Identification of highly oscillatory ultrasonic NDE echoes through evolutionary algorithm

The performance of ultrasonic non-destructive estimation (NDE) is strongly dependent on the axial resolution of ultrasonic echoes. However, the highly band-pass signals commonly used in ultrasonic NDE systems pose significant limitations upon the time resolution enhancement due to seriously oscillatory cost function under classic estimation methods. In the present paper, a novel approach to improve the fidelity of sinusoidal NDE signals is presented which incorporating the pre-whitening of the non-linear least-squares (LS) cost function with the evolutionary optimization to facilitate the efficient search of global minimum. Simulation results are presented, demonstrating the feasibility of the proposed algorithm.