Reconfigurable Multi-Channel Gas-Sensor Array for Complex Gas Mixture Identification and Fish Freshness...

Area of Science:

• Chemical Sensors
• Machine Learning Applications
• Food Safety Technology

Background:

• Metal-oxide semiconductor gas sensors offer advantages like low cost and rapid response.
• Limited selectivity in complex gas mixtures due to cross-sensitivity is a significant challenge.
• Existing sensor systems struggle with accurate identification in diverse gas environments.

Purpose of the Study:

• To develop a reconfigurable sensor-array system for enhanced gas mixture analysis.
• To investigate the use of cross-sensitivity in metal-oxide sensors for improved selectivity.
• To create a practical platform for identifying specific gas biomarkers, such as those indicating fish spoilage.

Main Methods:

• Designed a reconfigurable chemiresistive sensor array supporting up to 12 sensors with advanced control features.
• Applied machine learning classifiers including random forest (RF), convolutional neural network (CNN), and support vector machine (SVM) after principal component analysis (PCA) preprocessing.
• Optimized sensor channels through correlation analysis and feature importance assessment to reduce system complexity.

Main Results:

• The random forest (RF) model initially achieved 94% classification accuracy for fish-spoilage biomarkers.
• Optimizing the sensor array from 12 to 8 sensors improved accuracy to 96% while simplifying the system.
• The study demonstrated that leveraging sensor cross-sensitivity generates an information-rich odor fingerprint.

Conclusions:

• A reconfigurable sensor-array system effectively addresses selectivity limitations in complex gas mixtures.
• The deliberate use of cross-sensitivity, combined with machine learning, provides a powerful approach for gas sensing.
• This platform offers a practical solution for complex gas identification and real-time food freshness assessment.