Using sensor arrays to decode NOx/NH3/C3H8 gas mixtures for automotive exhaust monitoring

Abstract

An array of four mixed-potential sensors is employed to identify and quantify gases in complex mixtures of unknown composition which mimic diesel engine exhaust. The sensors use dense metal and metal oxide electrodes with a porous ceramic electrolyte, yttria-stabilized zirconia (YSZ). Since the sensors exhibit cross-specificity toward target gases, we develop a computational model for predicting gas concentrations in the mixtures. Our model is based on fundamental principles of gas-sensor interactions and, furthermore, takes into account the non-linearity of the observed sensor voltage response. Our approach enables accurate predictions of gas concentrations from the voltage output of the sensor array exposed to an extensive set of mixtures involving C3H8, NH3, NO and NO2. We find that our predictions remain accurate even if the model is trained using a reduced set of mixtures, or if the number of sensors is decreased to three or two. Our experimental and computational framework can be used to decipher contents of complex gas mixtures of unknown composition in numerous industrial, automotive, and national security settings.