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

Oxide semiconductor gas sensors have been widely adopted owing to their low cost, rapid response, small footprint, and facile integration. However, in complex mixed-gas environments their selectivity is impaired by inherent cross-sensitivity. To address this limitation, we develop a reconfigurable gas-sensor array system capable of integrating up to 12 chemiresistive sensors in either four- or six-electrode configurations, with independent thermal control and user-defined gas distribution and exhaust paths. Fish freshness detection is used as an application exemplar. After Principal Component Analysis (PCA)-based preprocessing, Convolutional Neural Network (CNN), Random Forest (RF), and Support Vector Machine optimized by Particle Swarm Optimization (PSO-SVM) models were evaluated, with RF achieving the highest classification accuracy. Correlation analysis and feature-importance evaluation were then used to down-select channels, yielding an optimized 8-sensor array that lowered system complexity and increased recognition accuracy to 96%. These results highlight the feasibility of leveraging cross-sensitivity within a carefully designed array, providing a practical platform for complex gas-mixture identification and food freshness assessment.