新型无监督聚类算法监测与评估桥梁结构健康状况

A Novel Unsupervised Clustering Algorithm for Monitoring and Evaluating Bridge Structural Health

近年来,因车辆超载、结构设计缺陷、施工质量问题等原因导致的城市高架桥梁倒塌事故时有发生。因此,首次提出了一种可实时监测高架桥梁运行状况、通过改进K-means聚类算法实现对结构损伤实时检测、对数据驱动结构损伤进行检测的高效方法。该方法主要是对钢结构桥梁模型在结构完好状态下的振动数据进行采集,然后通过深度研究从这些数据中提取有效的结构损伤敏感特征值,最后利用改进的无监督聚类算法训练奇异值检测模型。试验结果表明,采用桥梁结构完好状况下的损伤敏感特征值作为训练数据,对数学模型加以训练后,可以有效检测并识别出桥梁结构在不同损伤状况下的测试结果。该新型检测方法可实现城市高架桥梁在长期运营阶段的结构健康实时监测。

In recent years,accidents of urban elevated bridge collapse have occurred frequently due to vehicle overloading,structural design defects,construction quality issues and other problems.Therefore,an efficient method has been proposed that can monitor the operation status of elevated bridges in real time,conduct real-time detection of structural damage by improved K-means clustering algorithm,and detect data-driven structural damage for the first time.This method mainly collects vibration data of steel structure bridge models under intact structural conditions.The effective structural damage sensitive feature values are extracted from these data by deep research.Finally,an improved unsupervised clustering algorithm is used to train the singular value detection model.The experimental results show that damage sensitive characteristic values under intact bridge structures serve as training data to train the mathematical model.It can effectively detect and identify the test results of bridge structures under various damage conditions.This new detection method can monitor real-time the structural health of urban elevated bridges during long-term operation.