摘要
基于薄板成形回弹正交试验的设计,利用Dynaform仿真软件对薄板成形回弹进行数值模拟,仿真结果表明:薄板弯曲成形高度随着模具间隙以及弯曲半径的增大而逐渐减小,随着冲压速度以及摩擦系数的增加而不断增大。以模具间隙、弯曲半径、冲压速度以及摩擦系数为输入层,将薄板弯曲成形高度作为输出层,建立4-12-1的3层BP神经网络。基于正交试验数据进行BP神经网络的训练与测试,BP神经网络预测值与有限元模拟值的误差为2.053%。此外,利用薄板成形模具进行试验验证,试验值与BP神经网络预测值的误差为11.87%,从而验证了BP神经网络的可靠性。
Based on the orthogonal experiment design for springback of sheet metal forming, the springback of sheet metal forming was simulated by finite element software Dynaform. The simulation results show that the forming height of sheet metal decreases with the in- crease of die clearance and bending radius, while it shows an increasing trend with the rising of stamping speed and friction coefficient. Therefore, taking die clearance, bending radius, stamping speed and friction coefficient as the input layer, forming height as the output layer, three - layer BP neural network of 4 - 12 - 1 was established. Training and testing of BF neural network were carried out based on data from the orthogonal experiment, and the error between the predicted value and the simulation value of BP neural network is 2. 053%. In addition, forming die of sheet metal is designed to verify the predicted value, and the error between the experimental value and the predicted value of BP neural network is 11.87%. Therefore, the reliability of BP neural network is proved.
出处
《锻压技术》
CAS
CSCD
北大核心
2016年第6期146-149,160,共5页
Forging & Stamping Technology
基金
连云港市中小企业技术创新项目(CK1411)
关键词
薄板冲压
成形回弹
BP神经网络
正交试验
数值模拟
stamping of sheet metal
forming springback
BP neural network
orthogonal experiment
numerical simulation
