Enhanced VOCs gas mixture recognition with sensor array and SSA-BP neural network

Abstract

Accurate identification of VOCs in mixtures is essential for monitoring toxic and explosive gases, serving industrial, environmental, as well as military and defense applications. In this study, a Salp Swarm Algorithm-Back Propagation (SSA-BP) neural network was proposed in combination with a MEMS-based nano-SnO₂ sensor array to enhance mixed-gas detection. The nano SnO2 sensors offer high sensitivity, while the SSA-BP neural network optimizes data processing based on noise filtering, feature extraction, and a robust nonlinear learning model, thus improving recognition accuracy, thereby improving recognition accuracy. This combination achieved excellent classification results for mixed gases, with a classification accuracy of up to 99% for various indoor toxic gas mixtures, including acetone, ethanol, and methanol. Additionally, it attained an R-squared score of 0.95 for accurately predicting gas concentrations. Based on the experimental results, we also propose reducing the number of sensors required while maintaining system performance. This integration shows great potential for real-time gas monitoring, as well as for portable VOCs detection systems and safety applications.