Intelligent Structural Health Monitoring with Ultrasonic Lamb Waves

Abstract

Structural health monitoring is a branch of machine learning where we automatically interpret the output of in situ sensors to assess the structural integrity and remaining useful lifetime of engineered systems. Sensors can often be permanently placed in locations that are inaccessible or dangerous, and thus not appropriate for traditional nondestructive evaluation techniques where a technician both performs the inspection and interprets the output of the measurement. Ultrasonic Lamb waves are attractive because they can interrogate large areas of structures with a relatively small number of sensors, but the resulting waveforms are challenging to interpret even though these guided waves have the property that their propagation velocity depends on remaining wall thickness. Wavelet fingerprints provide a method to interpret these complex, multi-mode signals and track changes in arrival time that correspond to thickness loss due to inevitable corrosion, erosion, etc. Guided waves follow any curvature of plates and shells, and will interact with defects and structural features on both surfaces. We show results on samples from aircraft and naval structures.