Discriminative Detection for Multiple Volatile Organic Compounds via Dynamic Temperature Modulation Based on Mixed Potential Gas Sensor

Abstract

Gas sensors combined with artificial intelligence capable of distinguishing multiple odors hold great promise in volatile organic compounds (VOCs) discriminative detection. However, various issues such as large size, high expenses, and mutual interference have limited the utilization of sensor array with conventional single-output sensors. Herein, a novel method for multicomponent gas detection was proposed based on pulsed heating (PH) with single-sensor operation. This strategy involved rapid and continuous dynamic temperature modulation to stimulate the sensor for generating feature-rich response signals toward isoprene, n-propanol, acetone, and their gas mixtures. First, the heating pulse was optimized to show the best sensing performance and reflect the maximum difference between diverse categories of gas compositions. Then the discrete wavelet transform (DWT) was utilized to further magnify the difference on signal curves toward target gases. Subsequently, multivariate features from the signals can be extracted, which were input into the machine learning algorithm for classification. By virtue of the proposed strategy, it showed the highest accuracy of 98.94% in the identification experiments of seven groups of VOC components. The results demonstrated that the PH strategy with feature engineering contributed to efficient identification with a limited sensor. It offers the chance to apply simple, miniaturized, and highly efficient multivariable gas sensor instead of multisensor array for artificial olfaction.