Digital beamforming methods in plate–like structures are widely exploited for Lamb waves-based damage imaging. Among them, the Delay and Sum (DAS) imaging technique is the most popular thanks to its low computational cost and ease of implementation. However, the imaging outputs are low–quality due to the high levels of side lobes and limited off-axis signal rejection, which leads to limited image resolution and contrast. Recently, the Delay Multiply and Sum (DMAS) beamforming has been applied to NDT field as a promising DAS alternative able to enhance the imaging reconstruction in terms of contrast and damage detectability. However, DMAS is still affected by high levels of artefacts. To tackle this aspect, literature offers a beamforming algorithm called Double Stage DMAS (DS–DMAS), firstly introduced in photoacoustic imaging and medical ultrasound imaging. In this paper, the DS–DMAS performance is analysed for Lamb waves inspection, in order to provide an exhaustive comparison between DAS, DMAS and DS–DMAS. As a further step, a filtering process addressed as Fresnel Zone Filtering (FZF) is used to restrict the beamforming partial sums in a physical way to the area around the scattering point. The proposed approach is an adaptation of a well–established technique in seismic data processing called Fresnel migration, able to suppress artefacts and enhance the quality of the imaging. The algorithms have been compared and characterized by exploiting an online free data set for guided waves inspection (http://openguidedwaves.de/) which collects piezo pitch–catch signals travelling through a quasi-isotropic Carbon Fiber Reinforced Plate (CFRP) at different actuated frequencies and damage positions.

Experimental characterization of Lamb waves in plate-like structures overcomes the intrinsic limits of a priori semianalytical finite element simulations, where material inaccuracies and nonidealities cannot be easily considered. Unfortunately, the experimental extraction of guided wave dispersion curves, and especially their polar representation along different directions of propagation at a given frequency, is not trivial. In nonisotropic materials, such analysis is a key aspect for a reliable and robust characterization of the behavior of waves. In this work, by exploiting scanning laser Doppler vibrometer measurements with narrowband excitation, two different signal processing methods for the extraction of the wavenumber polar representation at the excitation frequency are investigated and characterized. The first method is based on a distance regularized level set (DRLSE) algorithm, widely used in image processing and computer vision but, to the best of the author’s knowledge, never used in the Lamb waves’ field. The second method is based on the 2-D sparse wavenumber analysis which exploits the wavefield sparse representation in the wavenumber domain. With a precise and reliable extraction of the wavenumber characteristic in the ${k}$ -space, the polar representations at the excitation frequency of phase and group velocities can be estimated. The former, by exploiting the well-known wavenumber–frequency relation, the latter, instead, by computing numerical derivative among wavenumbers at multiple frequencies. The methodology has been validated on three different composite plates with different degrees of nonisotropy properties. The results show the effectiveness of the two methods, highlighting the advantages and disadvantages of both.