AI-Based Electronic Nose System for Real-Time Decomposition Stage Classification Using Multi-Gas Sensors

Abstract

Odour identification using conventional single gas sensors can be challenging because different odour conditions may produce overlapping sensor responses. This work presents a low-cost AI-based electronic nose using an ESP32 microcontroller and an array of MQ135, MQ2, MQ3, and MQ4 gas sensors, along with a DHT22 sensor for temperature and humidity measurement. The collected multi-sensor data are used to train a machine-learning classification model for identifying four odour conditions: FRESH, EARLY, ACTIVE, and ADVANCED. The developed model achieved 100% classification accuracy on the test dataset of 80 samples, with precision, recall, and F1-score of 1.00 for all four classes. The proposed system demonstrates that low-cost gas sensors combined with machine learning can provide an affordable and practical approach for real-time odour-condition classification using an ESP32-based platform.