A study on transient and steady state sensor data for identification of individual gas concentrations in their gas mixtures

In this study, a comparative study was performed for the quantitative identification of individual gas concentrations (trichloroethylene and acetone) in their gas mixtures using transient and steady state sensor responses. For this purpose, three neural network (NN) structures were used. The quartz crystal microbalance (QCM) type sensors were selected as gas sensors. One of the neural networks was used for quantitative identification using only steady state response. The other two neural networks were used for quantitative identification using both transient and steady state responses. One of them was a neural network with tapped time delays, and this NN used sensor frequency responses and past values of these responses. The other NN structure used sensor frequency responses and slope values of these sensors frequency responses to quantify the components in the binary mixture. Levenberg–Marquardt training algorithm was performed as the training method of the neural network structure. Quantitative analysis of trichloroethylene (TCE) and acetone was evaluated in terms of neural network structures and sensor responses.