Detection of CWA simulants by electronic nose based on low-powered MEMS gas sensor array

Given the severe threats of chemical warfare agents (CWAs) to public safety, gas sensing technology development is vital. Metal oxide semiconductor (MOS) based gas sensors are popular for detection of CWA simulants, which have been widely used in the safe study of CWAs detection due to their similar structure and properties with actual CWAs. But the reported MOS sensors still suffer from high power consumption and poor selectivity of discriminating multiple simulants. In this work, an electronic nose (e-nose) driven by artificial intelligence (AI) algorithms is established to detect and identify five CWA simulants. The e-nose consists of 24 MOS based micro-electro-mechanical system (MEMS) gas sensors that own advantages of small size (0.9 mm×0.9 mm), low-power (<10 mW in average), and exhibits good sensing performances of relatively high response, low detection limit, wide concentration detection range, considerable reversibility and stability towards CWA simulants. Methods are proposed to extract three types of features from sensing performances (the steady-state response, maximum response and response time), initial response rate coefficient, and global change trend in the response stage, respectively. Subsequently, linear discriminant analysis (LDA) was utilized and back propagation network (BPN) was optimized to identify the simulants, obtaining preliminary separated classes in 3D LDA space and high prediction accuracy (>99.0%) with BPN, respectively. These results suggest that the AI driven e-nose strategy has highly potential to be applied as a portable equipment for rapid, on-site detection and identification of CWA simulants and even actual CWAs.