基于数据融合的两阶段损伤识别方法

A two-stage damage recognition method based on data fusion

因无需测量环境激励、对结构损伤敏感及较强噪声鲁棒性等优点,基于互相关函数的结构损伤识别方法受到了广泛关注。当应用于大型复杂结构时,结构单元众多且测量数据有限,导致识别结果精度和计算效率严重影响。针对这一问题,提出了基于数据融合的两阶段损伤识别方法。首先利用多类测量数据融合进行响应重构,基于结构损伤前后多类数据间的互相关函数变化量构建了损伤指标,结合D-S证据理论定位结构损伤;继而基于多类数据间的互相关函数构建目标函数,采用粒子群-梯度算法量化结构损伤。数值模拟与试验验证结果表明,在环境激励未知的情况下,本文所提出的基于数据融合的两阶段损伤识别方法可有效提高损伤识别结果精度和计算效率,且具有很好的噪声鲁棒性。

Due to advantages of not requiring ambient excitation measurements,having high sensitivity to structural damage and stronger noise robustness,the structural damage recognition method based on cross-correlation function receives extensive attention.However,when it is applied in large and complex structures,many structural elements and limited measurement data seriously affect the accuracy of recognition results and computation efficiency.Here,aiming at this problem,a two-stage damage recognition method based on data fusion was proposed.Firstly,multi-type measured data fusion was used to perform response reconstruction,and damage indexes were constructed based on changes in cross-correlation functions among multi-type data before and after structural damage.Combined with D-S evidence theory,structural damages were positioned.Then,based on cross-correlation functions among multi-type data,an objective function was constructed,and the particle swarm-gradient algorithm was used to quantify structural damages.Numerical simulation and experimental verification results showed that under the condition of unknown environment excitation,the proposed two-stage damage recognition method based on data fusion can effectively improve the accuracy of damage recognition results and computation efficiency,and have good noise robustness.