Rz} of the model.

[0212] Feasibility verification of clamping mechanics: Estimate the horizontal cutting force and verify whether the frictional force generated by the clamping force provided by the top pressure plate is sufficient to resist the cutting force and prevent the workpiece from sliding horizontally.

[0213] Machining space interference verification: Using a 3D model simulation, check whether the clamping components such as the pressure plate and support column collide with the movement trajectory of the end mill when machining the right side wall.

[0214] If any verification fails (for example, if the pressure plate is found to potentially interfere with the tool holder at a certain position), the system will generate specific verification feedback information (such as "the gap between pressure plate A and the tool holder is less than 5mm at the end of its travel in the negative Y-axis direction"), and return this feedback to the large model to guide it to make targeted corrections to the solution (such as adjusting the installation position or shape of the pressure plate). This "generate-verify-correct" cycle continues until the solution passes all verifications.

[0215] Step S7: Solution Output:

[0216] Once the fixture design scheme has passed all engineering feasibility checks, the final tooling fixture design scheme with engineering reliability is output. This scheme is described in a structured form, including a clear fixture structure, the type and location of positioning and clamping elements, key dimensional parameters, and other information, which can be directly used for subsequent detailed design, modeling, and manufacturing.

[0217] like Figure 3 As shown in the figure, an artificial intelligence-based tooling fixture design system provided by an embodiment of the present invention includes:

[0218] The acquisition unit is used to acquire the 3D model data, machining process information, and manufacturing constraints of the part to be processed.

[0219] A construction unit is used to construct a clamping state model that characterizes the clamping requirements of the part to be processed during the processing, based on the three-dimensional model data and the processing technology information.

[0220] The fusion unit is used to fuse the clamping state model with the manufacturing constraints to obtain a design semantic graph;

[0221] The processing unit is used to retrieve similar historical fixture cases from the historical fixture case library based on the design semantic graph, and extract reusable constraint patterns and configuration rules from the retrieved historical fixture cases to form a structured case context.

[0222] The processing unit is also used to input the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme for the part to be processed. The tooling fixture design scheme is a structured scheme that includes fixture structure, positioning scheme, clamping configuration and key parameters.

[0223] This invention provides an artificial intelligence-based tooling fixture design device, comprising a memory and a processor; the memory is used to store computer programs; the processor is used to implement the artificial intelligence-based tooling fixture design method as described above when the computer programs are executed.

[0224] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the artificial intelligence-based tooling and fixture design method described above.

[0225] 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 tooling fixture design method based on artificial intelligence, characterized in that, include: Obtain the 3D model data, machining process information, and manufacturing constraints of the part to be processed; Based on the three-dimensional model data and the processing technology information, a clamping state model is constructed to characterize the clamping requirements of the part to be processed during the processing. The clamping state model is fused with the manufacturing constraints to obtain a design semantic graph; Based on the design semantic graph, similar historical fixture cases are retrieved from the historical fixture case library, and reusable constraint patterns and configuration rules are extracted from the retrieved historical fixture cases to form a structured case context. The design semantic graph and the case context are input into a preset large model to generate a tooling fixture design scheme for the part to be processed. The tooling fixture design scheme is a structured scheme that includes fixture structure, positioning scheme, clamping configuration and key parameters.

2. The tooling and fixture design method based on artificial intelligence according to claim 1, characterized in that, The step of constructing a clamping state model based on the three-dimensional model data and the machining process information to characterize the clamping requirements of the part to be processed during the machining process includes: Geometric analysis is performed on the three-dimensional model data to obtain geometric features, which include a positioning reference plane, a clamping area, and a key force-bearing area. Based on the processing technology information, determine the process characteristic parameters and constraint requirements; The geometric features, process feature parameters, and constraint requirements are modeled in a unified manner to obtain a structured clamping state model.

3. The tooling and fixture design method based on artificial intelligence according to claim 1, characterized in that, The process of fusing the clamping state model with the manufacturing constraints to obtain a design semantic map includes: Extract part structural feature data and process feature data from the clamping state model, and extract engineering constraint data from the manufacturing constraints; Based on the clamping logic information in the clamping state model and the engineering constraint information in the manufacturing constraints, the positioning relationship, clamping relationship and constraint relationship are determined. The part structural feature data, the process feature data, and the engineering constraint data are used as semantic nodes, and the positioning relationship, the clamping relationship, and the constraint relationship are used as semantic edges. Based on the semantic nodes and semantic edges, the design semantic graph is constructed.

4. The tooling and fixture design method based on artificial intelligence according to claim 1, characterized in that, The step of retrieving similar historical fixture cases from the historical fixture case library based on the design semantic graph includes: The design semantic graph is subjected to multimodal feature analysis to obtain geometric modal features, process modal features and engineering constraint modal features; The geometric modal features, the process modal features, and the engineering constraint modal features are fused using a preset fusion method to obtain a joint feature vector; The joint feature vector is matched with the corresponding feature vector of each case in the historical fixture case library for similarity, and one or more similar historical fixture cases are determined based on the similarity matching results.

5. The tooling and fixture design method based on artificial intelligence according to claim 4, characterized in that, The geometric modal features include positioning reference plane, clamping area, hole features, and key stress area information; the process modal features include processing method, processing direction, number of clamping operations, and machine tool type information; the engineering constraint modal features include degree-of-freedom constraint set, allowable clamping force range, spatial interference restrictions, and interface constraint information.

6. The tooling and fixture design method based on artificial intelligence according to claim 1, characterized in that, After inputting the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme, the process further includes: Based on the preset inspection method, the engineering feasibility of the tooling fixture design scheme is verified, and the verification results are obtained. If the verification result is unsuccessful, verification feedback information containing the specific constraint violation type is generated and fed back to the preset large model to guide the correction of the tooling fixture design scheme and to re-execute the step of verifying the engineering feasibility of the tooling fixture design scheme until the tooling fixture design scheme meets the engineering constraints.

7. The tooling and fixture design method based on artificial intelligence according to claim 6, characterized in that, The preset inspection methods include freedom constraint integrity verification, clamping mechanics feasibility verification, and machining space interference verification.

8. The tooling and fixture design method based on artificial intelligence according to claim 1, characterized in that, The manufacturing constraints include at least one of the following: machining accuracy requirements, clamping stiffness requirements, clamping time requirements, machine tool interface constraints, and fixture manufacturing cost constraints.

9. A tooling and fixture design system based on artificial intelligence, characterized in that, include: The acquisition unit is used to acquire the 3D model data, machining process information, and manufacturing constraints of the part to be processed. A construction unit is used to construct a clamping state model that characterizes the clamping requirements of the part to be processed during the processing, based on the three-dimensional model data and the processing technology information. The fusion unit is used to fuse the clamping state model with the manufacturing constraints to obtain a design semantic graph; The processing unit is used to retrieve similar historical fixture cases from the historical fixture case library based on the design semantic graph, and extract reusable constraint patterns and configuration rules from the retrieved historical fixture cases to form a structured case context. The processing unit is also used to input the design semantic graph and the case context into a preset large model to generate a tooling fixture design scheme for the part to be processed. The tooling fixture design scheme is a structured scheme that includes fixture structure, positioning scheme, clamping configuration and key parameters.

10. A tooling fixture design device based on artificial intelligence, 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 tooling and fixture design method based on artificial intelligence as described in any one of claims 1 to 8.

Images