Characterization and neural modeling of a microwave gas sensor for oxygen detection aimed at healthcare applications

The studied sensor consists of a microstrip interdigital capacitor covered by a gas sensing layer made of titanium dioxide (TiO2). To explore the gas sensing properties of the developed sensor, oxygen detection is considered as a case study. The sensor is electrically characterized using the complex scattering parameters measured with a vector network analyzer (VNA). The experimental investigation is performed over a frequency range of 1.5 GHz to 2.9 GHz by placing the sensor inside a polytetrafluoroethylene (PTFE) test chamber with a binary gas mixture composed of oxygen and nitrogen. The frequency-dependent response of the sensor is investigated in detail and further modelled using an artificial neural network (ANN) approach. The proposed modelling procedure allows mimicking the measured sensor performance over the whole range of oxygen concentration, going from 0% to 100%, and predicting the behavior of the resonant frequencies that can be used as sensing parameters.