人工神经网络在覆盖件拉深筋设计中的应用

Application of ANN in the drawbead design of automobile panel

文章引入流入量的概念作为衡量板料进料阻力的依据,提出用拉深筋嵌入程度和材料流经拉深筋时弯曲程度2个因素来评价拉深筋对进料阻力的影响。采用BP神经网络分别建立单筋和双筋与流入量之间关系的神经网络模型,对神经网络建模的关键技术进行讨论分析,建立了性能良好的映射模型。结果表明,所建立的网络模型具有优良的性能,能充分反映流入量与筋参数之间的映射关系,为后续有限元模拟和实际生产提供了有价值的参考。

The length of flown-in material （LFM） is introduced to scale the restraining force, and the embedded extent of drawbead and the bending degree as the material flows through the drawbead are used to estimate the influence of drawbead on restraining force. The models for single drawbead and double drawbead are constructed by applying the artificial neural network（ANN）. The experimental research on the networks has been made by the back propagation arithmetic, and the key techniques for model construction are discussed. The result indicates that models are good enough to sufficiently reflect the mapped relationship between the length of flown-in material （LFM） and the parameters of drawbead, so they are valuable for FE simulation and production.