包装件非线性特性识别的进化神经网络混合方法

吉林大学学报(工学版) ›› 2001, Vol. ›› Issue (1): 66-70.

包装件非线性特性识别的进化神经网络混合方法

杨晓伟¹, 梁艳春², 龚文英³, 周春光⁴

1. 华南理工大学应用数学系, 广东广州 510640;
2. 中国科学院现代制造CAD/CAM技术开放实验室, 辽宁沈阳 110003;
3. 吉林大学前卫校区数学系, 吉林长春 130012;
4. 吉林大学前卫校区计算机科学系, 吉林长春 130012

收稿日期:2000-06-15 出版日期:2001-01-25
基金资助:
国家自然科学基金资助项目(19872027);符号计算与知识工程国家教育部开放研究实验室资助项目

A Hybrid Evolutionary Neural Network Method in Identifying Nonlinear Characteristics of Packaging

YANG Xiao-wei¹, LIANG Yan-chun², GONG Wen-ying³, ZHOU Chun-guang⁴

1. Dept of Applied Mathematics, South China University of Technology, Guangzhou 510640, China;
2. Open Lab of CAD/CAM Technology for Modern Manu facturing, Academia Sinica, Shenyang 110003, China;
3. Dept of Mathematics, Jilin University, Qianwei Campus, Changchun 130012, China;
4. Dept of Computer Science, Jilin University, Qianwei Campus, Changchun 130012, China

Received:2000-06-15 Online:2001-01-25

摘要/Abstract

摘要： 提出了一种将遗传算法和BP算法相结合的关于结构化神经网络的混合训练方法,并将其用于解决包装件缓冲垫层非线性特性识别问题。在该方法中,提出了一种新的自适应变异操作技术及将遗传算法与BP算法进行自适应切换的实施方案。用于两种典型的包装件缓冲垫层材料的模拟识别结果表明:应用此方法可以有效地解决包装件缓冲垫层非线性特性识别的问题,同时也为神经网络的混合训练提供了一种新的可行的途径。

关键词: 遗传进化, 缓冲包装, 非线性特性, BP算法, 结构化神经网络, 模型识别

Abstract: A hybrid method using the genetic algorithm(GA) and the error back propagation (BP) algorithm to train structural neural networks is presented and applied to the identification of nonlinear characteristics of packaging cushioning in this paper.A novel adaptive mutation technique and a scheme to shift the training of the network from the GA to the BP algorithm are proposed.The simulated results on the two typical models of packaging cushioning materials show that the nonlinear charateristics can be identified perfectly.This method also provides a feasible way of hybrid training of neural networks.

Key words: genetic evolution, packaging cushioning, nonlinear characteristics, BP algorithm, structural neural networks, model identification

中图分类号:

O327

杨晓伟, 梁艳春, 龚文英, 周春光. 包装件非线性特性识别的进化神经网络混合方法[J]. 吉林大学学报(工学版), 2001, (1): 66-70.

YANG Xiao-wei, LIANG Yan-chun, GONG Wen-ying, ZHOU Chun-guang. A Hybrid Evolutionary Neural Network Method in Identifying Nonlinear Characteristics of Packaging[J]. 吉林大学学报(工学版), 2001, (1): 66-70.

参考文献

[1] Liang Yanchun, Yang Xiaowei, Zhou Chunguang, et al. Application of neural networks to identification of nonlinear characteristics in cushioning packaging［J］. Mechanics Research Communications, 1996, 23(6):607-613.
[2] Liang Yanchun, Wang Zheng, Yang Xiaowei, et al. Identification of nonlinear characteristics in cushioning packaging using neural networks with fi.Jzzy adaptive control［J］. Mechanics Research Communications, 1997, 24(4):447-455.
[3] 梁艳春,王政,杨晓伟,等.基于神经网络方法的包装件非线性恃性识别的研究［J］.力学学报, 1997, 29(4):497-500.
[4] Masri S F, Chassiakas A G, Caughey T K. Identification of nonlinear dynamic systems using neural networks［J］. ASME Journal of Applied Mechanics, 1993, 60(1):123-133
[5] Liang Yanchun, Zhou Chunguang, Wang Zaishen. Identification of restoring forces in non-Iinear vibration systems based on neural networks［J］. Journal of Sound and Vibration, 1997, 206(1):103-108.
[6] 杨建刚,戴德成.利用结构化神经网络识别振动系统非线性特性［J］振动工程学报, 1995, 8(1):62-66.
[7] Goldberg D E. Genetic algorithms in search, optimization and machine learning［M］, Addison-Wesley, Reading, MA, 1989.
[8] 梁化楼,戴贵亮.人工神经网络与遗传算法的结合:进展及展望［J］电子学报, 1995, 23(10),194-220.
[9] Yoon B, Holmes D J, Langholz G, et al. Efficient genetic algorithms for training layered feed-forward neural networks［J］. Information Sciences, 1994, 76(1):67-85.
[10] Kitano H. Empirical studies on the speed of convergence of neural network trainig using genetic algorithms［A］. Proceedings of AAAI90［C］, 1990:789-795.
[11] Pal S K, Bhandari. Selection of optimal set of weights in a layered network using genetic algorithms［J］. Information Sciences, 1994, 80(3-4):213-234.
[12] Fogel D B. An introduction to simulated evolutionary optimization［J］. IEEE Transactions on Neural Networks, 1994, 5(1):3-14.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed