Chemical substance classification by electronic noses

Abstract

Normally, an electronic nose project uses two researches areas which are hardware for developing sensors to detect substance smell and software using pattern matching theorem for recognizing substance. The operation begins with sensors hit the smell of chemical substance. The result is converted from analog to digital representation. An artificial intelligence is a tool of a thinking system which can create knowledge as if a human does. The objective of this research is to classify chemical substance by using electronic noses. We used eight types of chemical substance in the experiment which are 1) Acetone, 2) Benzene, 3) Propanal, 4) Butanol, 5) Chloroform, 6) Ethanol, 7) Methane and 8) Tetrahydrofuran. We compared nine structures of neural network to classify the chemical substance data. The precision of correctness is equal to 94.64 for a neural network structure as 54 input-layer nodes, 216 hidden-layer1 nodes, 8 hidden-layer2 nodes and 8 output-layer nodes.