Wan-Li Xing and Xi-Wen He
An array of nine piezoelectric quartz crystal sensors, each coated with a different crown ether derivative, was constructed for multicomponent analysis of organic vapors. The usefulness of this array was evaluated by quantifying known mixture samples in two three- and one four-component cases; the first sample set consisted of aniline, butan-1-ol and acetonitrile, the second of pyridine, pyrrole and tetrahydrofuran and the third of formic acid, acetic acid, propionic acid and acrylic acid. Three chemometric techniques, viz, artificial neural networks (ANN), partial least squares (PLS) and non-linear partial least squares (NPLS), were used in the data analysis. The results showed that better prediction was achieved with ANN in all cases. Also, as the number of components present in the mixture increased, the sensors tended to exhibit non-linear sensory behavior. The effect of the number of neurons in the hidden layer on the performance of the network is also discussed.