A modeling-based technique for nondestructive evaluation of metal powders undergoing microwave sintering

A microwave imaging technique is proposed for reconstruction of geometrical parameters (a spatial position and a radius) of a metal sphere emerging inside the sample of metal powder in the course of microwave sintering. The technique relies on a numerical inversion realized with an artificial neural network backed by data on S-parameters obtained from FDTD simulation and measurement. Numerical experiments are performed in the frequency range from 2 to 3 GHz for a cylindrical sample of iron powder in a two-port waveguide system. The results show that the spheres of not less than 2 mm radius are reconstructed with the average errors of 0.2-2.2%.