Workpiece electromagnetic forming numerical simulation modeling method and system

Abstract

The invention belongs to the field of electromagnetic forming, and particularly discloses a workpiece electromagnetic forming numerical simulation modeling method and system.The method comprises the steps that a trained first full-connection deep neural network model is adopted, and a tangential magnetic field component is calculated according to workpiece deformation real-time surface coordinates and normal vectors; dynamic magnetic pressure is generated in combination with a magnetic pressure correction coefficient output by the trained second full-connection deep neural network model; and applying the magnetic pressure to a mechanical model of the workpiece as a surface load, and iteratively solving a bidirectional coupling effect of workpiece deformation and an electromagnetic field. The problem that an existing workpiece electromagnetic forming modeling method cannot be compatible in the aspects of calculation efficiency, geometric adaptability and physical consistency is solved.

Description

Technical Field

[0001] The present application belongs to the field of electromagnetic forming (EMF), and more specifically, relates to a numerical simulation modeling method and system for electromagnetic forming of a workpiece. Background Art

[0002] In the existing technology, electromagnetic forming modeling methods include analytical modeling, semi-analytical modeling, and numerical modeling. Analytical and semi-analytical modeling describe the coupling relationship between electromagnetic and mechanical fields by establishing closed-form physical equations. They have the advantages of fast computational speed and clear physical meaning, but they are only applicable to simple geometric structures and their accuracy is limited due to idealized assumptions. Numerical modeling uses the finite element method or boundary element method to solve Maxwell's equations and the governing mechanical equations. Although it can accurately characterize the multi-physics field coupling effects, it has bottlenecks such as high computational cost, convergence, and stability issues.

[0003] With the increasing demand for lightweight manufacturing, electromagnetic forming processes are developing towards the design of complex three-dimensional components, multi-material composite forming, and high-efficiency processes. This places higher demands on modeling methods, requiring them to simultaneously meet key criteria such as three-dimensional versatility, wide parameter adaptability, and an efficiency-accuracy balance. However, existing technologies cannot fully meet these requirements. Analytical modeling is limited by strong assumptions and geometric simplifications, while numerical modeling is inefficient. Therefore, there is an urgent need to develop new modeling methods to address these technical challenges. Summary of the Invention

[0004] In response to the defects of the existing technology, the purpose of this application is to provide a numerical simulation modeling method and system for electromagnetic forming of workpieces, aiming to solve the incompatibility problems of existing modeling methods for electromagnetic forming of workpieces in terms of computational efficiency, geometric adaptability and physical consistency.

[0005] To achieve the above objectives, in a first aspect, the present application provides a numerical simulation modeling method for electromagnetic forming of a workpiece, comprising the following steps: The first fully connected deep neural network model that has been trained is used to calculate the tangential magnetic field component based on the real-time surface coordinates and normal vector of the workpiece deformation, and the dynamic magnetic pressure is generated by combining it with the magnetic pressure correction coefficient output by the second fully connected deep neural network model that has been trained. The magnetic pressure is applied as a surface load to the mechanical model of the workpiece, and the bidirectional coupling effect between the workpiece deformation and the electromagnetic field is iteratively solved. The first fully connected deep neural network model takes as input the real-time surface coordinates of the workpiece deformation, and outputs the tangential magnetic field components at each surface coordinate. The second fully connected deep neural network model takes as input the parameter space dataset, and outputs the magnetic pressure correction coefficient. The parameter space data set is a parameter combination in a four-dimensional parameter space constructed by the resistivity, thickness, discharge frequency, and average deformation velocity of the workpiece; and the workpiece is a metal workpiece.

[0006] Further preferably, the training method of the first fully connected deep neural network model is: Based on the electromagnetic theory of the mirror method, a coil-workpiece mirror model is established and subjected to static finite element simulation to generate magnetic field distribution data of the drive coil in free space; wherein the drive coil is located above the workpiece; Randomly sampling several groups of workpiece deformation real-time surface coordinates and their corresponding magnetic field components in the spatial domain as training sets for the first fully connected deep neural network model, and training the first fully connected deep neural network model; The coil-workpiece mirror model is used to obtain the theoretical magnetic field distribution value of the driving coil.

[0007] Further preferably, the training method of the second fully connected deep neural network model is: obtaining the expression of the magnetic pressure correction coefficient based on the equivalent circuit of the workpiece, combining the parameter space data set, generating the true value of the correction coefficient through frequency domain finite element simulation, obtaining the training set of the second fully connected deep neural network model, and training the second fully connected deep neural network model.

[0008] Further preferably, the trained first fully connected deep neural network model and the second fully connected deep neural network model are embedded in the Abaqus explicit dynamics simulation model, and the first fully connected deep neural network model and the second fully connected deep neural network model are called through Fortran user subroutines.

[0009] Further preferably, the first fully connected deep neural network model includes an input layer, four hidden layers and an output layer; the hidden layer adopts a hyperbolic tangent activation function, and the output layer adopts a linear activation function. The Adam optimizer is used to minimize the mean square error between the predicted tangential magnetic field and the tangential magnetic field in the training data set. After the training is completed, real-time spatial interpolation of the tangential magnetic field distribution is realized.

[0010] Further preferably, the second fully connected deep neural network model includes an input layer, a hidden layer and an output layer; the hidden layer adopts a Tanh activation function, and the output layer adopts a linear function to constrain the output range; batch normalization and early stopping strategies are used during the training process to prevent overfitting, thereby realizing a nonlinear mapping of parameter space data to magnetic pressure correction coefficients.

[0011] In a second aspect, the present application provides a numerical simulation modeling system for electromagnetic forming of a workpiece, comprising: A magnetic field calculation module is used to calculate the tangential magnetic field component based on the real-time surface coordinates and normal vector of the workpiece deformation using the trained first fully connected deep neural network model; A magnetic pressure generation module is used to generate dynamic magnetic pressure based on the tangential magnetic field component obtained by the magnetic field calculation module and the magnetic pressure correction coefficient output by the trained second fully connected deep neural network model; The coupling effect solving module is used to apply magnetic pressure as a surface load to the mechanical model of the workpiece and iteratively solve the bidirectional coupling effect between the workpiece deformation and the electromagnetic field; The first fully connected deep neural network model takes as input the real-time surface coordinates of the workpiece deformation, and outputs the tangential magnetic field components at each surface coordinate. The second fully connected deep neural network model takes as input the parameter space dataset, and outputs the magnetic pressure correction coefficient. The parameter space data set is a parameter combination in a four-dimensional parameter space constructed by the resistivity, thickness, discharge frequency, and average deformation velocity of the workpiece; and the workpiece is a metal workpiece.

[0012] Further preferably, the workpiece electromagnetic forming numerical simulation modeling system further includes a first training module, and the first training module includes: A first data generation unit is configured to establish a coil-workpiece mirror model based on the mirror electromagnetic theory and perform static finite element simulation on the model to generate magnetic field distribution data of the drive coil in free space; wherein the drive coil is located above the workpiece; A first model training unit is used to randomly sample a plurality of groups of workpiece deformation real-time surface coordinates and their corresponding tangential magnetic field components in the spatial domain as training sets for a first fully connected deep neural network model, and train the first fully connected deep neural network model; The coil-workpiece mirror model is used to obtain the theoretical magnetic field distribution value of the driving coil.

[0013] Further preferably, the workpiece electromagnetic forming numerical simulation modeling system further includes a second training module, which includes: The second data generation unit is used to obtain a magnetic pressure correction coefficient expression based on an equivalent circuit of the workpiece, and generate a true value of the correction coefficient through a frequency domain finite element simulation in combination with a parameter space data set; The second model training unit is used to train the second fully connected deep neural network model based on the data obtained by the second data generation unit.

[0014] Further preferably, the workpiece electromagnetic forming numerical simulation modeling system also includes a model calling module, which is used to embed the trained first fully connected deep neural network model and the second fully connected deep neural network model into the Abaqus explicit dynamics simulation model, and call the first fully connected deep neural network model and the second fully connected deep neural network model through a Fortran user subroutine.

[0015] Further preferably, the first fully connected deep neural network model includes an input layer, four hidden layers and an output layer; the hidden layer adopts a hyperbolic tangent activation function, and the output layer adopts a linear activation function. The Adam optimizer is used to minimize the mean square error between the predicted tangential magnetic field and the tangential magnetic field in the training data set. After the training is completed, real-time spatial interpolation of the tangential magnetic field distribution is realized.

[0016] Further preferably, the second fully connected deep neural network model includes an input layer, a hidden layer and an output layer; the hidden layer adopts a Tanh activation function, and the output layer adopts a linear function to constrain the output range; batch normalization and early stopping strategies are used during the training process to prevent overfitting, thereby realizing a nonlinear mapping of parameter space data to magnetic pressure correction coefficients.

[0017] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0019] In a fifth aspect, the present application provides a computer program product, which, when executed on a processor, enables the processor to execute the method described in the first aspect or any possible implementation of the first aspect.

[0020] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0021] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the existing technologies: This application provides a numerical simulation modeling method for electromagnetic forming of workpieces. Through a deep learning model with physical prior constraints, the simulation time of multi-physics field dynamic coupling of three-dimensional models is greatly shortened. At the same time, the fully connected deep neural network model adopts Tanh activation function and hidden structure, which greatly improves the prediction accuracy of magnetic field spatial distribution through training, and reduces the average error of magnetic pressure correction coefficient in a wide parameter range. At the same time, it supports three-dimensional asymmetric coils (ellipsoid, rectangle, triangle) and extreme parameters (such as conductivity 50×10 -8Ω·m, frequency 20 kHz), and the maximum deviation of the formed height from the experimental measurement was low. This method provides an efficient and universal simulation tool for electromagnetic forming process design, significantly reducing the computational cost of complex mold optimization and multi-parameter inverse identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a numerical simulation modeling method for electromagnetic forming of a workpiece provided in an embodiment of the present application; Figure 2 This is a flow chart of step S1 provided in an embodiment of the present application; Figure 3 Schematic diagram of the process of step S2 provided in the embodiment of the present application; Figure 4 Schematic diagram of the process of step S3 provided in the embodiment of the present application; Figure 5 is a structural diagram of an electronic device provided in an embodiment of the present application; Throughout the drawings, the same reference numerals are used to denote the same elements or structures, wherein: 810 is a processor; 820 is a communication interface; 830 is a memory; and 840 is a communication bus. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0024] The term "and / or" as used herein describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. The symbol " / " as used herein indicates that the related objects are in an "or" relationship, for example, A / B means either A or B.

[0025] The terms "first" and "second" and the like in the description and claims herein are used to distinguish different objects rather than to describe a specific order of the objects.

[0026] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0027] In the description of the embodiments of the present application, unless otherwise specified, “plurality” means two or more.

[0028] This application provides a numerical simulation modeling method for electromagnetic forming of workpieces, which aims to solve the defects of existing modeling methods in terms of computational efficiency, geometric adaptability and physical consistency by integrating electromagnetic field analytical theory, finite element simulation and deep neural network technology. The three-dimensional modeling method of electromagnetic forming provided by this application decomposes the traditional dynamic coupling simulation into offline training and online prediction by constructing a multi-stage hybrid drive framework, realizing efficient coupling analysis of electromagnetic-mechanical fields in three-dimensional complex coil scenarios, and covering a wide parameter range (voltage 6kV~8kV, conductivity ~ , frequency 0.5kHz~10kHz, workpiece thickness 0.5mm~4mm), significantly reducing computing costs while maintaining high accuracy.

[0029] More specifically, by combining physical prior knowledge with data-driven methods, the physical consistency of the model is guaranteed and the computational efficiency is improved; secondly, a staged modeling approach is adopted to transform the computationally intensive electromagnetic field solution process into offline training and online prediction, greatly reducing the computational burden of real-time simulation; thirdly, this method supports the simulation requirements of three-dimensional asymmetric coils and a wide parameter range, and has good versatility and adaptability; experimental results show that compared with the existing technology, this application can improve the computational efficiency by 1 to 2 orders of magnitude while maintaining accuracy, providing an effective solution to the long-standing modeling problems in the field of electromagnetic forming.

[0030] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0031] like Figure 1 As shown, the present application provides a numerical simulation modeling method for electromagnetic forming of a workpiece, comprising the following steps: Step S1: Based on the mirror electromagnetic theory, a coil-workpiece mirror model is established. Static finite element simulation is performed using COMSOL Multiphysics finite element software to generate magnetic field distribution data of the drive coil in free space. Subsequently, 100,000 sets of coordinate points (the input of a two-dimensional axisymmetric scene is the rz coordinate) and their corresponding magnetic field components (Br, Bz) are randomly sampled in the spatial domain to construct the first fully connected deep neural network model I. The input is the real-time surface coordinates of the workpiece deformation, and the output is the tangential magnetic field component at each coordinate point. The coil-workpiece mirror model is used to obtain the theoretical magnetic field distribution value of the coil. The workpiece is a metal workpiece. The first fully connected deep neural network model I consists of an input layer, four hidden layers, and an output layer. The input layer has two neurons, and each hidden layer contains 30 neurons. The output layer also has two neurons, corresponding to the magnetic field component. The hidden layer uses the hyperbolic tangent (Tanh) activation function, and the output layer uses a linear activation function. The Adam optimizer is used to minimize the mean square error between the predicted magnetic field and the simulated data. After training, real-time spatial interpolation of the magnetic field distribution is achieved. Step S2: deriving a magnetic pressure correction coefficient expression based on the equivalent circuit model of the workpiece, combining the parameter space data set, and generating the true value of the correction coefficient through frequency domain finite element simulation to complete the acquisition of the second training set; The parameter space data set includes but is not limited to the parameter combination in the four-dimensional parameter space constructed by the workpiece resistivity, thickness, discharge frequency and average deformation velocity; Specifically, the Latin hypercube sampling method was used to select 2000 parameter combinations in the four-dimensional parameter space (workpiece resistivity, thickness, discharge frequency, and average deformation velocity). The COMSOL finite element software was used to perform frequency domain analysis to obtain the true value of the correction coefficient. Based on this data, a second fully connected deep neural network model II was trained. Its input layer consists of 4 neurons (corresponding to the four-dimensional parameters), 4 hidden layers each contain 10 neurons, and the output layer consists of 1 neuron (correction coefficient). The hidden layers use the Tanh activation function, and the output layer uses a linear function to constrain the output range (0.5-2.0). Batch normalization and early stopping strategies were used during training to prevent overfitting, ultimately achieving a nonlinear mapping from parameters to correction coefficients. Step S3: Embed the trained first fully connected deep neural network model and the second fully connected deep neural network model into the Abaqus explicit dynamics simulation process: call the first fully connected deep neural network model I through the Fortran user subroutine, calculate the tangential magnetic field component according to the real-time surface coordinates and normal vector of the workpiece deformation, and generate dynamic magnetic pressure in combination with the correction coefficient output by the second fully connected deep neural network model II; wherein, the magnetic pressure is the product of the magnetic pressure correction coefficient and the magnetic field component, and the magnetic pressure is applied to the mechanical model of the workpiece as a surface load, and the bidirectional coupling effect between the workpiece deformation and the electromagnetic field is iteratively solved; wherein, the mechanical model uses the Cowper-Symonds constitutive relationship to describe the material strain rate effect, and the grid density is Unit, which realizes data interaction between deep learning model and finite element solver through Python script; In step S1, a static electromagnetic field simulation model of the driving coil is established using COMSOL Multiphysics finite element software. The coil geometric parameters, including the number of turns and cross-sectional dimensions, are set, and the boundary conditions are set to open boundaries. The static electromagnetic field simulation model of the driving coil is used to obtain the actual magnetic field distribution value of the coil. In step S1, the Sobol sequence sampling method is used to generate 100,000 spatial coordinate points in the simulation domain, and the actual magnetic field component value of each point is recorded; In step S1, a first fully connected neural network model I is constructed, and the network architecture is input layer- Neuron hidden layer - output layer, the hidden layer activation function uses the tanh function, and the output layer uses the linear activation function linear; In step S1, the adaptive moment estimation optimizer (Adam) is used for training, with the initial learning rate set to 0.001, the batch size to 1000, and the training cycle to 10,000 times; In step S2, the parameter scanning range is set during the frequency domain finite element simulation: resistivity (0.5~50 ), thickness (0.5~5mm), frequency (0.5~20kHz), speed (0~120m / s); In step S2, the Latin hypercube sampling method is used to generate 2000 sets of parameter combinations, and the magnetic pressure correction coefficient under each working condition is calculated through frequency domain simulation; In step S2, the second fully connected deep neural network model II is designed, with an input layer of 4 nodes and a hidden layer of The neuron structure uses the tanh activation function and the output layer uses the linear function; In step S2, an early stopping strategy (patience = 100) is used during training; In step S3, the workpiece mechanical model is established in Abaqus / Explicit, and the material constitutive model adopts the Cowper-Symonds model with parameters C=6500 and P=4; In step S3, the trained first fully connected deep neural network model I is called through a Fortran user subroutine to calculate the magnetic pressure distribution on the workpiece surface in real time; In step S3, a Python script is used to implement data interaction between Abaqus and the second fully connected deep neural network model II, and the correction coefficient is updated once per iteration.

[0032] The above method is suitable for high to low conductivity workpieces (resistivity ~ ), axisymmetric to three-dimensional asymmetric coils (ellipsoid, rectangle and triangle) and a wide range of process parameters (frequency 0.5kHz~20 kHz, thickness 0.5mm~5mm).

[0033] Example This application provides a numerical simulation modeling method for electromagnetic forming of a workpiece, comprising the following steps: Step S1: constructing a first fully connected deep neural network model I; obtaining a three-dimensional magnetic field distribution through theoretical calculation combined with finite element simulation, and obtaining a nonlinear mapping from spatial coordinates to spatial magnetic fields through the first fully connected deep neural network model I; like Figure 2 As shown, in step S1, the method for constructing the first fully connected deep neural network model I includes the following steps: Step S1.1: Use COMSOL Multiphysics software to establish an axisymmetric static magnetic field simulation model. Set the coil geometry parameters to include an inner diameter of 20 mm, an outer diameter of 30 mm, a height of 10 mm, and 20 turns. The solution domain is a cylindrical region with a radius of 100 mm and a height of 150 mm. The boundary condition is set to magnetic insulation. A parametric sweep is performed to obtain the magnetic field distributions Br and Bz at different spatial locations. A Sobol sequence is used to generate 100,000 sampling points, and the coordinates and magnetic field components are normalized. Step S1.2: Construct the first fully connected deep neural network model I, with normalized (r, z) coordinates as the input layer, four hidden layers with 30 neurons each, the hyperbolic tangent function (Tanh) as the activation function, and a linear activation function in the output layer to predict the Br and Bz components. Train the model 10,000 times using the Adam optimizer (learning rate 0.001), the mean squared error (MSE) as the loss function, a batch size of 1,000, and an early stopping strategy to prevent overfitting. Step S2: deriving the expression of the magnetic pressure correction coefficient by combining the equivalent circuit model, generating a parameter space data set through frequency domain finite element simulation, and constructing the second fully connected deep neural network model II; like Figure 3 As shown, in step S2, the method for constructing the second fully connected deep neural network model II specifically includes the following steps: Step S2.1: Establish a frequency domain electromagnetic field simulation model and set the resistivity (0.5~50 ), thickness (0.5-4 mm), frequency (0.5-10 kHz), and speed (0-120 m / s), Latin hypercube sampling was used to generate 2000 sets of parameter combinations; the magnetic pressure correction coefficient was calculated using the frequency domain model; Step S2.2: Design a second fully connected neural network model II. The input layer contains four parameters (resistivity, thickness, frequency, and speed). Each of the four hidden layers contains 10 neurons. The activation function uses the tanh function. The output layer uses a linear activation function to predict the correction coefficient. Training uses the mean square error loss function, a batch size of 128, a learning rate of 0.001, and 20,000 training cycles to improve generalization ability. Step S3: combining the trained first fully connected neural network model I, the second fully connected neural network model II and the finite element software ABAQUS to realize electromagnetic forming simulation; Figure 4 As shown, the specific steps include: Step S3.1: Establish a mechanical model of the workpiece in Abaqus / Explicit. The material is AA5052 aluminum alloy, and the Cowper-Symond model is used with parameters C = 6500 and P = 4. Step S3.2: The trained first fully connected neural network model I is called in real time through the Fortran user subroutine, the theoretical static magnetic field size is calculated in real time according to the node coordinates of the workpiece surface, the second fully connected neural network model II is called to obtain the correction coefficient, and the theoretical magnetic field value is corrected and applied to the mechanical model.

[0034] In summary, compared with the prior art, this application has the following advantages: (1) Through the deep learning model with physical prior constraints, the three-dimensional dynamic coupling simulation time is shortened from 16 hours of traditional numerical models to 0.6 hours, and the efficiency is improved by 26 times; (2) The fully connected neural network adopts Tanh activation function and 4-layer hidden structure, and the prediction error of the magnetic field spatial distribution is less than 3% under 100,000 training samples, and the average error of the correction coefficient is less than 5% in a wide parameter range; (3) It supports three-dimensional asymmetric coils (ellipsoids, rectangles and triangles) and extreme parameters (such as conductivity). , frequency 20kHz), the maximum deviation of the forming height from the experimental measurement value does not exceed 8%. This method provides an efficient and versatile simulation tool for electromagnetic forming process design, significantly reducing the computational cost of complex mold optimization and multi-parameter inverse identification.

[0035] The coupling effect acquisition system for electromagnetic forming of a workpiece provided in the present application is described below. The coupling effect acquisition system for electromagnetic forming of a workpiece described below and the coupling effect acquisition method for electromagnetic forming of a workpiece described above can be referenced to each other.

[0036] The present application provides a numerical simulation modeling system for electromagnetic forming of a workpiece, comprising: A magnetic field calculation module is used to calculate the tangential magnetic field component based on the real-time surface coordinates and normal vector of the workpiece deformation using the trained first fully connected deep neural network model; A magnetic pressure generation module is used to generate dynamic magnetic pressure based on the tangential magnetic field component obtained by the magnetic field calculation module and the magnetic pressure correction coefficient output by the trained second fully connected deep neural network model; The coupling effect solving module is used to apply magnetic pressure as a surface load to the mechanical model of the workpiece and iteratively solve the bidirectional coupling effect between the workpiece deformation and the electromagnetic field; The first fully connected deep neural network model takes as input the real-time surface coordinates of the workpiece deformation, and outputs the tangential magnetic field components at each surface coordinate. The second fully connected deep neural network model takes as input the parameter space dataset, and outputs the magnetic pressure correction coefficient. The parameter space data set is a parameter combination in a four-dimensional parameter space constructed by the resistivity, thickness, discharge frequency, and average deformation velocity of the workpiece; and the workpiece is a metal workpiece.

[0037] Further preferably, the coupling effect acquisition system for electromagnetic forming of a workpiece further includes a first training module, and the first training module includes: A first data generation unit is configured to establish a coil-workpiece mirror model based on the mirror electromagnetic theory and perform static finite element simulation on the model to generate magnetic field distribution data of the drive coil in free space; wherein the drive coil is located above the workpiece; A first model training unit is used to randomly sample a plurality of groups of workpiece deformation real-time surface coordinates and their corresponding tangential magnetic field components in the spatial domain as training sets for a first fully connected deep neural network model, and train the first fully connected deep neural network model; The coil-workpiece mirror model is used to obtain the theoretical magnetic field distribution value of the driving coil.

[0038] Further preferably, the workpiece electromagnetic forming acquisition system further includes a second training module, and the second training module includes: The second data generation unit is used to obtain a magnetic pressure correction coefficient expression based on an equivalent circuit of the workpiece, and generate a true value of the correction coefficient through a frequency domain finite element simulation in combination with a parameter space data set; The second model training unit is used to train the second fully connected deep neural network model based on the data obtained by the second data generation unit.

[0039] Further preferably, the workpiece electromagnetic forming acquisition system also includes a model calling module, which is used to embed the trained first fully connected deep neural network model and the second fully connected deep neural network model into the Abaqus explicit dynamics simulation model, and call the first fully connected deep neural network model and the second fully connected deep neural network model through a Fortran user subroutine.

[0040] Further preferably, the first fully connected deep neural network model includes an input layer, four hidden layers and an output layer; the hidden layer adopts a hyperbolic tangent activation function, and the output layer adopts a linear activation function. The Adam optimizer is used to minimize the mean square error between the predicted tangential magnetic field and the tangential magnetic field in the training data set. After the training is completed, real-time spatial interpolation of the tangential magnetic field distribution is realized.

[0041] Further preferably, the second fully connected deep neural network model includes an input layer, a hidden layer, and an output layer; the hidden layer uses a Tanh activation function, and the output layer uses a linear function to constrain the output range; batch normalization and early stopping strategies are used during training to prevent overfitting and achieve a nonlinear mapping from parameter space data to magnetic pressure correction coefficients. It is understood that the detailed functional implementation of each of the above units / modules can be found in the description of the aforementioned method embodiments and will not be repeated here.

[0042] It should be understood that the above-mentioned system is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the system are similar to those described in the above-mentioned method. The working process of the system can refer to the corresponding process in the above-mentioned method and will not be repeated here.

[0043] Based on the method in the above embodiment, Figure 5 As shown, an embodiment of the present application provides an electronic device, which may include: a processor (Processor) 810, a communication interface (Communications Interface) 820, a memory (Memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call the logic instructions in the memory 830 to execute the method in the above embodiment.

[0044] In addition, the logic instructions in the aforementioned memory 830 can be implemented in the form of a software functional unit and, when sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0045] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0046] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0047] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0048] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC.

[0049] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions. When loaded and executed on a computer, the computer program instructions fully or partially produce the processes or functions described in the embodiments of this application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive (SSD)).

[0050] It will be understood that the various numerical numbers involved in the embodiments of the present application are merely distinctions for the convenience of description and are not intended to limit the scope of the embodiments of the present application.

[0051] It is easy for those skilled in the art to understand that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A numerical simulation modeling method for electromagnetic forming of a workpiece, characterized in that: The following steps are involved: The first fully connected deep neural network model that has been trained is used to calculate the tangential magnetic field component based on the real-time surface coordinates and normal vector of the workpiece deformation, and the dynamic magnetic pressure is generated by combining it with the magnetic pressure correction coefficient output by the second fully connected deep neural network model that has been trained. The magnetic pressure is applied as a surface load to the mechanical model of the workpiece, and the bidirectional coupling effect between the workpiece deformation and the electromagnetic field is iteratively solved. The first fully connected deep neural network model takes as input the real-time surface coordinates of the workpiece deformation, and outputs the tangential magnetic field components at each surface coordinate. The second fully connected deep neural network model takes as input the parameter space dataset, and outputs the magnetic pressure correction coefficient. The parameter space data set is a parameter combination in a four-dimensional parameter space constructed by the resistivity, thickness, discharge frequency, and average deformation velocity of the workpiece; and the workpiece is a metal workpiece.

2. The numerical simulation modeling method for electromagnetic forming of a workpiece according to claim 1, characterized in that: The training method of the first fully connected deep neural network model is: Based on the electromagnetic theory of the mirror method, a coil-workpiece mirror model is established and subjected to static finite element simulation to generate magnetic field distribution data of the drive coil in free space; wherein the drive coil is located above the workpiece; Randomly sampling several groups of workpiece deformation real-time surface coordinates and their corresponding magnetic field components in the spatial domain as training sets for the first fully connected deep neural network model, and training the first fully connected deep neural network model; The coil-workpiece mirror model is used to obtain the theoretical magnetic field distribution value of the driving coil.

3. The numerical simulation modeling method for electromagnetic forming of a workpiece according to claim 1 or 2, characterized in that: The training method of the second fully connected deep neural network model is: obtain the magnetic pressure correction coefficient expression based on the equivalent circuit of the workpiece, combine the parameter space data set, generate the true value of the correction coefficient through frequency domain finite element simulation, obtain the training set of the second fully connected deep neural network model, and train the second fully connected deep neural network model.

4. The numerical simulation modeling method for electromagnetic forming of a workpiece according to claim 1, characterized in that: The trained first fully connected deep neural network model and the second fully connected deep neural network model are embedded in the Abaqus explicit dynamics simulation model, and the first fully connected deep neural network model and the second fully connected deep neural network model are called through Fortran user subroutines.

5. The numerical simulation modeling method for electromagnetic forming of a workpiece according to claim 2, characterized in that: The first fully connected deep neural network model includes an input layer, four hidden layers and an output layer; the hidden layer uses a hyperbolic tangent activation function, and the output layer uses a linear activation function. The Adam optimizer is used to minimize the mean square error between the predicted tangential magnetic field and the tangential magnetic field in the training dataset. After training is completed, real-time spatial interpolation of the tangential magnetic field distribution is achieved.

6. The numerical simulation modeling method for electromagnetic forming of a workpiece according to claim 3, characterized in that: The second fully connected deep neural network model includes an input layer, a hidden layer, and an output layer; the hidden layer uses the Tanh activation function, and the output layer uses a linear function to constrain the output range; During the training process, batch normalization and early stopping strategies are used to prevent overfitting and realize the nonlinear mapping from parameter space data to magnetic pressure correction coefficients.

7. A numerical simulation modeling system for electromagnetic forming of a workpiece, characterized in that: include: A magnetic field calculation module is used to calculate the tangential magnetic field component based on the real-time surface coordinates and normal vector of the workpiece deformation using the trained first fully connected deep neural network model; A magnetic pressure generation module is used to generate dynamic magnetic pressure based on the tangential magnetic field component obtained by the magnetic field calculation module and the magnetic pressure correction coefficient output by the trained second fully connected deep neural network model; The coupling effect solving module is used to apply magnetic pressure as a surface load to the mechanical model of the workpiece and iteratively solve the bidirectional coupling effect between the workpiece deformation and the electromagnetic field; The first fully connected deep neural network model takes as input the real-time surface coordinates of the workpiece deformation, and outputs the tangential magnetic field components at each surface coordinate. The second fully connected deep neural network model takes as input the parameter space dataset, and outputs the magnetic pressure correction coefficient. The parameter space data set is a parameter combination in a four-dimensional parameter space constructed by the resistivity, thickness, discharge frequency, and average deformation velocity of the workpiece; and the workpiece is a metal workpiece.

8. The workpiece electromagnetic forming numerical simulation modeling system according to claim 7, characterized in that: The first training module includes: A first data generation unit is configured to establish a coil-workpiece mirror model based on the mirror electromagnetic theory and perform static finite element simulation on the model to generate magnetic field distribution data of the drive coil in free space; wherein the drive coil is located above the workpiece; A first model training unit is used to randomly sample a plurality of groups of workpiece deformation real-time surface coordinates and their corresponding tangential magnetic field components in the spatial domain as training sets for a first fully connected deep neural network model, and train the first fully connected deep neural network model; The coil-workpiece mirror model is used to obtain the theoretical magnetic field distribution value of the driving coil.

9. The workpiece electromagnetic forming numerical simulation modeling system according to claim 7, characterized in that: Also included is a second training module, the second training module including: The second data generation unit is used to obtain a magnetic pressure correction coefficient expression based on an equivalent circuit of the workpiece, and generate a true value of the correction coefficient through a frequency domain finite element simulation in combination with a parameter space data set; The second model training unit is used to train the second fully connected deep neural network model based on the data obtained by the second data generation unit.

10. The workpiece electromagnetic forming numerical simulation modeling system according to any one of claims 7 to 9, characterized in that: It also includes a model calling module for embedding the trained first fully connected deep neural network model and the second fully connected deep neural network model into the Abaqus explicit dynamics simulation model, and calling the first fully connected deep neural network model and the second fully connected deep neural network model through a Fortran user subroutine.

Images