基于小波变换与神经网络的结构损伤检测

Structural damage detection based on wavelet transform and neural network

对BP网络和小波分析理论做了简要的概述,并给出了其应用于结构损伤检测的方法.将固有频率进行归一化处理,作为神经网络的输入参数进行结构损伤位置的检测,然后利用小波包技术对损伤结构的振动信号进行分解,求出各频带内的能量作为网络输入参数,进行损伤程度的评估.悬臂梁损伤诊断与实际损伤情况比较结果表明,该方法合理、有效,可用于实际结构的损伤检测.

A brief introduction is given to the theories of BP network and wavelet transform and methods of their application to detection of structural damage. The normalized natural frequency is taken as the input parameter of the neural network for detection of the location of the structural damage. Then, the wavelet packet technique is adopted to decompose the vibration signal of damage structures, and the energy derived for each frequency band is taken as the input parameter of the network for evaluation of the degree of structural damage. By calculation, the damage diagnosis for cantilever beams is made. The comparison of the calculated result with the reality of structural damage shows that the method is reasonable and effective, and can be applied to actual damage detection of structures.