Reflectivity Estimation Using Expectation Maximization Algorithm in Ultrasonic Nondestructive Evaluation

The finite frequency bandwidth of ultrasonic transducer is the most significant factor that limits the time resolution of the ultrasonic non-destructive evaluation (NDE) signals. Consequently, in order to remove the transducer effect from the received signal and minimize the effects of noise, deconvolution techniques should be applied. In this study we apply a model-based deconvolution approach on received signals from thin layer sheets to estimate the time of arrival (TOA) of the echoes and subsequently obtain an accurate estimation of thickness. In this approach the transducer wavelet is modeled by some Gaussian echoes as preliminary stage and then deconvolution of superimposed echoes is carried out using expectation maximization (EM) algorithm as a parameter estimation problem. A simulation study was performed to assess the merit of this algorithm in separating the superimposed echoes and finally the algorithm was applied on experimental data to estimate the reflectivity accurately for thickness measurement.