Machine learning-optimised carbon nanotube sensors for simultaneous monitoring of multiple environmental contaminants

Bhagavan Konduri, Suraya Mubeen, Ch.V. Raghavendran, R. Deepa, Satish Sampatrao Salunkhe, Ajith Sundaram,
Machine learning-optimised carbon nanotube sensors for simultaneous monitoring of multiple environmental contaminants,
Microchemical Journal,
Volume 218,
2025,
115284,
ISSN 0026-265X,
https://doi.org/10.1016/j.microc.2025.115284.
(https://www.sciencedirect.com/science/article/pii/S0026265X25026323)
Abstract: This study presents a machine learning (ML)-optimised carbon nanotube (CNT) sensor platform for the simultaneous detection and quantification of multiple environmental contaminants, including heavy metals, organic pollutants, and gaseous compounds. The platform integrates a hierarchical 16-element CNT array with a multi-task deep learning model, utilizing a Synthetic Signature Subtraction (S3) algorithm to minimize cross-reactivity and interference. The system employs a combination of multi-functional CNTs and ML-driven signal processing to achieve high classification accuracy (94.2 %), low detection limits (0.05–1.2 ppb), and fast response times (30–120 s) across complex environmental matrices, including river water, groundwater, and seawater. Real-world validation demonstrates that the system maintains 87.3–96.8 % accuracy in these challenging conditions, with cross-reactivity reduction factors of 13–28× compared to conventional CNT sensors. The platform is designed for real-time operation with low power consumption (315 mW average), achieving approximately 35 h of continuous operation on a 3.7 V, 3000 mAh battery with a 32-s measurement cycle. The wireless-enabled system provides a 90 % cost reduction compared to conventional analytical methods while enabling field deployment for environmental monitoring applications. This work demonstrates the potential of ML-optimised CNT sensors for comprehensive environmental surveillance, with practical advantages in analysis speed, cost-effectiveness, and multi-analyte detection capability for regulatory compliance and pollution monitoring.
Keywords: Carbon nanotubes; Machine learning; Environmental sensors; Multi-analyte detection; Cross-reactivity mitigation; Sensor arrays