Performance improvement: A lightweight gas information classification method combined with an electronic nose system

Abstract

Electronic nose (e-nose) system can recognize gas information by imitating the perception pattern of the human olfactory system, and it has been widely used as a fast and effective nondestructive testing technology. An effective gas information classification method can promote the engineering transformation of the e-nose system. In this work, a lightweight gas information classification is proposed to effectively classify the e-nose system's detection data. First, a gas feature attention mechanism (GFAM) is proposed based on the data characteristics of the e-nose system, which combines the peak factor (PeF.), integral value (InV.), and steady mean (StM.) to focus on the key features of the deep gas information. Second, a lightweight convolutional neural network (CNN) structure is designed and combined with the GFAM to establish a gas information classification model (GFAM-Net). Finally, the effectiveness of GFAM-Net is verified based on different datasets of the e-nose system. The results show that GFAM-Net not only has a small number of parameters and calculations but also achieves the best classification performance in the comparison results of multi-learning models.