Plate blanking process parameter optimization method based on automatic simulation and BP-GA algorithm

Abstract

The invention discloses a plate blanking process parameter optimization method based on automatic simulation and a BP-GA algorithm, and belongs to the field of plate stamping forming. The method comprises the steps that firstly, a blanking simulation model is established with the help of DEFORM simulation software, and setting of all pretreatment parameters is completed; performing multi-group finite element simulation by using a DEFORM-3D automatic simulation system to obtain corresponding wear depth values under different process parameter combinations; a BP neural network is constructed through sample data obtained through DEFORM automatic simulation, and a mapping relation between each blanking process parameter and the mold abrasion loss is established; and finally, using a genetic algorithm to obtain the minimum wear depth value of the die in the plate blanking process as an optimization target, and further obtaining an optimal blanking process parameter combination. According to the method, the service life of the die is prolonged, automatic, rapid and efficient optimization of the plate blanking process is achieved through automatic analog simulation of DEFORM software in combination with the BP-GA algorithm, and the process shaping speed is increased.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of sheet stamping forming, and particularly relates to a sheet blanking process parameter optimization method based on DEFORM software automatic simulation and neural network-genetic algorithm. BACKGROUND

[0002] Blanking is a kind of stamping process in which metal sheet is separated by a stamping die, including punching and blanking. Punching is a process of cutting off the sheet in one stroke, removing unnecessary materials and retaining small holes; and blanking is a process of obtaining a finished product from a parent material, and the process is completely consistent. In the blanking process, the die and the sheet are in contact and relative displacement occurs between them, resulting in severe friction on the contact surface, which causes die surface wear. In order to improve the service life of the blanking die, the blanking process parameters can be optimized to improve the die wear condition.

[0003] At present, in actual production, the relationship between the die wear depth and the blanking process parameters is studied by experimental method, which requires a large amount of economic cost. The finite element software can be used to conveniently and quickly obtain the die wear amount corresponding to the process parameters. DEFORM is a professional metal plastic forming simulation software, and its secondary development function provides great convenience and flexibility for users. Through secondary development, DEFORM text mode is called for automatic simulation, which can effectively reduce the workload of simulation experiment and improve the calculation efficiency of the software. SUMMARY

[0004] In order to solve the problems in the prior art, the application aims to provide a sheet blanking process optimization method based on DEFORM automatic simulation and BP-GA algorithm. The established DEFORM-3D automatic simulation system is used to complete the simulation of the sheet blanking process under multiple process parameters, and the neural network-genetic algorithm is used to analyze the simulation data to obtain the optimal process, thereby effectively improving the service life of the blanking die.

[0005] The technical scheme adopted by the application is as follows: a sheet blanking process parameter optimization method based on automatic simulation and BP-GA algorithm, comprising the following steps:

[0006] Step 1: a sheet punching finite element model is established, and the blank basic attribute setting, Archard wear model establishment, mesh division, simulation control and other pretreatment operations are performed in the DEFORM software, and the pretreatment setting is saved as a KEY file;

[0007] Step 2: Through the MATLAB software design interface, add buttons, text areas, menu bars and other controls and write callback functions to build the DEFORM-3D automated simulation system, realize the improvement of DEFORM software simulation efficiency, the specific functions of the system are: editing the KEY file saved in step 1, modifying the pre-processing parameters through the system, and generating a new KEY file; generating a bat batch file that can call the DEFORM software text mode for simulation; running multiple bat files in turn, displaying in the status bar whether the simulation calculation is completed, if not, display the name of the bat file being executed;

[0008] Step 3: Select the orthogonal test design method to design and distribute the mold gap, mold hardness, and stamping speed three process parameters, and generate KEY files and corresponding batch files through the DEFORM-3D automated simulation system. Run the bat file one by one, call the DEF_PRE.COM and DEF_ARM_CTL.COM programs of DEFORM software, realize the automatic simulation of the simulation process, and obtain the wear degree of the blanking die under each parameter combination;

[0009] Step 4: Import the sample data obtained by finite element simulation into MATLAB software, use the mold gap, mold hardness and stamping speed as the input variables of the neural network model, and use the punch wear depth as the output variable of the neural network model to establish a BP neural network. Compare the neural network output value and the finite element simulation value, use the relative error and the fitting coefficient to evaluate the neural network, and determine the accuracy of the constructed BP neural network;

[0010] Step 5: Use genetic algorithm to use the constructed neural network function as a function function to find the global optimal value (minimum wear depth) of the mapping function and the corresponding three input values (mold gap, mold hardness and stamping speed).

[0011] Further, the step 3 includes:

[0012] According to the process requirements of the actual production of the experiment: the range of the mold gap is C1~C2mm, the range of the mold hardness is H1~H2HRC, and the range of the stamping speed is V1~V2mm / s. The orthogonal table of the orthogonal experiment method is represented by the expression L n (m k ), in which, n is the number of experiments, k is the experimental factor, and m is the number of levels of each factor; In this experiment, the three input parameters of the BP neural network are the experimental factors, which are the mold gap, the mold hardness and the stamping speed, and five levels are selected for each process parameter, that is, L 25 (5 3 ).

[0013] Further, the calculation formula of the relative error evaluation in step 4 is:

[0014]

[0015] In the formula, Z is the relative error, Y is the predicted value of the BP neural network, and X is the actual value of the finite element simulation.

[0016] The beneficial effects of the present application are that: a blanking simulation model is established by means of DEFORM simulation software to complete the setting of each pretreatment parameter; a plurality of finite element simulations are carried out by using a DEFORM-3D automatic simulation system to obtain corresponding wear depth values under different process parameter combinations; a BP neural network is constructed by using sample data obtained by DEFORM automatic simulation to establish a mapping relationship between each blanking process parameter and die wear amount; finally, a genetic algorithm is used to obtain the minimum die wear depth value in the plate blanking process as an optimization objective, and then the optimal blanking process parameter combination is obtained. The present application takes improving the service life of the die as the target, realizes automatic, rapid and efficient optimization of the plate blanking process by automatic simulation and emulation of DEFORM software and combination of BP-GA algorithm, and speeds up the process standardization speed.

[0017] The neural network can take a plurality of process parameters as inputs, automatically extract potential features and relationships between the parameters through its complex network structure, and the genetic algorithm can search for the optimal parameter combination in a high-dimensional parameter space, thereby effectively processing the multi-parameter process optimization problem. For the data obtained by DEFORM software automatic simulation, a mapping function relationship between the objective function and the variables is established by using the neural network, and then global optimization is carried out on the basis of the neural network model by using the genetic algorithm, which can quickly perform process parameter optimization and obtain the best blanking process parameter combination.

[0018] The present application constructs a simulation model of the blanking process by DEFORM finite element software, replaces the blanking physical experiment, reduces the trial and error cost, and establishes a DEFORM-3D automatic simulation system to obtain blanking simulation to improve the simulation efficiency. The mapping relationship between each blanking process parameter and die wear amount is established by the neural network, the global search ability of the genetic algorithm is used for optimization and solution, the traditional algorithm is avoided to fall into local optimum, the minimum value of the die wear amount and the corresponding blanking process parameter combination are found out to reduce the die wear amount.

[0019] The efficiency of obtaining die wear amount data corresponding to different blanking process parameter combinations is high by applying the DEFORM-3D automatic simulation system for blanking simulation; the blanking process parameter optimization is carried out by combining the neural network and the genetic algorithm, and the method can jump out of the local optimal solution and has strong global search ability. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 Finite element model for punching of sheet metal.

[0021] Figure 2 Interface of DEFORM-3D automatic simulation system.

[0022] Figure 3 Structure diagram of BP neural network.

[0023] Figure 4 Comparison diagram of predicted value of neural network and simulated value of finite element.

[0024] Figure 5 Curve diagram of fitting degree of BP neural network.

[0025] Figure 6 Curve diagram of optimal fitness.

[0026] Figure 7 Simulation result of die wear depth corresponding to optimal process parameters. DETAILED DESCRIPTION

[0027] The present application will be further described in the following examples. It is necessary to point out that the following examples are only used to further illustrate the present application, but not limited to the present application, and unless otherwise specified.

[0028] The present application will be further described in the following examples. It is necessary to point out that the following examples are only used to further illustrate the present application, but not limited to the present application, and unless otherwise specified.

[0029] The present application will be further described in the following examples. It is necessary to point out that the following examples are only used to further illustrate the present application, but not limited to the present application, and unless otherwise specified.

[0030] Step 1: a quarter three-dimensional model of punch, pressure ring, sheet metal and lower die is respectively established, and then they are combined into a punching model, as shown in Figure 1 .

[0031] The DEFORM software is used to set the punch material as AISI-D2 die steel, and the sheet metal material as CP800 complex phase steel, and both are divided into 90,000 grids. The classic Archard wear model is used, which is based on the metal volume plastic forming theory, and the calculation formula is as follows:

[0032]

[0033] In the formula, W is the wear depth; P is the surface normal stress between the sheet metal and the die contact surface; v is the relative sliding speed of the die and the sheet metal; H is the hardness of the die; a=1, b=1, c=2, which are constant factors related to the material; K is the proportional coefficient, and the value is K=2×10 -6t is the contact time of the die and the sheet. The simulation step is set to 0.08 mm, the simulation number is set to 100 steps, the friction coefficient is set to 0.12, the die gap is set to 0.23 mm, the initial hardness of the die is set to 56HRC, and the stamping speed is set to 170 mm / s. After determining that each parameter is correctly set, the pre-processing setting is saved in the KEY file format.

[0034] Step 2: Through the MATLAB software design interface, add buttons, text areas, menu bars and other controls and write callback functions to build a DEFORM-3D automated simulation system to improve the simulation efficiency of DEFORM software. The specific functions of the system are: editing the KEY file saved in step 1, modifying the pre-processing parameters through the system, and generating a new KEY file; generating a bat batch file that can call the DEFORM software text mode for simulation; running multiple bat files in sequence, and displaying in the status bar whether the simulation calculation is completed, if not, displaying the name of the bat file being executed;

[0035] Step 3: According to the process requirements of the actual production of this experiment: the range of die gap is 0.23~0.35mm, the range of die hardness is 56~60HRC, and the range of stamping speed is 170~210mm / s. Select the die gap, the initial hardness of the die, and the stamping speed as the main influencing factors of the punch wear depth, and design an orthogonal experiment with 5 levels for each factor, as shown in Table 1.

[0036] Table 1 Orthogonal experiment factor level table

[0037]

[0038] In this embodiment, in order to improve the calculation efficiency of the finite element software, an DEFORM-3D automated simulation system is built to complete the orthogonal experiment, as shown in Figure 2 In the pre-processing module of the interface, according to the orthogonal experiment scheme, the specific values of the die gap, die hardness and stamping speed are input to realize the setting of the related parameters. In the simulation calculation module of the automated simulation system, the KEY file corresponding batch processing bat file can be generated. After clicking run, the bat file can be run one by one, calling the DEF_PRE.COM and DEF_ARM_CTL.COM programs of DEFORM software, realizing the automatic calculation of the simulation process, and the running state of the simulation process can be monitored in real time in the running state bar. The use of DEFORM-3D automated simulation system can simplify the repeated pre-processing parameter modification step into one operation, thereby effectively improving the efficiency of numerical simulation. The calculation results are shown in Table 2.

[0039] Table 2 Orthogonal experiment scheme and results

[0040]

[0041] Step 4: Using the 25 sets of experimental data from Step 3, construct a neural network containing an input layer, hidden layers, and an output layer. The input layer has 3 nodes, representing the 3 factors affecting die wear (die clearance, die hardness, and stamping speed); the output layer has 1 node, representing the wear depth of the punch; the number of nodes in the hidden layer is determined using an empirical formula.

[0042]

[0043] In the formula: m is the number of neurons in the input layer; n is the number of neurons in the output layer; The value is a constant between [1, 10]. With the number of hidden layer neurons p set to 10, the topology of the BP neural network is as follows: Figure 3 As shown, the data is randomly divided proportionally, with training data, validation data, and test data accounting for 70%, 15%, and 15% of the total data, respectively.

[0044] The BP neural network outputs a predicted value of wear depth, and the reliability of the BP neural network model is measured by the relative error. Figure 4 The image shows a comparison between 25 sets of simulated experimental values ​​and the prediction results of the BP neural network. The maximum error in the test results is 7.79%, indicating that the prediction values ​​of the established BP neural network match the simulated experimental values ​​well, and the prediction results are quite ideal. Figure 5 The figure shows the fitting analysis between the predicted values ​​and simulation experimental values ​​of the (3-10-1) structured BP neural network model. The results show that the established BP neural network model has high fitting accuracy. As can be seen from the figure, the fitting regression R value of the training set is 0.984, the fitting regression R value of the validation set is 0.967, the fitting regression R value of the test set is 0.937, and the overall fitting regression R value is 0.969, indicating a good fitting effect.

[0045] Step 5: Using a genetic algorithm, the neural network established in Step 4 is used as the function to find the global optimum of the mapping function (minimum wear depth) and its corresponding three input values ​​(die clearance C, die hardness H, and stamping speed V). The crossover probability is set to 0.8, the mutation probability to 0.1, the number of generations to 100, and the population size to 100. The fitness value iteration process is as follows... Figure 6 As shown, with the increase of the number of iterations, the fitness value gradually decreases and eventually tends to stabilize. After 79 iterations, the iterations stopped, and the minimum wear depth of the punching die was found to be 2.55 × 10⁻⁶. -6 The corresponding optimal parameter combination is a die clearance of 0.35 mm, a die hardness of 60 HRC, and a stamping speed of 180 mm / s.

[0046] The optimized process parameters were imported into DEFORM-3D software for numerical simulation verification of the punching process. The simulation results are as follows: Figure 7 As shown, the wear location is mainly at the cutting edge that contacts the workpiece, and the maximum wear depth in a single punching operation is 2.66 × 10⁻⁶. -6 mm, the same as the prediction result of the genetic algorithm is 2.55×10. -6 The results showed that the parameters were close to the target value (mm), with a relative error of 4.13%, proving the feasibility of optimizing stamping process parameters through intelligent algorithms.

[0047] The above embodiments are only used to illustrate the present invention. Any equivalent transformations and improvements made on the basis of the technical solutions of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for optimizing the blanking process parameters based on automatic simulation and BP-GA algorithm, characterized in that, The method comprises the following steps: Step 1: Establishing a blank punching finite element model, setting the pre-processing parameters in DEFORM software, and saving the pre-processing settings as a KEY file; Step 2: Building a DEFORM-3D automatic simulation system in MATLAB, editing the KEY file saved in step 1, modifying the pre-processing parameters through the system, and generating a new KEY file; and generating a bat batch file that can call the text mode of DEFORM software for simulation; Running multiple bat files in sequence, and displaying in the status bar whether the simulation calculation is completed, and if not, displaying the name of the bat file being executed; Step 3: Designing an orthogonal experiment table, and designing and distributing three process parameters of die gap, die hardness and punching speed. Through the DEFORM-3D automatic simulation system, the model established in step 1 is pre-processed, which specifically includes: modifying the process parameters according to the orthogonal experiment table, generating KEY files and bat batch files corresponding to each group of process parameters, and performing finite element simulation calculation to obtain the wear degree of the punching die under each parameter combination; Step 4: Based on the sample data obtained in step 3, taking the die gap, die hardness and punching speed as the input variables of the neural network model, and taking the punch wear depth as the output variable of the neural network model to establish a BP neural network; Step 5: Using genetic algorithm to take the constructed neural network function as a functional function to find the minimum wear depth value and the corresponding punching process parameter combination.

2. The method of claim 1, wherein, In step 1, the pre-processing parameters include blank basic attribute setting, Archard wear model establishment, mesh division, simulation step, simulation step number, die gap, die hardness and punching speed.

3. The method of claim 1, wherein: In step 2, the DEFORM-3D automatic simulation system is built, DEFORM software is called by running the system, and the automatic simulation and emulation of the blank punching process are realized.

4. The method of claim 1, wherein, In step 3, the process parameters are determined according to the actual production process requirements: the die gap range is C1~C2mm, the die hardness range is H1~H2HRC, and the punching speed range is V1~V2mm / s.

5. The method of claim 1, wherein: In step 3, the bat files are run in sequence, the DEF_PRE.COM and DEF_ARM_CTL.COM programs of DEFORM software are called, the automatic simulation of the simulation process is realized, and the simulation results of the blank punching process are obtained.

Images