Nonlinear ultrasonics for health monitoring of aerospace structures using active sparse sensor networks

In order to maintain the integrity of aerospace structures to their best extent and avoid structural failure, it is highly desirable to identify damage in them as accurately and as early as possible. Traditional nondestructive inspection (NDI) and structural health-monitoring (SHM) techniques based on guided waves have been well developed over the years, and are now on the verge of maturity for real-world applications. Nonetheless, the effectiveness and accuracy of the majority of available techniques, which use linear signal features such as the time of flight (TOF) of a particular waveform, have been demonstrated effectively for characterization of gross damage only. With this motivation, nonlinear attributes of guided elastic waves, as typified by second harmonics, have been increasingly studied and employed in NDI and SHM for aerospace structures because of their higher sensitivity to small-scale damage (eg, fatigue cracks) compared with linear counterparts. In this chapter, a quantitative damage detection technique using nonlinear Lamb wave features and active sensor networks is proposed, toward implementation of SHM for plate-like engineering structures. Advanced signal processing techniques are applied in the time, spectral, and time-frequency domains, respectively, to understand damage-related signal features such as the relative acoustic nonlinearity parameter (RANP). Damage indices based on these features are then developed for diagnostic imaging through a probabilistic data fusion scheme. To apply the proposed methodology a self-contained, integrated SHM system is developed, along with the conception of "decentralized standard sensing" for configuring active sensor networks. Finally, two case studies are performed. In the first one, fatigue damage in aluminum structures is characterized, demonstrating the accuracy and efficiency of the approach for small-scale damage localization in a quantitative manner, in comparison to the results obtained from traditional, linear techniques. In the second case, the practical concern of uncertainties involved in RANP is addressed, by developing a probabilistic model and using it in repetitive experiments for detection of barely visible impact damage in carbon fiber laminates.