A Review of the High-Performance Gas Sensors Using Machine Learning

High-performance gas sensors are of great importance to accurately identify/detect pollutant gases and monitor their concentrations in the environment to ensure human safety in daily life and production. Machine-learning techniques have been used to successfully improve gas sensing performances of gas sensors leveraging large onsite data sets generated by them. A simple process is introduced to show the typical approach to collect the features from sensing response curves and conduct a machine-learning algorithm to further analyze the data set. The improved gas sensing performances of the machine-learning-enabled sensors reported recently are summarized and compared, especially regarding selectivity and long-term stability (drift compensation). Furthermore, the expanded applications of a gas sensor or sensor array under machine-learning algorithms were discussed and reviewed. In addition, the possible challenges/prospects are emphasized and discussed as well. Our review further indicated that machine-learning techniques are effective strategies to successfully improve the gas sensing behavior of a single gas sensor or sensor array.