基于GS理论与神经网络的汽车覆盖件成形优化
发布时间：2018-09-18 07:26
【摘要】：车身覆盖件由于形状复杂、精度要求高,其冲压成形过程是一个复杂的弹塑性、大变形力学过程。随着数值模拟技术的高速发展,已有大量学者利用数值模拟技术对板料在成形过程中可能出现的缺陷进行了预测,利用该技术既可以减少试模次数,还能够优化模具的结构设计和工艺设计。然而,由于汽车覆盖件外形结构复杂,对其质量要求较高,这使得单纯使用数值模拟技术具有较高的盲目性。花费了大量仿真时间,实验结果仍不理想。如何合理使用各种优化算法,并将多种算法结合起来对冲压过程中的各个成形工艺参数进行合理调整,并获得满足质量要求的工艺参数组合,长期以来都是汽车制造领域的热点和难点。因此,本文基于GS理论和神经网络遗传算法函数寻优法,利用非线性有限元分析软件Dynaform,将一汽某汽车制造有限公司生产的轻量型卡车左后侧围外板为研究对象,对其冲压成形中工艺参数寻优,以提高其成形质量。首先,借助正交试验法,初步获取在不同工艺参数组合下的最大减薄率数值;接着,基于GS理论,对获取的数据进行灰色关联度分析,分析得出影响左后侧外板最大减薄率的两个最主要因素即冲压速度和压边力;之后,使用拉丁超立方抽样法对选出的两个主要因素在既定的范围内进行随机抽样,并借助非线性有限元软件Dynaform对抽样结果逐一仿真分析获得各自的最大减薄率;然后,基于神经网络遗传算法函数寻优模型,将拉丁超立方抽样获得数据中的冲压速度和压边力作为输入,最大减薄率作为输出,用输入输出数据训练BP神经网络;用遗传算法寻优把训练后的BP神经网络预测结果作为个体适应度值计算的初始值,并进行个体适应度值计算即预测得到最大减薄率最小的结果,并得到对应输入值。最后,运用Dynaform软件对输出的最优工艺参数组合进行仿真验证,借助该公司的实验平台进行实验验证。通过对比优化前后的有限元仿真结果和实验结果可知,优化后的拉延成形工艺参数能够改善板料的成形性能。采用此方法可以有效的预测和优化汽车覆盖件成形工艺参数。
[Abstract]:Due to complex shape and high precision, the stamping process of body panel is a complicated elastoplastic and large deformation mechanical process. With the rapid development of numerical simulation technology, a large number of scholars have made use of numerical simulation technology to predict the possible defects in sheet metal forming process. Also can optimize the mold structure design and process design. However, due to the complexity of the shape and structure of the automobile panels, the quality of the panels is very high, which makes the use of numerical simulation technology more blind. A lot of simulation time has been spent and the experimental results are still not satisfactory. How to reasonably use all kinds of optimization algorithms and combine them together to adjust each forming process parameter reasonably and obtain the combination of process parameters to meet the quality requirements. For a long time, it has been a hot and difficult point in the field of automobile manufacturing. Therefore, based on GS theory and neural network genetic algorithm function optimization method, this paper uses nonlinear finite element analysis software Dynaform, to study the left rear side outer panel of light weight truck produced by FAW Automobile Manufacturing Co., Ltd. The process parameters are optimized to improve the forming quality. Firstly, the maximum thinning rate of different process parameters is obtained by orthogonal test. Then, based on GS theory, grey correlation degree analysis of the obtained data is carried out. The results show that the two most important factors affecting the maximum thinning rate of the left and posterior outer plate are the stamping speed and the blank holder force, and then the Latin hypercube sampling method is used to sample the selected two main factors randomly within a given range. With the help of the nonlinear finite element software Dynaform, the sample results are simulated one by one to obtain the maximum thinning rate, and then, based on the neural network genetic algorithm function optimization model, The punching speed and blank holding force in the data obtained by Latin hypercube sampling are taken as input, the maximum thinning rate is taken as output, and the input and output data are used to train BP neural network. Genetic algorithm is used to optimize the predicted results of the trained BP neural network as the initial value of the individual fitness calculation, and the individual fitness value is calculated, that is, the maximum thinning rate and the minimum thinning rate are predicted and the corresponding input values are obtained. Finally, Dynaform software is used to simulate the optimal process parameters, and the experiment platform of the company is used to verify the optimal process parameters. By comparing the finite element simulation results and experimental results before and after optimization, it can be seen that the optimized drawing process parameters can improve the forming performance of sheet metal. This method can effectively predict and optimize the forming process parameters of automobile panels.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG386;U466
[Abstract]:Due to complex shape and high precision, the stamping process of body panel is a complicated elastoplastic and large deformation mechanical process. With the rapid development of numerical simulation technology, a large number of scholars have made use of numerical simulation technology to predict the possible defects in sheet metal forming process. Also can optimize the mold structure design and process design. However, due to the complexity of the shape and structure of the automobile panels, the quality of the panels is very high, which makes the use of numerical simulation technology more blind. A lot of simulation time has been spent and the experimental results are still not satisfactory. How to reasonably use all kinds of optimization algorithms and combine them together to adjust each forming process parameter reasonably and obtain the combination of process parameters to meet the quality requirements. For a long time, it has been a hot and difficult point in the field of automobile manufacturing. Therefore, based on GS theory and neural network genetic algorithm function optimization method, this paper uses nonlinear finite element analysis software Dynaform, to study the left rear side outer panel of light weight truck produced by FAW Automobile Manufacturing Co., Ltd. The process parameters are optimized to improve the forming quality. Firstly, the maximum thinning rate of different process parameters is obtained by orthogonal test. Then, based on GS theory, grey correlation degree analysis of the obtained data is carried out. The results show that the two most important factors affecting the maximum thinning rate of the left and posterior outer plate are the stamping speed and the blank holder force, and then the Latin hypercube sampling method is used to sample the selected two main factors randomly within a given range. With the help of the nonlinear finite element software Dynaform, the sample results are simulated one by one to obtain the maximum thinning rate, and then, based on the neural network genetic algorithm function optimization model, The punching speed and blank holding force in the data obtained by Latin hypercube sampling are taken as input, the maximum thinning rate is taken as output, and the input and output data are used to train BP neural network. Genetic algorithm is used to optimize the predicted results of the trained BP neural network as the initial value of the individual fitness calculation, and the individual fitness value is calculated, that is, the maximum thinning rate and the minimum thinning rate are predicted and the corresponding input values are obtained. Finally, Dynaform software is used to simulate the optimal process parameters, and the experiment platform of the company is used to verify the optimal process parameters. By comparing the finite element simulation results and experimental results before and after optimization, it can be seen that the optimized drawing process parameters can improve the forming performance of sheet metal. This method can effectively predict and optimize the forming process parameters of automobile panels.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG386;U466
【相似文献】
相关期刊论文 前10条
1 覃建周,吴伯杰;汽车覆盖件冲压方向的优化[J];模具工业;2002年02期
2 沈启，
本文编号：2247208
本文链接：https://www.wllwen.com/kejilunwen/jiagonggongyi/2247208.html
