An electronic nose for CO concentration prediction based on GL-TCN

Abstract

The air has a direct impact on both life and health. In most situations, the damage caused by excessive inhalation of carbon monoxide (CO) is irreversible. To prevent further catastrophes, the monitoring of CO concentrations deserves our attention and effective action. According to the available research, electronic nose (E-nose) consistently exhibits great performance in additional sectors while offering fresh approaches to problem-solving. However, the neural networks currently used in E-nose are still somewhat constrained, and the time-series dataset cannot be processed to its fullest potential by using conventional neural networks. In this paper, we propose an improved temporal convolutional network (TCN) to complete reliable training on high-dimensional time series datasets. GL-TCN exhibits a better fit even after decreasing the data, when compared to recurrent neural networks (RNN), long short-term memory (LSTM), TCN and gate recurrent unit (GRU).