Monitoring structural integrity of large planar structures that aims at detecting and localizing impact or damage at any point of the structure requires normally a relatively dense network of uniformly distributed ultrasonic sensors. 2-D ultrasonic phased arrays, due to their beam-steering capability and all azimuth angle coverage are a very promising tool for structural health monitoring (SHM) of plate-like structures using Lamb waves (LW). Linear phased arrays that have been proposed for that purpose, produce mirrored image characterized by azimuth dependent resolution, which prevents unequivocal damage localization. 2D arrays do not have this drawback and they are even capable of mode selectivity when generating and receiving LWs. Performance of 2D arrays depends on their topology as well as the number of elements (transducers) used and their spacing in terms of wavelength. In this paper we propose a consistent methodology for three-step: theoretical, numerical and experimental investigation of a diversity of 2D array topologies in SHM applications. In the first step, the theoretical evaluation is performed using frequency-dependent structure transfer function (STF). STF that defines linear propagation of different LWs modes through the dispersive medium enables theoretical investigation of the particular array performance for a predefined tone-burst excitation signal. A dedicated software tool has been developed for the numerical evaluation of 2D array directional characteristics (beampattern) in a specific structure. The simulations are performed using local interaction simulation
approach (LISA), implemented using NVIDIA CUDA graphical computation unit (GPU), which enables time-efficient
3D simulations of LWs propagation. Beampatterns of a 2D array can be to some extend evaluated analytically and using numerical simulations; in most cases, however, they require experimental verification. Using scanning laser vibrometer is proposed for that purpose, in a setup where LWs, excited by PZT transmitters of the investigated array are sensed in multiple points corresponding to the locations of the 2D array receiving elements. A virtual receiving sub-array is created in this way and the performance of various array architectures in the reception mode can be evaluated experimentally without the need of physical prototype; a change of topology requires only straightforward modification of the measurement points distribution at the tested structure. For illustration, beampatterns of three symmetrical 2D topologies, i.e., circular, star-shaped and spiralshaped, will be examined in the paper and compared in terms of their beam-width and side-lobes level. The effect of apodization applied to the array elements will be also investigated.