Investigation into the efficacy of a neural network approach to concentration measurements of ethanol in water based on multianalyzer inputs

This study explores the quantitative capabilities of the multianalyzer tens iograph for applications such as distilling. Two standard tensiograph trace features (from the many on offer) were selected in this study of pure etha nol-water solutions for a comparative study with a neural net approach base d on the entire data set. The advantage of the latter approach is that it c an learn to extract good estimates of an output variable based on (possibly interfering) multivariate inputs such as the measurands found in complex p roducts. The conclusion of this study is that, quantitatively, the multiana lyzer has the potential of approaching the sensitivities required for the f ast qualitative and quantitative analysis of whiskies. This could make it u seful as a supplement to the slow gas chromatograph-mass spectroscopic tech niques currently employed in the industry.