Cent Eur J Public Health 2001, 9(2):95-101

Pattern recognition of respirable dust particles by a back-propagation artificial neural network

Wippel R.¹, Pichler-Semmelrock F.P.², Köck M.², Kosmus W.¹

¹ Institute for Chemistry, Karl-Franzens-Universität Graz, Austria

² Environmental Protection Center of Styria, Graz, Austria

A back-propagation neural network was used as a pattern recognition tool for LAMMA mass spectral data. Standard EPA source profiles were used as training and test data of the net. The elemental patterns (10 elements) of the sum of 100 mass spectra of fine dust particles were presented to the trained nets and satisfactory recognition (> 50 %) was obtained.

Published: May 1, 2001  Show citation