Electromagnetic ultrasonic signal processing and imaging for debonding detection of bonded structures

This paper investigates an electromagnetic ultrasonic method for debonding detection of adhesive structures. Considering the non-stationary and nonlinear characteristics of the electromagnetic ultrasonic reflection signals, a joint adaptive noise reduction and reconstruction (JANRR) method, namely, complete ensemble empirical mode decomposition(CEEMDAN) and singular spectrum analysis(SSA) based on continuous mean square Error(CMSE) are proposed, in which, mirror extension is used to suppress the endpoint effect of decomposition, and the order of SSA noise reduction is determined based on singular entropy theory. For two side debonding detection of the adhesive layer, a back propagation (BP) neural network classification model is developed, and five characteristics such as signal skewness factor are taken as the network inputs based on Chi square test. Experiments were carried out for validation, which show that the proposed method can distinguish both signals of the upper and lower interfaces of the adhesive layer under different bonding states.