A method for designing vacuum isothermal forging dies and cavity design methods

Abstract

This invention relates to the field of forging die design technology, and discloses a vacuum isothermal forging die design method and cavity design method, including: acquiring a target part drawing, identifying and deleting machining features, repairing and adding forging allowance to obtain a final forging drawing; using the geometric features of the final forging drawing as cavity parameters of the final forging die, matching the forging process scheme, and reverse-deriving the preceding forging drawing, using its geometric features as the corresponding die cavity parameters. This invention uses the target part drawing in the provided design drawings to reverse-derive the final forging drawing based on machining characteristics, uses the geometric features of the final forging drawing as the cavity design parameters of the final forging die, selects forging process parameters from a forging process database to match the forging process scheme of the final forging drawing, reverse-derives the preceding forging drawing before the final forging process, and uses the geometric features of the preceding forging drawing as the cavity design parameters of the preceding forging die, thus realizing die cavity design for the entire process.

Description

Technical Field

[0001] This invention relates to the field of forging die design technology, specifically to a vacuum isothermal forging die design method and cavity design method. Background Technology

[0002] Vacuum isothermal forging is a core forming process for key hot-end components such as powder metallurgy high-temperature alloy discs. It places extremely high demands on the precision of mold cavity design, the transition shape of each process, and the temperature-stress-structure coupling evolution during the forming process.

[0003] Mold design is highly dependent on engineers' experience, has a long development cycle, high costs, and makes it difficult to ensure the simultaneous optimization of multiple objectives (filling degree, load, wear, stress, material utilization, etc.). Existing mold design typically falls into two categories: One approach is the traditional "experience-driven + manual trial molding" method. Designers use a 3D model of the part to deduce the final forging shape based on empirical values, judge the process plan based on experience, and then CAD engineers manually model the part. Single-process simulations are then performed using CAE software for verification. If the specifications are not met, dimensions are manually modified and iteratively repeated. This method heavily relies on the experience of senior engineers, resulting in design cycles that can last for weeks or even months. The designs for each process are fragmented, easily leading to local optima but overall suboptimal results. Multi-objective trade-offs are based entirely on intuition, lacking quantitative basis. Repetitive manual operations are performed before and after simulation, resulting in low efficiency and a high risk of errors. Existing solutions cannot be directly reused for modified parts.

[0004] The second approach is a semi-automated mold optimization method based on secondary development of commercial CAE software. This method parameterizes the key dimensions of the mold, uses optimization platforms such as Isight to construct a surrogate model, and employs a multi-objective evolutionary algorithm for optimization. However, this method still requires manual completion of geometric feature recognition, process scheme decision-making, and CAD parametric modeling, making it difficult to achieve end-to-end automation from part to mold; the number of parameterized variables is fixed, resulting in poor adaptability to new structural parts, requiring a new template to be built for each new part; the evolutionary algorithm converges slowly, making the computational cost prohibitive for expensive 3D full-process simulation; it lacks formal utilization of engineering experience and "soft knowledge" from professional books; and the automation chain is incomplete, often requiring manual coordination between preprocessing scripts, solver calls, and post-processing data extraction.

[0005] In summary, existing technologies, when dealing with the design of vacuum isothermal forging die cavities, lack a design process from finished parts to the die cavities of each process, resulting in the need for multiple manual trial runs to complete the die design. Summary of the Invention

[0006] The purpose of this invention is to provide a vacuum isothermal forging die design method and cavity design method to solve the technical problem in the prior art that the lack of a design process from finished parts to the mold cavities of each process leads to the need for multiple manual trial moldings to complete the die design.

[0007] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: A method for designing the cavity of a vacuum isothermal forging die includes the following steps: Step 100: Based on the product design drawings, obtain the target part drawing formed by die forging and then machining from the design drawings; Step 200: Based on the machining characteristics of the target part drawing, identify the machining features formed by machining on the target part drawing; Step 300: Delete the machining features in the target part drawing, repair the missing features and inherent features of the forging formed after deletion, and add forging allowance for machining to obtain the final forging drawing. Step 400: Extract the geometric features of the final forging drawing and use the geometric features as the cavity design parameters of the final forging die; Step 500: Based on the geometric features of the final forging drawing, match it with the forging process parameters in the forging process database to obtain the forging process scheme corresponding to the final forging drawing. Step 600: Based on the forging process parameters in the forging process plan, the forging process drawing is reverse-engineered from the final forging drawing to obtain the preliminary forging drawing before the final forging process. Step 700: Use the geometric features of the preceding forging drawing as the cavity design parameters for the corresponding forging die.

[0008] Further, in step 200, the method for identifying machining features formed by machining on the target part drawing includes: Based on the target part drawing, extract all surface regions that can be processed independently as candidate feature surfaces; The feature indicators of each candidate feature surface are quantified, and the feature indicators include at least the local radius of curvature, the number of adjacent surfaces, the maximum feature coverage ratio, the metal flow angle, and the minimum thickness. If all the aforementioned feature indicators meet the forging requirements, then it is marked as a qualified candidate feature surface; otherwise, it is marked as a candidate feature surface to be determined. Based on the feature indicators of the candidate feature surfaces to be determined, they are matched with the machining parameters in the machining database to obtain the machining determination result of the candidate feature surfaces to be determined; The local radius of curvature is the radius of curvature value along the direction of minimum principal curvature on the candidate feature surface, which is used to characterize the degree of curvature of the surface region; The number of adjacent faces is the number of faces that are directly adjacent to the candidate feature face; The large feature coverage ratio is the ratio of the fitted area of ​​the basic geometric surface to the total area of ​​the candidate feature surface; wherein, the basic geometric surface includes at least a plane, a cylindrical surface, and a conical surface; The metal flow angle is the angle between the metal flow direction and the mold opening direction along the main forging direction. The minimum thickness is the thickness value at the thinnest point of the candidate feature surface.

[0009] Furthermore, in step 300, the inherent features of the forging include at least a draft angle.

[0010] Furthermore, in step 300, the method for repairing the missing features and inherent features of the forging formed after deletion, and supplementing the forging allowance for machining, includes: Based on the target part drawing, suppress the feature tree nodes corresponding to the missing features and the inherent features of the forging; The retained surfaces adjacent to the suppressed area in the target part drawing are extended outward along their respective normal directions until the extended curved surfaces completely cover the gap left by the suppressed area, forming a transition area. For the transition region formed after extension, a C1 continuous smooth fit is performed using a cubic B-spline surface to form an image of the repaired forging; Based on the image of the repaired forging, and according to the functional requirements of different parts after repair, the basic value of the machining allowance is set; Based on the formula: final allowance value = base value × material coefficient × size coefficient, the machining allowance of each part of the repair forging image is calculated to obtain the final forging drawing.

[0011] Furthermore, it also includes a method for feature extraction of the forging process scheme corresponding to the final forging drawing: Extract all geometric features of the final forging drawing, identify forging features among all geometric features based on the die forging characteristics, and count the number N of the forging features; When N≤6, the forging process is determined to be: billet preparation-final forging process. The preceding forging drawing corresponds to the billet preparation drawing, and the geometric features of the billet preparation drawing are used as the cavity design parameters of the billet preparation mold. When N > 6, the forging process is: billet preparation - pre-forging - final forging process. The preceding forging drawing corresponds to the pre-forging drawing, and the geometric features of the pre-forging drawing are used as the cavity design parameters of the pre-forging die. Furthermore, based on the geometric features of the pre-forged drawing, it is matched with the forging process parameters in the forging process database to obtain the forging process scheme corresponding to the pre-forged drawing; By reverse engineering the pre-forging drawing, a blanking drawing is obtained before the pre-forging process, and the geometric features of the blanking drawing are used as the cavity design parameters of the blanking mold.

[0012] Furthermore, in step 500, the forging process parameters include at least the geometric parameters of the forging surface, forging flash, and draft angle.

[0013] Furthermore, in step 600, the method for reverse engineering the final forging drawing includes: When the forging process scheme is a billet-final forging process, the billet drawing is directly derived from the final forging drawing; When the forging process scheme is billet preparation-pre-forging-final forging process, the pre-forging drawing is derived from the final forging drawing, and then the billet preparation drawing is derived from the pre-forging drawing; During the reverse calculation process, the rate of change of cross-sectional area at the corresponding positions of adjacent processes is less than or equal to 25%.

[0014] Furthermore, after step 700, the following steps are also included: Step 800: Based on the design parameters of the blanking mold cavity, the design parameters of the pre-forging mold cavity, and the design parameters of the final forging mold cavity, design the blanking mold, the pre-forging mold, and the final forging mold respectively, and perform full-process simulation according to the forging process scheme to obtain multiple physical parameters of the three molds after simulation. Based on multiple physical parameters after simulation of the three molds, a weighted normalized multi-objective function is constructed, and a gradient optimization algorithm is used to iteratively optimize the design parameters of the three mold cavities. The design parameters after each iteration are then fed back to step 600. Step 900, repeat steps 600 to 800 until the multi-objective function meets the convergence condition, and output the optimal mold cavity design parameters for each process as the optimal design parameters for the corresponding mold cavity. The convergence condition is that the gradient norm of the objective function is less than a preset first threshold, or the difference between two consecutive objective function values ​​is less than a preset second threshold, or the number of iterations reaches a preset upper limit, or the binary penalty terms are all zero and the continuous objective improvement is less than a preset third threshold.

[0015] Furthermore, in step 800, the method of constructing a weighted normalized multi-objective function and performing iterative optimization using a gradient optimization algorithm includes: Extract multiple physical parameters from each mold simulation, including at least forging filling degree, maximum forming load, maximum equivalent stress of mold, mold wear, and material utilization rate. Each physical parameter is normalized to map it to a value range of 0 to 1. Assign a weight coefficient to each objective, and the sum of all weight coefficients is 1; Multiply each normalized physical parameter by its corresponding weight coefficient and sum them to construct a multi-objective function; The gradient optimization algorithm is used to calculate the gradient of the multi-objective function with respect to the mold cavity design parameters, and the design parameters are updated along the gradient descent direction; The updated design parameters were verified by full-process simulation, and the physical parameters were re-extracted and new multi-objective function values ​​were calculated. Repeat the process of updating design parameters and extracting physical parameters until the convergence condition is met, and output the current optimal mold cavity design parameters.

[0016] A vacuum isothermal forging die cavity design method based on the above-mentioned vacuum isothermal forging die cavity design method includes, in step 900, a specific method for designing the die according to the optimal design parameters of the die cavity, including: Based on the billet drawing, the pre-forging drawing, and the final forging drawing, the demolding angle is set respectively; Separate the upper and lower halves along the neutral planes of the billet drawing, the pre-forging drawing, and the final forging drawing, and extend them outward to generate the billet mold body, the pre-forging mold body, and the final forging mold body, respectively. Bridges, slots, flash grooves, and positioning structures are added to the outside of the blanking mold cavity, the pre-forging mold cavity, and the final forging mold cavity, respectively, to obtain the blanking mold drawing, the pre-forging mold drawing, and the final forging mold drawing, respectively. The geometric features of the billet forming die drawing, the pre-forging die drawing, and the final forging die drawing are used as the design parameters for the billet forming die, the pre-forging die, and the final forging die, respectively.

[0017] Compared with the prior art, the present invention has the following advantages: This invention uses the target part drawing in the provided design drawings to reverse-engineer the final forging drawing based on the machining characteristics. The geometric features of the final forging drawing are used as the cavity design parameters of the final forging die. Furthermore, a forging process scheme matching the forging process parameters of the final forging drawing is selected from the forging process database. The preceding forging drawing before the final forging process is reverse-engineered, and the geometric features of the preceding forging drawing are used as the cavity design parameters of the preceding forging die, thus realizing the die cavity design for the entire process. Attached Figure Description

[0018] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the mold design method provided in this application embodiment.

[0020] Figure 2 A flowchart illustrating the design method executed by the mold design system, provided for an embodiment of this application.

[0021] Figure 3 This is a schematic diagram illustrating the parsing of a 3D input workpiece geometry file provided in an embodiment of this application.

[0022] Figure 4 This is a schematic diagram illustrating the parsing of a 2D input workpiece geometry file provided in an embodiment of this application.

[0023] Figure 5 This is a shape drawing of a machined product provided in an embodiment of this application.

[0024] Figure 6 This is a flowchart of the geometry recognition and processing sub-process of the design system provided in the embodiments of this application.

[0025] Figure 7 This is a feature structure location identification diagram of a machined workpiece provided in an embodiment of this application.

[0026] Figure 8 This is a schematic diagram of the geometric features of the final forging provided in an embodiment of this application.

[0027] Figure 9 This is a flowchart for determining the process plan in an embodiment of this application.

[0028] Figure 10 Comparison of forging morphology before and after billet preparation, provided for embodiments of this application.

[0029] Figure 11 The above and below mold structure design drawings are provided for the blanking process in the embodiments of this application.

[0030] Figure 12 The flowchart of the Deform dual-channel drive sub-module of the full-process simulation module of the design system provided in this application embodiment.

[0031] Figure 13 This is a schematic diagram of the full-process simulation model of Deform provided in the embodiments of this application.

[0032] Figure 14 A flowchart of the SolidWorks+Deform multi-objective optimization iteration process for the multi-objective optimization module provided in the embodiments of this application.

[0033] Figure 15 A block diagram illustrating the composition of the design system provided in the embodiments of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] like Figure 1 , Figure 14 As shown, the present invention provides a vacuum isothermal forging die design method and cavity design method, which are executed by a vacuum isothermal forging die design system.

[0036] like Figure 1 As shown, the cavity design method includes the following steps: Step 100: Based on the product design drawings, obtain the target part drawing formed by die forging and then machining from the design drawings. The target part drawing is a three-dimensional model of the final machined product, such as the final geometric drawing of a powder metallurgy turbine disk after machining.

[0037] Step 200: Based on the machining characteristics of the target part drawing, identify the machining features formed by machining on the target part drawing. Machining characteristics refer to the geometric shapes formed on the surface by processes such as cutting, drilling, and milling.

[0038] Step 300: Delete the machining features in the target part drawing, repair the missing features and inherent features of the forging after deletion, and add forging allowance for machining to obtain the final forging drawing. Inherent features of the forging include draft angles, and forging allowance is a metal layer reserved for subsequent cutting.

[0039] Step 400: Extract the geometric features of the final forging drawing and use these features as the cavity design parameters for the final forging die. The cavity of the final forging die is consistent with the outer contour of the final forging drawing, directly determining the shape of the final forging.

[0040] Step 500: Based on the geometric features of the final forging drawing, match it with the forging process parameters in the forging process database to obtain the forging process scheme corresponding to the final forging drawing. The forging process database stores the process tree of typical parts, such as billet preparation-pre-forging-final forging or billet preparation-final forging.

[0041] Step 600: Based on the forging process parameters in the forging process plan, the forging drawing before the final forging process is derived by reverse engineering from the final forging drawing. The reverse engineering is carried out by progressively deducing the blank shape of the previous process according to the principle of constant volume.

[0042] Step 700: Use the geometric features of the preceding forging drawing as the cavity design parameters for the corresponding forging die.

[0043] In this method, the final forging drawing is derived by reverse engineering based on the machining characteristics of the target part drawing in the provided design drawings. The geometric features of the final forging drawing are used as the cavity design parameters of the final forging die. Furthermore, the forging process parameters are selected from the forging process database to match the forging process scheme of the final forging drawing. The preceding forging drawing before the final forging process is derived by reverse engineering. The geometric features of the preceding forging drawing are used as the cavity design parameters of the preceding forging die, thereby realizing the die cavity design of the entire process.

[0044] For this method, such as Figure 15 As shown, to facilitate the implementation of this method, the present invention also provides a design system for vacuum isothermal forging dies. This system includes: an AI agent (with a built-in large language model and communication connections to a knowledge base and database), an input module, a geometry recognition module, a process judgment module, a die design module, and a simulation and optimization module. The AI ​​agent acts as the scheduling core, achieving decoupled transmission between modules through parameter files, geometry files, and result files. The knowledge base (forging process database) stores forging knowledge and historical process data; the database (quadruple database) stores the design parameters, geometry files, simulation results, and optimization target values ​​for each design task, archived in quadruple form (design parameters, geometry files, simulation input / output, result indicators) for knowledge migration in new projects.

[0045] This method reverse-engineers the final part back into a forging blank, automatically removing machining features and adding allowances, avoiding manual mold repair. By matching geometric features with a database, the process plan is determined, making decisions quantifiable and reproducible. The entire reverse engineering process requires no manual intervention, significantly shortening the design cycle and eliminating instability caused by experience differences.

[0046] Furthermore, the steps of this method are coordinated and completed by various functional modules under the unified scheduling of an AI agent. The following uses a powder metallurgy high-temperature alloy turbine disk as an example, along with accompanying drawings, to provide a detailed explanation of the specific implementation of each step.

[0047] Example 1: Steps 100 to 200 are completed by the AI ​​agent in conjunction with the input module and the geometric recognition module.

[0048] Step 100: Based on the product design drawings, obtain the target part drawing formed by die forging and then machining from the design drawings.

[0049] In this step, such as Figure 3 , Figure 4As shown, the AI ​​agent receives two types of input files: 3D solid input in the native SolidWorks format (SLDPRT / SLDASM, including a complete feature tree, feature parameters, and sketch constraints); and 2D contour input in DXF format (including front view, sectional view, and key dimension annotations). The AI ​​agent reads the 3D feature tree through the SolidWorks COM interface and the 2D dimension chain through the DXF parsing library, and cross-validates the two to ensure recognition reliability.

[0050] Step 200: Based on the machining characteristics of the target part drawing, identify the machining features formed by machining on the target part drawing. The method for identifying the machining features formed by machining on the target part drawing includes: Based on the target part drawing, extract all surface regions that can be processed independently as candidate feature surfaces; The feature indicators of each candidate feature surface are quantified, and the feature indicators include at least the local radius of curvature, the number of adjacent surfaces, the maximum feature coverage ratio, the metal flow angle, and the minimum thickness. If all the aforementioned feature indicators meet the forging requirements, then it is marked as a qualified candidate feature surface; otherwise, it is marked as a candidate feature surface to be determined. Based on the feature indicators of the candidate feature surfaces to be determined, they are matched with the machining parameters in the machining database to obtain the machining determination result of the candidate feature surfaces to be determined; The local radius of curvature is the radius of curvature value along the direction of minimum principal curvature on the candidate feature surface, which is used to characterize the degree of curvature of the surface region; The number of adjacent faces is the number of faces that are directly adjacent to the candidate feature face; The large feature coverage ratio is the ratio of the fitted area of ​​the basic geometric surface to the total area of ​​the candidate feature surface; wherein, the basic geometric surface includes at least a plane, a cylindrical surface, and a conical surface; The metal flow angle is the angle between the metal flow direction and the mold opening direction along the main forging direction. The minimum thickness is the thickness value at the thinnest point of the candidate feature surface.

[0051] In this step, such as Figure 5 As shown, for disc-shaped parts with central symmetry characteristics, the AI ​​agent first performs a rotational scan along the main axis to determine the overall symmetry; if the structure is centrally symmetric but has asymmetrical features in some areas (typically, formed parts with blades), the area where the blades are located is identified as a "machining area" to be deleted; the central stepped holes / countersunk holes are uniformly changed to central through holes; the outer edge teeth, pin holes, oil passage holes, machining bosses, etc. are deleted and the surfaces are smoothed according to the "central symmetry equivalent surface".

[0052] like Figure 6 As shown, the machining feature recognition adopts a dual-channel strategy of "geometric heuristic + deep learning semantic classification".

[0053] The specific implementation of the dual-channel strategy is as follows: 1. Geometric Heuristic Channel: Each candidate feature surface is screened based on the 5 quantitative indicators in the table below.

[0054]

[0055] If all five indicators meet the "matrixability threshold", the item is considered retained; otherwise, proceed to the next step of semantic classification.

[0056] 2. Deep Learning Channel: The candidate regions are input into the pre-trained PointNet / Graph-CNN model in the form of B-Rep face / edge / vertex topology and corresponding surface sampling point cloud. The model is pre-trained on historical labeled data of about 2,000 powder high-temperature alloy discs and outputs the probability distribution of four types of labels: "retain / delete / supplement / add surplus".

[0057] 3. Dual-channel result adjudication: If the results from the two channels are consistent, they will be adopted directly; if they are inconsistent, the geometric heuristic result will prevail and the face will be marked with a "manual review mark". The subsequent module will provide a prompt when outputting the parameter table.

[0058] like Figure 7 As shown, the blade profile, central stepped hole, and tooth / small fillet area circled in red are all identified as machining areas and need to be deleted.

[0059] Example 2: Step 300 is completed by the AI ​​agent in conjunction with the geometric recognition module.

[0060] Step 300: Delete the machining features in the target part drawing, repair the missing features and inherent features of the forging after deletion, and add forging allowance for machining to obtain the final forging drawing. The inherent features of the forging include at least the draft angle. During the repair process, it is necessary to ensure that the draft angle area is not damaged.

[0061] The methods for repairing missing features and inherent features of forgings resulting from deletion, and for supplementing forging allowances for machining, include: Based on the target part drawing, suppress the feature tree nodes corresponding to the missing features and the inherent features of the forging; The retained surfaces adjacent to the suppressed area in the target part drawing are extended outward along their respective normal directions until the extended curved surfaces completely cover the gap left by the suppressed area, forming a transition area. For the transition region formed after extension, a C1 continuous smooth fit is performed using a cubic B-spline surface to form an image of the repaired forging; Based on the image of the repaired forging, and according to the functional requirements of different parts after repair, the basic value of the machining allowance is set; Based on the formula: final allowance value = base value × material coefficient × size coefficient, the machining allowance of each part of the repair forging image is calculated to obtain the final forging drawing.

[0062] In this step, the AI ​​agent removes feature structures from the workpiece image and fills in missing areas.

[0063] A three-step method of "feature volume suppression + adjacent surface extension + local B-spline fitting" is adopted.

[0064] The first step is to suppress the feature tree node corresponding to this feature in SolidWorks via API; The second step is to extend the adjacent preserved surfaces along their normal direction until they cover the suppressed area. The third step is to use a cubic B-spline to perform C1 continuous smoothing on the extended transition region to ensure the curvature continuity with the surrounding preserved surfaces.

[0065] Furthermore, in this step, the AI ​​agent dynamically determines the additional forging allowance.

[0066] The dynamic determination of forging allowance is divided into three steps: Step 1: Define the basic range according to the part category: 0.5~1.5 mm for stress surfaces and high-precision parts (basic value 1.0 mm); 1.5~3 mm for free surfaces (basic value 2.0 mm); 1.5~2.5 mm for transition areas such as wheel rims / outer circles (basic value 2.0 mm).

[0067] The second step is to adjust the coefficient according to the deformation of the material after heat treatment: For FGH96, since the deformation after hot isostatic pressing is small, multiply by a coefficient of 0.8 to take the lower limit; for thin-walled spokes, since the warping after heat treatment is large, multiply by a coefficient of 1.2 to take the upper limit; for mold materials such as K403 / N3, there is no deformation adjustment.

[0068] Step 3: Adjust the size according to geometric dimensions: based on the maximum outer diameter of the workpiece. For reference, when The base value is taken at that time; Multiply by 1.05; Multiply by 1.10.

[0069] Final allowance value = base value × material coefficient × size coefficient. All values ​​are recorded in the "Basis of Value" field of the "Mold Design Parameter Table" for easy auditing.

[0070] like Figure 8As shown, local features generated after machining, such as tooth profile, small fillet, and mounting holes, are removed, and forging allowance is added according to the above dynamic rules to finally obtain the shape of the final forging, thereby outputting the geometric file (final forging drawing) and feature list of the final forging.

[0071] Example 3: Steps 500-600 are completed by the AI ​​agent in conjunction with the process judgment module.

[0072] Step 500: Based on the geometric features of the final forging drawing, match it with the forging process parameters in the forging process database to obtain the forging process scheme corresponding to the final forging drawing. The forging process scheme is obtained jointly through feature quantity-driven and RAG-assisted decision-making.

[0073] like Figure 9 As shown, all geometric features of the final forging drawing are extracted, forging features are identified among all geometric features based on the die forging characteristics, and the number N of the forging features is counted.

[0074] When N≤6, the forging process is determined to be a two-process scheme of billet preparation and final forging, and the preceding forging drawing corresponds to the billet preparation drawing.

[0075] When N > 6, the forging process is a three-step scheme of billet preparation, pre-forging, and final forging. The preceding forging drawing corresponds to the pre-forging drawing. Furthermore, based on the pre-forging drawing, the corresponding forging process scheme is matched, and the billet drawing is derived in reverse.

[0076] When N is in the boundary range of 5 to 7 or the feature complexity is ambiguous, activate RAG-assisted decision-making: 1) Feature vectorization: The AI ​​agent extracts 5-dimensional feature vectors from the geometric features of the final forging, namely "maximum outer diameter, minimum thickness, aspect ratio, maximum projected area, and contour complexity coefficient". 2) Knowledge base retrieval: Use cosine similarity to retrieve the top K=10 most similar historical cases in the RAG knowledge base (the similarity threshold is 0.8 by default). 3) Majority voting: Count the number of two-process and three-process cases in the search results, and make the final decision based on the majority principle; 4) Output Case Number: Write the historical case numbers involved in the decision-making process into the JSON of the process plan for easy auditing.

[0077] The forging process parameters include at least the geometric parameters of the forging surface, forging flash, and draft angle. These parameters are read from the forging process database (i.e., knowledge base) during matching.

[0078] Step 600: Based on the forging process parameters in the forging process plan, the preceding forging drawings are derived by reverse engineering from the final forging drawing. The method for reverse engineering the process from the final forging drawing includes: When the forging process scheme is a billet-final forging process, the billet drawing is directly derived from the final forging drawing; When the forging process scheme is billet preparation-pre-forging-final forging process, the pre-forging drawing is derived from the final forging drawing, and then the billet preparation drawing is derived from the pre-forging drawing; During the reverse calculation process, the rate of change of cross-sectional area at the corresponding positions of adjacent processes is less than or equal to 25%.

[0079] Specifically: 1. Automatic determination of process plan.

[0080] 1. First-level decision (feature quantity driven): when In this case, a two-step process of "blanking-final forging" is directly adopted. In the blanking stage, in addition to free forging with a flat anvil, die forging (with flat upper and lower dies and a central spherical / curved cavity) is also permitted, allowing the direct formation of 1-2 curved surface features during the blanking stage. The initial blank is a cylinder (without a central through hole) or an annulus with a central hole (with a through hole), such as... Figure 10 As shown.

[0081] when At that time, a three-process scheme of "bill making-pre-forging-final forging" is adopted. The AI ​​agent selects 2 to 3 deep features from the feature list of the final forging and assigns them to the pre-forging part according to the "feature depth / diameter ratio ≥ 0.15". The selected features are then subjected to 30% to 60% transition filling on the pre-forging part.

[0082] II. RAG (Knowledge Base) retrieval combined with thresholds.

[0083] When N is in the boundary range of 5 to 7 or the feature complexity is ambiguous, activate RAG-assisted decision-making: 1) Feature vectorization: The AI ​​agent extracts 5-dimensional feature vectors from the geometric features of the final forging, namely "maximum outer diameter, minimum thickness, aspect ratio, maximum projected area, and contour complexity coefficient". 2) Knowledge base retrieval: Use cosine similarity to retrieve the top K=10 most similar historical cases in the RAG knowledge base (the similarity threshold is 0.8 by default). 3) Majority voting: Count the number of two-process and three-process cases in the search results, and make the final decision based on the majority principle; 4) Output Case Number: Write the historical case numbers involved in the decision-making process into the JSON of the process plan for easy auditing.

[0084] III. Determining the threshold for quantitative indicators.

[0085] The thresholds for each quantitative indicator are determined through a combination of the following three methods: 1) Empirical values: Recommended ranges were extracted from monographs such as "Forging Handbook" and "Powder Metallurgy High Temperature Alloy Forging" as initial thresholds; 2) Statistical calibration: Statistical analysis is performed on samples of whether the process plan is successful on the enterprise's historical trial dataset (no less than 500 pieces), and the optimal segmentation threshold is determined using decision trees or logistic regression. 3) Continuous correction: After each new project is completed, the threshold is updated in reverse through the quadruple database in step 500 based on the actual trial feedback.

[0086] 4) Reverse estimation of billet volume: ,in, Take 5% to 15% (depending on structural complexity). Take 0.5% to 2% (depending on the heating time); then deduce the billet diameter and height by reverse calculation based on the equipment mold cavity ratio.

[0087] like Figure 10 As shown, the morphology of the forging before and after billet preparation is compared (feature comparison), and the upper surface curved surface features formed by forging with the die after billet preparation are shown.

[0088] Example 4: Steps 400 and 700 are completed by the AI ​​agent in conjunction with the mold design module.

[0089] Step 400: Extract the geometric features of the final forging drawing and use these geometric features as the cavity design parameters for the final forging die. That is, directly extract the geometric features of the final forging drawing as the cavity design parameters for the final forging die.

[0090] Step 700: Use the geometric features of the preceding forging drawing as the cavity design parameters for the corresponding forging die. Use the geometric features of the preceding forging drawing as the cavity design parameters for the corresponding forging die (such as a pre-forging die or a billet die).

[0091] In mold design, based on the design parameters of the mold cavity, the specific methods for designing the mold include: Based on the billet drawing, the pre-forging drawing, and the final forging drawing, the demolding angle is set respectively; Separate the upper and lower halves along the neutral planes of the billet drawing, the pre-forging drawing, and the final forging drawing, and extend them outward to generate the billet mold body, the pre-forging mold body, and the final forging mold body, respectively. Bridges, slots, flash grooves, and positioning structures are added to the outside of the blanking mold cavity, the pre-forging mold cavity, and the final forging mold cavity, respectively, to obtain the blanking mold drawing, the pre-forging mold drawing, and the final forging mold drawing, respectively. The geometric features of the billet forming die drawing, the pre-forging die drawing, and the final forging die drawing are used as the design parameters for the billet forming die, the pre-forging die, and the final forging die, respectively.

[0092] Specifically, in step 300, the final forging drawing output by the AI ​​agent includes geometric feature parameters and geometric files.

[0093] After the process determination is completed, the initial mold design is carried out in the following 4 steps: 1. Reverse deduction of geometric features of forgings in each process: Based on the process plan, reverse deduction from the final forging to the front: final forging → pre-forging (if any, remove some detailed features from the final forging and increase the fillet radius) → blank (remove all details and retain the main curved surface).

[0094] The transition shape of each stage is designed according to the principle of uniform metal flow: ensuring that the cross-sectional area change rate at corresponding positions of adjacent processes is ≤25%.

[0095] 2. Draft angle setting: For all local edges on the forging morphology with an angle ≥80° to the mold opening direction, a draft angle of 3~7° is uniformly set; the smaller value (3°~5°) is used for the outer surface, and the larger value (5°~7°) is used for the deep cavity of the inner surface.

[0096] 3. Upper and Lower Mold Geometry Generation: First, separate the forging into an upper and lower half along the neutral plane (maximum projection plane); then extend them outwards to form the mold body; the mold thickness is processed according to 30% to 70% of the maximum dimension of the corresponding workpiece's horizontal cross-section (30% is suitable for light-load, small-sized discs, 70% is suitable for heavy-load, large-sized discs); finally, add a bridge (0.5~2 mm in height), a slot opening (3~6 times the width of the bridge), a flash groove, and a positioning structure to the outside of the cavity, such as... Figure 11 As shown.

[0097] 4. Automatic SolidWorks model building: 1) Template recording: The SolidWorks modeling process of several typical molds is recorded in advance to obtain a VBA / VBS script template library containing basic operations such as sketching, extrusion, cutting, filleting, and arraying; 2) Template retrieval: When a new part needs to be modeled, the AI ​​agent finds the most similar template through feature vector retrieval; 3) AI Parametric Replacement: The AI ​​agent reads the current forging die parameter table (including die outer diameter, thickness, fillet radius, demolding angle, etc.), calls the LLM generator to replace the sketch dimensions, stretch length, fillet radius, demolding angle, etc. in the template, and generates a new VBS script; 4) COM Interface Execution: This VBS script is executed through the SolidWorks COM automation interface to automatically complete the modeling and output SLDPRT / STEP files; 5) Self-check feedback: After the script is executed, the AI ​​agent reads the SolidWorks error log and geometric dimensions, cross-compares them with the mold parameter table, and if they are inconsistent, the script is regenerated and executed (maximum 3 retries).

[0098] The AI ​​agent also outputs a "forging die parameter table", which lists the source of each feature (automatically generated / experience-based default / manually specified), the basis for the value, and the parameter range.

[0099] Example 5: Steps 800-900 are completed sequentially by the AI ​​agent in conjunction with the full-process simulation module, the multi-objective optimization module, and the parameter output module.

[0100] Step 800: Based on the design parameters of the blanking mold cavity, the design parameters of the pre-forging mold cavity, and the design parameters of the final forging mold cavity, design the blanking mold, the pre-forging mold, and the final forging mold respectively, and perform full-process simulation according to the forging process scheme to obtain multiple physical parameters of the three molds after simulation. Based on multiple physical parameters after simulation of the three molds, a weighted normalized multi-objective function is constructed, and a gradient optimization algorithm is used to iteratively optimize the design parameters of the three mold cavities. The design parameters after each iteration are then fed back to step 600. Step 900, repeat steps 600 to 800 until the multi-objective function meets the convergence condition, and output the optimal mold cavity design parameters for each process as the optimal design parameters for the corresponding mold cavity. The convergence condition is that the gradient norm of the objective function is less than a preset first threshold, or the difference between two consecutive objective function values ​​is less than a preset second threshold, or the number of iterations reaches a preset upper limit, or the binary penalty terms are all zero and the continuous objective improvement is less than a preset third threshold.

[0101] I. CAE Dual-Channel Drive Simulation.

[0102] like Figure 12 As shown, the AI ​​agent uses a dual-channel collaborative drive of "KEY file parameter direct modification + RPA image recognition" to drive the Deform software to complete the simulation of the entire process.

[0103] While general-purpose RPA platforms (such as UiPath and Automation Anywhere) provide basic image recognition capabilities, they lack robustness against the dynamic interface changes, version differences, and complex parameter dialog boxes of professional CAE software like Deform. The specific method of this invention is as follows: 1. MO template reuse: For two-process and three-process workflows, corresponding Deform MO model templates (pre-made solidified blank-mold object structure, contact pairs, temperature boundaries, press motion control) are pre-made. 2. Channel 1 (Direct Modification of KEY File Parameters): The AI ​​agent performs field-level parsing of the MO's KEY file and directly modifies parameters such as material grade and rheological curve index, initial temperature, mold temperature, press speed, friction coefficient (viscoplastic friction factor m is 0.3~0.7), solution step size, maximum number of steps, and convergence tolerance. 3. Channel 2 (RPA Image Recognition + Keyboard and Mouse): For operations that cannot be textualized through the KEY file (mesh generation, initial mold positioning, coordinate system alignment, reference point selection), RPA executes the pre-recorded workflow; 4. RPA Robust Design: 1) The image template is matched by "main icon + auxiliary anchor point". The main icon is positioned in the center of the button, and the auxiliary anchor point (button text, adjacent controls) is used for secondary verification. 2) The template matching threshold is 0.85 by default. If there are 3 consecutive unmatched attempts, the threshold will be reduced to 0.75 and a self-check of the interface resolution will be triggered. 3) If it still fails, it will revert to the keyboard shortcut path (most Deform operations have backup keyboard shortcuts). 4) If it fails again, save a screenshot of the failure and a snapshot of the parameters, and pause the process while waiting for manual intervention; 5) After each RPA operation is completed, compare the screenshot of the interface status with the expected status to confirm that the operation has taken effect; 6) Workflow Prefabricated Units: Import new geometric model, move mold to designated position, set mold movement direction and speed, set mesh generation density and re-meshing criteria, select mold reference point, set contact pairs, start solving, save and export results. The AI ​​agent controls the internal values ​​of each unit through parameterization (e.g., the 20 in "move 20 mm" is generated and replaced by AI).

[0104] like Figure 13 As shown, the Deform full-process simulation model consists of an upper mold (moving), a lower mold (fixed), a blank (including a mesh), contact pairs, and temperature and friction boundary conditions.

[0105] In step 800, the method for constructing a weighted normalized multi-objective function and performing iterative optimization using a gradient optimization algorithm includes: Extract multiple physical parameters from each mold simulation, including at least forging filling degree, maximum forming load, maximum equivalent stress of mold, mold wear, and material utilization rate. Each physical parameter is normalized to map it to a value range of 0 to 1. Assign a weight coefficient to each objective, and the sum of all weight coefficients is 1; Multiply each normalized physical parameter by its corresponding weight coefficient and sum them to construct a multi-objective function; The gradient optimization algorithm is used to calculate the gradient of the multi-objective function with respect to the mold cavity design parameters, and the design parameters are updated along the gradient descent direction; The updated design parameters were verified by full-process simulation, and the physical parameters were re-extracted and new multi-objective function values ​​were calculated. Repeat the process of updating design parameters and extracting physical parameters until the convergence condition is met, and output the current optimal mold cavity design parameters.

[0106] II. Extraction of multiple physical quantities from simulation results.

[0107] Refer to the table below for the extraction specifications of the seven physical quantities in the simulation results.

[0108]

[0109] III. Construct a weighted normalized multi-objective function.

[0110] The AI ​​agent constructs a weighted multi-objective function based on the above seven physical quantities.

[0111] Weighted multi-objective function: ; Symbol meaning: : Mold parameter vector, containing geometric feature dimensions such as fillet radius, bridge height, slot width, flash groove size, and demolding angle for each process upper and lower mold cavity; The normalized maximum load sub-target of the mold should be as small as possible; The normalized maximum wear quantum target for the mold should be as small as possible; : A binary penalty term for incomplete forging; 1 for incomplete forging, 0 for complete forging. : Binary penalty term for forging folding, 1 for folding, 0 otherwise; : Binary penalty term for forging damage, 1 if the damage threshold is exceeded, 0 otherwise; The normalized maximum equivalent stress sub-objective for forgings should be as small as possible; The normalized streamline effect sub-target (spacing of the densest streamline regions) indicates that the smaller the value, the more severe the local deformation; therefore, the target value should be small. ~ The weight coefficients corresponding to each sub-objective. The default value is specified by the engineer in the configuration file, and is automatically suggested by the AI ​​agent based on historical project data; to avoid fatal flaws, / / The weights of the three binary penalties are significantly higher by default than those of the continuous indicator items.

[0112] Formula Source / Derivation: This formula is an application of the classic multi-objective optimization method, the Weighted Sum Method (belonging to the category of existing mathematical methods). Its basis is that, in the Pareto optimal sense, for a convex Pareto front, any Pareto optimal solution can be obtained by minimizing the weighted sum of the corresponding set of non-negative weights. The creative contribution of this invention lies not in the formula itself, but in: incorporating three critical process defects into the continuous objective function in the form of binary penalties; providing the deformation post-processing extraction specifications for each physical quantity; and providing the basis for the values ​​of the normalization endpoints.

[0113] Normalization method: Min-Max normalization is used for each continuous sub-target. ; in, The normalized endpoints for this physical quantity are determined by the RAG knowledge base, which provides suggested ranges based on material, equipment tonnage, and forging shape. For example, the maximum load reference range for an 800 mm diameter FGH96 disc is [reference range to be inserted here]. .

[0114] IV. Optimization and Iteration.

[0115] like Figure 14 As shown, the sequential quadratic programming (SQP) gradient method is used for iterative optimization.

[0116] 1. Parameter range adaptive: The optimization variable range of each mold geometry parameter is adaptively set to ±10% ~ ±80% of the initial value; the parameter complexity is weighted by "the number of geometric surfaces affected by the parameter + the number of adjacent features", with a wider range for high complexity and a narrower range for low complexity.

[0117] 2. Gradient calculation: Forward difference is used, with a step size of 5% of the current parameter value.

[0118] 3. Convergence Criterion (Stop if any one criterion is met): 1) ; 2) ; 3) The maximum number of iterations has been reached. (Default 30); 4) All binary penalty terms are 0, and continuous target improvement is <1%.

[0119] 4. Algorithm switching: When the objective function is highly nonlinear and the SQP oscillation does not decrease, switch to a surrogate model to assist optimization: construct a Kriging / RBF surrogate model by sampling 30 points using Latin hypercube, perform a global search on the surrogate model to obtain candidate solutions, and then verify them with real simulation.

[0120] V. Parameter Output and Quadruple Database.

[0121] Final output: 1. Final mold design parameters (CSV / Excel, source of each parameter indicated); 2. Mold geometry files (SLDPRT + STEP / IGES); 3. Current output parameter report (including 7 physical quantity values ​​and corresponding Deform post-processing cloud plots / curves); 4. Iterative process convergence curve (PNG + CSV); 5. Archive the "Parameters-Geometry-Simulation-Results" quadruple.

[0122] The database (quadruple database) stores "design parameter X, CAD geometry file path, Deform input KEY and output DAT file, and normalized result index" in SQLite or JSON Lines format. "Four types of related data; when a new project starts, historical records with a geometric similarity greater than a preset threshold (default 0.8) are retrieved as initial optimization values, and the Kriging proxy model is pre-trained using historical data to achieve knowledge transfer and enterprise-level design experience accumulation."

[0123] In summary, the design method of this invention aims to transform the traditional "experience-driven, fragmented" mold design process into a computable engineering process based on "geometric reverse engineering." Its core principle can be broken down into the following four stages: The method reverse-engineers the final forging from the machined part: The machined workpiece is the final product, and its surface has many local features (such as tooth profiles, oil holes, mounting surfaces, etc.) formed only by machining processes such as cutting, drilling, and milling. These features do not exist in the forging stage. This method first identifies and deletes these "machining features", then uses the surface filling method to restore the missing forging blank shape, and then, according to the forging process rules such as material thermal shrinkage rate, draft angle, and subsequent machining allowance, uniform or partitioned forging allowances are superimposed on the outer contour to obtain the true geometric features of the final forging (including geometric image files: two-dimensional contour, three-dimensional model; feature data).

[0124] Compared to existing technologies, this method: Reverse engineering process: The final forging usually cannot be formed in one forging and requires multiple processes (such as billet preparation → pre-forging → final forging). The method is based on the forging knowledge base (including the process tree and deformation rules of typical disc and shaft parts), and reverses the process from the final forging to deduce the billet shape of the previous process, i.e., "incremental reverse deformation". The shape of the forging in each process is obtained by adding the local deformation contributed by the process (such as fillet filling and cross-section expansion) to the shape of the forging in the next process, thus generating a complete process chain forging model.

[0125] Parametric mold generation: The forging model for each process is divided into upper and lower halves along the parting surface (neutral surface). Then, the cavity is offset outward (considering thermal shrinkage and elastic deformation compensation). At the same time, bridges, flash grooves, positioning and ejection structures are added to generate the 3D parametric model of the mold for that process. All geometric dimensions (corner radius, draft angle, rib thickness, bridge height, etc.) are parameterized for easy subsequent optimization.

[0126] Closed-loop optimization based on multi-objective gradients: Traditional mold design relies on "trial and error + local adjustment," making it difficult to simultaneously consider multiple mutually exclusive indicators such as filling degree, forming load, mold wear, and material utilization. This method combines full-process finite element simulation (such as Deform) with mathematical optimization: Each simulation outputs physical quantities such as mold stress, wear, forging filling degree, and load curves after each process. These physical quantities are dimensionless (minmax normalization) and weighted according to engineering importance to synthesize a single objective function F. Gradient algorithms such as Sequential Quadratic Programming (SQP) are used to automatically calculate the sensitivity of the objective function to each mold geometric parameter, update the parameters along the descent direction, and then feed them back into the mold model for the next round of simulation. After iterative convergence, the optimal solution on the Pareto front is obtained.

[0127] The entire process forms a closed loop of "geometry → process → mold → simulation → optimization → geometry". The cycle of steps 600 to 800 does not change the shape of the final forging, but only adjusts the mold cavity details of each process until the overall performance is optimal.

[0128] Therefore, compared with existing methods such as "experience + trial and error" or "single-point CAE optimization", this invention has the following outstanding advantages: Completely eliminate manual geometric reconstruction: Automatically delete machined features and fill in repairs, eliminating the need for CAD engineers to manually repair the model, reducing the final forging generation time from days to minutes.

[0129] Scientific decision-making for process planning: Based on forging knowledge base matching, it avoids blindly selecting two or three processes based on experience, thus reducing the number of trial moldings.

[0130] Multi-objective synchronous optimization: Incorporate mutually constraining indicators such as fill degree, load, stress, wear, and material utilization rate into a unified weighting function, and use gradient optimization algorithm to automatically find the comprehensive optimal solution, avoiding the one-sidedness of manual compromise.

[0131] Traceable and reproducible: The parameters, simulation results, and objective function values ​​of each optimization are recorded, forming an auditable digital twin archive, allowing newcomers to directly inherit the optimization strategy.

[0132] Reduce trial molding costs: Replacing physical trial molding with high-fidelity full-process simulation can reduce on-site mold repair work by more than 80%, which is especially suitable for the design of large isothermal forging dies for difficult-to-deform alloys (such as powder high-temperature alloy FGH96).

[0133] Highly adaptable: The method is not limited to a certain type of part. It can be quickly migrated to different products such as discs, rings, and shafts simply by updating the process templates and material database in the forging knowledge base.

[0134] Example 6: All modules are scheduled and executed by a unified AI agent.

[0135] like Figure 2 , Figure 15 As shown, the design system consists of six functional modules forming a complete closed loop: "Geometric Recognition (Module 1) → Process Judgment (Module 2) → Mold Initial Design (Module 3) → Full Process Simulation (Module 4) → Multi-Objective Optimization (Module 5) → Parameter Output (Module 6)". All modules are scheduled by a unified AI Agent (i.e., AI intelligent agent) and are decoupled and transmitted to each other through explicit file contracts such as parameter files (JSON), geometric files (SLDPRT / STEP), and result files (TXT / DAT / PNG).

[0136] The data contract between modules is as follows: Module 1 outputs "final forging B-Rep geometry file + feature list JSON"; Module 2 outputs "process plan JSON"; Module 3 outputs "upper and lower mold SLDPRT / STEP geometry + design parameter table CSV for each process"; Module 4 outputs "Deform post-processing result dataset"; Module 5 feeds back "a new set of design parameter CSV" to Module 3 in each iteration; Module 6 finally outputs "optimal parameter CSV + optimal mold geometry + convergence curve PNG / CSV".

[0137] When any module fails to execute (such as CAD script error, Deform solution not converging, RPA image matching failure), the AI ​​Agent handles it according to a three-level mechanism of "parameter stepping retry → template rollback → manual intervention alarm".

[0138] The AI ​​agent includes: a large language model, a knowledge base, a geometry recognition module, a process judgment module, a mold design module, a CAE-driven module, an optimization and iteration module, and a database.

[0139] The knowledge base is used to store knowledge and historical process data in the field of forging, i.e., a forging process database.

[0140] The database (quadruple database) is used to store the design parameters, geometry files, simulation results and optimization target values ​​for each design task. The design parameters, geometry files, simulation results and optimization target values ​​for each design process are stored in the database in the form of quadruples.

[0141] When performing mold design for a new workpiece, the AI ​​agent retrieves historical records from the database that have a geometric similarity to the current workpiece greater than a preset threshold, and uses the design parameters in the historical records as the initial values ​​for optimization iteration.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. All such modifications or substitutions should be covered within the protection scope of this application, and should not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for designing the cavity of a vacuum isothermal forging die, characterized in that, Includes the following steps: Step 100: Based on the product design drawings, obtain the target part drawing formed by die forging and then machining from the design drawings; Step 200: Based on the machining characteristics of the target part drawing, identify the machining features formed by machining on the target part drawing; Step 300: Delete the machining features in the target part drawing, repair the missing features and inherent features of the forging formed after deletion, and add forging allowance for machining to obtain the final forging drawing. Step 400: Extract the geometric features of the final forging drawing and use the geometric features as the cavity design parameters of the final forging die; Step 500: Based on the geometric features of the final forging drawing, match it with the forging process parameters in the forging process database to obtain the forging process scheme corresponding to the final forging drawing. Step 600: Based on the forging process parameters in the forging process plan, the forging process drawing is reverse-engineered from the final forging drawing to obtain the preliminary forging drawing before the final forging process. Step 700: Use the geometric features of the preceding forging drawing as the cavity design parameters for the corresponding forging die.

2. The method for designing the cavity of a vacuum isothermal forging die according to claim 1, characterized in that, In step 200, the method for identifying machining features formed by machining on the target part drawing includes: Based on the target part drawing, extract all surface regions that can be processed independently as candidate feature surfaces; The feature indicators of each candidate feature surface are quantified, and the feature indicators include at least the local radius of curvature, the number of adjacent surfaces, the maximum feature coverage ratio, the metal flow angle, and the minimum thickness. If all the aforementioned feature indicators meet the forging requirements, then it is marked as a qualified candidate feature surface; otherwise, it is marked as a candidate feature surface to be determined. Based on the feature indicators of the candidate feature surfaces to be determined, they are matched with the machining parameters in the machining database to obtain the machining determination result of the candidate feature surfaces to be determined; The local radius of curvature is the radius of curvature value along the direction of minimum principal curvature on the candidate feature surface, which is used to characterize the degree of curvature of the surface region; The number of adjacent faces is the number of faces that are directly adjacent to the candidate feature face; The large feature coverage ratio is the ratio of the fitted area of ​​the basic geometric surface to the total area of ​​the candidate feature surface; wherein, the basic geometric surface includes at least a plane, a cylindrical surface, and a conical surface; The metal flow angle is the angle between the metal flow direction and the mold opening direction along the main forging direction. The minimum thickness is the thickness value at the thinnest point of the candidate feature surface.

3. The method for designing the cavity of a vacuum isothermal forging die according to claim 1, characterized in that, In step 300, the inherent features of the forging include at least a draft angle.

4. The method for designing the cavity of a vacuum isothermal forging die according to claim 3, characterized in that, In step 300, the method for repairing the missing features and inherent features of the forging resulting from the deletion, and for supplementing the forging allowance for machining, includes: Based on the target part drawing, suppress the feature tree nodes corresponding to the missing features and the inherent features of the forging; The retained surfaces adjacent to the suppressed area in the target part drawing are extended outward along their respective normal directions until the extended curved surfaces completely cover the gap left by the suppressed area, forming a transition area. For the transition region formed after extension, a C1 continuous smooth fit is performed using a cubic B-spline surface to form an image of the repaired forging; Based on the image of the repaired forging, and according to the functional requirements of different parts after repair, the basic value of the machining allowance is set; Based on the formula: final allowance value = base value × material coefficient × size coefficient, the machining allowance of each part of the repair forging image is calculated to obtain the final forging drawing.

5. The method for designing the cavity of a vacuum isothermal forging die according to claim 1, characterized in that, It also includes a feature extraction method for the forging process scheme corresponding to the final forging drawing: Extract all geometric features of the final forging drawing, identify forging features among all geometric features based on the die forging characteristics, and count the number N of the forging features; When N≤6, the forging process is determined to be: billet preparation-final forging process. The preceding forging drawing corresponds to the billet preparation drawing, and the geometric features of the billet preparation drawing are used as the cavity design parameters of the billet preparation mold. When N > 6, the forging process is: billet preparation - pre-forging - final forging process. The preceding forging drawing corresponds to the pre-forging drawing, and the geometric features of the pre-forging drawing are used as the cavity design parameters of the pre-forging die. Furthermore, based on the geometric features of the pre-forged drawing, it is matched with the forging process parameters in the forging process database to obtain the forging process scheme corresponding to the pre-forged drawing; By reverse engineering the pre-forging drawing, a blanking drawing is obtained before the pre-forging process, and the geometric features of the blanking drawing are used as the cavity design parameters of the blanking mold.

6. The method for designing the cavity of a vacuum isothermal forging die according to claim 5, characterized in that, In step 500, the forging process parameters include at least the geometric parameters of the forging surface, forging flash, and draft angle.

7. The method for designing the cavity of a vacuum isothermal forging die according to claim 5, characterized in that, In step 600, the method for reverse engineering the final forging drawing includes: When the forging process scheme is a billet-final forging process, the billet drawing is directly derived from the final forging drawing; When the forging process scheme is billet preparation-pre-forging-final forging process, the pre-forging drawing is derived from the final forging drawing, and then the billet preparation drawing is derived from the pre-forging drawing; During the reverse calculation process, the rate of change of cross-sectional area at the corresponding positions of adjacent processes is less than or equal to 25%.

8. The method for designing the cavity of a vacuum isothermal forging die according to claim 7, characterized in that, Following step 700, the following is also included: Step 800: Based on the design parameters of the blanking mold cavity, the design parameters of the pre-forging mold cavity, and the design parameters of the final forging mold cavity, design the blanking mold, the pre-forging mold, and the final forging mold respectively, and perform full-process simulation according to the forging process scheme to obtain multiple physical parameters of the three molds after simulation. Based on multiple physical parameters after simulation of the three molds, a weighted normalized multi-objective function is constructed, and a gradient optimization algorithm is used to iteratively optimize the design parameters of the three mold cavities. The design parameters after each iteration are then fed back to step 600. Step 900, repeat steps 600 to 800 until the multi-objective function meets the convergence condition, and output the optimal mold cavity design parameters for each process as the optimal design parameters for the corresponding mold cavity. The convergence condition is that the gradient norm of the objective function is less than a preset first threshold, or the difference between two consecutive objective function values ​​is less than a preset second threshold, or the number of iterations reaches a preset upper limit, or the binary penalty terms are all zero and the continuous objective improvement is less than a preset third threshold.

9. The method for designing the cavity of a vacuum isothermal forging die according to claim 8, characterized in that, In step 800, the method for constructing a weighted normalized multi-objective function and performing iterative optimization using a gradient optimization algorithm includes: Extract multiple physical parameters from each mold simulation, including at least forging filling degree, maximum forming load, maximum equivalent stress of mold, mold wear, and material utilization rate. Each physical parameter is normalized to map it to a value range of 0 to 1. Assign a weight coefficient to each objective, and the sum of all weight coefficients is 1; Multiply each normalized physical parameter by its corresponding weight coefficient and sum them to construct a multi-objective function; The gradient optimization algorithm is used to calculate the gradient of the multi-objective function with respect to the mold cavity design parameters, and the design parameters are updated along the gradient descent direction; The updated design parameters were verified by full-process simulation, and the physical parameters were re-extracted and new multi-objective function values ​​were calculated. Repeat the process of updating design parameters and extracting physical parameters until the convergence condition is met, and output the current optimal mold cavity design parameters.

10. A vacuum isothermal forging die design method based on the vacuum isothermal forging die cavity design method according to claim 8 or 9, characterized in that, In step 900, the specific method for designing the mold based on the optimal design parameters of the mold cavity includes: Based on the billet drawing, the pre-forging drawing, and the final forging drawing, the demolding angle is set respectively; Separate the upper and lower halves along the neutral planes of the billet drawing, the pre-forging drawing, and the final forging drawing, and extend them outward to generate the billet mold body, the pre-forging mold body, and the final forging mold body, respectively. Bridges, slots, flash grooves, and positioning structures are added to the outside of the blanking mold cavity, the pre-forging mold cavity, and the final forging mold cavity, respectively, to obtain the blanking mold drawing, the pre-forging mold drawing, and the final forging mold drawing, respectively. The geometric features of the billet forming die drawing, the pre-forging die drawing, and the final forging die drawing are used as the design parameters for the billet forming die, the pre-forging die, and the final forging die, respectively.

Images