Abstract: Optimization of pickup position based on uniform design method，neural network and genetic algorithm is proposed，which is in order to decrease the structural deformation of the automobile panels while they are transported in the stamping production line.Uniform design method is used to optimize the BP neural network parameters.Meanwhile，the work for FEA has been decreased effectively by establishing the nonlinear relation between coordinates of pickup position and deformation.Finally，the optimal coordinates of pickup position are obtained by GA，and the simulation results prove the effectiveness of the proposed methods.

Key words: uniform design method, neural network, genetic algorithm, automobile panels, optimization of pickup position

摘要： 为了减少车身覆盖件在冲压输送过程中的结构变形，提出了基于均匀设计法、神经网络和遗传算法相结合的智能的夹持点位置的优化算法。充分利用均匀设计代表性好，实验次数少，实验效益高的特点，建立了神经网络的训练样本；利用神经网络学习算法建立起夹持点位置坐标与位移、应力等的非线性全局映射关系，极大地减少了有限元重分析的工作量；最后利用遗传算法的全局寻优功能得出最优夹持点位置坐标，用于指导实际生产。仿真实验结果表明所提出的算法是有效的。

关键词: 均匀设计法, 神经网络, 遗传算法, 车身覆盖件, 夹持位置优化