基于时间序列分析与高阶统计矩的结构损伤检测

基于时间序列分析,提出一种有效检测结构非线性损伤的方法.在分析基于AR模型残差均方差指标的基础上,指出传统方法存在损伤信息泄漏的缺陷,特别是对于非线性损伤检测.为提高传统方法损伤检测结果的可靠性,提出采用残差的高阶统计矩———偏度和峰度作为传统指标的补充,提出了3个指标的算术平均和几何平均共6种综合指标,并应用模糊聚类分析实现结构损伤检测.利用考虑了环境因素影响的三层建筑结构模型的非线性损伤实验数据验证了提出的方法.研究表明,6种综合指标对非线性损伤检测的可靠性均高于传统方法,其中以均方差和峰度的几何平均指标检测效果最佳.

Damage detection based on time series analysis and higher statistical moments

Based on time series analysis,a new efficient nonlinear damage detection method is proposed in this paper. The traditional methods,in which the standard deviation of autoregressive(AR) model residual error is defined as damage sensitive index,may have the shortage of damage information loss,especially for the detection of the nonlinear damage.To improve the reliability of the traditional methods for the detection of the nonlinear damage,the higher statistical moments of residual error,such as skewness and kurtosis,are further defined as the complementary damage features to the standard deviation.Six comprehensive damage indexes are developed based on the arithmetic and geometric mean of the statistical moments,and are classified by using fuzzy clustering method to achieve damage detection.A series of the experimental data from a three-story building structure considering the environmental variety are analyzed to validate the viewpoints in this paper.The results from all six new damage indexes show higher reliability than those from traditional methods,and the index obtained from geometric mean of standard deviation and kurtosis has the best performance for all the six damage indexes.