Abstract
Recent research efforts have focused on improving the performance of metal-oxide semiconductor (MOS) gas sensors through their thermal modulation using integrated heaters. This approach allows us to enhance the selectivity of measurements; however, the main challenge with this amelioration lies in interpreting the sensor response, which takes the form of complex patterns that require the application of advanced signal processing techniques. This study introduces a methodology for the quantitative determination of volatile compounds (ethanol and methanol at various concentrations ranging from 31 to 2000 ppm for each of these compounds) in mixtures using a single thermally modulated MOS gas sensor. The recorded responses of the detector were interpreted by combining two signal processing techniques: principal component analysis (PCA) for feature extraction, and artificial neural networks (ANNs) for predicting the levels of the tested volatile components. The proposed methodology demonstrated satisfactory performance achieving R(2) values at the level of 0.999 across all datasets (learning, test, validation) and low error metrics (RMSE = 11.6-14.4 ppm), thereby confirming the robustness and accuracy of the approach and its applicability in a wide range of fields where rapid, cost-effective, and precise detection of ethanol and methanol is essential.