Rapid Identification Method for CH₄/CO/CH₄-CO Gas Mixtures Based on Electronic Nose

Area of Science:

• Chemical Sensors
• Artificial Intelligence in Chemistry
• Gas Chromatography

Background:

• Semiconductor gas sensors suffer from cross-sensitivity, complicating the detection of mixed gases.
• Existing electronic nose (E-nose) methods often require lengthy analysis of the full sensor response using complex algorithms.

Purpose of the Study:

• To develop a rapid and accurate method for identifying methane (CH4), carbon monoxide (CO), and their mixtures using an E-nose.
• To overcome the time-consuming limitations of conventional E-nose analysis.

Main Methods:

• Designed an E-nose system with seven gas sensors.
• Proposed analyzing only the initial sensor response phase to reduce detection time.
• Developed polynomial fitting methods for feature extraction.
• Applied Linear Discriminant Analysis (LDA) for dimensionality reduction.
• Trained an XGBoost model on LDA-optimized features for gas identification.

Main Results:

• Significantly shortened the gas detection and identification time.
• Successfully extracted relevant gas features from the initial response stage.
• Achieved nearly 100% accuracy in identifying CH4, CO, and their mixtures.
• Demonstrated a reduced computational complexity for the identification model.

Conclusions:

• The proposed rapid E-nose method effectively identifies CH4, CO, and mixtures with high accuracy.
• Analyzing the initial sensor response stage is sufficient for accurate gas identification.
• The combination of LDA and XGBoost provides an efficient and accurate gas identification model.