基于新奇检测技术的桥梁结构损伤预警方法  

Damage Alarming of Bridge Structures Based on Novelty Detection Technique

作　　者：孙宗光[1] 倪一清[2] 高赞明[2] 

机构地区：[1]大连海事大学路桥所,辽宁大连116026 [2]香港理工大学土木系

出　　处：《公路交通科技》2007年第10期72-75,共4页Journal of Highway and Transportation Research and Development

基　　金：辽宁省自然科学基金资助项目(20022004);香港特区RGC资助项目(PolyU5052/99E)

摘　　要：结构损伤预警是实现损伤检测策略的第一步,是进一步进行结构健康诊断的基础。大量的研究证明新奇检测技术可以较好地应用于结构的损伤预警。探讨了应用前馈BP网络(Feed-forward back-propagation network)实现新奇检测技术的方法。以结构的自振频率为作为网络的基本输入,对斜拉桥结构进行了损伤预警模拟研究。该方法的优点是不依赖于数值模型。对神经网络的训练仅需要健康结构的若干实测频率。模拟研究表明,该方法具有较高的实用价值。导致频率2.5%以上变化的损伤情况,均可给予预警。模拟了14种损伤情况,每种损伤考虑2种损伤程度。通过多指标策略,预警同时还在一定意义上指示损伤程度。The alarming of structural damage occurrence is the first stage of the successive structural damage identification strategy. It has been proven that the novelty detection technique can perform the damage alarming well. A novelty detection technique in terms of feed-forward back-propagation neural network is proposed for damage alarming. By taking the natural frequencies as input parameters, damage alarming for a cable-stayed bridge is simulated.The method is independent of structural medel.In training stage of the network, only a series of measured natural frequencies of the intact structure are used.The simulation shows that the method is practical and the damage cases which cause the frequency change of 2.5% can be alarmed soundly. 14 damage cases are simulated and 2 damage extents are considered for each damage case. By multi-index strategy, alarming also indicates the damage extent.

关 键 词：桥梁工程 损伤预警 神经网络 斜拉桥 新奇检测 

分 类 号：O346.5[理学—固体力学]