Characterization of ceramic materials using ultrasonic technique in the frequency domain and artificial networks

The ceramic material characterization is very important to guarantee its mechanical properties. In the case of nuclear fuel (UO2) the adequate porosity ensures its thermal efficiency and its structural integrity that contribute to the safety at nuclear power plants. The Ultrasound Laboratory of the Nuclear Engineering Institute (LABUS/IEN) has developed a technique to measure the porosity in ceramic materials. This technique uses ultrasound signal in the frequency domain and creates spectrum patterns related to the material porosity. Trained artificial neural networks recognizes these patterns and associates them to the porosities. In this work 20 pellets of Alumina were used with porosities in the same range used in the nuclear fuel (0.70% to 4.25%). In this case the used network was able to recognize the patterns of the pellets and to associate to the porosities with 100% of precision. It was possible to distinguished pellets with a difference of 0.01% of the porosity. (author)