Defects Localization by a Temporal Maximum A Posteriori Deconvolution Method of Ultrasonic Signals

Abstract

In this work, the Maximum A Posteriori (MAP) method, is used for defect detection in aluminium by determining the reflectivity sequence of the ultrasonic signal. In numerical simulations, the MAP method is tested under two conditions, when a priori knowledge of the noise and the signal to be deconvolved is available and in the absence of this knowledge. MAP performance is therefore directly related to the signal-to-noise ratio. Correlation coefficients between the reconstructed reflectivity sequence and the original one, $r_{\hat{h}h}$, of 0.9155, 0.9880 have been obtained, for SNRs of 60 and 80 dB respectively in the case of the absence of the priori knowledge. Experimentally, the previously developed method has been applied to signals provided from an aluminum part. For this, two near defects of differents depths and positions has been considered. The deconvolution of experimental signals gives good results and allows to correctly locating the defects. However, the peaks of the estimated reflectivity are tinged with oscillations. Indeed, somes information characterizing the input signal is lost due to the limited bandwidth of the filter and because of the random noise affecting the measurements.