基于完备集合经验模态分解的结构损伤识别

Damage identification of the structure based on complete ensemble empirical mode decomposition

针对建筑结构在损伤后结构受外界环境影响、结构损伤指标不敏感等问题,应用基于完备集合经验模态分解的结构损伤识别方法,提出归一化特征固有模态函数能量比作为损伤指标,解决了传统损伤指标不敏感的问题。先将原信号与一正一负成对的白噪声叠加组成目标信号,再将目标信号分解为有限个固有模态函数分量。对每个测点的目标信号分解的各固有模态函数分量进行了能量计算。选取能量最大的固有模态函数作为特征固有模态函数,并将其进行了希尔伯特变换,获得瞬时频率图。通过瞬时频率的变化来判断结构是否受损,并对各测点特征固有模态函数分量进行了归一化处理,计算出正常与损伤情况下各测点的特征固有模态函数分量的归一化能量比。以结构损伤前、后信号的完备集合经验模态分解的特征固有模态函数分量归一化能量比差值作为结构损伤敏感因子,以此来识别结构损伤。模拟5层剪切结构在刚度折损后损伤分析,并与经验模态分解方法进行了比较。研究结果表明:完备集合经验模态分解算法的效果比经验模态分解算法的好,它能够准确地识别结构损伤的位置。

Aiming at the problems of the structure which is affected by external environment and insensitive to structural damage index after damage, a method of structural damage identification based on complete ensemble empirical mode decomposition is applied and normalized, and characteristic intrinsic mode function energy ratio is proposed as the damage index, which can solve the problem of insensitivity of traditional damage index. Firstly, the original signal is superimposed with a pair of positive and negative white noise to form the target signal, and then the target signal is decomposed into a serious of intrinsic mode functions. The energy of each intrinsic mode function component decomposed from the target signal of each measurement point is calculated, and then the largest energy intrinsic mode function is selected as the characteristic intrinsic mode function. The application of Hilbert transform on intrinsic mode functions can obtain instantaneous frequency figure, through the change of instantaneous frequency to judge the damaged structure, and the energy of the characteristic intrinsic mode function of each measuring point is normalized to compare the energy ratio of the characteristic intrinsic mode function component under normal and damaged conditions. The energy ratio of the characteristic intrinsic mode function normalized by complete ensemble empirical mode decomposition is used as a structural damage index to identify the damage location. The damage analysis of the simulated five-layered shear structure after stiffness reduction is compared with the empirical mode decomposition method. The results show that complete ensemble empirical mode decomposition algorithm is more effective than EMD algorithm, and it can accurately identify the damage location of structures.