摘要:
|
在弯曲成形智能化控制过程的4个要素中,材料性能参数的实时识别占有极其重要的地位。为提高实时识别的精度和效率,根据宽板V型自由弯曲成形的特点,建立了宽板V型自由弯曲成形智能化控制过程材料性能参数实时识别的LM神经网络模型。训练结果表明,与改进的BP网络模型比较,LM网络模型的收敛精度和收敛速度均有明显的提高,为实现弯曲成形过程的智能化控制奠定了基础。
|
In the four basic factors on the intellectualization of sheet metal forming,the real-time identification of the material performance parameter has a very important position.To improve accuracy and efficiency of the real-time identification,during the process of the intelligence control of wide V-shaped free bending,the real-time identification of the material performance parameter neural network model based on LM algorithm has been used based on the characteristic of V-shaped wide free bending.The training result proves that the converge accuracy and speed of the neutral network have been evidently boosted,compared with the improved neural network based on BP algorithm,it paves the way for the intelligent control of sheet metal forming.
|
基金项目:
|
河北省自然科学基金资助项目(501215)
|
作者简介:
|
|
参考文献:
|
[1]赵军,李硕本,吕炎.板材冲压成形的智能化控制技术[J].塑性工程学报,1999,6(4):10 21.
[2]赵军,秦泗吉,曹宏强,等.轴对称曲面件智能化拉深成形过程的解析定量描述[J].塑性工程学报,1998,5(4):47 58.
[3]焦李成.神经网络系统理论[M].西安:西安电子科技大学出版社,1992.
[4]张立明.人工神经网络的模型及其应用[M].上海:复旦大学出版社,1994.
[5]Pourboghrat F,Chu E,Prediction of spring-back and side-wall curl in 2D draw bending[J].Journal of Materials Pro-cessing Technology,1995,50:361 374.
[6]Long Wu Wu.Spring-back and residual stress analysis forsheet metal forming processing by dynamic relaxation in con-junction with finite element method[C].Internatonal con-ference technology plasticity,1993.
[7]Hagen T M,Menhaj M.B.Training feedforward networkswith the Levenberg-Marquardt algorithm[J].Institate ofElectrical and Electronics Ensineers Trans on Neural Net-works.1994,5(6):989 993.
[8]司捷,周贵安,李函,等.基于梯度监督学习的理论与应用(I)-基本算法[J].清华大学学报,1997,37(7):71 73.
[9]袁亚湘,孙文瑜.最优化理论与方法[M].北京:科学出版社,1997.
[10]张莹莹,王劭伯.基于LM的神经网络偏差补偿预测控制及其应用[J].福州大学学报(自然科学版),2001,29(1):43 46.
[11]左敦稳,朱纪军,王珉.基于LM算法的无氢类金刚石薄膜喇曼高斯分解[J].中国机械工程,2000,11(11):1296 1298.
[12]闻新.MATLAB神经网络应用设计[M].北京:科学出版社,2000.
[13]官英平.板料V型自由弯曲智能化控制技术的研究[D].秦皇岛.燕山大学,2004.
[14]Carpenter W C,Hoffman M E.Selecting the architecture of aclass of back-propagation neural networks used as approxima-tor[J].Artificial Intelligence for Engineering Design,A-nalysis and Manufacturing,1997,11(1):33 44.
|
服务与反馈:
|
【文章下载】【加入收藏】
|
|
|