基于Lamb导波深度学习的钢桥面板疲劳裂纹智能监测研究

Research on Intelligent Monitoring of Fatigue Cracks in Steel Bridge Decks Based on Deep Learning of Lamb Guided Waves

正交异性钢桥面板疲劳开裂问题突出,传统方法难以实现有效监测,可采用对钢结构裂纹高度敏感的Lamb导波信号进行裂纹监测。考虑Lamb导波在钢桥面板上的不同传播方式,建立钢桥面板有限元模型,并开展导波传播数值模拟;通过连续小波变换提取得到导波的主要特征,并运用深度学习技术挖掘导波特征中的疲劳裂纹信息,实现对钢桥面板疲劳裂纹的智能监测。结果表明:导波纵向和横向传播模式下的导波特征均能很好反映各类裂纹的影响,且2种传播模式之间可实现裂纹监测需求的有效互补;经过学习训练后的深度置信网络可实现对4mm及以上长度裂纹的高准度识别,对裂纹深度的测量误差也在1mm以内。研究成果为Lamb导波传感技术在钢桥面板疲劳裂纹监测中的应用提供了重要依据和方法参考。

Fatigue cracking in orthotropic steel bridge decks(OSDs)presents a severe problem;however,its effective identification using traditional methods is often difficult.Lamb guided waves are highly sensitive to cracks in steel structures and can be used in crack identification.In this study,finite element models of OSDs were established,and guided wave propagation simulations were conducted considering different propagation modes.A continuous wavelet transform was employed to extract the guided wave features.The fatigue-crack information was mined from these features using deep learning methods to achieve intelligent monitoring of fatigue cracks in OSDs.The results reveal that guided wave features can accurately reflect the effects of cracking under two propagation modes,whereby the requirements of crack monitoring can be effectively satisfied.The well-trained deep confidence network can accurately identify the cracks with length of 4mm or more,while maintaining the measurement error in the crack depth within 1mm.The findings provide a reference and methodological guidance for the application of Lamb guided waves in the fatigue-crack monitoring of OSDs.