为适应在线结构健康监测的要求,基于统计模式识别技术提出了一种新的结构损伤诊断方法.采集结构在健康和损伤两类工况下的动力响应数据,并对数据样本分段建立ARMA模型,对模型中的AR参数进行特征提取,获得主成分矩阵.计算健康、损伤状态主成分矩阵间的Mahalanobis距离.分析结果发现,损伤前后两状态Mahalanobis距离存在差异,因而提出以该距离的平方值作为损伤敏感特征DSPR,并建立基于假设检验t检验的方法辨识结构的状态.以环境激励下IASC-ASCE Benchmark结构的损伤试验为例,运用该方法进行了损伤诊断研究.试验表明,损伤敏感指标DSPR可有效辨识结构的健康与损伤状态,具备在线实时损伤诊断的应用价值.
A new method for structural damage diagnosis based on the technique of statistical pattern recognition is presented for the on-line structural health monitoring(SHM) system.First,the dynamic response data from pre-and post-damaged structure are obtained and modeled as auto-regressive moving-average(ARMA) models,while a principal-component matrix is derived from the AR parameters by the process of feature extraction.Then,the distance of Mahalanobis between the principal-component matrix of pre-and post-damaged is calculated.It was observed that the Mahalanobis distance was difference,thus a new damage sensitive index DSPR was established using the square value of Mahalanobis distance.At last,a hypothesis t-test was applied to obtain a decision of damage when the DSPR changed significantly.The ambient vibration test of the IASC-ASCE Benchmark structure was taken as an experimental study in this paper.Result shows that,the damage sensitive index DSPR is able to identify damage of structure and the proposed method can be applied to the on-line damage diagnosis in SHM.