Machine learning-assisted Ti3C2Tx-PANI-SnO2 heterojunctions for real-time monitoring of NH3 released from food spoilage.

Ammonia (NH3) released during food spoilage is an important indicator of food freshness, yet its accurate monitoring remains challenging because of high humidity and interference from coexisting volatile amines. In this study, an intelligent sensing platform integrating a Ti3C2Tx-polyaniline (PANI)-SnO2 multidimensional heterojunction sensor with Transformer-based signal analysis was developed for reliable, real-time NH3 monitoring. The synergistic combination of Ti3C2Tx, PANI, and SnO2 enabled efficient signal transport, selective NH3 adsorption, enhanced interfacial charge transfer, and improved structural stability. The optimized Ti3C2Tx-PANI(50)-SnO2(30) sensor exhibited a response of 12.6 to 200 ppm NH3 over a wide detection range of 0.2-200 ppm. It maintained stable sensing performance at 20%-90% relative humidity and retained approximately 80% of its initial response after 60 days of storage, demonstrating good humidity tolerance and long-term stability. To mitigate interference from dimethylamine and trimethylamine generated during food spoilage, a Transformer model was employed to analyze the dynamic sensing signals. The model achieved a classification accuracy of 97.51% for mixed-gas samples and enabled reliable quantitative determination of NH3 under complex interference conditions. The sensor and embedded model were further integrated into a portable device for real-time monitoring of NH3 released during the spoilage of shrimp, fish, and pork. The measured NH3 concentrations deviated from the actual values by less than 10%. This multifunctional intelligent sensing platform provides a practical approach for on-site, intelligent food freshness assessment in complex high-humidity environments.