Discriminant Analysis of Industrial Gases for Electronic Nose Applications

Abstract

This work is a part of ongoing research project for optimization of the Electronic Nose System (ENS) for its applications related to the identification of industrial gases. Two different experimental datasets of several gases are collected in a laboratory setup using two different sensor arrays. A dataset of six different gases (C3H8, Cl2, CO, CO2, SO2 and NO2 is collected using a commercially available array of seven Figaro gas sensors. Another dataset of three gases (C2H6O, CH4 and CO) is collected using a 4 × 4 tin-oxide sensors array which is built in the In-house foundry. In this paper some of the existing state of the art classification models are tested for the classification of experimentally acquired datasets. The existing classification models are used to analyze the behavior of the data acquired. The models that are tested for identification of gases are Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Regularized Discriminant Analysis (RDA), and K-Nearest Neighbor (KNN). Besides testing these classification models, fuzzy C means (FCM) clustering is also tested for the separation of clusters of gases.