基于自回归支持向量机的结构损伤诊断

Structural damage detection with autoregressive support vector machines

将自回归时间序列(AR)模型和支持向量机方法结合应用于结构的损伤诊断,以一个3层框架结构为分析对象,模拟两种损伤模式:初始线性结构发生质量变化和初始非线性结构发生质量变化.首先对实验中采集到的加速度信号建立AR模型,从而提取模型参数作为损伤特征,再利用支持向量机进行损伤诊断.结果表明,在小样本情况下基于自回归支持向量机进行结构非线性损伤诊断,能够得到很好的结果.

The autoregressive（AR）time series model and support vector machine（SVM） are applied to detect the damage pattern.With a three-story frame structure as an analysis object,two damage patterns i.e.initial linear structure with mass varying and initial nonlinear structure with mass varying are simulated.According to the need of damage detection,the AR model of acceleration signals collected from the experiments is established.Then the AR coefficients are extracted as damage features and SVM is used to detect damage.The AR-SVM method for the insufficient training samples is proved to be practical and efficient on structure nonlinear damage detection.