employing machine learning approach in cavity resonator sensors for characterization of lossy dielectrics

this work presents a novel microwave sensor that is specially designed for the retrieval of complex permittivity. the proposed sensor is designed to operate in the c band (4.54 ghz). by implementing a novel feeding structure, the proposed siw cavity design improves the coupling and achieves a better quality factor. several techniques are used to enhance sensitivity, including a photonic band gap (pbg), corner cut, and slow-wave vias. these techniques increase the interaction between the material under test and the electric field. by utilizing slow-wave vias, 35% size reduction is achieved. achieving simultaneous miniaturization and sensitivity enhancement in this study introduces a new possibility and application for sensor design. the values of complex permittivities are extracted from scattering parameters obtained from simulation of the structure in cst microwave studio (mws) using a machine learning approaches. our sensor has 0.8% sensitivity, which is better than that of other sensors. moreover, the maximum error rate in our method is lower than other existing methods. this ratio for the proposed method is 2.31% while for curve fitting and analytical solutions are 26% and 16%, respectively.