Abstract:

Caproic acid and acetic acid, two important organic acids in liquor, have obvious effects on the quality of liquor.
In order to accurately and quickly measure the contents of caproic acid and acetic acid in liquor base, calibration models
were established based on Fourier-transform near-infrared spectroscopy with artificial neural network by cross-validation
after the original spectra were analyzed and interpreted. The results showed that the caproic and acetic acid in liquor
base exhibited strong near infrared absorption and that the best spectral pretreatment methods for the two acids were first
derivative combined with deduction of a straight line and standard normal variate (SNV) transformation, respectively. Based
on the selected spectral ranges of 6 101.7–5 446 cm-1, 11 998.9–7 501.7 cm-1, 6 101.7–5 449.8 cm-1, and 11 998.9–7 497.9 cm-1,
a calibration model was established using partial least square regression and the internal cross-validation method. After
optimization of the model, the correlation coefficients (R2) between the chemical values of caproic acid and acetic acid in
calibration samples and near infrared prediction were 99.73% and 97.00%, internal cross-validation root mean square deviation
were 0.90 and 0.63 mg/100 mL, and prediction standard deviation were 1.00 and 1.73 mg/100 mL, respectively. The method
has been applied with satisfactory results to quick determination of caproic acid and acetic acid in the distilled spirit industry.

Key words: Fourier transform near infrared spectroscopy, liquor base, caproic acid, acetic acid, partial least squares