Damage identification in industrial products and manufacturing infrastructures at the earliest possible time is an issue that almost all industries are interested in. Such a need makes industrial sectors to employ some kinds of structural health monitoring configurations to provide better life safety and also to have access to economical benefits which are provided by SHM. Among the different types of damage identification approaches in aerospace structures, guided waves have attracted the most interest. Utilizing guided waves for damage identification in thin-walled metallic and composite materials and structures has created a promising outlook for on-line or real-time health monitoring of engineering structures. Depending up on the kind of signal processing method and pattern recognition technique, a variety of information can be obtained from guided wave signals which are classified in time, frequency and joint time-frequency domains based on the field of study. To increase our knowledge about existing damages in a structure, after localization and severity determination, it seems to be necessary to attain more possible information about their geometry. It looks that identification of crack shape as a subsequent level in damage assessment is important, considering this fact that cracks are not necessarily linear. To investigate the feasibility of crack quantification with curved geometry using Lamb waves, at first, in this study, relative reflection intensity from typical cracks is investigated in an aluminum beam as a 1/D waveguide for propagation of Lamb waves. It is then shown that RRI criteria have no possibility of providing an appropriate signal feature for a typical damage in a structure due to the same RRI values in circumstances of different crack shape. The feasibility of using energy envelop curve in a captured signal from the structure, is also evaluated in providing an appropriate feature from a damage signal. This method also seems to be not suitable due to the crucial need to wave signal reflections from boundaries. The concept of digital damage fingerprint has been developed to overcome the uncertainties associated with employing existent feature extraction methods in evaluation of crack parameters using artificial neural network (ANN) technique. Among different types of feature extraction procedures, DDF method is assumed as one of the best choices for extraction of an appropriate feature from Lamb wave signals. Next in this study, for the first time, the prediction of damage geometry and size for transverse cracks with curved geometry in beam-like structures is systematically scrutinized by using DDF method, based on an inverse analysis approach. By utilizing damage parameter database (DPD), comprising a sufficient number of DDFs, training of designated neural network is then carried out as a powerful tool in the artificial intelligence (AI) and damage parameter prediction in the beam under study. Results indicate that in comparison with signal amplitude, frequency content of received Lamb waves are more sensitive to crack shape and play an effective role in prediction of crack curvature as a main part of damage geometry. On the contrary, DDF including signal amplitudes in time domain, can predict size of crack with less error values in comparison with DDF extracted from frequency content of signals. Finally, it can be concluded that an acceptable quantification of crack size and curvature in an aluminum beam can be achieved by using ultrasonic guided waves, activated and received via piezoelectric wafer active sensor and actuator. keywords: structural health monitoring, Ultrasonic guided waves, digital damage fingerprints,