Method, device, equipment and medium for springback compensation of punch die profile

Abstract

The present disclosure provides a springback compensation method, device, equipment and medium for a stamping die surface, the springback compensation method comprising: performing stamping process simulation on an initial die surface to obtain simulation grid data; performing non-uniform offset processing on the simulation grid data to obtain offset grid; and reconstructing the offset grid into a compensated die surface by using a density clustering algorithm. The simulation grid data containing wall thickness distribution is obtained based on the stamping process simulation of the initial die surface, providing accurate physical basis for subsequent compensation, and the offset grid adapting to the non-uniform deformation of the sheet metal is generated through non-uniform offset processing, breaking through the limitation of traditional equidistant offset that cannot cope with complex springback of high-strength steel. Finally, the offset grid is reconstructed by using the density clustering algorithm, which compresses the number of control points of the curved surface to improve efficiency while accurately retaining the global deformation characteristics of the sheet metal, thereby effectively suppressing the stamping springback of high-strength steel and greatly improving the forming precision of the stamped part.

Description

Technical Field

[0001] This disclosure relates to the field of stamping die design technology, and in particular to a method, apparatus, equipment and medium for springback compensation of stamping die surface. Background Technology

[0002] During the stamping process, sheet metal undergoes both elastic and plastic deformation within the stamping die. Once the stamping pressure is released, the elastic stress within the sheet metal is released, causing the geometry and dimensions of the stamped part to partially return to their pre-forming state—a phenomenon known as springback. For high-strength steel, the springback is greater than that of traditional low-carbon steel, and the springback behavior is complex and difficult to predict, resulting in lower dimensional accuracy of the stamped parts.

[0003] Currently, to compensate for springback, the existing solutions typically involve designing the die profile based on the target profile of the stamped part, offsetting it at equal intervals according to the sheet metal thickness to generate the initial die profile. Subsequently, the die profile is repeatedly tested, measured, and modified based on the experience of technicians to compensate for springback. This approach not only relies on the personal experience of technicians, making it impossible to ensure consistent precision of stamped parts in mass production, but also results in long development cycles and high development costs for the stamping dies. Summary of the Invention

[0004] This disclosure provides a method, apparatus, equipment, and medium for springback compensation of stamping die surfaces; it can not only shorten the development cycle and cost of stamping dies, but also improve the consistency of stamping part precision.

[0005] The technical solution disclosed herein is implemented as follows: In a first aspect, this disclosure provides a springback compensation method for the surface of a stamping die, comprising: performing a stamping process simulation on the initial die surface to obtain simulated mesh data; performing non-equidistant offset processing on the simulated mesh data to obtain an offset mesh; and reconstructing the offset mesh into the compensated die surface using a density clustering algorithm.

[0006] Secondly, this disclosure provides a springback compensation device for the surface of a stamping die, comprising: a simulation module configured to perform stamping process simulation on an initial die surface to obtain simulated mesh data; a processing module configured to perform non-equidistant offset processing on the simulated mesh data to obtain an offset mesh; and a reconstruction module configured to reconstruct the offset mesh into a compensated die surface using a density clustering algorithm.

[0007] Thirdly, this disclosure provides a computer device including a memory and a processor, the memory for storing executable instructions, and the processor executing the executable instructions stored in the memory to implement the steps of the springback compensation method for the stamping die surface as described in the first aspect.

[0008] Fourthly, this disclosure provides a computer storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the springback compensation method for the stamping die surface as described in the first aspect.

[0009] Fifthly, this disclosure provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the springback compensation method for the stamping die surface as described in the first aspect.

[0010] This disclosure provides a springback compensation method for stamping die surfaces. The method involves performing stamping process simulation on the initial die surface to obtain simulated mesh data; then, the simulated mesh data is processed with non-equidistant offset to obtain an offset mesh; finally, a density clustering algorithm is used to reconstruct the compensated die surface. First, simulated mesh data containing wall thickness distribution is obtained based on the stamping process simulation of the initial die surface, providing a precise physical basis for subsequent compensation. Then, non-equidistant offset processing generates an offset mesh adapted to the uneven deformation of the sheet metal, overcoming the limitation of traditional equidistant offsets in handling the complex springback of high-strength steel. Finally, a density clustering algorithm is used to reconstruct the offset mesh, compressing the number of control points for surface reconstruction to improve efficiency while accurately preserving the global deformation characteristics of the sheet metal, thereby effectively suppressing springback in high-strength steel stamping and significantly improving the forming accuracy of stamped parts. Attached Figure Description

[0011] Figure 1 A flowchart of a springback compensation method for the surface of a stamping die provided in this disclosure.

[0012] Figure 2 This is a schematic diagram of the operation interface for the development of a drawing process mold for the lower control arm body provided in this disclosure.

[0013] Figure 3 This is a schematic diagram of a three-dimensional simulation cloud map provided in this disclosure.

[0014] Figure 4 This is a schematic diagram of the thickness strain distribution and STL mesh element of a stamped part provided in this disclosure.

[0015] Figure 5 This is a schematic diagram of an STL cell non-uniform thickness bias program interface provided in this disclosure.

[0016] Figure 6 This is a schematic diagram of a vertex offset provided in this disclosure.

[0017] Figure 7 This is a schematic diagram of a non-equidistant offset of a stamping die surface provided in this disclosure.

[0018] Figure 8 This is a schematic diagram showing the distribution of the 200 units with the largest absolute value of thinning rate on a stamped part, as provided in this disclosure.

[0019] Figure 9 This is a schematic diagram illustrating the principle of a density clustering algorithm provided in this disclosure.

[0020] Figure 10 This disclosure provides a flowchart for clustering biased grid cells based on a density clustering algorithm.

[0021] Figure 11 This is a schematic diagram of the surface reconstruction deviation distribution without using the DBSCAN clustering algorithm, as provided in this disclosure.

[0022] Figure 12 This is a schematic diagram of the surface reconstruction deviation distribution optimized using the DBSCAN clustering algorithm, as provided in this disclosure.

[0023] Figure 13 This is a schematic diagram of the composition of a springback compensation device for the surface of a stamping die provided in this disclosure.

[0024] Figure 14 A block diagram of the computing device provided in this disclosure. Detailed Implementation

[0025] The technical solutions of this disclosure will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure should fall within the protection scope of this disclosure.

[0026] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of this disclosure. The singular forms “a,” “the,” and “the” used in this disclosure are also intended to include the plural forms unless the context clearly indicates otherwise.

[0027] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0028] Furthermore, in the embodiments of this disclosure, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0029] To facilitate understanding of the technical solutions of the embodiments of this disclosure, the related technologies of the embodiments of this disclosure are described below. The following related technologies are optional solutions and can be combined with the technical solutions of the embodiments of this disclosure in any way, and they all fall within the protection scope of the embodiments of this disclosure.

[0030] In the automotive industry, advanced high-strength steels (such as duplex steel DP and transformation-induced plasticity steel TRIP) have been widely used to achieve lightweight vehicle bodies, improve fuel economy, and meet increasingly stringent safety standards. However, the high yield strength and high work hardening rate of these materials lead to a severe "springback" phenomenon during stamping. This means that after unloading, the shape and dimensions of the stamped part elastically recover, deviating from the target design and causing serious dimensional accuracy problems.

[0031] To address the springback problem, the industry currently relies heavily on experienced technicians for repeated manual trial molding, measurement, and mold repair. This method is not only time-consuming and costly, but the quality of mold repair is also highly dependent on individual skills, making it impossible to guarantee consistency in mass production due to factors such as material batch fluctuations. Building on this, while using Computer-Aided Engineering (CAE) for assisted design, generating the initial mold surface based on the target part shape and the nominal thickness of the sheet metal with equidistant offsets, this method ignores the physical fact that the actual thickness of the sheet metal varies significantly across different regions due to uneven plastic deformation during stamping. Since the root cause of springback is the uneven distribution of residual stress internally, and this stress distribution is closely related to thickness variations, equidistant offsets are inherently difficult to accurately and effectively compensate for springback. Furthermore, although CAE software can predict springback, converting complex simulation results (usually discrete mesh models with a massive number of vertices, such as STL format) into a smooth, compensated mold CAD model that can be directly used for CNC machining, such as a non-uniform rational B-spline (NURBS) surface, usually requires tedious manual reverse modeling. This is not only inefficient, but also prone to introducing new errors during the data conversion process, leading to a decrease in compensation accuracy.

[0032] Based on this, the embodiments of this disclosure perform stamping process simulation on the initial mold surface to obtain simulated mesh data that reflects the real physical deformation. The simulated mesh data is transformed into non-equidistant and non-uniform geometric compensation for the mold surface through non-equidistant offset processing to obtain an offset mesh. A small number of key points are extracted from the massive offset mesh through density clustering algorithm to ensure that the amount of information is compressed without losing the geometric information that is important for springback compensation. This can shorten the stamping die development cycle and cost, and also improve the consistency of stamping part precision.

[0033] See Figure 1 , Figure 1 The flowchart of a springback compensation method for a stamping die surface provided in this disclosure specifically includes steps S102-S106.

[0034] Step S102: Perform stamping process simulation on the initial mold surface to obtain simulated mesh data.

[0035] Specifically, the initial die surface is a reference die surface generated based on the target die surface of the stamped part, offset at equal distances according to the nominal thickness of the sheet metal. The target die surface is the geometric model of the part to be stamped, serving as the reference target for die design. It is typically stored in an industrial-grade CAD format (such as IGES, STEP) and includes the complete outline, dimensions, and surface topology information of the part (like the 3D model of a control arm). Stamping process simulation utilizes stamping CAE simulation software (such as PAM-STAMP, AutoForm, etc.) to numerically simulate the stamping process (including drawing, springback, etc.) of the initial die surface, replicating the elasto-plastic deformation process of the sheet metal within the die. The simulation mesh data is the discrete mesh data output from the stamping process simulation, reflecting the state of the sheet metal after stamping, and is typically in Standard Triangular Mesh Language (STL) format. The simulated mesh data includes the following information: Node data (NodeData) is used to record the three-dimensional spatial coordinates of each node; Element data (ElementData) is used to record the node index of the triangular facet and the thickness value of the element after forming; and the wall thickness distribution data of the stamped part reflects the thickness difference caused by uneven deformation in different areas of the sheet metal.

[0036] For example, firstly, the target shape of the stamping part (such as the IGES / STEP format CAD model of the control arm) and the sheet metal parameters (such as the initial thickness of 1.5mm of DP600 steel plate and material mechanical property parameters) are obtained; according to the traditional equidistant offset method, the target shape of the part is translated along the normal direction by half the thickness of the sheet metal to generate the initial die shape; the initial die shape, sheet metal parameters and stamping process parameters (such as blank holder force and friction coefficient) are imported into the stamping CAE simulation software, and the stamping forming and springback simulation is started; after the simulation is completed, the simulation mesh data containing node data, element data and wall thickness distribution data is exported.

[0037] like Figure 2 As shown, Figure 2 This is a schematic diagram of the operation interface for the development of a drawing process mold for the lower control arm body provided in this disclosure. Figure 2 The mold includes a punch 20, a die 22, and a sheet metal 21 to be stamped. The three-dimensional model intuitively presents the stamping deformation scene of the sheet metal when the mold is closed, which is used to simulate how the sheet metal is formed under the action of the mold in actual production.

[0038] like Figure 3 As shown, Figure 3 This is a schematic diagram of a three-dimensional simulation cloud map provided in this disclosure. Figure 3 The visualization results of the simulated mesh data obtained after the initial mold surface is simulated by the stamping process are shown, which can intuitively present the elastic-plastic deformation state of the sheet metal in the mold.

[0039] Step S104: Perform non-uniform offset processing on the simulated grid data to obtain an offset grid.

[0040] Specifically, non-uniform offset processing differs from the traditional "globally uniform distance" offset method. Based on the wall thickness distribution information in the simulated mesh data, it dynamically calculates the offset distance and offset direction of each vertex in the simulated mesh through a hybrid vertex offset algorithm, making the offset amount adaptable to the non-uniform deformation characteristics of different regions of the sheet metal. The offset mesh is an STL format discrete mesh model generated after non-uniform offset processing, which includes the offset element geometric coordinates, element topological relationships, and thickness characteristics with integrated springback pre-compensation logic.

[0041] For example, the simulated mesh data obtained in step S102 is imported, and a "vertex-cell" topological association is established to clarify the triangular cell to which each vertex belongs. A hybrid vertex offset algorithm of area weighting and cell thickness is used to perform calculations for each vertex of the simulated mesh: the offset vector of the vertex is determined by weighted averaging the normal vector of each cell to which the vertex belongs, the cell area, and the interior angle at the vertex; the projection distance of the thickness in the offset vector direction is calculated by combining the formed thickness value of each cell to which the vertex belongs, and then the non-equidistant offset distance of the vertex is determined by area weighted averaging; each vertex of the simulated mesh is moved according to the calculated offset vector and offset distance, and the vertices are reconnected according to the original cell topological relationship to generate an offset mesh in STL format.

[0042] like Figure 4 As shown, Figure 4 This is a schematic diagram of the thickness strain distribution and STL mesh element of a stamped part provided in this disclosure. Figure 4It demonstrates the differences in deformation, stress distribution, and springback in different regions of a stamped part. It enables visualization analysis of the spatial coordinates, deformation, and wall thickness characteristics of offset mesh elements before density clustering, providing data support for subsequent division of deformation feature clusters and extraction of local principal deformation features.

[0043] Step S106: Use density clustering algorithm to reconstruct the biased mesh into the compensated mold surface.

[0044] Specifically, density clustering algorithms include the density-based spatial clustering application (DBSCAN) algorithm. This algorithm, by setting the neighborhood radius (Eps) and the minimum number of neighborhood samples (MinPts) of the core point, divides spatially adjacent cells with similar deformation characteristics (based on the wall thickness change rate of the offset grid) into deformation feature clusters, while simultaneously identifying noisy cells, thus efficiently extracting key deformation areas of sheet metal. The compensated mold surface is a continuous curved surface model conforming to mold manufacturing standards, such as a NURBS surface. It is reconstructed from the offset grid after extracting key features through density clustering, and integrates springback compensation logic for non-equidistant offsets. It can be directly used for computer numerical control (CNC) machining programming or subsequent simulation verification.

[0045] For example, firstly, the wall thickness change rate of each element in the offset mesh is calculated, that is, the ratio of the difference between the thickness of the element after forming and the initial thickness of the sheet to the initial thickness, and an element spatial coordinate-wall thickness change rate dataset is constructed; then, DBSCAN density clustering is performed: set the clustering parameters (neighborhood radius, minimum number of neighborhood samples of the core point); cluster the element spatial coordinate-wall thickness change rate dataset, divide the offset mesh into multiple deformation feature clusters (each cluster represents a local area with similar deformation behavior), and filter out noisy elements; traverse each deformation feature cluster, select the element with the largest absolute value of the wall thickness change rate in the cluster (the element that best represents the deformation characteristics of the area), extract its geometric center point as the surface reconstruction control point, and compress the number of control points to less than 1% of the original number of offset mesh elements; with the extracted control points as the core input, under the framework of the original initial mold surface CAD reference data (such as IGES / STEP format continuous surface), NURBS surface fitting technology is used to perform reconstruction to generate a continuous and smooth compensated mold surface, ensuring that the surface inherits the springback compensation logic of the offset mesh and meets the mold processing accuracy requirements.

[0046] In this embodiment, the stamping process simulation based on the initial mold surface first obtains simulated mesh data containing wall thickness distribution, providing a precise physical basis for subsequent compensation. Then, the non-equidistant offset processing generates an offset mesh adapted to the uneven deformation of the sheet metal, overcoming the limitation of traditional equidistant offsets in dealing with the complex springback of high-strength steel. Finally, the density clustering algorithm is used to reconstruct the offset mesh, which, while compressing the number of control points for surface reconstruction to improve efficiency, accurately preserves the global deformation characteristics of the sheet metal, thereby effectively suppressing the springback of high-strength steel stamping and significantly improving the forming accuracy of stamped parts.

[0047] For example, the initial die surface includes the die surface obtained by equidistant offsetting the target surface of the stamped part.

[0048] Specifically, the equidistant offset processing is a method of generating the initial mold surface by globally translating along the normal of the target surface of the stamped part by a fixed distance (usually half the nominal thickness of the sheet metal). This method ignores the uneven deformation of the sheet metal during subsequent stamping processes and is only used as a benchmark for the simulation of the first stamping process.

[0049] For example, import the target surface of the stamped part of the lower control arm (STEP format) from the product design system, determine the sheet metal parameters as advanced high-strength steel DP600, and the nominal initial thickness as 1.5mm; calculate the equidistant offset distance: based on the nominal thickness of the sheet metal, determine the equidistant offset distance to be half the thickness, i.e., 0.75mm (because the mold needs to include a punch and a die, the punch surface is offset inward by 0.75mm along the normal of the target surface, and the die surface is offset outward by 0.75mm along the normal to ensure that the initial mold clearance matches the nominal thickness of the sheet metal); perform the equidistant offset: use the "equidistant surface function" of CAD software (such as UG / NX) to translate 0.75mm along the normal of each surface unit of the target surface to generate the punch and die surfaces of the initial mold, and export them as STL format as the initial geometric model for subsequent stamping process simulation.

[0050] Applying this embodiment, it is clear that the initial mold surface is generated by equidistant offset of the target surface of the stamped part, anchoring the reference starting point of the springback compensation process, and providing a comparison reference for subsequent non-equidistant offsets to ensure the continuity of the compensation logic.

[0051] For example, the non-equidistant offset processing includes: calculating the non-equidistant offset distance of shared vertices in the offset mesh based on a hybrid vertex offset algorithm; and moving the shared vertices along the offset vector based on the non-equidistant offset distance to generate the offset mesh.

[0052] Specifically, the hybrid vertex offset algorithm calculates the offset of vertices in a simulated mesh by combining "element area weighting" and "element thickness data." It simultaneously determines the offset direction and distance of a vertex, balancing geometric smoothness and physical deformation adaptability. Shared vertices are vertices in the simulated mesh shared by two or more triangular elements (basic building blocks in STL format). Their offset results directly affect the topological relationships of multiple elements, requiring assurance of continuity in element connections after offsetting. Non-equidistant offset distances are dynamic offset distances calculated for shared vertices, varying with element thickness. The offset distance differs between regions with drastic thickness changes and those with gentler changes, adapting to uneven deformation of the sheet metal. The offset vector is a three-dimensional vector that determines the offset direction of shared vertices. It is calculated by weighting the geometric features (normal vector, area, interior angles) of the adjacent elements to which the shared vertex belongs, ensuring that the offset direction is consistent with the local surface shape.

[0053] For example, the simulated mesh data obtained after the initial mold surface is simulated by the stamping process is imported into the offset calculation module; the algorithm automatically identifies the adjacent triangular elements to which each shared vertex belongs in the simulated mesh, and constructs a mapping table of shared vertices and adjacent elements; based on the hybrid vertex offset algorithm, combined with the forming thickness value of adjacent elements, the non-equidistant offset distance of each shared vertex is calculated by weighted averaging of the projection distance and area of ​​the element thickness in the offset vector direction; combined with the normal vector of adjacent elements, element area, and interior angle at the shared vertex, the offset vector of each shared vertex is calculated by weighted averaging (weight is the ratio of interior angle to area); each shared vertex is moved by a non-equidistant offset distance according to the offset vector direction, and the vertex coordinates are updated; according to the element topology relationship of the original simulated mesh, the updated vertices are reconnected as triangular elements to generate an offset mesh in STL format.

[0054] like Figure 5 As shown, Figure 5 This is a schematic diagram of an STL cell non-uniform thickness bias program interface provided in this disclosure. The interface includes an "Import NodeData..." button, an "Import ElementData..." button, a "Non-uniform thickness bias coefficient" input box (current value 0.6), a "Bias Mode" area (containing two radio buttons, "Non-uniform thickness bias" and "Uniform thickness bias," with "Non-uniform thickness bias" selected), a "Uniform thickness bias amount" input box (current value 0), an "Output file name" input box, a "Data bias" button, and an "End" button. Through... Figure 5The interface shown allows you to import NodeData and ElementData data from STL meshes. You can select either "Non-uniform thickness offset" or "Uniform thickness offset" editing mode, set the non-uniform thickness offset coefficient and uniform thickness offset coefficient, specify the output file name, and then perform the data offset operation to achieve the offset processing of STL elements. This generates offset mesh data for the springback compensation of the subsequent stamping die surface. After the operation is complete, you can click the "End" button to exit the interface.

[0055] In this embodiment, the non-equidistant offset distance of shared vertices is calculated by a hybrid vertex offset algorithm and the offset mesh is generated by moving the offset mesh. This can accurately adapt to the uneven deformation of sheet metal after stamping. Compared with the traditional equidistant offset, it greatly improves the pre-compensation accuracy for springback and lays a precise geometric foundation for subsequent reconstruction of the compensation surface.

[0056] For example, the bias vector is determined by a weighted average of the normal vectors, areas, and interior angles of multiple adjacent cells sharing a vertex.

[0057] For example, for a shared vertex in the simulated mesh, such as vertex i, with coordinates P i ( x i , y i , z i Extract all adjacent triangular elements from the given set of elements, and calculate the element normal vector for each adjacent triangular element. Unit area S and interior angles Among them, the element normal vector It is calculated by vector cross product based on the coordinates of the three vertices of the unit. For the specific calculation method, please refer to the following formula (1).

[0058] (1) in, p 1, p 2, p 3 represents the three vertices that make up the triangular unit, corresponding to the node coordinates in the simulated grid data. x 1, y 1, z 1; x 2, y 2, z 2; x 3, y 3, z 3 are the vertices p 1, p 2, p The spatial coordinates of 3 belong to the node data in the simulated grid data. i , j, k It is the unit vector in a spatial rectangular coordinate system, used to represent the component form of the normal vector. a , b , c Is the normal vector in x , y , z The directional component, calculated from the determinant, is used to quantify the spatial orientation of the normal vector. a The calculation method is shown in the following formula (2). b The calculation method is shown in the following formula (3). c The calculation method is shown in equation (4) below.

[0059] (2) (3) (4) For example, the unit area is calculated using Heron's formula, and the specific calculation method is shown in equation (5) below.

[0060] (5) Among them, parameters c This is the perimeter of the triangle. The specific calculation method is shown in equation (6) below. Parameter l 1, l 2, l 3 represents the lengths of the three sides of the triangular element, calculated from the node coordinates in the simulated mesh data.

[0061] (6) For example, interior angles are calculated using the Law of Cosines, and the specific calculation method is shown in equation (7) below.

[0062] (7) For example, the ratio of the interior angle to the cell area is used as the weighting weight of the cell's normal vector, reflecting the degree of influence of the cell on the bias direction of the shared vertex. That is, the larger the interior angle and the larger the area of ​​the cell, the stronger its contribution to the bias direction. The normal vectors of all adjacent cells are weighted and summed according to the above weights, and then divided by the total weights to obtain the weighted normal vector of the shared vertex. This vector is the bias vector of the shared vertex. Bias vector The specific calculation formula is shown in equation (8) below. Repeat the above steps to calculate the offset vector for each shared vertex in the simulated mesh, ensuring that the offset direction of each vertex is adapted to the local surface shape. For example, the offset vector of the area with larger surface curvature is more in line with the tangent direction of the surface, so as to avoid mesh distortion after offset.

[0063] (8) in, i The node number representing the shared vertex. j This indicates the number of the triangular unit that shares this vertex.

[0064] In this embodiment, the bias vector is determined by a weighted average of the normal vector, area, and interior angle of the adjacent cells of the shared vertex. This makes the bias direction more consistent with the local surface geometry, ensuring that the bias mesh generated after the shared vertex is moved is geometrically continuous and distortion-free, avoiding mesh quality degradation and ensuring the accuracy of subsequent density clustering and surface reconstruction.

[0065] For example, the simulated mesh data includes the post-forming thickness values ​​of multiple cells; the non-equidistant offset distance is determined by an area-weighted average based on the projected distance of the post-forming thickness values ​​of multiple adjacent cells sharing a vertex in the offset vector direction.

[0066] Specifically, the post-forming thickness value is the actual average thickness value of each triangular element after the stamping process simulation, recorded in the element data file of the simulated mesh data. It is the core physical data reflecting the uneven deformation of the sheet metal (e.g., the post-forming thickness value of the stretched area element is less than the initial thickness, while the extruded area element is greater than the initial thickness). The projection distance is the projection length of the post-forming thickness value of the element in the direction of the shared vertex offset vector.

[0067] For example, the formed thickness value of each adjacent cell (the cell to which the shared vertex belongs) is read from the ElementData file. t ij When calculating the projected distance, the normal vector is first calculated using the dot product formula. With bias vector included angle The cosine of the included angle is given in equation (9) below. Then, the projection distance of the unit thickness in the offset vector direction is calculated. d ij See equation (10) below.

[0068] (9) (10) in, t ij Triangular unit j The thickness is input from the ElementData file.

[0069] For example, in calculating the projection distance d ij Then, using the cell area as the weight, a weighted average of the projected distances of all adjacent cells is calculated to generate the offset distance of the shared vertices. di See equation (11) below. Then, based on the material properties of the stamped parts (such as the springback tendency of DP steel) and the die clearance requirements, an adjustable offset coefficient k is introduced (usually ranging from 0.8 to 1.2, in this embodiment k=1.0), and the non-equidistant offset distance is calculated. See equation (12) below. Repeat the above steps to calculate the final non-equidistant offset distance for each shared vertex in the simulated mesh, ensuring that areas with drastic thickness changes (such as...) are properly offset. d ij Larger units have greater offset distances to their vertices, enabling targeted bounce compensation.

[0070] (11) in, A ij It is the first j The area of ​​each triangular unit.

[0071] (12) It should be noted that the offset coefficient k is set based on engineering practice experience. Its core function is to precisely fine-tune non-equidistant offset distances to adapt to the personalized needs of actual production scenarios. For example, when a specific fitting clearance needs to be reserved between the stamping sheet and the die (to reduce forming friction and extend the die's service life), k can be set to a value greater than 1.0; if it is necessary to improve the surface fitting accuracy and reduce clearance redundancy, k can be set to a value less than 1.0. The specific optimization of this coefficient needs to be comprehensively determined by combining the stamping die structure design, forming process parameters (such as blank holder force and friction coefficient), batch fluctuations of sheet metal, and other engineering factors. Its control logic is related to the stamping die design and process optimization system.

[0072] Furthermore, based on the bias vector Non-equidistant offset distance The coordinates of each vertex after offset are calculated as shown in equations (13) to (15).

[0073] (13) (14) (15) Among them, P i ( x i , y i , z i ) are the node coordinates before offset. These are the offset node coordinates. u i , vi , w i These are the bias vectors. In x , y , z Components in direction.

[0074] For example, by substituting the offset node coordinates back into the above equation (1), the normal vector of each triangular element after offset is calculated. The offset node coordinates of all triangular elements and the normal vectors calculated based on the offset node coordinates are output according to the standard data format of the STL file, and the offset mesh can be obtained.

[0075] Figure 6 This is a schematic diagram of a vertex offset provided in this disclosure. Figure 6 It shows the state of shared vertices moving along the bias vector during the bias mesh generation process. The arrows intuitively illustrate the process of shared vertices moving along the bias vector direction after the non-equidistant bias distance is calculated based on the hybrid vertex bias algorithm.

[0076] Figure 7 This is a schematic diagram of a non-equidistant offset stamping die surface provided in this disclosure. Figure 7 The diagram illustrates the morphological differences of the stamping die surface after non-equidistant offset processing. Point P is the shared vertex before offset, while point P' is the offset vertex obtained by moving along the offset vector after calculating the non-equidistant offset distance based on a hybrid vertex offset algorithm. The surrounding values, such as -0.041 and -0.068, reflect the differences in non-equidistant offset amounts in different regions. Points A and B in the upper right corner of the diagram mark the characteristic areas of the die surface, helping to illustrate the application effect of non-equidistant offset at these key locations. This visually reflects the technical logic of this embodiment, which uses non-equidistant offset to compensate for the springback of the stamping die surface, enabling the die surface shape to accurately adapt to the uneven deformation of the sheet metal.

[0077] In this embodiment, the non-equidistant offset distance is determined based on the projected distance of the thickness value after forming and the area-weighted average, which binds the offset distance calculation to the actual deformation depth of the sheet metal. This makes the compensation logic of the offset grid fully conform to the physical deformation law, further improving the accuracy of springback compensation and better adapting to the compensation of complex deformation materials such as high-strength steel.

[0078] For example, reconstructing the offset mesh into a compensated mold surface using a density clustering algorithm includes: performing density clustering on the spatial coordinates and wall thickness change rate of multiple cells in the offset mesh to divide the offset mesh into multiple deformation feature clusters; extracting local principal deformation features from the multiple deformation feature clusters; using the local principal deformation features as control points to perform surface fitting on the reference mold surface data of the initial mold surface to generate the compensated mold surface.

[0079] Specifically, deformation feature clusters are sets of spatially adjacent units with similar wall thickness change rates, partitioned using a density clustering algorithm. Each cluster represents a local deformation region on the sheet metal (e.g., stretching thinning clusters, extrusion thickening clusters). Local principal deformation features are the unit features within each deformation feature cluster that best represent the degree of deformation in that cluster. They are defined as "the unit with the largest absolute value of the wall thickness change rate within the cluster and its geometric center coordinates," serving as the source of key control points for surface reconstruction. The reference mold surface data for the initial mold shape is the industrial-grade CAD data (e.g., a continuous surface model in IGES format) corresponding to the initial mold shape. This data serves as the geometric framework for surface reconstruction, ensuring topological compatibility between the reconstructed compensation surface and the initial design. Surface fitting processing uses local principal deformation features as control points and employs NURBS surface fitting technology to generate a continuous, smooth surface based on the initial mold reference mold surface data.

[0080] For example, for the biased mesh, the formed thickness value of each cell is extracted. t ij With the initial thickness of the board t 0, based on the wall thickness variation rate of each unit (i.e., | t ij - t 0| / t 0) Construct a cell spatial coordinate-wall thickness change rate dataset. Clustering parameters were determined using the k-distance map method: neighborhood radius Eps = 0.8 mm (cell spatial distance threshold), minimum neighborhood sample size MinPts = 10 (a cell is considered a core point if its Eps neighborhood contains at least 10 cells). The DBSCAN algorithm was executed on the cell spatial coordinate-wall thickness change rate dataset to divide the offset mesh into 8 deformation feature clusters (including 2 stretching thinning clusters, 3 extrusion thickening clusters, and 3 slight deformation clusters), while filtering out 5 noisy cells. Each deformation feature cluster was traversed, and the cell with the largest absolute value of wall thickness change rate within the cluster was selected. The geometric center coordinates of this cell (calculated by averaging the coordinates of the cell's three vertices) were extracted as control points for surface reconstruction, resulting in 92 control points (the original offset mesh had 17673 cells, reducing the number of control points to less than 1% of the original number). 92 control points are imported into the CAD surface reconstruction module. The reference mold surface data (IGES format) of the initial mold shape is used as the geometric framework. NURBS surface fitting technology (least square approximation) is used to generate a continuous surface. During the fitting process, control point constraints are used to ensure that the non-equidistant compensation logic of the surface and the offset mesh is consistent (e.g., the compensation amount of the surface area corresponding to the stretching and thinning cluster is larger). Finally, the compensated mold surface (IGES format, which can be directly used for CNC machining programming) is generated.

[0081] For example, silhouette coefficients are used in the implementation of clustering algorithms. S iEvaluate the rationality of the clustering algorithm and parameter settings. S i For the specific calculation method, please refer to the following formula (16).

[0082] (16) in, For the sample i The average distance to other data points in the same cluster. b i For the sample i Inter-cluster similarity. and b i For the specific calculation method, please refer to equations (17) and (18) below.

[0083] (17) Equation (18) in, m This represents the number of data points contained in the cluster. n The total number of clusters, c for b i The cluster number it belongs to. For the sample i With the k Cluster similarity, i.e., sample similarity i To the k The average distance between all points in the cluster. For the specific calculation method, please refer to the following formula (19).

[0084] (19) in, For the first k The number of samples contained in a cluster. S i The closer to 1, the better the sample. i The average distance to a point within a cluster is much smaller than the distance to any other cluster, i.e., the sample i The intra-cluster similarity is much higher than the inter-cluster similarity, and the samples i The more reasonable the clustering, the better. Conversely, S i The closer to -1, the better the sample. i The average distance to points within a cluster is much greater than the distance to any other cluster, i.e., the sample i The intra-cluster similarity is much lower than the inter-cluster similarity, and the samples i The more unreasonable the clustering, the better. S i A value close to 0 indicates that the sample size is close to 0. i The average distance to a point within a cluster is close to the distance to another point in a cluster, i.e., the sample...i The intra-cluster similarity and the inter-cluster similarity are close, and the sample i is on the boundary of two clusters.

[0085] Figure 8 This is a distribution schematic diagram of the 200 units with the largest absolute value of thinning rate provided by the present disclosure on the stamping part. According to Figure 8 , the key units that can best represent the local deformation of the sheet material can be quickly identified, and then used as control points for surface fitting, and finally the die surface that precisely adapts to the springback compensation requirements is generated.

[0086] Figure 9 This is a schematic diagram of the principle of a density clustering algorithm provided by the present disclosure. The point A marked in the figure is the core point, and the points in its neighborhood, etc., form clusters that are mutually reachable because of their proximity in space position and the density satisfying the conditions; the point B belongs to a noise point (isolated point) because of the low density of the surrounding points. In this way, the algorithm can divide the units with proximity in space position and similar deformation characteristics into deformation feature clusters.

[0087] Figure 10 This is a flow chart for clustering offset grid units based on a density clustering algorithm provided by the present disclosure.

[0088] Step S1002: Process all triangular patch units p in the unit set D output by the forming analysis, and mark them as unprocessed.

[0089] Step S1004: Determine whether p has been assigned to a certain cluster or marked as noise. If so, continue to judge the next p; if not, execute step S1006.

[0090] Step S1006: Process all triangular patch units p in the unit set D output by the forming analysis, mark them as unprocessed, and check the neighborhood Nεps (p) of p.

[0091] Step S1008: Determine whether the number of objects included in Nεps (p) is < MinPts. If so, execute step S1010; if not, execute step S1012.

[0092] Step S1010: Mark p as a boundary point or a noise point, and return to continue judging the next unit p.

[0093] Step S1012: Mark p as a core point, establish a new unit cluster C, and assign all points in the neighborhood to C.

[0094] Step S1014: Process all unprocessed units q in Nεps (p).

[0095] Step S1016: Determine whether the number of objects contained in Nεps(q) is ≥ MinPts. If yes, proceed to step S1018; otherwise, proceed to step S1020.

[0096] Step S1018: Add all cells in Nεps(q) that are not assigned to any cluster to C.

[0097] Step S1020: Determine whether all units q have been processed. If not, return to step S1014; if yes, proceed to step S1022.

[0098] Step S1022: Determine whether all p have been processed. If not, return to step S1004; if yes, proceed to step S1024.

[0099] Step S1024: Process ends.

[0100] Figure 11 This is a schematic diagram of the surface reconstruction deviation distribution without using the DBSCAN clustering algorithm, as provided in this disclosure. Figure 12 This is a schematic diagram of the surface reconstruction deviation distribution optimized using the DBSCAN clustering algorithm, as provided in this disclosure. Figure 11 Although the peak values ​​of the deviation histogram are concentrated, the overall distribution is relatively dispersed. Figure 12 The peak values ​​of the deviation histogram are more concentrated around 0, indicating that the concentration of the deviation distribution is significantly improved after optimization. Without DBSCAN clustering, the uneven distribution of surface reconstruction deviations makes it difficult to achieve the required fit between the formed sheet and the target surface; after optimization using DBSCAN clustering, the precise control of deviations greatly improves the fit between the formed sheet and the target surface, and the final surface accuracy of the stamped part better meets the design requirements.

[0101] This embodiment extracts local principal deformation features through density clustering and performs surface fitting based on the initial mold reference data. This not only accurately extracts key deformation features to significantly reduce redundant data, but also ensures the topological compatibility of the compensation surface with the original design. At the same time, it inherits the non-equidistant compensation logic of the offset mesh, realizing efficient and high-precision conversion from discrete mesh to industrial-grade continuous mold surface, thus ensuring the accuracy of mold CNC machining and mass production.

[0102] For example, after reconstructing the offset mesh into a compensated mold surface using a density clustering algorithm, the method further includes: determining the fitting rate between the forming sheet and the target surface of the stamping part corresponding to the compensated mold surface; if the fitting rate is lower than a preset threshold, performing stamping process simulation on the compensated mold surface to obtain new simulated mesh data, and returning to perform non-equidistant offset processing on the simulated mesh data to obtain an offset mesh, until the fitting rate is not lower than the preset threshold.

[0103] Specifically, the fit rate is the geometric overlap between the formed sheet (the simulated part) and the target surface of the stamped part after the die surface has been simulated for stamping process. It is the core indicator for evaluating the compensation effect and is calculated as: (Area of ​​overlap between the formed sheet and the target surface / Total area of ​​the target surface) × 100%. The preset threshold is a pre-set standard for the fit rate. Based on the assembly requirements of high-strength steel stamped parts, the preset threshold is set to 99% (i.e., when the fit rate is ≥ 99%, the compensated die surface meets the design requirements).

[0104] For example, the generated compensated mold surface (IGES format) is imported into stamping CAE simulation software (such as PAM-STAMP). Using the sheet metal parameters and process parameters, a new round of stamping forming and springback simulation is performed, outputting the STL mesh data of the formed sheet metal. Using a 3D deviation analysis tool, the STL mesh of the formed sheet metal is aligned with the target surface of the stamped part (STEP format), and the deviation is calculated. The overlapping area (area with deviation ≤ 0.1 mm) is counted, and the fitting rate is calculated. If the fitting rate is less than 99%, iterative optimization is initiated. The new simulation mesh data (including new ElementData thickness values ​​after forming) output from the verification simulation is extracted and replaced with the initial simulation mesh data. The non-equidistant offset processing step is returned to generate a new offset mesh based on the new simulation mesh data. The new offset mesh is reconstructed into the mold surface after a new round of compensation using a density clustering algorithm. The "verification simulation → calculation of fitting rate → iterative optimization" process is repeated until the fitting rate is ≥ 99%. The final result is output and delivered to the mold manufacturing stage for CNC machining.

[0105] By applying this embodiment, the compensation mold surface is continuously optimized through closed-loop iterative verification with the fitting rate as a quantitative indicator, ensuring that the fitting rate between the final formed sheet and the target surface meets the standard. This completely solves the problem that traditional methods rely on experience and that one compensation is difficult to meet the accuracy requirements. It realizes automated and standardized iteration of springback compensation, significantly improving the mold development efficiency and the consistency of the forming accuracy of the final part.

[0106] and Figure 13 Corresponding to the method embodiments shown, this disclosure also provides embodiments of a springback compensation device for the shape surface of a stamping die. Figure 13 This is a schematic diagram illustrating the composition of a springback compensation device for the surface of a stamping die provided in this disclosure. Figure 13 As shown, the springback compensation device 1300 for the stamping die surface includes: Simulation module 1302: is configured to perform stamping process simulation on the initial mold surface to obtain simulated mesh data.

[0107] Processing module 1304 is configured to perform non-uniform offset processing on the simulated mesh data to obtain an offset mesh.

[0108] Reconstruction module 1306: is configured to reconstruct the biased mesh into a compensated mold surface using a density clustering algorithm.

[0109] For example, the initial die surface includes the die surface obtained by equidistant offsetting the target surface of the stamped part.

[0110] For example, the processing module 1304 is further configured to: calculate the non-equidistant offset distance of shared vertices in the offset mesh based on the hybrid vertex offset algorithm; and move the shared vertices along the offset vector based on the non-equidistant offset distance to generate the offset mesh.

[0111] For example, the bias vector is determined by a weighted average of the normal vectors, areas, and interior angles of multiple adjacent cells sharing a vertex.

[0112] For example, the simulated mesh data includes the post-forming thickness values ​​of multiple cells; the non-equidistant offset distance is determined by an area-weighted average based on the projected distance of the post-forming thickness values ​​of multiple adjacent cells sharing a vertex in the offset vector direction.

[0113] For example, the reconstruction module 1306 is further configured to: perform density clustering processing on the spatial coordinates and wall thickness change rate of multiple cells in the bias mesh, and divide the bias mesh into multiple deformation feature clusters; extract local principal deformation features from the multiple deformation feature clusters; use the local principal deformation features as control points to perform surface fitting processing on the reference mold surface data of the initial mold surface, and generate the compensated mold surface.

[0114] For example, the springback compensation device 1300 for the stamping die surface also includes a verification module configured to: determine the fitting rate between the forming sheet corresponding to the compensated die surface and the target surface of the stamped part; if the fitting rate is lower than a preset threshold, perform stamping process simulation on the compensated die surface to obtain new simulated mesh data, and return to perform non-equidistant offset processing on the simulated mesh data to obtain an offset mesh step, until the fitting rate is not lower than the preset threshold.

[0115] The above is a schematic scheme of a springback compensation device for the surface of a stamping die provided in this disclosure. The technical solution of this springback compensation device for the surface of a stamping die belongs to the same concept as the technical solution of the springback compensation method for the surface of a stamping die described above. For details not described in detail in the technical solution of the springback compensation device for the surface of a stamping die, please refer to the description of the technical solution of the springback compensation method for the surface of a stamping die described above.

[0116] Please refer to Figure 14This diagram illustrates a structural block diagram of a computing device provided in an exemplary embodiment of this disclosure. In some examples, the computing device 140 can be at least one of devices such as a smartphone, smartwatch, desktop computer, laptop, virtual reality terminal, augmented reality terminal, wireless terminal, and laptop computer. The computing device 140 has communication capabilities and can access wired or wireless networks. The computing device 140 can refer to one of multiple terminals, and those skilled in the art will understand that the number of such terminals can be more or less. In some examples, the computing device 140 can receive initial mold surface data based on the accessed wired or wireless network. It is understood that the computing device 140 undertakes the calculation and processing work of the technical solution of this disclosure, and this disclosure does not limit it in this regard.

[0117] like Figure 14 As shown, the computing device in this disclosure may include one or more of the following components: processor 1410 and memory 1420.

[0118] Optionally, the processor 1410 connects various parts within the computing device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1420, and by calling data stored in the memory 1420. Optionally, the processor 1410 can be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 1410 can integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), Neural-network Processing Unit (NPU), and baseband chip. Specifically, the CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required to be displayed on the touch screen; the NPU is used to implement Artificial Intelligence (AI) functions; and the baseband chip is used to handle wireless communication. It is understandable that the aforementioned baseband chip may not be integrated into the processor 1410, but may be implemented as a separate chip.

[0119] The memory 1420 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 1420 may include a non-transitory computer-readable storage medium. The memory 1420 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 1420 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data created according to the use of the computing device, etc.

[0120] In addition, those skilled in the art will understand that the structure of the computing device shown in the above figures does not constitute a limitation on the computing device. The computing device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the computing device may also include a display screen, camera assembly, microphone, speaker, radio frequency circuit, input unit, sensors (such as accelerometer, angular velocity sensor, light sensor, etc.), audio circuit, WiFi module, power supply, Bluetooth module, etc., which will not be described in detail here.

[0121] This disclosure also provides a computer-readable storage medium storing at least one instruction that is executed by a processor to implement the springback compensation method for the stamping die surface as described in the various embodiments above.

[0122] This disclosure also provides a computer program product including computer instructions stored in a computer-readable storage medium; a processor of a computing device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computing device to perform the springback compensation method for the stamping die surface described in the above embodiments.

[0123] Those skilled in the art will recognize that the functions described in this disclosure in one or more of the examples above can be implemented using hardware, software, firmware, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of a computer program from one place to another. Storage media can be any available medium accessible to a general-purpose or special-purpose computer.

[0124] It should be noted that the technical solutions described in this disclosure can be combined arbitrarily as long as they do not conflict.

[0125] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for springback compensation of a stamping die surface, characterized in that, include; A stamping process simulation was performed on the initial mold surface to obtain simulated mesh data; The simulated grid data is subjected to non-uniform offset processing to obtain an offset grid; The biased mesh is reconstructed into a compensated mold surface using a density clustering algorithm.

2. The springback compensation method for the stamping die surface according to claim 1, characterized in that, The initial mold surface includes the mold surface obtained after the target surface of the stamped part is offset by an equal distance.

3. The springback compensation method for the stamping die surface according to claim 1, characterized in that, The non-equidistant offset processing includes: Based on the hybrid vertex bias algorithm, the non-equidistant bias distance of shared vertices in the biased mesh is calculated; The shared vertex is moved along the bias vector based on the non-equidistant bias distance to generate the bias mesh.

4. The springback compensation method for the stamping die surface according to claim 3, characterized in that, The bias vector is determined by a weighted average of the normal vectors, areas, and interior angles of multiple adjacent cells of the shared vertex.

5. The springback compensation method for the stamping die surface according to claim 3, characterized in that, The simulated mesh data includes the thickness values ​​of multiple cells after forming; The non-equidistant offset distance is determined by an area-weighted average based on the projection distance of the post-forming thickness values ​​of multiple adjacent units of the shared vertex in the offset vector direction.

6. The springback compensation method for the stamping die surface according to claim 1, characterized in that, The step of reconstructing the biased mesh into a compensated mold surface using a density clustering algorithm includes: Density clustering is performed on the spatial coordinates and wall thickness change rate of multiple cells in the offset grid to divide the offset grid into multiple deformation feature clusters; Extract local principal deformation features from the plurality of deformation feature clusters; Using the local principal deformation features as control points, surface fitting processing is performed on the reference mold surface data of the initial mold surface to generate the compensated mold surface.

7. The springback compensation method for the surface of a stamping die according to any one of claims 1-6, characterized in that, The method further includes: Determine the fit rate between the formed sheet material corresponding to the compensated mold surface and the target surface of the stamped part; If the bonding rate is lower than a preset threshold, a stamping process simulation is performed on the compensated mold surface to obtain new simulated mesh data, and the process of performing non-equidistant offset processing on the simulated mesh data to obtain an offset mesh is repeated until the bonding rate is not lower than the preset threshold.

8. A springback compensation device for the surface of a stamping die, characterized in that, include; The simulation module is configured to perform stamping process simulation on the initial mold surface to obtain simulated mesh data; The processing module is configured to perform non-uniform offset processing on the simulated grid data to obtain an offset grid; The reconstruction module is configured to reconstruct the biased mesh into a compensated mold surface using a density clustering algorithm.

9. A computer device, characterized in that, include: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the springback compensation method for the stamping die surface as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the springback compensation method for the stamping die surface as described in any one of claims 1-7.

Images