基于转角模态小波神经网络的结构损伤识别方法  

The method of structural damage identification by the wavelet-neural network analysis of rotation mode

作　　者：管德清[1] 汤博文[1] 荣政[1] 

机构地区：[1]长沙理工大学土木与建筑学院,湖南长沙410004

出　　处：《交通科学与工程》2014年第1期44-48,共5页Journal of Transport Science and Engineering

基　　金：国家自然科学基金资助项目(51378079)

摘　　要：以小波分析为基础,结合神经网络技术,研究了结构的损伤识别问题,建立了一种基于转角模态小波神经网络的结构损伤识别方法。采用有限元理论,运用Lanczos法,分析结构的损伤,得到了结构的转角模态参数。然后,对其模态参数进行连续小波变换,得出了小波系数图,由小波系数模极大值可判断结构损伤的位置。利用BP神经网络模拟小波系数模极大值与损伤程度之间的非线性关系,从而由网络的输出结果可识别结构的损伤程度。通过对一简支梁的损伤识别计算分析,验证了该方法的有效性。该方法可供结构损伤诊断的工程应用参考。Based on the wavelet analysis, combined with neural network technology to solve the structure damage identification problem, a rotation modal based wavelet neural network method is set up for structural damage identification.Based on the finite element theory,using Lanczos method, the damage of the structure is analyzed,and the structural parameters of the rotation mode are obtained with continuous wavelet transform of model parameters and then the wavelet coefficients are obtained, the location of structure＇s damage is identified by the maximum of wavelet coefficients.Using BP neural network, the nonlinear relationship between the maximum of wavelet coefficients and the degree of damage is simuloated,and by the network output, the damage degree of the structure can be identified.Through the calculation and analysis of a simply supported beam＇s damage identification,the effective mess of the method is verified. This method can be used as a reference in the application of structural damage diagnosis.

关 键 词：小波变换 BP神经网络 转角模态 结构损伤诊断 

分 类 号：TU317.5[建筑科学—结构工程] TN911.6[电子电信—通信与信息系统]