Investigation of Signal Separation for Ultrasonic Nondestructive Evaluation

Ultrasonic array-based techniques are widely used in nondestructive evaluation (NDE). One critical aspect of these techniques is the duration of the data acquisition. An approach to reduce the duration of the data acquisition is to excite all transducers simultaneously. This paper focuses on signal separation to deal with a signal overlapping phenomenon that occurs naturally with simultaneous excitation. The key to successful separation is to select the excitation signals and the methods for the separation. Sinusoids and chirps are discussed here as excitation signals, and the correlation between the excitation signals and the received total signal is used as the separating method. How the length of the signal, its frequency bandwidth, frequency step, and windowing are affecting the separation performance is investigated. The results demonstrate that the arrival time of the separated signals can be accurately determined, and successful separation is achievable when the parameters of signals are suitable. The best choice of sinusoids is found in the signal group 3 having 0.5 ms duration and 10 kHz frequency step. The best choice of chirps is the signal group 10 with 0.5 ms duration, 10 kHz frequency step, and 5 kHz bandwidth signals. The best window is the Blackman window.