摘要
为了准确地提取结构损伤异常信息,消除小波奇异值分解时存在需要特定的小波基和分解层数以及经验模态分解(EMD)方法存在诸如虚假模态混叠等问题,提出一种基于改进的总体平均经验模态分解(EEMD)与快速独立分量分析(Fast ICA)相结合的提取结构损伤特征并进行识别与定位的新方法。首先,通过EEMD对结构动力响应信号进行预处理并用Fast ICA提取出包含损伤信息的特征分量对结构响应异常进行识别和初步定位;然后,计算归一化的源分布向量(NSDV)的最大值,并根据该最大值精确定位结构损伤。最后,通过框架数值算例和试验进行了所提方法的验证,结果表明该算法能够较好地进行结构损伤异常的识别与定位。
It is known that the wavelet decomposition requires specific wavelet basis functions and decomposition layers. Meanwhile, there exist some problems in the empirical mode decomposition （EMD） , such as, false modes. To avoid the disadvantages above, a method of structural damage detection and locating based on the ensemble empirical mode decomposition （EEMD） and the fast independent componentan analysis （FastICA） extract the structural damage novelty information. At first, the measured structural dynamic with EEMD, and then FastlCA algorithm was used to extract the to detect the structural response anomalies and preliminarily normalized source distribution vector （NSDV） was computed to numerical example and test were conducted, the results showed and locate them. was presented to accurately responses were preprocessed feature components involving the damage information so as locate damage. After ward, the maximum value of the accurately locate the structural damage. Finally, a frame that the proposed method can successfully detect damages
出处
《振动与冲击》
EI
CSCD
北大核心
2016年第1期203-209,共7页
Journal of Vibration and Shock
基金
国家自然科学基金(51278127
50878057)
国家十二五科技支撑计划(2012BAJ14B05)
