论文部分内容阅读
为准确进行结构损伤预警,提出基于前馈BP网络实现新奇检测技术的斜拉桥损伤预警方法.以斜拉索局部振动模态的基频作为网络的基本输入,定义新奇指标为网络输出,根据训练阶段和检测阶段的新奇指标间的比较指示结构的健康状态.相比一般的损伤检测,其不依赖于数值模型,避免了对数值模型高精度的要求,提高了实用价值.对汲水门大桥14种损伤的模拟结果表明,在1%的噪声水平下,该方法可达到较好的预警率.
In order to accurately predict structural damage, a pre-warning method based on feedforward BP neural network to detect novelty of cable-stayed bridge is put forward. The basic frequency of local vibration mode of cable is taken as the basic input of the network and the new index is defined as the network output. The comparison between the novelty indexes in the training phase and the testing phase indicates the health status of the structure.Compared with the common damage detection, it does not depend on the numerical model, avoids the high precision of the numerical model and enhances the practical value.Based on the analysis of the Kapok waterfront bridge The simulation results of 14 kinds of damage show that the method can achieve a better warning rate under the noise level of 1%.