铁道技术标准(中英文)

为实现桥梁结构精准损伤识别，本文基于三维车-路-连续梁桥耦合模型模拟损伤，提出以挠度差值最大值为识别指标。选取测点计算不同损伤工况下的挠度响应，经加噪、归一化处理构建数据集，采用支持向量机和BP神经网络建立识别模型，探究BP神经网络参数对识别效果的影响。结果表明，所提指标敏感性高；BP神经网络模型识别正确率与抗噪能力优于支持向量机，隐含层神经元数为6、传递函数与训练函数选用tansig和trainrp时效果最优。将方法推广至公铁两用斜拉桥，验证其良好通用性。

In order to realize the precise identification of bridge structure damage, this paper simulates the damage based on the 3-D vehicle-road-continuous beam bridge coupling model, and proposes the maximum deflection difference as the identification indicator. The measuring points are selected to calculate the deflection response under different damage conditions. The data set is constructed by adding noise and normalization. The support vector machine and BP neural network are used to establish the recognition model, and the influence of BP neural network parameters on the recognition effect is explored. The results show that the proposed indicator is highly sensitive; the recognition accuracy and anti-noise ability of BP neural network model are better than those of support vector machine. When the number of hidden layer neurons is 6, tansig and trainrp are selected as transfer function and training function, the effect is the best. The method is to be extended to the highway-railway cable-stayed bridge to verify its good versatility.