【摘要】 环境激励下的桥梁结构响应易受噪声干扰，导致信号中的各特征分量无法有效辨别。针对现有降噪方法的局限性，将噪声辅助分析理论引入经验小波变换(EWT)进行改进。首先，基于能量准则与残量频域概率密度曲线特征对改进EWT的关键参数进行自适应确定；其次，根据能量密度与平均周期乘积、JS(Jensen-Shannon)散度构造筛分系数，实现对信号分量的特征筛选；最后，设定频域误差指标进行信号分量集成，对迭代所得信号分量重构进而实现降噪效果。为验证改进EWT的降噪能力，先后以仿真信号和某斜拉桥监测信号为研究对象，通过各降噪指标及时域、频域、时频域图形对各方法的降噪性能进行对比，结果表明，改进EWT的降噪能力更强，在抑制噪声的同时可有效提取特征分量，避免了过度分割及整体集成造成的数据冗余，能够用于实际桥梁动力响应分析。更多还原

【Abstract】 Response of a bridge structure under environmental excitation is easily disturbed by noise to cause inability to effectively distinguish feature components in its vibrational signals. Here, aiming at limitations of existing denoising methods, the noise aided analysis theory was introduced into empirical wavelet transform(EWT) for improvement. Firstly, key parameters of the improved EWT were adaptively determined based on energy criterion and characteristics of residual frequency domain probability density curve. Secondly, screening coefficient was constructed according to product of energy density and average period as well as Jensen-Shannon(JS) divergence to realize feature screening of signal components. Finally, the frequency domain error index was set to integrate signal components, and signal components obtained with iteration were reconstructed to acquire the denoising effect. In order to verify the denoising ability of the improved EWT, simulation signals and a certain cable-stayed bridge’s monitoring signals were taken as study objects, denoising performances of various methods were compared using various denoising indexes, time domain graphs, frequency domain graphs and time-frequency domain graphs. The results showed that the improved EWT can have stronger denoising ability, effectively extract feature components while suppressing noise, and avoid data redundancy caused by excessive segmentation and overall integration; it can be applied in dynamic response analysis of actual bridges.更多还原