Machine Learning-Assisted Volatile Organic Compound Gas Classification Based on Polarized Mixed-Potential Gas Sensors

Abstract

The performance of electrochemical gas sensors depends on the reactions at the three-phase boundary. In this work, a mixed-potential gas sensor containing a counter electrode, a reference electrode, and a sensitive electrode was constructed. By applying a bias voltage to the counter electrode, the three-phase boundary can be polarized. The polarization state of the three-phase boundary determined the gas-sensitive performance. Taking 100 ppm ethanol vapor as an example, by regulating the polarization state of the three-phase boundary, the response value of the sensor can be adjusted from -170 to 40 mV, and the sensitivity can be controlled from -126.4 to 42.6 mV/decade. The working temperature of the sensor can be reduced after polarizing the three-phase boundary, lowering the power consumption from 1.14 to 0.625 W. The sensor also showed good stability and short response-recovery time (3 s). Based on this sensor, the Random Forest algorithm reached 99% accuracy in identifying the kind of VOC vapors. This accuracy was made possible by the ability to generate several signals concurrently. The above gas-sensitive performance improvements were due to the polarized three-phase boundary.