Anomaly-sensitive dictionary learning for structural diagnostics from ultrasonic wavefields

This paper proposes a strategy for the detection and triangulation of localized anomalies, such as defects, inclusions, or damage zones, in solid and structural media. The method revolves around the construction of sparse representations of the structure's ultrasonic wavefield response, which are obtained by learning instructive dictionaries that form a suitable basis for the response data. The resulting sparse coding problem is cast as a modified dictionary learning task with additional spatial sparsity constraints enforced on the atoms of the learned dictionaries, which provide them with the ability to unveil anomalous regions in the physical domain. The proposed methodology is model-agnostic, i.e., it forsakes the need for a physical model and requires virtually no a priori knowledge of the material properties. This characteristic makes the approach especially powerful for anomaly identification in systems with unknown or highly heterogeneous property distribution, for which a material model is unsuitable or unreliable. The method is tested against synthetically generated data as well as experimental data acquired using a scanning laser Doppler vibrometer.