Identification and quantification of gases and their mixtures using GaN sensor array and artificial neural network

Accurate identification and quantification of gas mixtures are almost unattainable utilizing only a metal-oxide/GaN sensor because of its cross-sensitivity to many gases. Here, an array of sensors has been formed consisting of Ag and Pt incorporated ZnO, In2O3 and TiO2 coated two terminal GaN photoconductors. The common environmental toxic gases, such as SO2, NO2, H-2, ethanol and their mixtures have been chosen as the gas analytes. All the gas responses have been obtained at 20 degrees C under UV illumination. Temporal responses have been post-processed to develop the training and test dataset. Then, four different artificial neural network models have been analyzed and optimized for gas classification study, which is done for the first time on GaN sensors. Statistical and computational complexity results indicate that back-propagation neural network (NN) stands out as the optimal classifier among the considered algorithms. Then, ppm concentrations of the identified gases have been estimated using the optimal model. Furthermore, implementation of the developed sensor array in combination with NN algorithm for real-time gas monitoring applications has been discussed.