Selectivity enhancement for metal oxide (MOX) based gas sensor using thermally modulated datasets coupled with golden...

The ever-increasing demand for smart sensors for internet of things applications drove the change in outlook toward smart sensor system design. This paper focuses on using low-cost gas sensors [Metal Oxide (MOX)] for detection of more than one gas, which is otherwise complex due to poor selectivity of MOX sensors. In this work, detection of two gases, namely, ammonia (NH₃) and carbon monoxide (CO), using a single metal oxide (pristine tin oxide) sensor is demonstrated. Furthermore, chemometric based algorithms have been used to classify and quantify both gases. The present investigation uses the temperature modulated gas sensor response obtained at different concentrations for the mentioned gases. The golden section based optimization technique has been employed to obtain two different ranges of temperatures for both gases. After applying certain pre-processing techniques, the acquired data from the sensors were fed to various classification techniques, such as partial least squares (PLS) discriminant analysis, k-means, and soft independent modeling by class analogy, and 100% classification results were obtained. Furthermore, PLS regression (PLS-R) was used to perform quantitative analysis on the data using the optimized temperature ranges for both gases, and R² regression coefficients, 0.999 25 for NH₃ and 0.9399 for CO, were obtained. The results obtained from both the qualitative and quantitative analyses make our approach low-cost and smart to mitigate the cross-selectivity of metal oxide semiconductor based smart sensor design.