论文部分内容阅读
为了研究一种基于神经网络与奇异分析技术的结构损伤检测方法来剔除不利温变影响,以1座能代表中小跨桥梁性能的基准结构的有限元模型为例,分析温变和多级损伤对结构频率的影响,研究该检测方法的有效性和可靠性。以温变条件下无损结构的温度数据和前10阶竖向模态频率训练BP神经网络(BPNN),来构建温度和频率间的量化模型;然后用该网络来预测不同温度条件下结构的频率,计算频率预测误差来消除温变影响;以该预测误差的欧式范数为奇异指标,用结构待检状态奇异指标序列均值与健康状态的相对变化率来指示损伤的存在。研究结果表明:该方法不仅能可靠地检测温变条件下结构的损伤,而且能定性地区分结构整体损伤程度的大小,具有很强的噪声鲁棒性。
In order to study a structural damage detection method based on neural network and singularity analysis technology to eliminate adverse temperature change effects, taking a finite element model of a benchmark structure capable of representing the performance of medium and small span bridges as an example, Structure frequency of the test method to study the effectiveness and reliability. The BP neural network (BPNN) is trained by the temperature data of the nondestructive structure under the condition of temperature change and the BP neural network (BPNN) of the first 10 vertical modes to construct a quantitative model of temperature and frequency. Then the network is used to predict the structure frequency under different temperature conditions , The frequency prediction error is calculated to eliminate the influence of temperature change. The European norm of the prediction error is regarded as a singular index, and the relative change rate of the mean of the singular index sequence and the state of health is used to indicate the existence of the damage. The results show that this method can not only detect the structure damage under the condition of temperature change reliably, but also can distinguish the damage degree of the whole structure qualitatively and have strong noise robustness.