用改进的BP神经网络评判管道的腐蚀类型
Judgment of Pipeline Corrosion Patterns with Improved BP Neural Network
浏览(2139) 下载(1)
- DOI:
- 作者:
- 陈明1, 褚家荣2, 蒲家宁1, 黄开阳1
CHEN Ming1, CHU Jia-rong2, PU Jia-ning1, HUANG Kai-yang1
- 作者单位:
- 1. 中国人民解放军后勤工程学院, 重庆 400016;
2. 73808部队, 浙江宁波 315803
Logistics Engineering Institute of CPLA, Chongqing 400016, China
- 关键词:
- BP神经网络;共轭梯度优化算法;管道;腐蚀类型;评判
BP neural network;conjugated gradient optimum algorithm;pipeline;corrosion pattern;judgment
- 摘要:
- 用BP神经网络分析评判管道的腐蚀类型,可以避开寻找各种因素对腐蚀类型影响规律的难题,方便准确地分析评判出管道的腐蚀类型,但是传统的BP神经网络存在收敛速度较慢和容易陷入局部极小点两个问题,为此文章提出了将传统的BP神经网络与共轭梯度优化算法相结合,以优化网络权值和阈值的计算,同时确定了相应的计算方法。将改进后的BP神经网络应用到管道腐蚀类型的评判中,取得了良好的效果。计算结果表明,改进后的BP神经网络具有更好的学习能力,可以在更少的迭代次数和时间内,得到高精度的输出结果。
The analysis and judgment of pipeline corrosion patterns with improve d BP neural network can overcome difficulties of searching for various factors a ffecting corrosion patterns. The pipeline corrosion patterns can be analyzed and judged conveniently and accurately. But conventional BP neural network has two problems, i.e. lower convergent speed and partial minimum point. This paper com bines conventional BP neural network with conjugated gradient optimum algorithm to optimize the calculations of network weighting value and threshold value,and furthermore,it offers the relevant calculation method. The application of improv ed BP neural network to pipeline corrosion pattern judgment achieves good result s. The network possesses better study ability and gives more accurate output wit h less iterative frequency and less time.2005,3(3): 4-7 收稿日期：2004-2-23分类号：TE980.4作者简介：陈明(1982-),男,重庆人,2002年毕业于中国人民解放军后勤工程学院油气储运工程专业,现在中国人民解放军后勤工程学院攻读油气储运工程硕士学位.参考文献：
[1]侯祥林,陈长征,虞和济,等.神经网络权值和阈值的优化方法[J].东北大学学报,1999,20(4):447-450.
[2]李永强,刘杰,侯祥林,等.人工神经网络的混合算法及其工程应用[J].机械工程学报,2004,40(1):127-130.
[3]傅玉华,周汉平,周吐清,等.石油化工设备腐蚀与防治[M].北京:机械工业出版社,1997.
[4]邓先礼.最优化技术[M].重庆:重庆大学出版社,1998.
[5]周东华,叶银忠.现代故障诊断与容错控制[M].北京:清华大学出版社,2000.
[6]闻新,周露,李翔,等.Matlab神经网络仿真与应用[M].北京:科学出版社,2003.