基于声发射信号改进经验小波分析的钢桥面板疲劳裂纹定量监测方法研究

The orthotropic steel bridge decks of existing medium- and long-span modern bridges all over the world are particularly vulnerable to fatigue. However, there is no satisfactory solution for the in-service diagnosis and monitoring of fatigue cracks, especially the crack size quantification, in orthotropic bridge decks. Taking advantages of the acoustic emission (AE) technique in crack detection, this project investigates the fatigue crack monitoring in orthotropic steel bridge decks based on improved empirical wavelet transform (EWT), according to their loading features, cracking modes, geometrical properties and operational noise. In order to extract the time-frequency features of AE signals, EWT will be improved by developing an adaptive wavelet spectrum segmentation method. Then, an anti-noise robust classification index based on the energy distribution in multiple time-frequency scales will be established to distinguish between the crack propagation-induced AE signals and the crack closure-induced AE signals with a high accuracy. Following the AE wave classification, a novel method to estimate the fatigue crack size is proposed by making use of crack closure-induced AE signals and applied to the crack monitoring of orthotropic steel bridge decks. Aiming at fatigue crack size estimation of orthotropic steel bridge decks, this project develops advanced AE signal processing algorithms based on theoretical analysis and experimental study. The outcomes of this project are of great significance for safety and usability performance evaluation.

国内外在役大中型现代桥梁的正交异性钢桥面板面临着严重的疲劳问题，然而针对其运营状态下疲劳裂纹的诊断和监测，特别是裂纹尺寸估计，还没有切实有效的方法。本项目利用声发射技术在裂纹监测中的优势，充分考虑钢桥面板的受力特点、开裂特征、几何特征和运营噪声，研究基于声发射信号改进经验小波分析的钢桥面板疲劳裂纹定量监测方法。以裂纹扩展和闭合两类声发射信号的时频特征提取为出发点，提出基于自适应频带划分技术的改进经验小波变换，进而根据信号的多尺度时频能量分布特征建立具有抗噪性和鲁棒性的分类指标，提高两类声发射信号分离的准确度。在此基础上，提出一种基于裂纹闭合声发射信号的疲劳裂纹尺寸估计新方法，用于正交异性钢桥面板的疲劳裂纹定量监测。本项目采用理论与试验相结合的研究方案，利用先进的声发射信号处理方法解决正交异性钢桥面板疲劳裂纹监测中裂纹尺寸估计的难题，研究成果对在役桥梁结构的安全和使用性能评价具有重要意义。

正交异性钢桥面板在国内外大中型桥梁上被广泛应用，不仅作为主梁的一部分参与结构整体受力，并且直接承受车轮荷载的反复作用，存在着严重的疲劳损伤风险，其运营状态下的疲劳裂纹诊断和监测问题亟待解决。本项目充分考虑正交异性钢桥面板的受力特点、开裂模式、几何特征和运营噪声，研究基于声发射信号改进经验小波分析的钢桥面板疲劳裂纹定量监测方法。首先，针对疲劳裂纹声发射信号的时频特征，提出基于自适应频带划分的改进经验小波变换方法。然后，为降低声发射信号在正交异性钢桥面板中传播的频散、反射、衍射和噪声影响，提出基于时频分析和长短时记忆神经网络的裂纹定位方法。其次，根据不同损伤机理声发射信号的频率成份差异，建立基于多尺度时频能量分布的信号分类指标，提出基于时频分析和深度卷积神经网络的声发射信号分类方法。最后，分别利用不同类型的声发射信号，提出基于裂纹闭合声发射信号的疲劳裂纹尺寸估计新方法，改进传统基于裂纹扩展声发射信号的疲劳裂纹尺寸估计方法，并结合两种方法提升裂纹定量监测的准确性。本项目采用理论推导、数值模拟、实验室和现场实验相结合的研究方案，解决正交异性钢桥面板疲劳裂纹识别、定位和尺寸估计的难题，研究成果对在役桥梁的安全和使用性能评价具有重要意义。