【摘要】 本文基于双链量子遗传(DCQGA)算法对BP神经网络的权值和阈值进行了优化,提出了BPN-DCQGA算法,建立了板料激光弯曲中弯曲角度的预测模型。利用实验结果作为样本数据对网络进行训练和校验,结果表明该算法能有效解决网络的收敛速度慢和易陷入局部最优解的问题。基于新的算法建立了板料激光弯曲工艺参数优化系统,该系统具有较好的鲁棒性和较准确的工艺参数优化和变形角度预测功能,为实际生产和加工提供了实用途径,同时推动了板料激光弯曲技术的实用化进程。更多还原

【Abstract】 Aiming at the disadvantage of classical GA,this paper improved GA and put forwards a new algorithm named DCQGA. BP network was improved based on DCQGA and then we get BPN-DCQGA net. The predicted model of laser bending angle was set up in the process of laser bending for sheet metal. The BPN-DCQGA net was trained and verified through the sample data result from experimental data. It proves that the improved net has enhanced the rate of convergence,and gained high train efficiency and better robustness. Moreover,based on the mentioned model,the optimization system was found and it can well solve the problem of manufacture parameters optimization in the laser bending process. This system will largely benefit the manufacture and drive the application of laser bending.更多还原