基于小波神经网络的框架结构损伤识别

以含损伤的框架结构为研究对象,对损伤位置和损伤程度进行识别。运用有限元分析原理,采用Lanczos法得到框架结构的转角模态,对其转角模态进行连续小波变换可以得到结构的小波系数,再由小波系数模极大值确定损伤的位置。以损伤后结构的固有频率作为神经网络输入参数构造神经网络,从而实现对框架结构损伤程度的识别。通过对一平面框架结构的损伤识别计算分析,验证了方法的有效性。

Damage identification of frame structure by wavelet and neural network analysis

The frame structure with damages is taken as the research obj ect in order to have a clear identification of the damage location and damage degree.Based on the principle of FEA and the Lanczos method,the modal parameters of frame structure can be calculated. And with a continuous wavelet transform of the modal parameters according to the rotation modal wavelet analysis theory,the wavelet coefficient can be got.The damaged construction can be located by the maximum modulus of wavelet coefficients.The neural network can be formed with neural network input parameter as its natural frequency of the structure.So the identification and positioning of structural damage can be made.The method is verified use-ful by computational analysis of the frame structure damage.