The
Fourier transform mid-infrared (FT-MIR) spectroscopy technique,
with multivariate calibration by the partial least-squares (PLS) regression
method, was used for real-time monitoring of the ethanol molar fraction
in the gas phase during ethanol fermentation. The PLS model was obtained
using a calibration set constructed of gas streams with different
molar fraction proportions and the corresponding FT-MIR spectra. The
molar fraction compositions were calculated by combining thermodynamic
assumptions and mass balances. The proposed model presented excellent
performance when the spectra were submitted to pretreatment using
middle center (MC), smoothed moving average (SMA), and second derivative
(secondD). The determination coefficient for the model revealed an
excellent fit to the experimental data (R
2 = 0.999). The root-mean-square error of cross-validation (RMSECV)
and the root-mean-square error of prediction (RMSEP) were less than
1.6 and 1.9% of the range of values, respectively, indicative of excellent
predictive capacity. The high values of the range error ratio (RER
= 52.52) and the ratio of the standard error of performance to the
standard deviation (RPD = 17.54) provided further evidence of the
excellent resolution of the model, which was used for carrying out
quantitative analyses of ethanol in the gas phase. A set of fermentations
performed in batch and fed-batch modes was used to evaluate the predictive
capacity of the model. The data for the molar fraction of ethanol
in the gas phase were used to estimate the ethanol and substrate concentrations
in the liquid phase. The findings demonstrated that analysis of the
gas phase using the FT-MIR/PLS technique is a reliable strategy for
real-time monitoring of ethanol fermentation under different operational
modes, without any sample manipulation.