遗传算法结合神经网络实现拉延筋优化设计

传统的冲压模具设计中,拉延筋设计和布置主要依靠经验,这使得模具设计和制造周期延长.以某汽车侧壁外板的拉深工序为例,讨论了神经网络技术与遗传算法在拉延筋优化设计中的综合应用问题.建立了反映板料成形参数与拉延筋阻力之间非线性映射关系的BP网络模型.利用该训练好的神经网络可以实现拉延筋的优化设计.由于相对于进行工艺试验来说数值仿真比较省时省力,因此,利用Dynaform模拟汽车侧壁外板的拉深成形过程,建立训练样本.在网络的训练方法上利用遗传算法进行了优化,有效地提高了神经网络的模拟精度.

Optimized design of drawbead using improved genetic algorithm combined with BP artificial neural network

In traditional process, the design and arrangement of drawbeads largely depend on experience, which makes the designing and manufacturing period of die set to be prolonged. The application of ANN and GA in drawbead optimized design was discussed. A neural network model based on car sidewall drawing was designed to simulate the nonlinear mapping relation between the drawbead force and sheet forming parameters. The drawbead optimized design could be realized by using the ANN. Because the finite element simulation costed less time than process test, the car sidewall drawing process with Dynaform was simulated and the training sample of ANN was created. The training method of the ANN was optimized in GA. It was helpful to improving the simulating degree of ANN.