Ambient noise correlation offers an elegant way of passively caracterize a medium. The principle is that in diffuse field condition, Green’s function between two sensors can be retrieved from the cross-correlated received signals. Here, the technique is applied in conditions where convergence toward Green’s function is not ensured. Since the obtained information is sensible to the medium state, it is still exploitable for damage detection provided a baseline version is available for comparison. This requires the identification of the acoustic source characteristics at the instant of measurement, which could be achieved using the signal measured at a “reference” sensor.

The identification of Lamb mode amplitude variation as a function of the damage evolution is still the most difficult step in the process of damage monitoring using embedded Lamb wave-based systems. The aim of this paper is to propose a simple system based on the generation of two different frequencies in order to better identify Lamb mode amplitude and to avoid false data interpretation in plates containing a hole of variable diameter. This identification is based on a simple relation between the short-time Fourier transform and the two-dimensional Fourier transform. Experimentally, a 3 mm thick aluminium plate is used and the two frequencies have been chosen equal to 400 kHz and 600 kHz in order to generate the two first fundamental Lamb waves.

The point source response of a reverberant solid plate is modeled through a nonstationary Poisson process based on the image-source method. The theoretical expectation of the envelope is then derived, taking into account the dispersive nature of plate waves, and validated by numerical results. Least-square curve-fitting applied to an ensemble average over N realizations can then be used to identify useful parameters such as wave velocity, plate surface, or source-receiver distance. It is shown that even values of N down to 1 (no averaging) allow a satisfying identification. Application to the estimation of the source-receiver distance using a single sensor is finally highlighted to illustrate the promising potentialities of the measurement principle proposed.