摘要
本文对一个四层钢框架连接损伤进行数值模拟,比较了两种改进算法的测试结果,即阈值和权值同步学习BP算法(改进算法2)以及L-M算法(改进算法1)。从而证明了改进算法2更精确、更优越。
In this paper,through numerical veritification on connection damage of a four-story steel frame,the test results of two kinds of improved BP algorithms,which are BP algorithm with weight and threshold synchro-learning ability(shorted for improved BP algorithm 1) and Levenberg-Marquardt algorithm(shorted for improved BP algorithm 2) are compared.As a result,it is proved that the improved BP algorithm 2 is more accurate and more superior.
引文
[1]李学良,霍达,腾海文.基于权值、阈值同步学习BP算法的结构损伤检测[J].世界地震工程,2005,21(2):52~56
[2]朱宏平,千力.利用振动模态测量值和神经网络方法的结伤识别研究[J].计算力学学报,2003,22(2):193~196
[3]Cawley P,Adams R D.The location of detects in structuresfrom measurements of natural frequencies[J].Journal ofStrain Analysis,1979,14:49~57
[4]Lam H F,Ko J M,Wong C W.Localization of damagedstructural connections based on experimental modal andsensitivity analysis[J].Journal of sound and Vibration,1998,210:91~115
[5]沈东强,徐礼华,龚建伍.基于BP网络的结构损伤识别[J].武汉理工大学学报,2004,26(9):48~50
[6]杜永峰,陈文元.小波分析与神经网络在结构损伤检测中的应用[J].兰州理工大学学报,2005,31(5):121~124
[7]飞思达科技产品研发中心.MATLAB6.5辅助神经网络分析与设计[M].北京:中国电子工业出版社