Analysis of Conductometric Microsensor Responses in a 36-Sensor Array by Artificial Neural Networks Modeling

Published

December 1, 2002

Author(s)

Zvi Boger, Richard E. Cavicchi, Stephen Semancik

Abstract

The ability of large-scale artificial neural networks (ANN) for gas classification and concentration prediction is demonstrated using microsensor responses. The data used were generated from a 36-element microhotplate (MHP) SnO2 sensor array which included different catalytic metals additives and had its sensing elements operated in a pulsed, ramped temperature mode. Mixtures of acetone and methanol at 0 ppm to 150 ppm concentrations in air were tested. Recursive ANN pruning and re-training techniques were used for the identification of the more relevant inputs, out of the original 1260 inputs (different sensors, materials, and temperatures). ANN models with 4 to 10 inputs gave good prediction values for the concentration of each of the two analytes in the mixtures.

Citation

International Symposium on Olfaction & Electronic Noses

Pub Type

Journals

Keywords

array, catalytic additive, conductance, gas microsensor, gas mixtures, knowledge extraction, microhotplate, neural network, tin oxide

Citation

Boger, Z. , Cavicchi, R. and Semancik, S. (2002), Analysis of Conductometric Microsensor Responses in a 36-Sensor Array by Artificial Neural Networks Modeling, International Symposium on Olfaction & Electronic Noses (Accessed December 7, 2024)