基于提升小波包和神经网络的结构损伤检测

提出一种提升小波包分解、多传感器特征融合和神经网络模式分类相结合的结构损伤诊断方法.首先,对多个传感器采集的振动响应信号进行提升格式小波包分解,定义标准化相对能量并计算每个频带上的相对能量；然后,把这些传感器信号的小波包相对能量融合作为神经网络分类器的输入特征向量,实现损伤的诊断和评价.数值仿真结果表明,提升小波包分解的频带能量分布能够较好地反映结构的损伤特征；特征融合能够使不同传感器的信息相互补充,减小了损伤检测信息的不确定性,使诊断信息具有较高的精度和可靠性.

Damage Detection of Wing-Box Based on Lifting Wavelet Packet Transform and Neural Network

In order to diagnose the position and degree of the damage of complex wingbox structures accurately, a method is presented by means of lifting wavelet packet decomposition, multi-sensor feature fusion theory and neural network pattern classification. Firstly, vibration signals gathered from sensors are decomposed by lifting wavelet packet transform. Secondly, the relative energy of decomposed frequency band is calculated. Thirdly, the input feature vectors of neural network classifier are built by fusing wavelet packet relative energy distributi on of these sensors. Finally, with the trained classifier, damage diagnosis and assessment are realized. The results indicate that: the frequency band energy decomposited by lifting wavelet packet transform could perfectly reflect the damage condition; the fused feature can make different information complementary, and reduce the uncertainty of damage detection information, so the damage information has much more precision and reliability, and the diagnosis accuracy can be improved.