Ambient noise monitoring of nonlinear defects on elastic plates

International audienceSolid structures used in industry damage with time due to their usage in sometimes extreme stressing conditions. Rapidly detecting the apparition of defects such as cracks, delamination or debonding and localizing them is crucial for user safety and machine performance. In this study, we investigate contact acoustic nonlinearities, or non-linear defects. When excited by low-frequency (LF) vibrations, non-linear defects generate harmonics or, more generally, signals of higher frequencies than the excitation, due for example to clapping or friction between two surfaces in contact. A periodic measurement of these high- frequency signals on structures could allow an efficient monitoring of a nonlinear damage apparition and evolution. Recent work has shown that passive methods (e.g. noise correlation) taking advantage of the vibration of the structures induced by the ambient noise can be used to localize linear defects (Chehami et al., 2015). The main advantage of these methods is that they are less power consuming than active methods. We present a signal processing technique applied on the high-frequency noise generated by nonlinear defects excited by ambient noise on plate-like structures. To test the method, we develop an experimental setup composed of a thin aluminium plate on which different types of non-linear defects are created (e.g., unbolted screw, Hertzian or frictional contact). The thin plate is excited by a shaker vibrating at low frequencies (< 500 Hz). The high-frequency acoustic signals generated by the defects are recorded by a network of piezoelectric sensors glued on the plate. The influence of the sensor network geometry, of LF excitation amplitude and of the type of nonlinear defect on the detection are discussed. It is shown that defect detection is possible even when the LF excitation is a realistic jet engine noise