An adhesive bond state classification method for a composite skin-to-spar joint using chaotic insonification

The combination of chaotically amplitude-modulated ultrasonic waves and time series prediction algorithms has shown the ability to locate and classify various bond state damage conditions of a composite bonded joint. This study examines the ability of a new two-part supervised learning classification scheme not only to classify disbond size but also to classify whether a bond for which there is no baseline data is undamaged or has some form of disbond. This classification is performed using data from a similarly configured composite bond for which baseline data are available. The test structures are analogous to a wing skin-to-spar bonded joint. An active excitation signal is imparted to the structure through a macro fiber composite (MFC) patch on one side of the bonded joint and sensed using an equivalent MFC patch on the opposite side of the joint. There is an MFC actuator/sensor pair for each bond condition to be identified. The classification approach compares features derived from an autoregressive (AR) model coefficient vector cross-assurance criterion.