Through the improvement of nanomaterial technologies, a gas sensor was developed for detecting ppm or ppb levels of gas. Our SnO2 nanosheet gas sensor can detect 50 ppb of acetone without the requirement of a novel metal catalyst by exposing the (101) facet containing the Sn2+ state. Despite the high performance, the fluctuation of the gas response value based on operating conditions, even at the same concentration, is a critical problem in gas sensors. Thus, the alarm criteria of the sensor are typically determined by a safety factor. However, this method is not suitable for application in ultrasensitive sensors that require distinguishing minute differences in extremely low concentrations for medical examination or odor analysis. Therefore, we suggest a self-adaptive system that is based on operating conditions in collaboration with the data prediction model. The sensor system is based on a predictive model obtained by the response surface methodology. When the system detects a change in conditions, the alarm criteria are changed appropriately through the calculated values from the predictive model. To prepare a database for an effective predictive model, the gas responses of the SnO2 nanosheet sensor were measured with 20 treatments with 3 independent variables, namely, the temperature, flow rate, and concentration. Our prediction model achieved its best performance on training data with R2 = 0.9299 and less than 5% error in the prediction of unseen data.
Keywords: SnO2; alarm criteria; data prediction; fluctuation; gas sensors; nanosheets.