Ultrasonic echoes estimation method using genetic algorithm

Ultrasonic echoes reflected from the inspected object contain information relative to the location, size and characteristics of defects. Accurate estimation of the echoes parameters is important in ultrasonic nondestructive evaluation (NDE). Model based method is often used to estimate ultrasonic echo signals, which models ultrasonic backscattered signal as superimposed Gaussian echoes corrupted by noise. This method requires good initial guesses for parameters and may get the local minima in the error function search space, instead of the global the minimum. To overcome the shortcomings, this research uses genetic algorithm (GA) to improve the expectation maximization (EM) algorithm. The improved algorithm using GA doesn´t need doing the initial guessing work manually, and can even acquire the estimation result directly. Simulation tests were carried out to validate the algorithm, which proved the effectiveness of the method.