Detection and Characterization of Buried Macroscopic Cracks Inside Dielectric Materials by Microwave Techniques and Artificial Neural Networks

The detection and characterization of macroscopic cracks inside dielectric materials is an important practical issue. Thus, there is a need to establish evaluation techniques, which can be used to characterize buried cracks; indeed, the knowledge of the geometrical configuration of a hidden crack is a key factor for fatigue crack engineering. Therefore, a microwave method for nondestructive characterization of macroscopic cracks inside dielectric materials is presented in this paper. This nondestructive and noncontact technique is based on the determination of the near-field reflection coefficient of an open-ended rectangular waveguide. The measurements are achieved by means of a microwave six-port-based system that operates at 35 GHz. We show that relatively small defects are detectable and demonstrate that the association of signal processing tools to this characterization method enables the retrieval of the crack profile in an acceptable manner. The reconstruction of a 1-D buried crack profile is performed by means of a multiple-multilayer-perceptron (MLP) approach. Several cases are investigated to demonstrate the capabilities of the method.