Abstract :
|
Time-frequency techniques are applied for mixing signals from a material inspected by a multiple sensor detection system. Ultrasonic sensors are located at the perimeter of a rectangular shaped material evaluated in a pulse-echo scheme. The resulting mixed signal becomes a high-resolution time-frequency image of the material. Different kinds of classification techniques can be applied to this image in order to obtain the defects in the material. In this paper, a simulation and experimental evaluation of the proposed approach are presented. Several time-frequency transforms and the fuzzy c-means are used. In the simulations, backscattering of the material grain is modelled by using Gaussian and K distributions with different signal to noise ratio parameters. The validity of the presented method is assessed through the detection and spatial location of artificial defects in a material with a rectangular shape. The performance of the classification technique in discerning defects buried in the backscattering from the material grain microstructure, is also discussed.
|
|
