C. Bessant et S. Saini, Simultaneous determination of ethanol, fructose, and glucose at an unmodified platinum electrode using artificial neural networks, ANALYT CHEM, 71(14), 1999, pp. 2806-2813
Dual pulse staircase voltammetry (DPSV)-a combination of pulsed electrochem
ical detection and staircase voltammetry-is investigated for the simultaneo
us determination of glucose, fructose, and ethanol in mixtures. Each analyt
e is found to elicit a distinctive response at a platinum electrode in an a
lkaline solution. A method is devised for visualizing the electrochemical r
esponses of numerous mixtures of the three compounds simultaneously, and ap
plication of this method reveals that the mixed responses contain character
istics of the individual analytes approximately in proportion to their conc
entrations but that the combination of the individual responses is not a si
mple summation. Extraction of individual analyte concentrations from combin
ed DPSV responses is subsequently achieved using artificial neural networks
(ANNs), The effects of the amount of training data, the number of hidden n
eurons, the hidden neuron transfer function, and the network training time
are investigated. Large amounts of training data and a hidden layer with lo
g-sigmoidal transfer functions are found to give the best results. Networks
with relatively small hidden layers and relatively little training are fou
nd to produce the most generalized models, giving the most accurate concent
ration predictions when tested on analyte concentrations not present in the
training data, The lowest rms errors achieved were 40 mu M, 40 mu M, and 0
.5 mM for fructose, glucose, and ethanol, respectively, over a range of app
roximately 0-700 mu M for the sugars and a range of 0-12 mM for ethanol, Th
e success of this novel combination of DPSV and ANNs opens new possibilitie
s for the simultaneous detection of mixtures of aliphatic compounds, which
are traditionally considered difficult to detect.