Chemiresistive sensor array for quantitative prediction of CO and NO₂ gas concentrations in their mixture using...

Area of Science:

• Materials Science
• Chemical Sensing
• Computational Chemistry

Background:

• Single gas sensors lack selectivity in mixed environments, hindering accurate real-time detection.
• Detecting specific gases like carbon monoxide (CO) and nitrogen dioxide (NO2) in mixtures is crucial for safety and monitoring.
• Interference from other gases limits the precision of traditional gas detection methods.

Purpose of the Study:

• To develop a sensor array system for quantitative estimation of CO and NO2 concentrations in their binary mixtures.
• To apply machine learning algorithms to sensor signals for improved gas concentration prediction.
• To enhance the accuracy of gas detection by segmenting sensor data based on concentration regimes.

Main Methods:

• A sensor array composed of zinc oxide and graphene-cobalt sulfide sensors was fabricated.
• The sensor array was exposed to 29 different proportions of CO and NO2 binary mixtures at room temperature.
• Machine learning algorithms were employed, with sensor signals divided into three levels to improve prediction accuracy.

Main Results:

• Initial ML models showed inaccurate predictions when using all sensor signals.
• Segmenting sensor signals into three concentration levels significantly improved classification accuracy (85.13% ± 3.2%) and prediction.
• The developed system demonstrated effective quantitative estimation of CO and NO2 concentrations.

Conclusions:

• The sensor array combined with ML, when data is stratified by concentration, offers a robust method for detecting CO and NO2 in mixtures.
• This computational framework is adaptable for detecting additional gases and has potential applications in automotive, industrial, and environmental monitoring.
• The study highlights the effectiveness of a multi-level ML approach for enhancing gas sensing accuracy in complex environments.