论文部分内容阅读
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 design of rebound orthogonal test of sheet forming, the numerical simulation of sheet forming rebound was carried out by using Dynaform simulation software. The simulation results show that the sheet bending forming height decreases with the increase of die gap and bending radius, Speed and increase the friction coefficient increases. Taking the die gap, bending radius, punching speed and friction coefficient as the input layer, the sheet bending forming height was taken as the output layer, and a 4-12-1 BP neural network was established. Based on the orthogonal test data, BP neural network training and testing, BP neural network prediction and finite element simulation of the error is 2.053%. In addition, the experimental verification of sheet metal forming molds shows that the error between the experimental value and the BP neural network prediction value is 11.87%, which verifies the reliability of the BP neural network.