Research on Structural Damage Identification Method Based on Vibration Response and SVM

针对工程实际中一些梁结构需要进行长期结构健康监测的需求，本文提出了一种基于应变模态柔度及其衍生指标以及支持向量机的在线损伤识别方法。通过损伤试验，选取等截面梁为研究对象，使用协方差驱动的随机子空间法识别其应变模态参数，构建损伤指标。最后，运用机器学习中的支持向量机分类算法实现损伤识别，并对四种基于应变模态柔度及其衍生指标的方法进行比较和评估，验证了该方法的可行性。结果表明，采用应变模态柔度及其衍生指标与支持向量机的识别方法能够用较少的试验数据样本实现梁结构的损伤定位，其中以应变模态柔度曲率差作为损伤指标的效果最佳。In response to the demand for long-term structural health monitoring of certain beam structures in practical engineering applications, this paper proposes an online damage identification method that relies on strain modal flexibility, its derived indicators, and support vector machine (SVM) techniques. Experimental damage tests were conducted on beams with equal cross-sections as the research subjects. The covariance-driven random subspace method was employed to identify the strain modal parameters and construct damage indicators. Finally, the SVM classification algorithm from machine learning was utilized for damage identification. A comparative evaluation of four methods based on strain modal flexibility and its derived indicators was conducted to validate the feasibility of the proposed approach. The results demonstrate that the damage identification method utilizing strain modal flexibility and its derived indicators in conjunction with SVM can achieve effective damage localization in beam structures with a reduced number of experimental data samples. Among the investigated indicators, the strain modal flexibility curvature difference emerges as the most optimal damage indicator.