Noise Analysis-Resonant Frequency-Based Combined Approach for Concomitant Detection of Unknown Vapor Type and Concentration

This paper introduces a novel measurement procedure for the concomitant detection of type and concentration of unknown target species employing a combination of noise spectra analysis and resonant frequency characteristics. An algorithm was proposed incorporating a machine learning phase (for the creation of a database) followed by a detection phase. In the machine learning phase, with the exposure to the known target species with known concentrations, three distinctive features of the device were measured, namely: 1) randomness on the estimation in capacitance; 2) impedance variation (with respect to the air) profile; and 3) capacitive response magnitude (CRM). In this regard, reduced graphene oxide–titanium dioxide nanotubes hybrid sensor was employed as the test device, and consequently, low-frequency noise (i.e., randomness) on the capacitance estimation with the exposure to different known target species (methanol, ethanol, and 2-propanol as test species) was measured, and the distinctive signature frequency for each one of them was obtained. A shift in the resonant frequency toward the higher frequency regime was also observed with the increase in target species concentrations, by which the concentration of target species was predicted effectively. Furthermore, a principal component analysis was performed on the CRM for cross-validation of obtained target species type and concentration. In the detection phase, the above-mentioned parameters of the device were measured for an unknown species and compared with the existing database. Moreover, a multivariate data analysis was carried out for the proposed method, and the maximum error was found to be within 13.25% at higher concentrations.