A method and device for springback compensation of a stamped part based on a machine learning model

Abstract

The present application relates to the technical field of sheet stamping forming, and provides a stamping part springback compensation method and device based on a machine learning model, which comprises the following steps: setting process parameter conditions for a stamping part, setting initial values for the set process parameter conditions, and obtaining initial process parameter conditions; performing finite element sampling on the stamping part, performing CAE forming simulation springback prediction and compensation calculation, and obtaining sample data under the initial process parameter conditions; traversing parameters in the initial process parameter conditions, cyclically adjusting parameter values of each parameter, and obtaining a sample data set; preprocessing the sample data set, training a machine learning model by using the preprocessed sample data set, and obtaining a springback compensation big data model; and taking measured deviation vector values of the stamping part as input, and obtaining predicted springback compensation vector values by the springback compensation big data model. The present application can efficiently, accurately and at low cost complete stamping die surface compensation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sheet metal stamping forming, and particularly relates to a stamping part springback compensation method and device based on a machine learning model. BACKGROUND

[0002] Sheet metal stamping forming is a metal plastic forming method for obtaining stamping parts with certain shape, size and performance by using a die and a stamping device to apply pressure to a metal sheet. It is widely used in the industrial fields of aerospace, automotive vehicles, electrical machinery and appliances, and daily necessities manufacturing. Since plastic deformation and elastic deformation occur simultaneously during the forming process of the sheet metal, a certain springback deformation will occur after the formed part is taken out of the die due to unloading of the load, which seriously affects the forming quality and dimensional accuracy of the stamping part. The springback amount of aluminum alloy and high-strength steel workpieces is large, which is a difficult problem in stamping manufacturing. With the continuous development of the trend of automobile lightweight, the application of aluminum alloy and high-strength steel is increasingly popular, and the springback problem of sheet metal stamping has attracted more and more attention from the stamping industry, and the mold surface reverse compensation technology needs to be used to solve it.

[0003] There are two traditional springback compensation technologies. One is the actual part detection-based physical processing trial-and-error method, that is, a multi-round repeated "part deviation detection, mold surface adjustment, mold processing and production, part deviation detection again, mold surface adjustment again" trial-and-error method is adopted according to the detection data of the actual part after forming. This method not only requires technicians to predict the mold surface compensation amount based on process experience, but also consumes a large amount of manpower and resources. The other is the numerical iteration trial-and-error method based on finite element numerical simulation, that is, a multi-round repeated "part deviation prediction after springback, mold surface adjustment, stamping forming CAE simulation, part deviation prediction again after springback, mold surface adjustment again" virtual trial-and-error method is adopted according to the springback prediction results of the stamping forming CAE simulation. The effectiveness of this method depends entirely on the calculation accuracy of the stamping forming simulation. Therefore, when the actual measurement results of the stamping part size are out of tolerance and need to be processed for springback, the technicians need to rely on process experience to carry out physical processing trial-and-error based on the measured deviation amount of the part; or correct the finite element numerical prediction accuracy again through stamping forming CAE simulation technology to carry out virtual springback compensation trial-and-error. In summary, the traditional springback compensation method for stamping parts is highly dependent on the process experience of technicians and the stamping forming simulation technology. Different process experience, different stamping forming simulation software and different mastery of the software will lead to different trial-and-error processes, which requires a large mold development cost and a long production cycle.

[0004] With the rapid development of artificial intelligence technology in recent years, machine learning methods have been applied to the field of engineering and manufacturing. Machine learning methods can process and analyze a large amount of data, find patterns and rules in the data, and give corresponding results and decisions, thereby improving production efficiency. Currently, there is no report on the application of machine learning algorithms to the springback compensation technology of stamping parts.

[0005] Therefore, how to use machine learning algorithms to establish a data-driven springback compensation model for stamping parts, improve the efficiency of die design, and ensure the dimensional accuracy of stamping parts, has become a technical problem to be solved. SUMMARY

[0006] Therefore, in order to overcome the shortcomings of the prior art, the present application aims to provide a springback compensation method and device for stamping parts based on a machine learning model.

[0007] According to a first aspect of the present application, a springback compensation method for stamping parts based on a machine learning model is provided, the method comprising:

[0008] Setting process parameter conditions for the stamping parts, setting initial values for the set process parameter conditions, and obtaining initial process parameter conditions;

[0009] Performing finite element sampling on the stamping parts, performing CAE forming simulation springback prediction and compensation calculation, and obtaining sample data under the initial process parameter conditions;

[0010] Iterating through the parameters in the initial process parameter conditions, cyclically adjusting the parameter values of each parameter, and obtaining a sample data set;

[0011] Preprocessing the sample data set, training the machine learning model using the preprocessed sample data set, and obtaining a springback compensation big data model;

[0012] Taking the measured deviation vector value of the stamping parts as input, and obtaining the predicted springback compensation vector value through the springback compensation big data model.

[0013] Preferably, in the springback compensation method for stamping parts based on the machine learning model, the process parameter conditions include forming process parameter conditions and material parameter conditions, the forming process parameters include forming blank holder force, lubrication condition parameters, and forming speed parameters, and the material parameter conditions include part size parameters.

[0014] Preferably, in the stamping part springback compensation method based on the machine learning model, finite element sampling is performed on the stamping part, CAE forming simulation springback prediction and compensation calculation are performed, sample data under initial process parameter conditions are obtained, including: taking the element nodes of the finite element model of the stamping part as sampling position points, performing CAE forming simulation springback prediction and compensation calculation, obtaining the size deviation vector value and the springback compensation vector value of each sampling position point of the stamping part after springback, and taking the obtained size deviation vector value and the springback compensation vector value of each sampling position point of the stamping part after springback as the sample data under the initial process parameter conditions.

[0015] Preferably, in the stamping part springback compensation method based on the machine learning model, parameters in the initial process parameter conditions are traversed, the parameter values of each parameter are adjusted in a loop, and a sample data set is obtained, including:

[0016] Traversing the parameters in the initial process parameter conditions, adjusting the parameter values of each parameter in a loop, obtaining the adjusted corresponding process parameter conditions;

[0017] Under each corresponding process parameter condition, finite element sampling is performed on the stamping part, CAE forming simulation springback prediction and compensation calculation are performed, and sample data under the corresponding process parameter conditions are obtained;

[0018] The sample data under the initial process parameter conditions and the sample data under each corresponding process parameter condition are taken as the sample data set.

[0019] Preferably, in the stamping part springback compensation method based on the machine learning model, the sample data set is preprocessed, the machine learning model is trained by using the preprocessed sample data set, and a springback compensation big data model is obtained, including:

[0020] The sample data set is preprocessed by normalization, the sample data set after normalization is subjected to data enhancement, and the enhanced sample data set is taken as training input data;

[0021] The machine learning model based on the convolutional neural network algorithm is used to train the training input data, the deviation between the model prediction value and the expected springback compensation vector value is taken as a loss function, and a springback compensation big data model is constructed.

[0022] According to a second aspect of the present application, a stamping part springback compensation device based on a machine learning model is provided, the device comprising a springback compensation server, the springback compensation server being configured to: set a process parameter condition for a stamping part, set an initial value for the set process parameter condition, and obtain an initial process parameter condition; perform finite element sampling on the stamping part, perform CAE forming simulation springback prediction and compensation calculation, and obtain sample data under the initial process parameter condition; traverse parameters in the initial process parameter condition, cyclically adjust parameter values of each parameter, and obtain a sample data set; pre-process the sample data set, train a machine learning model using the pre-processed sample data set, and obtain a springback compensation big data model; and input a measured deviation vector value of the stamping part, and obtain a predicted springback compensation vector value through the springback compensation big data model.

[0023] Preferably, in the stamping part springback compensation device based on the machine learning model, the springback compensation server comprises:

[0024] a data collection module configured to set a process parameter condition for a stamping part, set an initial value for the set process parameter condition, and obtain an initial process parameter condition; perform finite element sampling on the stamping part, perform CAE forming simulation springback prediction and compensation calculation, and obtain sample data under the initial process parameter condition; traverse parameters in the initial process parameter condition, cyclically adjust parameter values of each parameter, and obtain a sample data set;

[0025] a model construction module configured to pre-process the sample data set, train a machine learning model using the pre-processed sample data set, and obtain a springback compensation big data model;

[0026] a springback compensation prediction module configured to input a measured deviation vector value of the stamping part, and obtain a predicted springback compensation vector value through the springback compensation big data model.

[0027] Preferably, in the stamping part springback compensation device based on the machine learning model, the data collection module is specifically configured to:

[0028] set a process parameter condition for a stamping part, set an initial value for the set process parameter condition, and obtain an initial process parameter condition;

[0029] obtain a measured deviation vector value of the stamping part, and obtain a predicted springback compensation vector value through the springback compensation big data model.

[0030] The parameters in the initial process parameter condition are traversed, the parameter value of each parameter is cyclically adjusted, and an adjusted corresponding process parameter condition is obtained; under each corresponding process parameter condition, finite element sampling is performed on the stamping part, CAE forming simulation springback prediction and compensation calculation are performed, and sample data under the corresponding process parameter condition is obtained; and the sample data under the initial process parameter condition and the sample data under each corresponding process parameter condition are taken as a sample data set.

[0031] Preferably, in the stamping part springback compensation device based on the machine learning model, the process parameter condition includes a forming process parameter condition and a material parameter condition, the forming process parameter includes a forming blank holder force, a lubrication condition parameter and a forming speed parameter, and the material parameter condition includes a part size parameter.

[0032] Preferably, in the stamping part springback compensation device based on the machine learning model, the model construction module is specifically configured to: perform normalization preprocessing on the sample data set, perform data enhancement on the sample data set after the normalization processing, and take the enhanced sample data set as training input data; and train the training input data by using a machine learning model based on a convolutional neural network algorithm, and construct a springback compensation big data model by taking the deviation between a model prediction value and an expected springback compensation vector value as a loss function.

[0033] According to a third aspect of the present application, a computer device is provided, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method according to the first aspect of the present application when executing the program.

[0034] The stamping part springback compensation method and device based on the machine learning model can automatically establish the rule of the part deviation value and the springback compensation value based on the finite element numerical model of the physical model, generate the springback compensation big data model, directly predict the optimal springback compensation result according to the measured part deviation, avoid the trial-and-error iteration process in the traditional method, and thus can quickly, effectively and at low cost complete the stamping die surface compensation. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0036] Figure 1A schematic diagram of a device for a stamping part springback compensation method based on a machine learning model according to an embodiment of the present application;

[0037] Figure 2 A schematic diagram of an architecture of a springback compensation server in a stamping part springback compensation device based on a machine learning model according to an embodiment of the present application;

[0038] Figure 3 A step flowchart of a stamping part springback compensation method based on a machine learning model according to an embodiment of the present application;

[0039] Figure 4 An execution flowchart of a stamping part springback compensation method based on a machine learning model according to an embodiment of the present application;

[0040] Figure 5 A stamping part and a stamping process according to an embodiment of the present application;

[0041] Figure 6 A schematic diagram of an architecture of a convolutional activation function used in the method according to an embodiment of the present application;

[0042] Figure 7 A schematic diagram of a prediction case performed by a springback compensation big data model according to the method of an embodiment of the present application;

[0043] Figure 8 A cloud diagram of a measured deviation vector value input in the method according to an embodiment of the present application;

[0044] Figure 9 A cloud diagram of a springback compensation vector value output by a springback compensation big data model using the method according to an embodiment of the present application;

[0045] Figure 10 A structural schematic diagram of an apparatus provided by the present application. DETAILED DESCRIPTION

[0046] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0047] It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict; and all other embodiments obtained by those skilled in the art based on the embodiments in the present disclosure without creative labor are within the scope of protection of the present disclosure.

[0048] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on this disclosure, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this apparatus and / or practice this method.

[0049] Figure 1 An exemplary device for a method for compensating for springback of stamped parts based on a machine learning model applicable to an embodiment of the present invention is shown. Figure 1 As shown, the apparatus may include a rebound compensation server 101, a communication network 102 and / or one or more rebound compensation clients 103. Figure 1 The example in FIG. 1 is a plurality of rebound compensation clients 103 .

[0050] The springback compensation server 101 can be any suitable server for storing information, data, programs, and / or any other suitable type of content. In some embodiments, the springback compensation server 101 can perform appropriate functions. For example, in some embodiments, the springback compensation server 101 can be used to perform springback compensation on stamped parts based on a machine learning model. As an optional example, in some embodiments, the springback compensation server 101 can be used to perform springback compensation on stamped parts by obtaining a springback compensation big data model. For example, the springback compensation server 101 can be used to set process parameter conditions for stamping parts, set initial values ​​for the set process parameter conditions, and obtain initial process parameter conditions; perform finite element sampling on stamping parts, and obtain sample data under the initial process parameter conditions through CAE forming simulation springback prediction and compensation calculation; traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain a sample data set; preprocess the sample data set, and use the preprocessed sample data set to train the machine learning model to obtain a springback compensation big data model; use the measured deviation vector value of the stamping part as input, and obtain the predicted springback compensation vector value through the springback compensation big data model.

[0051] Figure 2 This is an example diagram of the architecture of the springback compensation server in the springback compensation device for stamping parts based on the machine learning model according to an embodiment of the present invention. Figure 2 As shown, the rebound compensation server of the embodiment of the present invention includes:

[0052] The data acquisition module is configured to set a process parameter condition for the stamping part, set an initial value for the set process parameter condition, and obtain an initial process parameter condition; perform finite element sampling on the stamping part, perform CAE forming simulation springback prediction and compensation calculation, and obtain sample data under the initial process parameter condition; and traverse parameters in the initial process parameter condition, cyclically adjust parameter values of each parameter, and obtain a sample data set.

[0053] As an optional example, the data acquisition module in the embodiment of the present application is specifically configured to set a process parameter condition for the stamping part, set an initial value for the set process parameter condition, and obtain an initial process parameter condition; take element nodes of a finite element model of the stamping part as sampling position points, perform CAE forming simulation springback prediction and compensation calculation, obtain size deviation vector values and springback compensation vector values of the stamping part after springback at each sampling position point, take the obtained size deviation vector values and springback compensation vector values of the stamping part after springback at each sampling position point as sample data under the initial process parameter condition, traverse parameters in the initial process parameter condition, cyclically adjust parameter values of each parameter, obtain a corresponding process parameter condition after adjustment, perform finite element sampling on the stamping part under each corresponding process parameter condition, perform CAE forming simulation springback prediction and compensation calculation, and obtain sample data under the corresponding process parameter condition; and take the sample data under the initial process parameter condition and the sample data under each corresponding process parameter condition as a sample data set.

[0054] The model construction module is configured to pre-process the sample data set, train a machine learning model by using the pre-processed sample data set, and obtain a springback compensation big data model.

[0055] As an optional example, the model construction module in the embodiment of the present application is specifically configured to perform normalization preprocessing on the sample data set, perform data enhancement on the sample data set after normalization processing, take the enhanced sample data set as training input data, train the training input data by using a machine learning model based on a convolutional neural network algorithm, take a deviation between a model prediction value and an expected springback compensation vector value as a loss function, and construct a springback compensation big data model.

[0056] The springback compensation prediction module is configured to take a measured deviation vector value of the stamping part as input, and obtain a predicted springback compensation vector value by using the springback compensation big data model.

[0057] As another example, in some embodiments, the springback compensation server 101 can send the stamping part springback compensation method based on the machine learning model to the springback compensation client 103 for use by a user according to a request of the springback compensation client 103.

[0058] As an optional example, in some embodiments, the springback compensation client 103 is configured to provide a visual springback compensation interface for receiving a user's selection input operation of springback compensation of the stamping part based on the machine learning model, and for obtaining and displaying, in response to the selection input operation, a springback compensation interface corresponding to the option selected by the selection input operation from the springback compensation server 101, the springback compensation interface at least displaying information of springback compensation of the stamping part based on the machine learning model and operation options for the information of springback compensation of the stamping part based on the machine learning model.

[0059] In some embodiments, the communication network 102 can be any suitable combination of one or more wired and / or wireless networks. For example, the communication network 102 can include any one or more of the following: the Internet, an intranet, a wide-area network (WAN), a local-area network (LAN), a wireless network, a digital subscriber line (DSL) network, a frame relay network, an asynchronous transfer mode (ATM) network, a virtual private network (VPN), and / or any other suitable communication network. The springback compensation client 103 can connect to the communication network 102 through one or more communication links (e.g., communication link 104), which can link to the springback compensation server 101 via one or more communication links (e.g., communication link 105). The communication links can be any communication links suitable for communicating data among the springback compensation client 103 and the springback compensation server 101, such as network links, dial-up links, wireless links, hard-wired links, any other suitable communication links, or any suitable combination of such links.

[0060] The springback compensation client 103 can include any one or more clients that present interfaces related to springback compensation of the stamping part based on the machine learning model in a suitable form for use and operation by a user. In some embodiments, the springback compensation client 103 can include any suitable type of device. For example, in some embodiments, the springback compensation client 103 can include a mobile device, a tablet computer, a laptop computer, a desktop computer, and / or any other suitable type of client device.

[0061] Although the springback compensation server 101 is illustrated as one device, in some embodiments, any suitable number of devices can be used to perform the functions performed by the springback compensation server 101. For example, in some embodiments, multiple devices can be used to implement the functions performed by the springback compensation server 101. Alternatively, cloud services can be used to implement the functions of the springback compensation server 101.

[0062] Based on the above device, the embodiment of the application provides a stamping part springback compensation method based on a machine learning model, which is described below through the following embodiment.

[0063] Figure 3 A step flowchart of the stamping part springback compensation method based on the machine learning model of the embodiment of the application. Figure 4 An execution flowchart of the stamping part springback compensation method based on the machine learning model of the embodiment of the application. The stamping part springback compensation method based on the machine learning model of the embodiment can be executed on a springback compensation server, as shown in Figure 3 and Figure 4 The stamping part springback compensation method based on the machine learning model includes the following steps:

[0064] Step S201: setting process parameter conditions for a stamping part, setting initial values for the set process parameter conditions, and obtaining initial process parameter conditions.

[0065] As an optional example, in the method of the embodiment, the process parameter conditions include forming process parameter conditions and material parameter conditions, the forming process parameters include forming blank holder force, lubrication condition parameters, and forming speed parameters, and the material parameter conditions include part size parameters.

[0066] In the stamping forming process, the coupling of various process parameters has a complex influence on the quality of the stamping part, and reasonable process parameters are the fundamental guarantee for producing qualified parts. In actual production, each process parameter such as forming blank holder force, lubrication condition, and forming speed will produce a certain range of numerical fluctuations. Different numerical combinations of these process parameters can be used as a set of process parameter conditions for sample data collection. In addition, the thickness of the plate and the fluctuation of the material performance can also be used as change conditions for sample data collection. Figure 5 A stamping part and a stamping process example of the embodiment of the application.

[0067] Step S202: performing finite element sampling on the stamping part, and obtaining sample data under the initial process parameter conditions through CAE forming simulation springback prediction and compensation calculation.

[0068] As an optional example, in the method of the embodiment, the element nodes of the finite element model of the stamping part are used as sampling position points, the size deviation vector values and the springback compensation vector values of the stamping part at each sampling position point after springback are obtained through CAE forming simulation springback prediction and compensation calculation, and the size deviation vector values and the springback compensation vector values of the stamping part at each sampling position point after springback are used as sample data under the initial process parameter conditions.

[0069] It should be noted that in the method of the embodiment, the CAE forming simulation is realized by AutoForm, Dynaform and other CAE forming simulation software. Those skilled in the art can also select other suitable forming technical means to realize the method of the present application according to the actual application scene when implementing the method of the present application, which is not limited in the embodiment. In the method of the embodiment, the unit nodes of the finite element model of the stamped part are taken as the sampling position points, and the size deviation vector values and the springback compensation vector values after springback of each sampling position point are collected as the sample data under the initial process parameter condition.

[0070] Step S203: traversing the parameters in the initial process parameter condition, cyclically adjusting the parameter values of each parameter, and obtaining a sample data set.

[0071] In the method of the embodiment, after collecting the sample data under the initial process parameter condition, each process parameter needs to be traversed, and the value of the process parameter is cyclically adjusted within a reasonable value range. The finite element sampling of the stamped part is performed, the springback prediction and compensation calculation of the CAE forming simulation are performed, the sample data under the corresponding process parameter condition is obtained, and the sample data collection is ended.

[0072] As an optional example, in the method of the embodiment, the parameters in the initial process parameter condition are traversed, the parameter values of each parameter are cyclically adjusted, and the adjusted corresponding process parameter condition is obtained. The finite element sampling of the stamped part is performed under each corresponding process parameter condition, the springback prediction and compensation calculation of the CAE forming simulation are performed, the sample data under the corresponding process parameter condition is obtained, and the sample data set is obtained by taking the sample data under the initial process parameter condition and the sample data under each corresponding process parameter condition.

[0073] Step S204: preprocessing the sample data set, training the machine learning model by using the preprocessed sample data set, and obtaining a springback compensation big data model.

[0074] In the method of the embodiment, the sample data set is used as the training data of the machine learning model, and the machine learning model is trained in the manner of "taking the size deviation vector value after springback of the sampling position point as the input and taking the springback compensation vector value of the same position point as the answer".

[0075] As an optional example, in the method of the embodiment, the sample data set is preprocessed before the machine learning model is trained by using the sample data set. For example, the sample data set is normalized and preprocessed, the sample data set after normalization is data enhanced, and the enhanced sample data set is taken as the training input data. For example, in the method of the present application, the sample data set after normalization can be data enhanced by interpolation or coordinate conversion according to the sample capacity.

[0076] After obtaining the training input data, the method of the embodiment adopts a machine learning model based on a convolutional neural network algorithm to train the training input data, and constructs a springback compensation big data model by taking the deviation between the model prediction value and the expected springback compensation vector value as a loss function.

[0077] In the method of the embodiment, a machine learning model based on a convolutional neural network algorithm is adopted to output the training result through convolution activation, pooling operation and a full connection layer.

[0078] As an optional example, the convolution activation function adopted in the method of the embodiment is as follows:

[0079]

[0080] In the formula, V is an output matrix, is an activation function, conv2() is a convolution operation, W is a convolution kernel matrix, X is an input matrix, and b is a bias. Figure 6 FIG. 1 shows an architecture example of the convolution activation function adopted in the method of the embodiment. The convolution activation function in the method of the embodiment adopts the architecture as shown in FIG. 1, which can improve the training accuracy. Figure 6

[0081] As an optional example, the loss function adopted in the method of the embodiment is as follows:

[0082]

[0083] In the formula, y is a prediction value, y i is a true value, and MSE is a mean squared error loss function.

[0084] As an optional example, the method of the embodiment divides the preprocessed data into a training set, a validation set and a test set, and gradually reduces the loss function curve by using a convolutional neural network algorithm model architecture and hyperparameters, thereby improving the deep learning prediction accuracy and generalization ability of the machine learning model.

[0085] Figure 7 FIG. 2 shows a prediction case example of the springback compensation big data model according to the method of the embodiment, and the embodiment makes two groups of predictions. The measured deviation vector is taken as the input of the springback compensation big data model, and the springback compensation vector value obtained by the springback compensation big data model is taken as the output. The prediction result is consistent with the accurate value. Figure 7

[0086] Step S205: Taking the measured deviation vector value of the stamped part as the input, the predicted springback compensation vector value is obtained through the springback compensation big data model.

[0087] Figure 8 ​​The cloud chart for the measured deviation vector value input in the method of the embodiment, Figure 9 The cloud chart for the springback compensation vector value output predicted by the springback compensation big data model using the method of the embodiment. As shown in Figure 8 and Figure 9 The springback compensation big data model of the method of the embodiment can quickly and accurately predict the springback compensation vector value. It should be noted that new sample data accumulated in the implementation process of production and measurement in the method of the embodiment can also be continuously added to the training input data for re-iteration and training to continuously update the big model parameters based on deep learning, improve the model accuracy and adaptability.

[0088] The traditional springback compensation method for stamping parts adopts a springback compensation trial-and-error method based on the measured deviation of the parts or the CAE simulation prediction, which requires a lengthy and complex trial-and-error process. The springback compensation method and device for stamping parts based on the machine learning model of the embodiment generate data based on the finite element numerical model of the physical model, collect data through finite element simulation prediction, automatically establish the rules of the part deviation value and the springback compensation value, generate a springback compensation big data model, and directly predict the optimal springback compensation result according to the measured deviation of the parts, which can avoid the trial-and-error process in the traditional method, thereby quickly, effectively and low-costly completing the stamping die surface compensation.

[0089] Specifically, the springback compensation method and device for stamping parts based on the machine learning model of the embodiment generate data based on the finite element numerical model of the physical model, collect the size deviation vector value of the stamping part finite element model after springback at each sampling position point and the springback compensation vector value of the corresponding position point under multiple different stress boundary conditions through finite element simulation prediction, and use the same as sample data to train the springback compensation big data model from the part deviation vector value to the springback compensation vector value through the machine learning method. When the size measurement result of the stamping part is out of tolerance and needs to be processed for springback, the optimal springback compensation result can be directly predicted by using the springback compensation big data model established by the application according to the measured deviation of the part. Compared with the traditional springback compensation method for stamping parts, the method and device of the application process and analyze a large amount of data through machine learning, and give the corresponding decision by establishing the rules of the part deviation value and the springback compensation value, which not only effectively avoids the lengthy and complex trial-and-error process in the traditional method, but also greatly reduces the dependence of technical personnel on process experience, thereby quickly, effectively and low-costly completing the stamping die surface compensation and ensuring the die manufacturing cycle and the size precision of the stamping parts.

[0090] As shown in Figure 10As shown, the present application also provides a device including a processor 310, a communication interface 320, a memory 330 for storing a processor-executable computer program, and a communication bus 340. The processor 310, the communication interface 320, and the memory 330 complete communication with each other through the communication bus 340. The processor 310 realizes the stamping part springback compensation method based on the machine learning model by running the executable computer program.

[0091] The computer program in the memory 330 can be implemented in the form of a software functional unit and sold or used as an independent product. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0092] The system embodiments described above are only schematic, and the units illustrated as separate components can or can not be physically separate, and the components illustrated as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected based on actual needs to achieve the purpose of the embodiments. Those skilled in the art can understand and implement without creative labor.

[0093] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions essentially or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of the embodiments or some parts of the embodiments.

[0094] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for compensating for springback of stamped parts based on a machine learning model, characterized in that: The method comprises: Setting process parameter conditions for stamping parts, setting initial values ​​for the set process parameter conditions, and obtaining initial process parameter conditions; Finite element sampling is performed on stamping parts, and springback prediction and compensation calculation are performed through CAE forming simulation to obtain sample data under initial process parameter conditions; Traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain a sample data set; Preprocess the sample data set, and use the preprocessed sample data set to train the machine learning model to obtain a rebound compensation big data model; The measured deviation vector value of the stamped part is used as input, and the predicted springback compensation vector value is obtained through the springback compensation big data model; Finite element sampling is performed on the stamping parts, and sample data under the initial process parameter conditions are obtained through CAE forming simulation springback prediction and compensation calculation, including: using the unit nodes of the finite element model of the stamping parts as sampling position points, and obtaining the dimensional deviation vector value and springback compensation vector value of each sampling position point of the stamping parts after springback through CAE forming simulation springback prediction and compensation calculation, and using the obtained dimensional deviation vector value and springback compensation vector value of each sampling position point of the stamping parts after springback as the sample data under the initial process parameter conditions.

2. The method for compensating for springback of stamped parts based on a machine learning model according to claim 1, characterized in that: The process parameter conditions include forming process parameter conditions and material parameter conditions. The forming process parameters include forming blank holding force, lubrication condition parameters, and forming speed parameters. The material parameter conditions include part size parameters.

3. The method for compensating for springback of stamped parts based on a machine learning model according to claim 1, characterized in that: Traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain a sample data set, including: Traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain the adjusted corresponding process parameter conditions; Under each corresponding process parameter condition, finite element sampling is performed on the stamping parts, and the sample data under the corresponding process parameter conditions are obtained through CAE forming simulation springback prediction and compensation calculation; The sample data under the initial process parameter conditions and the sample data under each corresponding process parameter condition are taken as the sample data set.

4. The method for compensating for springback of stamped parts based on a machine learning model according to claim 1, characterized in that: Preprocess the sample data set, use the preprocessed sample data set to train the machine learning model, and obtain a rebound compensation big data model, including: Perform normalization pre-processing on the sample data set, perform data enhancement on the normalized sample data set, and use the enhanced sample data set as training input data; A machine learning model based on the convolutional neural network algorithm is used to train the training input data, and a rebound compensation big data model is constructed by taking the deviation between the model prediction value and the expected rebound compensation vector value as the loss function.

5. A stamping part springback compensation device based on a machine learning model, characterized in that: The device includes a springback compensation service end, which is used to: set process parameter conditions for stamping parts, set initial values ​​for the set process parameter conditions, and obtain initial process parameter conditions; Finite element sampling is performed on stamping parts, and springback prediction and compensation calculation are performed through CAE forming simulation to obtain sample data under initial process parameter conditions; Traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain a sample data set; The sample data set is preprocessed and used to train a machine learning model to obtain a springback compensation big data model. The measured deviation vector value of the stamped part is used as input to obtain the predicted springback compensation vector value through the springback compensation big data model. Finite element sampling is performed on the stamping parts, and sample data under the initial process parameter conditions are obtained through CAE forming simulation springback prediction and compensation calculation, including: using the unit nodes of the finite element model of the stamping parts as sampling position points, and obtaining the dimensional deviation vector value and springback compensation vector value of each sampling position point of the stamping parts after springback through CAE forming simulation springback prediction and compensation calculation, and using the obtained dimensional deviation vector value and springback compensation vector value of each sampling position point of the stamping parts after springback as the sample data under the initial process parameter conditions.

6. The stamping part springback compensation device based on the machine learning model according to claim 5 is characterized in that: The rebound compensation server includes: The data acquisition module is used to set process parameter conditions for stamping parts, set initial values ​​for the set process parameter conditions, and obtain the initial process parameter conditions; perform finite element sampling on the stamping parts, and obtain sample data under the initial process parameter conditions through CAE forming simulation springback prediction and compensation calculation; traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain a sample data set; A model building module is used to preprocess the sample data set and use the preprocessed sample data set to train the machine learning model to obtain a rebound compensation big data model; The springback compensation prediction module is used to take the measured deviation vector value of the stamping part as input and obtain the predicted springback compensation vector value through the springback compensation big data model.

7. The stamping part springback compensation device based on the machine learning model according to claim 6 is characterized in that: Data acquisition module, specifically used for: Setting process parameter conditions for stamping parts, setting initial values ​​for the set process parameter conditions, and obtaining initial process parameter conditions; The unit nodes of the finite element model of the stamping part are used as sampling points. Through CAE forming simulation springback prediction and compensation calculation, the dimensional deviation vector value and springback compensation vector value of each sampling point of the stamping part after springback are obtained. The dimensional deviation vector value and springback compensation vector value of each sampling point of the stamping part after springback are used as sample data under the initial process parameter conditions. Traverse the parameters in the initial process parameter conditions, cyclically adjust the parameter value of each parameter, and obtain the adjusted corresponding process parameter conditions; Under each corresponding process parameter condition, finite element sampling is performed on the stamping parts, and the sample data under the corresponding process parameter conditions are obtained through CAE forming simulation springback prediction and compensation calculation; The sample data under the initial process parameter conditions and the sample data under each corresponding process parameter condition are taken as the sample data set.

8. The stamping part springback compensation device based on the machine learning model according to claim 6 is characterized in that: The process parameter conditions include forming process parameter conditions and material parameter conditions. The forming process parameters include forming blank holding force, lubrication condition parameters, and forming speed parameters. The material parameter conditions include part size parameters.

9. The stamping part springback compensation device based on a machine learning model according to claim 6, characterized in that: The model building module is specifically used to: perform normalization pre-processing on the sample data set, perform data enhancement on the normalized sample data set, and use the enhanced sample data set as training input data; use a machine learning model based on the convolutional neural network algorithm to train the training input data, and construct a rebound compensation big data model by using the deviation between the model prediction value and the expected rebound compensation vector value as the loss function.

Images