Konferensartikel

Classification of Gases and Estimation of Gas Flow Rate Based on Unsupervised and Supervised Learning Respectively

Maths Halstensen
Department of Electrical Engineering, IT and Cybernetics, University of South-Eastern Norway

Simen S. Halstensen
Department of Electrical Engineering, IT and Cybernetics, University of South-Eastern Norway

Ladda ner artikelhttps://doi.org/10.3384/ecp20176451

Ingår i: Proceedings of The 61st SIMS Conference on Simulation and Modelling SIMS 2020, September 22-24, Virtual Conference, Finland

Linköping Electronic Conference Proceedings 176:64, s. 451-457

Visa mer +

Publicerad: 2021-03-03

ISBN: 978-91-7929-731-2

ISSN: 1650-3686 (tryckt), 1650-3740 (online)

Abstract

In this research project, acoustic chemometrics was assessed as a new method for both classification and prediction of flow rate of five selected gas types. The gas types were selected to span different densities as much as possible while at the same time being relatively safe to use. The five gas types were Argon, Helium, Carbon dioxide, Nitrogen and Air. The research questions were 1) Can measurements of the vibrations in a gas control valve in combination with signal processing and unsupervised learning be used to classify the five gases mentioned above? 2) Can the vibrations in the gas valve in combination with supervised learning be used to determine the flow rate of the five gases? 3) Can a simple low cost piezo disk provide signals comparable to that of an industrial accelerometer? The results show that it is possible to classify the five gas types based on principal component analysis with three components. The gas flow rate could also be predicted for all five gases based on partial least squares regression with an average error of 2-5%. The Piezo disk could not be used for gas classification, but for prediction of gas flow rates it was comparable to the accelerometer. All the prediction models were validated based on independent data.

Nyckelord

supervised learning, unsupervised learning, partial least squares regression, classification, principal component analysis, accelerometer

Referenser

Inga referenser tillgängliga

Citeringar i Crossref