Monitoring the Flue Gas Using a Fuzzy Neural Network Based Artificial Olfactory System

Abstract

The potential of artificial olfactory technique for on-line monitoring of waste flue gas at different incineration temperatures was examined based on a metal oxide gas sensors array. The sensors signals from 120 samples of flue gas were collected at temperatures from 650 to 950°C. Statistical methods used in this study were principal component analysis (PCA), Linear discriminant analysis (LDA) and Fuzzy neural network (FNN). PCA and LDA were used to reduce the dimensionality and visualization of datasets. The FNN model was achieved with a high discrimination accuracy rate of 85%. Thus, an effective way to discriminate flue gas under different incineration temperatures was put forward.