Simulation program, press forming simulation method, and press forming simulation apparatus

Abstract

This invention provides a press forming simulation method that can predict the shape of a press-formed product after springback, regardless of the tensile strength of the blank material. [Solution] Shape data and strain data of the press-formed product at the bottom dead center are calculated by forming simulation based on mold data and blank material shape data. Actual formed outer shape data showing the outer shape after springback is obtained. Based on the shape data at the bottom dead center and the obtained actual formed outer shape data, the stress of the press-formed product at the bottom dead center is calculated. The relationship between the strain data and stress of the press-formed product at the bottom dead center is machine-learned into a learning model. The mold data is corrected. Shape data and strain data of the press-formed product at the bottom dead center are calculated by forming simulation based on the corrected mold. The calculated strain data of the press-formed product is input into the learning model to calculate the stress and calculate the shape after springback.

Description

Technical Field

[0001] The present disclosure relates to a simulation program, a press forming simulation method, and a press forming simulation apparatus.

Background Art

[0002] In the press forming of a blank material such as a metal plate, springback may occur due to the stress of the formed product at the bottom dead center, and a formed product with the target shape may not be obtained. Springback refers to a phenomenon in which, after a material undergoes plastic deformation by press forming, when the external force is removed, the formed product at the bottom dead center tries to return to its original shape due to the stress.

[0003] As a countermeasure against springback, by analysis using the finite element method (FEM analysis), the shape and stress of the press-formed product at the bottom dead center are simulated, and the difference between the target shape of the formed product and the shape after springback is reflected in the die shape (for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, there is a technical problem that the greater the tensile strength of the blank material, the more difficult it is to predict springback. When the prediction accuracy of the shape after springback decreases, it becomes difficult to determine the die shape considering springback.

[0006] The purpose of this disclosure is to provide a simulation program, a press forming simulation method, and a press forming simulation apparatus that can predict the shape of a press-formed product after springback, regardless of the tensile strength of the blank material. [Means for solving the problem]

[0007] A simulation program according to one aspect of this disclosure includes the steps of: calculating shape data and strain data of a press-formed product at the bottom dead center by performing a forming simulation on a computer based on die data relating to the die face shape, blank material shape data, and material property data relating to the blank material; acquiring actual formed outer shape data showing the outer shape of a press-formed product after springback obtained by actual press forming of the blank material; calculating stress of the press-formed product at the bottom dead center based on the shape data at the bottom dead center or after springback or the blank material shape data and the acquired actual formed outer shape data; and performing a pressing simulation at the bottom dead center based on the strain data and stress data of the press-formed product at the bottom dead center. The process involves generating a machine learning model that outputs the stress of a press-formed product at the bottom dead center when strain data of a molded product is input; modifying the die face; calculating the shape data and strain data of the press-formed product at the bottom dead center through a molding simulation based on the mold data after die face modification, blank material shape data, and material property data related to the blank material; inputting the calculated strain data of the press-formed product into the learning model to calculate the stress of the press-formed product; and calculating the shape of the press-formed product after springback based on the calculated shape data and stress of the press-formed product at the bottom dead center.

[0008] A press forming simulation method according to one aspect of the present disclosure includes the steps of: calculating shape data and strain data of a press-formed product at the bottom dead center by forming simulation based on die data relating to the die face shape, blank material shape data, and material property data relating to the blank material; acquiring actual formed outer shape data showing the outer shape of a press-formed product after springback obtained by actual press forming of the blank material; calculating stress of the press-formed product at the bottom dead center based on the shape data at the bottom dead center or after springback or blank material shape data and the acquired actual formed outer shape data; and calculating stress of the press-formed product at the bottom dead center based on strain data and stress data of the press-formed product at the bottom dead center The method includes the steps of: generating a machine learning model that outputs the stress of a press-formed product at the bottom dead center when strain data of the press-formed product is input; modifying the die face; calculating the shape data and strain data of the press-formed product at the bottom dead center by performing a forming simulation based on the mold data, blank material shape data, and material property data after the die face modification; inputting the calculated strain data of the press-formed product into the machine learning model to calculate the stress of the press-formed product; and calculating the shape of the press-formed product after springback based on the calculated shape data and stress of the press-formed product at the bottom dead center.

[0009] A press forming simulation apparatus according to one aspect of the present disclosure is a press forming simulation apparatus comprising a processing unit, wherein the processing unit calculates shape data and strain data of a press-formed product at the bottom dead center by performing a forming simulation based on die data relating to the die face shape, blank material shape data and material property data relating to the blank material, obtains actual formed outer shape data showing the outer shape of the press-formed product after springback obtained by actual press forming of the blank material, calculates the stress of the press-formed product at the bottom dead center based on the shape data at the bottom dead center or after springback or the blank material shape data and the obtained actual formed outer shape data, and calculates the strain data of the press-formed product at the bottom dead center. Based on the die and stress data, a machine learning model is generated that outputs the stress of the press-formed product at the bottom dead center when strain data of the press-formed product at the bottom dead center is input. The die face is modified, and the shape data and strain data of the press-formed product at the bottom dead center are calculated by a forming simulation based on the die face modified, blank material shape data, and material property data. The calculated strain data of the press-formed product is input to the machine learning model to calculate the stress of the press-formed product, and the shape of the press-formed product after springback is calculated based on the calculated shape data and stress of the press-formed product at the bottom dead center. [Effects of the Invention]

[0010] According to this disclosure, the shape of the press-formed product after springback can be predicted regardless of the tensile strength of the blank material. [Brief explanation of the drawing]

[0011] [Figure 1] This is a schematic diagram illustrating an example of the configuration of a press forming simulation apparatus according to Embodiment 1. [Figure 2] This flowchart shows the processing procedure for the press forming process simulation according to Embodiment 1. [Figure 3] This is a conceptual diagram of the press forming process simulation according to Embodiment 1. [Figure 4] This is a flowchart showing the procedure for generating a learning model according to Embodiment 1. [Figure 5] This is a conceptual diagram illustrating a method for calculating stress at the bottom dead center during actual press forming. [Figure 6] This is a conceptual diagram showing the first example of input and output data for a learning model. [Figure 7] This is a conceptual diagram showing a second example of input and output data for a learning model. [Figure 8] This is a flowchart showing the procedure for generating a learning model according to Embodiment 2. [Figure 9] This is a conceptual diagram illustrating a method for calculating stress at the bottom dead center during actual press forming. [Modes for carrying out the invention]

[0012] A simulation program, press forming simulation method, and press forming simulation apparatus according to embodiments of this disclosure will be described below with reference to the drawings. This disclosure is not limited to these examples, but is intended to include all modifications within the meaning and scope of the claims, as indicated by the claims. Furthermore, at least some of the embodiments described below may be combined in any way.

[0013] (Embodiment 1) Figure 1 is a schematic diagram illustrating an example configuration of a press forming simulation device 1 according to Embodiment 1. The press forming simulation device 1 is a computer that implements the press forming simulation method according to Embodiment 1, and comprises a processing unit 11, a display unit 12, an operation unit 13, a data input unit 14, and a storage unit 15. Each unit is connected by a bus. The press forming simulation method according to Embodiment 1 is a method that can calculate the shape after springback with sufficient accuracy, even for blank materials with high tensile strength such as high-tensile steel, by learning the relationship between the CAE calculation results of the press forming simulation and the press-formed product obtained using an actual machine.

[0014] Incidentally, the press forming simulation device 1 may be a stand-alone computer or a server device connected to a network. Further, the press forming simulation device 1 may be a computer in an on-premises environment or a computer such as a server in a cloud environment. The press forming simulation device 1 may be configured by a plurality of computers for distributed processing, may be realized by a plurality of virtual machines provided in one server, or may be realized using a cloud server.

[0015] The processing unit 11 is a processor having an arithmetic circuit such as a CPU (Central Processing Unit), internal storage devices such as a ROM (Read Only Memory) and a RAM (Random Access Memory), I / O terminals, a timing unit, etc. The processing unit 11 preferably has a configuration having a plurality of arithmetic cores. The processing unit 11 executes a simulation program (program product) 151 stored in the storage unit 15 described later to implement the press forming simulation method according to the first embodiment. Each functional unit of the press forming simulation device 1 may be realized software-wise or partially or entirely hardware-wise.

[0016] The display unit 12 is, for example, a display device such as a liquid crystal panel or an organic EL (Electro Luminescence) display.

[0017] The operation unit 13 is, for example, an input device such as a hardware keyboard, a pointing device, a touch panel, etc. The user of the press forming simulation device 1 can input arbitrary information to the press forming simulation device 1 using the operation unit 13. Incidentally, the operation unit 13 may be integrally configured with the display unit 12.

[0018] The data input unit 14 is an interface through which data is externally input. The data input unit 14 is, for example, a USB port, a wired communication port, a wireless communication port, or the like. In the present embodiment, actual molding outer shape data indicating the outer shape of the press-molded product after springback obtained by actual press molding is input. The actual molding outer shape data is, for example, image data obtained by imaging the press-molded product with an imaging device, distance measurement data obtained by measuring the distance to the surface of the press-molded product with a distance measurement sensor, 3D data (for example, point cloud data) obtained by scanning the press-molded product with a 3D scanner, or data indicating the outer shape of the press-molded product obtained by other measuring instruments. The processing unit 11 can acquire the actual molding outer shape data of the press-molded product via the data input unit 14.

[0019] The storage unit 15 has, for example, a main storage unit and an auxiliary storage unit. The main storage unit temporarily stores data necessary for the processing unit 11 to execute arithmetic processing. The auxiliary storage unit stores the simulation program 151 and the learning model 152 executed by the processing unit 11. The storage unit 15 stores the simulation program 151 and the learning model 152 executed by the processing unit 11.

[0020] The simulation program 151 is a program for causing a computer to execute a process of learning the relationship between the result of the press molding simulation and the shape of the press-molded product obtained using the actual machine and simulating the shape after springback. Note that the simulation program 151 includes a program for simulating the press molding of the blank material and a program for simulating the springback of the press-molded product.

[0021] The learning model 152 is a machine learning algorithm, such as a neural network (NN), that learns the relationship between the results of press forming simulations and press-formed products obtained using actual machines. The learning model 152 primarily learns the relationship between the strain and stress of the press-formed product at the bottom dead center. The data required to train the learning model 152 is sufficient if it is actual forming data from one press-formed product. Details of the learning model 152 will be described later.

[0022] The auxiliary storage unit may be an external storage device connected to the press forming simulation apparatus 1. The simulation program 151 may be written to the storage unit 15 during the manufacturing stage of the press forming simulation apparatus 1, or it may be distributed by an external server and acquired by the press forming simulation apparatus 1 via communication and stored in the storage unit 15. The simulation program 151 may also be recorded in a readable form on a recording medium 10 such as a magnetic disk, optical disk, or semiconductor memory, and a reader may read it from the recording medium 10 and store it in the storage unit 15.

[0023] Figure 2 is a flowchart showing the processing procedure of the press forming process simulation according to Embodiment 1, and Figure 3 is a conceptual diagram of the press forming process simulation according to Embodiment 1.

[0024] The processing unit 11 of the press forming simulation apparatus 1 acquires mold data representing the shape of the die face according to the target shape of the molded product via the data input unit 14 (step S11), and acquires graphic data representing the shape of the blank material and material property data (step S12). The graphic data of the blank material is 2D or 3D CAD data. In this embodiment, the blank material is plate-shaped. The material property data includes information related to the physical properties of the material, such as a stress-strain curve representing the relationship between stress and strain applied to the blank material. The processing unit 11 also acquires press forming conditions such as press force, press speed, and mold temperature (step S13).

[0025] Next, the processing unit 11 defines the geometric outline of the blank material as an analysis domain based on the blank material geometric data acquired in step S12, and creates a finite element model of a mesh structure by dividing the analysis domain into multiple finite elements (step S14). Hereinafter, the finite element model of the blank material will be referred to as the blank material shape data.

[0026] Next, the processing unit 11 performs a press forming simulation based on the mold data, blank material shape data, and material property data (step S15). The press forming simulation calculates the shape data of the press-formed product at the bottom dead center and the strain data for each finite element. The bottom dead center is the position or state where the vertically moving slide of the press machine reaches its lowest point.

[0027] On the other hand, the user manufactures a die with a die face corresponding to the target shape (step S1), and then test-presses a blank material using the manufactured die (step S2). This test pressing is an actual press forming process using a press machine. When the press-formed product is removed from the die, springback occurs as the stress on the product at the bottom dead center is released. Some of the deformation applied to the product is restored. Hereinafter, the product that has undergone actual press forming and springback will be referred to as the actual panel.

[0028] Next, the user measures the outer shape of the actual panel after springback (step S3). By measuring the outer shape of the actual panel, actual panel outer shape data (actual molded outer shape data) showing the outer shape of the actual panel after springback is obtained, and the actual panel outer shape data is input to the press forming simulation device 1.

[0029] The processing unit 11 acquires the actual panel outline data input to the press forming simulation device 1 (step S16).

[0030] Then, the processing unit 11 generates a machine learning model 152 that outputs the stress of the press-formed product at the bottom dead center when strain data of the press-formed product at the bottom dead center is input (step S17).

[0031] In the press forming simulation method according to this embodiment, the results of CAE calculations using the finite element method are corrected by machine learning the relationship between the results of CAE calculations and press-formed products obtained using an actual machine.

[0032] The concept of this embodiment will now be explained. Regarding the shape and strain of the press-formed product at the bottom dead center, there is no difference between the product obtained by simulation using the finite element method and the product obtained by actual press forming, and it is assumed that the stress at the bottom dead center is different. The problem lies in how to determine the stress at the bottom dead center of a molded product obtained by actual press forming. In this embodiment, a finite element model of the actual panel after springback is created, and the stress generated in each finite element is determined by forcibly elastically deforming the actual panel so that the shape represented by this finite element model becomes the shape of the molded product at the bottom dead center. Note that the mesh structure of the finite element model of the actual panel must be geometrically consistent with the mesh structure of the finite element model of the blank material. Next, the relationship between strain and stress in the press-formed part at the bottom dead center is learned using machine learning. In the press-formed simulation after machine learning, the stress is calculated using the learned model 152 from the strain of the molded part at the bottom dead center obtained by CAE calculation, and the shape of the molded part after springback is calculated. The method for generating the learned model 152 is described in detail below.

[0033] Figure 4 is a flowchart showing the generation process procedure for the learning model 152 according to Embodiment 1, and Figure 5 is a conceptual diagram showing the method for calculating stress at the bottom dead center of actual press forming.

[0034] The processing unit 11 creates springback-resolved actual panel shape data (finite element model) by deforming the shape represented by the bottom dead center shape data (finite element model) using a mesh morphing method to match the shape shown by the actual panel outline data (step S31). The mesh morphing method is a method of changing the shape by directly deforming only the nodes and element shapes of the mesh model. The mesh structure of the actual panel shape data is consistent with the mesh structure of the finite element model of the blank material and is geometrically consistent. The elements at each position of the actual panel and the elements at each corresponding position of the blank material correspond to each other and are assigned the same element number.

[0035] Next, the processing unit 11 calculates the stress by elastically deforming the finite element model representing the actual panel after springback to the shape at the bottom dead center (step S32). It is assumed that no stress is generated in the finite elements indicated by the actual panel shape data before elastic deformation. In step S32, stress is calculated for each of the multiple finite elements indicated by the actual panel shape data.

[0036] Then, the processing unit 11 generates a learning model 152 based on the strain and stress data of each finite element of the finite element model at the bottom dead center, relating to the actual panel shape data (step S33). If the finite element model relating to the actual panel shape data is composed of N finite elements, then N sets of training data are obtained. A single dataset is obtained in which the strain and stress of each finite element are associated. The strain is obtained by CAE calculation using the finite element method, and the stress is obtained by steps S31 and S32 described above.

[0037] The processing unit 11 trains the learning model 152 based on the training data so that when strain data of the press-formed product at the bottom dead center is input, the stress of the press-formed product at the bottom dead center is output.

[0038] The details of machine learning are as follows: The processing unit 11 inputs the strain from the training data to the learning model 152, whose weights have not yet been adjusted. The learning model 152 outputs stress corresponding to the input strain. The processing unit 11 adjusts the parameters of the learning model 152 so that the error between the stress output by the learning model 152 and the stress value associated with the input strain is reduced. For example, the weights are adjusted using backpropagation.

[0039] The processing unit 11 generates a learning model 152 by repeatedly performing the above process using multiple datasets included in the training data and adjusting the weights of the learning model 152. The weights of the learning model 152 are stored in the storage unit 15 of the press forming simulation apparatus 1.

[0040] While a neural network was described as an example of learning model 152, other nonlinear regression machine learning models such as SVM (Support Vector Machine) and (gradient) boosting are also acceptable, and the specific method used is not restricted.

[0041] Figure 6 is a conceptual diagram showing a first example of input and output data for the learning model 152. In this embodiment, the finite element model of the blank material shape is not divided into elements in the thickness direction of the blank material. That is, each finite element has surfaces corresponding to the top and bottom surfaces of the plate-shaped blank material.

[0042] The data input to the learning model 152 includes the normal strain and shear strain of the top surface of the finite element, the normal strain and shear strain of the neutral surface of the finite element, and the normal strain and shear strain of the bottom surface of the finite element. The data output from the learning model 152 includes the normal stress and shear stress of the top surface of the finite element, the normal stress and shear stress of the neutral surface of the finite element, and the normal stress and shear stress of the bottom surface of the finite element.

[0043] Preferably, the data input to the learning model 152 includes at least one of the curvature, pressure, and frictional force on the upper or lower surface of the finite element. The curvature, pressure, and frictional force are information that suggests the state of the finite element in the blank material or molded product, and can improve the accuracy of the stress output from the learning model 152.

[0044] Note that there are two coordinate systems for quantifying strain and stress: a global coordinate system and a local coordinate system. You can use either coordinate system as appropriate to quantify strain and stress. The global coordinate system establishes a common reference coordinate system for the entire finite element model, representing the strain and stress of each finite element. A local coordinate system represents strain and stress using a coordinate system defined for each finite element. For example, the axis where stress and strain are zero could be defined as the z-axis.

[0045] Figure 7 is a conceptual diagram showing a second example of input and output data for the learning model 152. The data input to the learning model 152 shown in Figure 7 includes the perpendicular strain and shear strain of the upper surface of the finite element, the perpendicular strain and shear strain of the neutral surface of the finite element, and the perpendicular strain and shear strain of the lower surface of the finite element, as well as the curvature, thickness reduction rate, and curvature ρ2 on the upper or lower surface of the finite element. The curvature, thickness reduction rate, and curvature ρ2 are also information that suggests the state of the finite element in the blank material or molded product, and can improve the accuracy of the stress output from the learning model 152.

[0046] Preferably, the data input to the learning model 152 includes at least one of the following: the minimum principal strain on the upper surface of the finite element, the maximum principal strain on the lower surface, the first principal strain and the second principal strain on the upper surface, and the second principal strain on the lower surface. These features are information related to wrinkles in the molded product and can improve the accuracy of the stress output from the learning model 152.

[0047] Preferably, the data input to the learning model 152 includes the vertical strain and shear strain of the upper surface of the surrounding elements of the finite element of interest, and the vertical strain and shear strain of the lower surface of the surrounding elements. Since the stress occurring in a finite element is related to the strain and stress occurring in the surrounding finite elements, including this information in the input data can improve the accuracy of the stress output from the learning model 152.

[0048] Returning to Figure 2, steps S11 through S17 constitute the creation process of the learning model 152, and steps S18 onward are the simulation process using the learning model 152. In step S18, the processing unit 11 modifies the mold data. In other words, the processing unit 11 creates mold data with the die face modified.

[0049] After the creation of the learning model 152, the mold data can be modified based on the difference between the target shape for press forming and the shape indicated by the actual panel outline data acquired in step S16. In the iterative simulation process using the learning model 152, the mold data is modified based on the difference between the shape of the molded product after springback, obtained in steps S19 to S22, and the target shape.

[0050] Next, the processing unit 11 calculates the shape data and strain data of the press-formed product at the bottom dead center by performing a press-forming simulation based on the modified mold data, blank material shape data, and material property data, similar to step S15 (step S19).

[0051] Next, the processing unit 11 calculates the stress of each finite element at the bottom dead center by inputting the calculated strain of each finite element at the bottom dead center into the learning model 152 (step S20).

[0052] Next, the processing unit 11 corrects the relationship between strain and stress at the bottom dead center (step S21), and then performs a springback simulation based on the bottom dead center shape data and the corrected stress (step S22).

[0053] Next, the processing unit 11 compares the shape of the press-formed product after springback, obtained from the simulation results, with the target shape, and determines whether or not a predetermined target accuracy has been achieved (step S23). If it is determined that the target accuracy has not been achieved (step S23: NO), the processing unit 11 returns to step S18. If it is determined that the target accuracy has been achieved (step S23: YES), the processing unit 11 terminates the process.

[0054] According to the press forming simulation apparatus 1, simulation program 151, etc., configured in this manner according to this embodiment 1, it is possible to predict the shape of the press-formed product after springback, regardless of the tensile strength of the blank material. For example, when using high-tensile steel sheets with a tensile strength of approximately 340 MPa to 790 MPa, the springback is greater than when using mild steel sheets, making it difficult to predict. Furthermore, when using ultra-high-tensile steel sheets with a tensile strength of 980 MPa or more, predicting the springback becomes even more difficult. Even in such cases, according to this embodiment 1, it is possible to predict the shape of the press-formed product after springback. On the other hand, even when using blank material with a tensile strength of less than 340 MPa, the press forming simulation device 1 according to this embodiment 1 can more accurately predict the shape of the molded product after springback, thus enabling more efficient mold modification.

[0055] In this embodiment 1, an example was described in which the actual panel shape data (finite element model) after springback is created based on the bottom dead center shape data. However, the actual panel shape data (finite element model) after springback may also be created based on the shape data after springback calculated by CAE calculation. In other words, the molded product after springback, obtained by CAE calculation of springback simulation, may be deformed using a mesh morphing method to create the actual panel shape data (finite element model).

[0056] Alternatively, the actual panel shape data (finite element model) after springback may be created based on the blank material shape data. In other words, the finite element model of the blank material shape data may be deformed using a mesh morphing method to create the actual panel shape data (finite element model).

[0057] Furthermore, the data input and output to the learning model 152 is just an example; it can also be configured so that at least the strain of a finite element is input and the stress in that finite element is output. Furthermore, although an example was described in which the mesh in the thickness direction is one element as a finite element model for the blank material and molded product, calculations may also be performed using a model in which elements are divided in the plate thickness direction. The learning model 152 should be configured so that, when the strain of each element is input, the stress of that element is output. Furthermore, the finite element models of molds, blanks, etc., may be configured to perform shell calculations as 2D models, or as 3D models for solid calculations. Both the mold and the blank may be 2D models, both 3D models, or one may be a 2D model and the other a 3D model for the simulation.

[0058] (Embodiment 2) The press forming simulation apparatus 1, simulation program 151, etc., according to Embodiment 2 differ from those in Embodiment 1 in the method of generating the learning model 152. The other configurations and processes of the press forming simulation apparatus 1, etc., are the same as those of the press forming simulation apparatus 1, etc., according to Embodiment 1, so the same reference numerals are used for the same parts, and detailed explanations are omitted.

[0059] Figure 8 is a flowchart showing the generation process procedure for the learning model 152 according to Embodiment 2, and Figure 9 is a conceptual diagram showing the method for calculating stress at the bottom dead center of actual press forming.

[0060] The processing unit 11 of the press forming simulation apparatus 1 according to Embodiment 2 creates an upper die (first actual product die) representing a die face that conforms to the upper surface (first press surface) of the actual panel, and a lower die (second actual product die) representing a die face that conforms to the lower surface (second press surface) of the actual panel, based on the actual panel outline data (step S51).

[0061] Next, the processing unit 11 creates the actual panel shape data (finite element model) after springback by sandwiching the molded product, represented by the bottom dead center shape data, between the upper and lower molds of the actual panel shape (step S52).

[0062] The following steps are performed in the same manner as in steps S32 and S33 of Embodiment 1 to calculate the stress at the bottom dead center and generate the learning model 152 (steps S53, S54).

[0063] According to the press forming simulation apparatus 1, simulation program 151, etc., of this second embodiment, similar to the first embodiment, it is possible to predict the shape of the press-formed product after springback, regardless of the tensile strength of the blank material.

[0064] In this embodiment 2, an example was described in which the actual panel shape data (finite element model) after springback is created based on the bottom dead center shape data. However, the actual panel shape data (finite element model) after springback may also be created based on the shape data after springback calculated by CAE calculation. In other words, the actual panel shape data (finite element model) may be created by sandwiching the molded product after springback, which is obtained by CAE calculation of the springback simulation, between the upper and lower molds of the actual panel.

[0065] Alternatively, the actual panel shape data (finite element model) after springback may be created based on the blank material shape data. In other words, the actual panel shape data (finite element model) may be created by sandwiching the blank material between the upper and lower dies of the actual panel.

[0066] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The technical features described in each embodiment can be combined with each other, and the scope of the present invention is intended to include all modifications within the claims and equivalents thereof. The sequences shown in each embodiment are not limiting, and within the bounds of consistency, the order of each processing step may be changed, and multiple processes may be executed in parallel. The processing entity for each process is not limiting, and within the bounds of consistency, the processing of each device may be executed by other devices.

[0067] The matters described in each embodiment can be combined with each other. Furthermore, the independent and dependent claims described in the claims can be combined with each other in any combination, regardless of the form of reference. In addition, the claims use a form in which claims referencing two or more other claims (multi-claim form), but are not limited to this. A form in which multi-claims referencing at least one multi-claim (multi-multi-claim) may also be used. [Explanation of Symbols]

[0068] 1: Press forming simulation device 10: Recording media 11: Processing Section 12: Display section 13:Operation section 14: Data Entry Section 15: Storage section 151: Simulation Program 152: Learning Model

Claims

1. On the computer, The process involves calculating the shape data and strain data of the press-formed product at the bottom dead center through a forming simulation based on die face shape data, blank material shape data, and material property data related to the blank material. The steps include: acquiring actual formed external shape data showing the external shape of a press-formed product after springback, obtained by actually press-forming the blank material; A stress calculation step that calculates the stress of the press-formed product at the bottom dead center based on the shape data or blank material shape data at the bottom dead center or after springback, and the acquired actual formed outer shape data. The steps include generating a machine learning model that outputs the stress of a press-formed product at the bottom dead center when strain data for the press-formed product at the bottom dead center is input, based on strain data and stress data for the press-formed product at the bottom dead center, and The steps include modifying the die face, The process involves calculating the shape data and strain data of the press-formed product at the bottom dead center through a forming simulation based on the die face modification mold data, blank material shape data, and material property data related to the blank material. The steps include: inputting the calculated strain data of the press-formed product into the learning model to calculate the stress of the press-formed product; A springback shape calculation step calculates the shape of the press-formed product after springback, based on the calculated shape data and stress of the press-formed product at the bottom dead center. A simulation program to run the program.

2. The blank material shape data and the shape data each include data for a mesh structure obtained by dividing the shape of the blank material and the shape of the press-formed product into a plurality of finite elements, The stress calculation step described above is: The steps include creating shape data that includes data for a mesh structure in which the shape of a press-formed product after springback is divided into a plurality of finite elements, by modifying the mesh structure indicated by the shape data or blank material shape data using a mesh morphing method so that the outer shape indicated by the shape data or blank material shape data at the bottom dead center or after springback matches the outer shape indicated by the actual formed outer shape data, The steps include: calculating the stress generated when the shape of the press-formed product after springback is elastically deformed to the shape of the press-formed product at the bottom dead center, based on the shape data of the press-formed product after springback; and A simulation program according to claim 1, including the following:

3. The blank material shape data and the shape data each include data for a mesh structure obtained by dividing the shape of the blank material and the shape of the press-formed product into a plurality of finite elements, The stress calculation step described above is: The steps include creating a first actual product mold and a second actual product mold that represent die faces that conform to the first and second press surfaces of an actual press-formed product, respectively, based on actual molding outline data, The steps include creating shape data that includes data for a mesh structure in which the shape of a press-formed product after springback is divided into a plurality of finite elements by sandwiching a press-formed product represented by shape data at the bottom dead center or after springback, or a blank material represented by the blank material shape data, between the first and second actual product molds, thereby changing the mesh structure indicated by the shape data; The steps include: calculating the stress generated when the shape of the press-formed product after springback is elastically deformed to the shape of the press-formed product at the bottom dead center, based on the shape data of the press-formed product after springback; and A simulation program according to claim 1, including the following:

4. The shape data includes data for a mesh structure obtained by dividing the shape of a press-formed product into a plurality of finite elements. The data input to the aforementioned learning model is: This includes the vertical strain and shear strain of the upper surface of the finite element and the vertical strain and shear strain of the lower surface of the finite element. The data output from the aforementioned learning model is: Includes stress on the upper and lower surfaces of the finite element. A simulation program according to any one of claims 1 to 3.

5. The data input to the aforementioned learning model is: The curvature, pressure, and frictional force on the upper or lower surface of the finite element are included. The simulation program according to claim 4.

6. The data input to the aforementioned learning model is: The above includes the perpendicular strain and shear strain of the neutral surface of the finite element, and at least one of the curvature, pressure, and frictional force on the upper or lower surface of the finite element. The data output from the aforementioned learning model is: The stress at the neutral plane of the finite element includes The simulation program according to claim 4.

7. The data input to the aforementioned learning model is: The plate thickness reduction rate, curvature ρ², the minimum principal strain on the upper surface of the finite element, the maximum principal strain on the lower surface, the first principal strain and second principal strain on the upper surface, and the second principal strain on the lower surface are all included. The simulation program according to claim 4.

8. The data input to the learning model is: The vertical strain and shear strain of the upper surface of the peripheral element of the finite element, and the vertical strain and shear strain of the lower surface of the peripheral element, are included. The simulation program according to claim 4.

9. The process involves calculating the shape data and strain data of the press-formed product at the bottom dead center through a forming simulation based on die face shape data, blank material shape data, and material property data related to the blank material. The steps include: acquiring actual formed external shape data showing the external shape of a press-formed product after springback, obtained by actually press-forming the blank material; A stress calculation step that calculates the stress of the press-formed product at the bottom dead center based on the shape data or blank material shape data at the bottom dead center or after springback, and the acquired actual formed outer shape data. The steps include generating a machine learning model that outputs the stress of a press-formed product at the bottom dead center when strain data for the press-formed product at the bottom dead center is input, based on strain data and stress data for the press-formed product at the bottom dead center, and The steps include modifying the die face, The process involves calculating the shape data and strain data of the press-formed product at the bottom dead center through a forming simulation based on mold data after die face modification, blank material shape data, and material property data. The steps include: inputting the calculated strain data of the press-formed product into the learning model to calculate the stress of the press-formed product; A springback shape calculation step calculates the shape of the press-formed product after springback, based on the calculated shape data and stress of the press-formed product at the bottom dead center. A press forming simulation method including the following.

10. A press forming simulation apparatus equipped with a processing unit, The aforementioned processing unit, Based on mold data relating to the die face shape, blank material shape data, and material property data relating to the blank material, a forming simulation is performed to calculate the shape data and strain data of the press-formed product at the bottom dead center. Actual formed external shape data is obtained showing the external shape of the press-formed product after springback, obtained by actually press-forming the blank material. Based on the shape data at the bottom dead center or after springback, or the blank material shape data, and the acquired actual molded outer shape data, the stress of the press-formed product at the bottom dead center is calculated. Based on strain and stress data of the press-formed product at the bottom dead center, a machine learning model is generated that outputs the stress of the press-formed product at the bottom dead center when strain data for the press-formed product at the bottom dead center is input. The die face is modified, Based on mold data after die face modification, blank material shape data, and material property data, forming simulations are performed to calculate the shape data and strain data of the press-formed product at the bottom dead center. By inputting the calculated strain data of the press-formed product into the learning model, the stress of the press-formed product is calculated. Based on the calculated shape data and stress of the press-formed part at the bottom dead center, the shape of the press-formed part after springback is calculated. A press forming simulation device configured as follows.

Images