【摘要】 随着大跨度桥梁数量及重要性的增加,桥梁损伤识别成为学界的研究热点。文中应用小波分析提取桥面的一阶振型,推导了梁式结构由不等间距位移模态向曲率模态转化的乘子矩阵DTOC,由损伤前后曲率模态的变异特征构造了梁式结构单损伤情况的普适概率神经网络,使之可以识别无训练样本时的损伤位置。通过数值仿真计算了某斜拉桥有限元模型桥面单元刚度折减前后在模拟脉动风荷载下的桥面板关键点的振动时程,讨论了小波参数识别技术和该普适概率神经网络的联合应用在不同级别的测量噪声影响下的斜拉桥桥面板损伤识别效果,说明在振动测量信号的信噪比达到一定要求时该方法具有实际效果。更多还原

【Abstract】 With the increase of the quantity and significance of the long-span bridges,the damage identification for bridges becomes the hot point in scientific field.In this paper,wavelet analysis technology has been used to obtain the first mode shape,and the convert matrix DTOC which transfers unequally spaced displacement mode shapes to curvature mode shapes has been deduced.Using the character of the difference between the curvatures mode shapes before and after damage occurs as damage index,a widely used probability neural network(PNN) for beam-like structures has been constructed to identify the damage mode lack of training samples.and the efficient of the networks are checked by the damage index extracted from the vibration time history with noises.At the last,in numerical simulation the simulated fluctuating wind induced vibration of the key points of bridge are extracted from a finite element model of a cable stayed bridge,and it is discussed that the effect of the collaboration of the wavelet mode parameter identification technology and PNN for detecting the damage of deck for cable stayed bridges under differenct signal-to-noise ratios.The discussion shows that the method may be used as the signal-to-noise ratio is high enough.更多还原