Abstract:

Many works published about identification of acoustic properties of inhomogeneous media present methods based on the assumption that these properties vary smoothly along the propagation direction. This work presents a comparison between two time-domain sequential inverse methods for evaluation of either smooth or nonsmooth impedance profiles of transversely infinite media excited by plane waves at normal incidence. Sequential methods, also known as layer stripping methods, despite having the disadvantage of being quite sensitive to noise in the measured signal, are faster than other approaches, such as those that use global optimization techniques. The first method in the comparison [R. A. Tenenbaum and M. Zindeluk, J. Acoust. Soc. Am. 92, 3364--3369 (1992)] assumes a smooth profile by using a transformation of variables, which neglects the refraction effects between two consecutive layers. The second one, more general, performs some more calculations since it takes these effects into account, paying the price of a higher CPU time, being suitable for non-smooth profiles. Numerical experiments are carried out in order to establish applicability limits for both algorithms as discontinuities are included in the impedance profile, and also to evaluate their sensitivities when the identification is performed using noisy data.