Classification and concentration estimation of CO and NO₂ mixtures under humidity using neural network-assisted...

Area of Science:

• Materials Science
• Chemical Sensing
• Artificial Intelligence

Background:

• Metal oxide sensors face cross-sensitivity issues, limiting their specificity.
• Accurate gas detection is crucial for environmental monitoring and industrial safety.
• Machine learning offers potential solutions for complex sensor data analysis.

Purpose of the Study:

• To develop a robust gas sensing system using an array of metal oxide sensors.
• To employ machine learning techniques to overcome sensor cross-sensitivity.
• To achieve high accuracy in identifying specific gases and their mixtures.

Main Methods:

• Fabrication of a sensor array using In2O3, Au-ZnO, Au-SnO2, and Pt-SnO2.
• Simultaneous exposure of sensors to various concentrations of nitrogen dioxide (NO2) and carbon monoxide (CO).
• Application of Principal Component Analysis (PCA) and deep neural networks (DNNs) for data analysis.
• Utilizing a Convolutional Neural Network (CNN) by treating sensor data as images.

Main Results:

• Achieved 100% accuracy in gas classification during cross-validation and testing.
• Demonstrated effective discrimination between NO2, CO, and their mixtures.
• Validated performance across different humidity levels (40% and 90% RH).

Conclusions:

• The proposed machine learning approach effectively resolves metal oxide sensor cross-sensitivity.
• Deep neural networks and CNNs provide accurate and reliable gas identification.
• This method eliminates the need for time-consuming manual feature extraction, enabling faster analysis.