基于支持向量机的梁桥多位置损伤识别研究

Multi position damage identification of beam-bridges based on support vector machine

为实现更准确的梁桥多位置损伤识别,从理论角度分析得出某位置的损伤会对周边位置的位移和曲率产生一定影响。曲率模态差的突变是损伤的典型特征,以曲率模态差变化率作为量化突变的参数,并将其归一化处理后输入支持向量机中进行桥梁多位置损伤识别。以某多跨连续刚构桥为例,将多处典型截面发生10%,30%和50%刚度折减情况下的前2阶竖向振动的曲率模态差变化率以及各位置的损伤状态作为特征向量去训练支持向量机。之后将预测集中包含20%和40%刚度折减的曲率模态差变化率输入训练好的支持向量机中去识别各位置的损伤情况,其识别准确率达到99.68%。

In order to identify multi position damage of beam bridge more accurately,through the theoretical analyzing,it is concluded that the damage of a certain position will have a certain effect on the displacement and curvature of the surrounding position.However,the sudden change of curvature modal difference is the typical feature of damage,and the change rate of curvature modal difference is used to quantize the sudden change.Then normalize and input them into the support vector machine to identify multi position damage.Taking a multi-span continuous rigid frame bridge as an example,typical cross section appears 10%,30%and 50%stiffness reduction.The change rate of curvature modal difference of the 1st and 2nd ordered vertical vibration and the damage state of each position are used as eigenvectors to train the support vector machine.After that,the change rate of curvature modal difference with 20%and 40%stiffness reduction in the prediction set is input into the trained support vector machine to identify the damage of each position,and the recognition accuracy is 99.68%.It is proved that this method has good accuracy and generalization in damage identification.